--- title: "Steel Surface Defects and Inspection Methods" description: "Learn to identify steel surface defects and compare inspection methods against acceptance standards." canonical: https://steelequivalents.com/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods language: en robots: index, follow og_type: article type: wiki-article language_alternates: en: https://steelequivalents.com/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods.md cs: https://steelequivalents.com/cs/wiki/materiálový-list/vady-ocelového-povrchu-a-metody-kontroly.md de: https://steelequivalents.com/de/wiki/werkstoffdatenblatt-lesen/stahloberflächenfehler-und-prüfverfahren.md es: https://steelequivalents.com/es/wiki/leer-una-ficha-técnica/defectos-de-la-superficie-del-acero-y-métodos-de-inspección.md fr: https://steelequivalents.com/fr/wiki/lire-une-fiche-technique/défauts-de-surface-de-l-acier-et-méthodes-d-inspection.md hu: https://steelequivalents.com/hu/wiki/anyagadatlap/acélfelületi-hibák-és-vizsgálati-módszerek.md it: https://steelequivalents.com/it/wiki/leggere-una-scheda-tecnica/difetti-delle-superfici-in-acciaio-e-metodi-di-ispezione.md nl: https://steelequivalents.com/nl/wiki/materiaaldatablad/oppervlaktedefecten-in-staal-en-inspectiemethoden.md pl: https://steelequivalents.com/pl/wiki/karta-materiałowa/wady-powierzchni-stali-i-metody-ich-kontroli.md pt: https://steelequivalents.com/pt/wiki/ler-uma-ficha-técnica/defeitos-de-superfícies-de-aço-e-métodos-de-inspeção.md ru: https://steelequivalents.com/ru/вики/характеристики-стали/дефекты-поверхности-стали-и-методы-контроля.md tr: https://steelequivalents.com/tr/wiki/teknik-veri-sayfası/çelik-yüzey-kusurları-ve-muayene-yöntemleri.md zh: https://steelequivalents.com/zh/维基/如何读懂材料数据表/钢材表面缺陷及检验方法.md ja: https://steelequivalents.com/ja/ウィキ/鋼材データシートの読み方/鋼材表面の欠陥と検査方法.md ar: https://steelequivalents.com/ar/ويكي/قراءة-ورقة-البيانات/عيوب-سطح-الفولاذ-وطرق-الفحص.md ko: https://steelequivalents.com/ko/위키/강재-물성표-읽는-법/강재-표면-결함-및-검사-방법.md pt-br: https://steelequivalents.com/pt-br/wiki/ler-uma-ficha-técnica/defeitos-em-superfícies-de-aço-e-métodos-de-inspeção.md hi: https://steelequivalents.com/hi/विकी/डेटाशीट-कैसे-पढ़ें/स्टील-की-सतह-के-दोष-और-निरीक्षण-विधियाँ.md content_hash: sha256:9f113f7e34086f38470d86536dbb6a481e7530315144038befb00460e504bef4 generated: 2026-09-24 --- > **Machine-readable Markdown rendering.** This is a Markdown representation of a web page, > served for AI agents and crawlers as a token-efficient alternative to the full HTML. > Full interactive HTML version: https://steelequivalents.com/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods > Canonical URL: https://steelequivalents.com/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods > Available languages: en, cs, de, es, fr, hu, it, nl, pl, pt, ru, tr, zh, ja, ar, ko, pt-br, hi > Change detection: `content_hash` is the SHA-256 of this document's Markdown body — everything > below this block, whitespace-trimmed, excluding the front-matter. Compare it to detect a real > content change; `generated` moves on every render and is not a change signal. ![Steel Surface Defects and Inspection Methods](/images/uploads/4884365e-b3c4-410c-b6f6-7827362d5b68/wiki-hero-a-steel-inspection-technician-examining-a-large-rolled-steel-plate-with-a-narrow-1920x823.jpg) Reading a Datasheet # Steel Surface Defects and Inspection Methods Learn to identify steel surface defects and compare inspection methods against acceptance standards. ![Portrait of Anders Bergström, steel industry reporter](/images/uploads/a7d8e5bc-7cdf-409b-913b-211e63e8c441/anders-bergstr-m-1920x1920.jpg) **[Anders Bergström](/news/author/anders-bergstrom "Anders Bergström")** Reading a Datasheet 60+ min read Updated Aug 16, 2026 Evidence-reviewed On this pageOn this page - [What Counts as a Steel Surface Defect?](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#what-counts-as-a-steel-surface-defect "What Counts as a Steel Surface Defect?") - [How Steel Surface Defects Form During Production](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#how-steel-surface-defects-form-during-production "How Steel Surface Defects Form During Production") - [Defect Geometry: Cracks, Linear Indications, and Nonlinear Features](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#defect-geometry-cracks-linear-indications-and-nonlinear-features "Defect Geometry: Cracks, Linear Indications, and Nonlinear Features") - [Visual Testing: More Than Looking at the Part](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#visual-testing-more-than-looking-at-the-part "Visual Testing: More Than Looking at the Part") - [Visual Examination of Steel Castings Under ASTM A802-19](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#visual-examination-of-steel-castings-under-astm-a802-19 "Visual Examination of Steel Castings Under ASTM A802-19") - [Visual Testing of Welds Under ISO 17637:2016](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#visual-testing-of-welds-under-iso-17637-2016 "Visual Testing of Welds Under ISO 17637:2016") - [Magnetic-Particle Testing for Ferromagnetic Steel](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#magnetic-particle-testing-for-ferromagnetic-steel "Magnetic-Particle Testing for Ferromagnetic Steel") - [Surface Acceptance by Magnetic Particle and Liquid Penetrant Inspection](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#surface-acceptance-by-magnetic-particle-and-liquid-penetrant-inspection "Surface Acceptance by Magnetic Particle and Liquid Penetrant Inspection") - [Choosing NDT Methods for Welded Steel](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#choosing-ndt-methods-for-welded-steel "Choosing NDT Methods for Welded Steel") - [Machine Vision and Automated Optical Inspection of Steel](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#machine-vision-and-automated-optical-inspection-of-steel "Machine Vision and Automated Optical Inspection of Steel") - [Inspection Planning by Product Form and Manufacturing Stage](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#inspection-planning-by-product-form-and-manufacturing-stage "Inspection Planning by Product Form and Manufacturing Stage") - [A Practical Workflow for Evaluating a Suspected Surface Defect](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#a-practical-workflow-for-evaluating-a-suspected-surface-defect "A Practical Workflow for Evaluating a Suspected Surface Defect") - [Reporting, Traceability, and Acceptance Decisions](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#reporting-traceability-and-acceptance-decisions "Reporting, Traceability, and Acceptance Decisions") - [Common Misinterpretations and Limits of Steel Surface Inspection](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#common-misinterpretations-and-limits-of-steel-surface-inspection "Common Misinterpretations and Limits of Steel Surface Inspection") ## What Counts as a Steel Surface Defect? A dark line, shallow groove, pit, scale patch, or change in color is not automatically a steel defect. It is first a physical feature on, or close to, the surface. Its significance depends on how the feature formed, what it indicates about the material, where it occurs, and which product or inspection requirement applies. A rolled plate, a machined shaft, a weld, and a steel casting can show similar-looking marks while being governed by different terminology and acceptance limits. This distinction matters because inspection methods do not all observe the same thing. A person may see a discontinuity directly. A liquid-penetrant test produces an indication from penetrant retained in an opening. Magnetic-particle testing produces an indication where magnetic flux leaks around a surface or near-surface discontinuity. Machine vision records image features, then classifies them through rules or trained models. None of those observations is, by itself, a final decision that the steel must be rejected. #### Inspection vocabulary Imperfection A departure from the intended surface condition; it is not necessarily harmful. Discontinuity An interruption in the expected structure or surface. Indication The response produced by an inspection method. Defect An imperfection or discontinuity judged unacceptable against a defined requirement. Failure mechanism The process by which required performance is lost, such as fatigue-crack growth or corrosion penetration. ### Imperfection, indication, discontinuity, and defect ISO 8785:1998 provides terminology, definitions, parameters, permissible imperfections, and measurement methods for surface imperfections. Strong evidence ISO 8785:1998 provides a standards-based vocabulary for surface imperfections, including terms, definitions, parameters, permissible imperfections, and measurement methods. In that framework, an **imperfection** is a departure from the intended surface condition. The departure may involve geometry, such as a depression or projection; texture, such as roughness; or appearance, such as a discoloration. The word is descriptive, not necessarily condemnatory. A permitted roll mark is still an imperfection if it departs from the specified surface condition, but it may have no effect on service. A **discontinuity** is a lack of continuity or an interruption in the expected structure or surface. Examples include a crack, lap, fold, seam, cavity, pore, inclusion breakout, or lack of fusion at a weld. The term describes physical form or structure. It does not state whether the feature is harmful. A shallow surface lap in a noncritical region may be removed and accepted; a crack of similar length at a highly stressed weld toe may require rejection or repair. An **indication** is the response produced by an inspection method. The distinction is especially important in nondestructive testing. ASNT’s 2024 explanation distinguishes visual testing from ordinary naked-eye inspection: visual testing directly observes visible discontinuities, while magnetic-particle and liquid-penetrant testing observe indications produced by test media. A penetrant line is therefore an indication that may correspond to an open crack, a seam, a scratch, a machining groove, or residue trapped in surface roughness. It requires interpretation. For magnetic-particle testing, ASTM E709-21 covers dry and wet techniques for detecting cracks and other surface or near-surface discontinuities in ferromagnetic materials, including steel. The component is magnetized, and ferromagnetic particles gather at magnetic-flux leakage fields. The resulting particle pattern is an indication. Its length, width, sharpness, orientation, and location must be evaluated; particle accumulation alone does not identify the metallurgical cause. A **defect** is an imperfection or discontinuity judged unacceptable against a defined requirement. That requirement may concern dimensions, function, appearance, manufacturability, pressure tightness, fatigue performance, or safety. Calling every visible line a defect collapses observation and judgment into one word. It also hides the question that controls disposition: unacceptable according to which specification? A **failure mechanism** is different again. It describes how a component loses required performance, such as fatigue-crack growth, hydrogen-assisted cracking, stress-corrosion cracking, wear, corrosion penetration, or brittle fracture. A surface imperfection can provide a starting site for a failure mechanism, but the two are not synonyms. A mark that has no meaningful stress concentration may never initiate failure. Conversely, a small crack that is barely visible may grow under cyclic loading. ### Why appearance alone does not determine significance Surface appearance supplies evidence, not a complete diagnosis. A bright linear mark may be a harmless handling scratch, a rolled-in scale impression, a seam, or a crack. Their visual geometry can overlap. Color is even less decisive: oxide, oil, heat tint, contamination, and local roughness can produce similar shades while having very different consequences. The visible surface also reveals only a limited geometric layer. It may show an opening, ridge, depression, or stain without revealing depth, continuity beneath the surface, or connection to an internal inclusion. A grinding mark can mask a crack. Scale can conceal a shallow lap. Conversely, a rough cast surface may produce numerous visual features that do not represent harmful discontinuities. Inspection physics changes the evidence. Visual testing requires adequate illumination, access, surface cleanliness, viewing distance, and observer qualification. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and can also apply before welding. Its scope does not turn visual examination into a substitute for volumetric or surface-sensitive tests. Liquid-penetrant testing is sensitive to discontinuities that are open to the surface and can retain penetrant. It cannot reliably reveal a fully subsurface flaw, and excessive roughness may create confusing background indications. Magnetic-particle testing can detect surface and near-surface discontinuities in ferromagnetic steel, but sensitivity depends on magnetization direction, field strength, particle application, surface condition, and discontinuity orientation. A flaw parallel to the effective magnetic field may create a weak response. No unified standard exists for defining and classifying steel surface defects. Limited evidence \[1\] \[1\] **Review of Surface Defect Detection of Steel Products Based on Machine Vision**. Review article, 2023. Review of Surface Defect Detection of Steel Products Based on Machine Vision, 2023Machine vision introduces another layer of interpretation. A 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects and identifies camera-based optical inspection as a key component of machine-vision systems. The University of Oulu’s 2021 thesis describes automated optical inspection for detecting and categorizing defects on flat steel products. These systems can measure repeatable image features at production speed, but a classification label such as “scratch” or “line defect” is not automatically an engineering disposition. Lighting, scale, camera angle, surface finish, training data, and the selected classification rules affect the result. Geometry must therefore be recorded before significance is assigned. Relevant observations can include length, width, depth, spacing, orientation, density, location relative to a weld or edge, and whether the feature is linear or nonlinear. A 2 mm pit and a 2 mm crack do not present the same risk. A group of rounded indications may be treated differently from aligned linear indications, even when their total displayed area is similar. Each inspection method observes a different type of evidence. | Inspection evidence | What it can show | What it cannot establish alone | |---|---|---| | Visual testing | Visible surface conditions and profile | Subsurface soundness or hidden flaws | | Liquid-penetrant testing | Surface-breaking openings that retain penetrant | Fully subsurface discontinuities | | Magnetic-particle testing | Surface and near-surface discontinuities in ferromagnetic steel | All internal discontinuities | | Machine vision | Repeatable optical features and classifications | Metallurgical identity or automatic rejection | ### The role of specification and acceptance criteria Acceptance criteria convert an observation into a decision. They may set maximum size, allowable number, spacing, length, depth, area, location, or indication type. They may also require removal and re-examination rather than immediate rejection. Without those criteria, an inspector can describe a feature but cannot consistently determine whether the product conforms. The governing document may be a material standard, purchase specification, drawing, weld quality level, casting standard, repair procedure, or customer acceptance document. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings. ASTM [A903](/materials/material-no/1.4438 " — composition, equivalents and standards")/A903M specifies surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. Those standards demonstrate why “visible defect” is too broad: visual acceptance and particle or penetrant acceptance are related activities, but they do not use identical evidence or limits. Weld inspection adds further control. ISO 17635:2025 provides rules for selecting nondestructive testing methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. The method is selected to answer a defined inspection question. Visual testing may identify surface profile, undercut, overlap, arc strikes, or visible cracking. Magnetic-particle testing may reveal surface and near-surface discontinuities in ferromagnetic welds and heat-affected zones. Liquid penetrant testing may reveal surface-breaking openings on suitable clean surfaces. A method’s ability to find a feature does not establish the acceptance level for that feature. #### Inspection sequence Describe the feature, identify the inspection response, determine the likely discontinuity, measure its relevant parameters, and compare those parameters with the applicable requirement. The correct sequence is consequently: describe the feature, identify the inspection response, determine the likely discontinuity, measure its relevant parameters, and compare those parameters with the applicable requirement. Only then should the term **defect** be used. A surface mark is an observation. An indication is a test response. A discontinuity is a physical interruption. A defect is a nonconforming condition. Keeping those categories separate makes inspection reports clearer and prevents both needless rejection and unsafe acceptance. Surface features can originate at successive stages from casting through finishing. ## How Steel Surface Defects Form During Production A mark on steel is not automatically a defect. It may be a harmless surface imperfection, a detected indication, a defect under a specified acceptance criterion, or evidence of a failure mechanism that could reduce service performance. These terms describe different stages of judgment. ISO 8785:1998 supplies vocabulary for surface imperfections, including definitions, parameters, permissible imperfections, and measurement methods, but it does not turn every visible feature into a rejectable condition. Process origin is therefore a more useful starting point than appearance alone. A dark line may be a seam, a rolled-in oxide stringer, a crack, or dirt. A depression may be a scale impression, a pit, a gouge, or an intentional product feature. The machine-vision review *Review of Surface Defect Detection of Steel Products Based on Machine Vision* (2023) states that no unified standard exists for defining and classifying steel surface defects. The categories below are process-origin groups used to explain formation, not universal classifications shared by every producer, product standard, or inspection system. #### Casting-related origins - **Longitudinal and transverse cracks** Can form during shell solidification, withdrawal, bending, straightening, or thermal contraction. - **Seams** Linear discontinuities inherited from elongated casting defects or incompletely fused regions stretched during rolling. - **Laps** Folded or projecting metal pressed onto the surface without metallurgical joining. - **Slivers** Thin, elongated fragments or partially attached layers linked to scabs, inclusions, cracks, or oxidized folds. - **Surface depressions** May begin as oscillation marks, shell irregularities, or local sticking in the mold. ### Solidification and casting-related origins Many surface problems begin before rolling, while liquid steel is solidifying in an ingot mold or continuous-casting machine. The outer shell must form quickly enough to contain the liquid core, yet remain in contact with the mold and grow uniformly. Uneven heat transfer, mold oscillation, lubrication problems, steel level fluctuations, or inadequate control of casting speed can disturb that shell. Longitudinal or transverse cracks can form when the partly solidified strand is pulled, bent, straightened, or exposed to thermal contraction stresses. A crack at this stage may remain open at the surface or become partly welded shut by later deformation. Corner cracks are associated with corner temperature conditions and strain during withdrawal and straightening. Improper secondary cooling can increase thermal gradients and tensile stress. A seam is often a linear surface discontinuity inherited from an elongated casting defect or from an incompletely fused region that is stretched during rolling. The term describes appearance and location more than one single metallurgical cause. A lap, by contrast, forms when a projecting or folded portion of metal is pressed onto the surface without being metallurgically joined. In cast feedstock, laps may arise from uneven solidification, surface folding, or a protruding shell that is subsequently flattened. Slivers are thin, elongated fragments or partially attached layers. They can develop from rolled-out casting scabs, nonmetallic inclusions, subsurface cracks, or oxidized folds. Their appearance can overlap with seams and scratches, so the manufacturing history matters. A sliver connected to an inclusion-rich zone is not interpreted in the same way as a shallow handling scratch. Nonmetallic inclusions also influence later surface quality. Oxides, sulfides, silicates, and alumina clusters can be trapped near the strand surface. Rolling may elongate them into streaks or expose them as open defects. Inclusions do not always create a visible mark at the casting stage; their significance depends on size, morphology, position, steel grade, and the loading expected in service. Surface depressions may begin as oscillation marks, shell irregularities, or local sticking between the strand shell and mold. If the steel tears away from the mold intermittently, a depression or crack can result. Deep defects may be removed by scarfing or grinding before rolling. Shallow ones can be carried forward and widened by deformation. Reheating and rolling conditions influence both defect formation and defect geometry. | Process condition | Possible surface result | |---|---| | Poor descaling | Rolled-in oxide or scale impressions | | Excessive reheating | Increased oxidation and possible surface melting risk | | Damaged or misaligned rolls | Repeated grooves, scratches, or dimensional variation | | Uneven temperature or reduction | Crown, wedge, camber, poor flatness, or cracking | | Folded excess metal | Laps at corners, fins, or overfilled regions | ### Hot rolling, reheating, and scale-related mechanisms Reheating prepares [cast steel](/categories/cast-steel "cast steel") for hot deformation, but the furnace also creates iron oxide scale. Scale consists mainly of wüstite, magnetite, and hematite, with the proportions changing with temperature, atmosphere, and time. If scale is not removed before or between rolling passes, it can be pressed into the steel. The result is rolled-in oxide: a dark, elongated, sometimes irregular feature made from oxide trapped at or just beneath the surface. Scale impressions have a related but distinct origin. A scale fragment can imprint a depression when it is forced against the workpiece, while detached scale may leave a shallow pit after the surface is cleaned. Excessive reheating, a strongly oxidizing furnace atmosphere, poor descaling-water coverage, or delays between descaling and rolling increase the opportunity for these defects. Steel temperature also matters. If the surface is too cold, deformation may not close shallow discontinuities; if it is too hot, oxidation and surface melting risks increase. Rolling changes defect geometry. A small casting crack can lengthen in the rolling direction. An inclusion can become a thin stringer. A folded edge can be trapped as a lap, and a groove from a damaged roll can be transferred repeatedly to the strip or bar. Roll bite, reduction schedule, friction, tension, and alignment determine whether a discontinuity opens, closes, or becomes elongated. Laps commonly result when excess metal folds over during billet, bloom, bar, or section rolling. They can form at corners, fins, or overfilled regions and may remain tight enough to escape casual visual inspection. Seams may be opened by rolling and then partly closed, producing a line that resembles a crack. The distinction requires attention to the lip shape, continuity, depth, and relation to the rolling direction; a photograph alone is often insufficient. Hot tears differ from ordinary rolling marks. They are separations produced while the steel has limited ductility, often during solidification or hot working when contraction and deformation impose tensile strain on a weak or partially coherent region. Their morphology, oxidation, and location help identify when they formed. A crack exposed during hot rolling may have a bright, newly opened surface, whereas an older casting crack may contain oxide and rolled-in material. Dimensional irregularities also have process origins. Uneven roll wear, thermal expansion, incorrect pass alignment, nonuniform temperature, or unstable strip tension can produce crown, wedge, thickness variation, camber, poor flatness, or out-of-round sections. These are not merely cosmetic conditions. They may prevent downstream forming or create local stress, even when no line or pit is visible. ### Cold working, finishing, and handling damage Cold rolling and drawing impose deformation without the recovery available at hot-working temperatures. Existing laps, seams, pits, and inclusions may be sharpened or opened rather than healed. Excessive reduction, poor lubrication, misaligned rolls, or an unsuitable pass schedule can create new tears and longitudinal cracks. A tear is a separation produced when local strain exceeds the material’s ductility; it may begin at an inclusion, edge crack, surface notch, or work-hardened zone. Edge cracking is common where trimming, rolling, or tension concentrates strain at a strip edge. A brittle or heavily work-hardened edge has less capacity to accommodate further deformation. During leveling, slitting, bending, or forming, a small edge imperfection can extend into a larger tear. Residual stresses from uneven cooling and reduction can also cause waviness, coil set, or shape instability. Finishing operations remove scale and modify roughness, but they can introduce their own marks. Grinding may leave longitudinal scratches, chatter marks, local overheating, or shallow grooves. Shot blasting can create a peened texture or expose subsurface pits. Pickling can remove oxide while leaving etch pits where scale or contamination protected portions of the steel. Excessive acid attack may enlarge pre-existing discontinuities. Scratches and gouges usually come from contact with guides, rolls, conveyors, hooks, lifting devices, slings, or adjacent products. A scratch is generally narrow and shallow; a gouge removes more material and produces a sharper depression with raised lips or displaced metal. Repeated contact can create parallel lines that resemble rolling defects. Handling damage commonly occurs after the principal deformation process, so it may be limited to one face, one edge, or a local impact zone. ASTM E709-21 covers dry and wet magnetic-particle techniques for detecting cracks and other surface or near-surface discontinuities in ferromagnetic materials, including steel. Strong evidence Inspection must match the suspected origin. Visual testing directly observes visible discontinuities, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media, as ASNT explained in 2024. ASTM E709-21 covers dry and wet magnetic-particle techniques for surface and near-surface discontinuities in ferromagnetic materials, including steel; particles gather where magnetization produces magnetic-flux leakage. That method cannot be treated as a substitute for visual examination of scale impressions, dimensional variation, or general finish. For castings, ASTM A802-19 addresses visual examination and surface-acceptance standards, while ASTM A903/A903M addresses acceptance criteria for magnetic-particle or liquid-penetrant examination and evaluation of relevant linear and nonlinear indications. ISO 17637:2016 specifies visual testing for fusion-welded joints and can also apply before welding. ISO 17635:2025 sets rules for selecting weld NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. The production route creates the feature; the applicable standard determines whether that feature is acceptable. ## Defect Geometry: Cracks, Linear Indications, and Nonlinear Features Morphology guides investigation but does not by itself identify the metallurgical cause. | Morphology | Typical appearance | Examples or concerns | |---|---|---| | Linear | Length substantially greater than width | Cracks, seams, laps, folds, machining marks, or grinding grooves | | Rounded | Compact spot, oval, or short irregular patch | Pits, exposed gas pores, isolated inclusions, or corrosion cavities | | Diffuse | Indistinct boundaries over a broad area | Scale residue, staining, oxidation, texture variation, or fine discontinuity networks | ### Linear versus rounded or diffuse morphology An observed mark is not automatically a crack, and an inspection indication is not automatically a rejectable defect. Geometry is the first clue, but diagnosis requires evidence about origin, continuity, depth, material condition, and the applicable standard. A **linear indication** has a substantially greater length than width and may appear as a straight, gently curved, jagged, or branched feature. A crack is the most serious possibility because its sharp tip can concentrate stress and extend under cyclic or tensile loading. A narrow indication may also result from a seam, lap, fold, machining mark, grinding groove, or a narrow inclusion exposed at the surface. These features can look similar under illumination or after magnetic-particle or liquid-penetrant testing. A **rounded indication** has a compact shape, such as a spot, oval, or short irregular patch. Pits, gas pores exposed at the surface, isolated inclusions, and small corrosion cavities commonly produce this morphology. Rounded features still require assessment: a cluster of pits can reduce section thickness, while a cavity with a sharp internal edge may be more damaging than its apparent diameter suggests. A **diffuse feature** has indistinct boundaries or covers a broad area. It may be a change in surface finish, scale residue, staining, shallow oxidation, texture variation, or a network of very fine discontinuities. Broad finish variations often affect visual appearance without forming a structurally significant discontinuity. They can nevertheless interfere with inspection by masking a tight crack or creating background contrast that resembles an indication. ISO 8785:1998 provides terminology, definitions, parameters, permissible imperfections, and measurement methods for surface imperfections. Its vocabulary is useful because “imperfection” describes a condition or feature without deciding whether it is harmful. The later decisions—whether the feature is relevant, whether it is a defect, and whether it is acceptable—depend on the product specification and acceptance standard. ASTM A903/A903M uses the categories linear indications and nonlinear indications when establishing surface acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection. Strong evidence ASTM A903/A903M uses the terms **linear indications** and **nonlinear indications** when establishing surface acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection. That distinction is not a substitute for diagnosis. It tells the examiner how to classify the indication for evaluation. A linear pattern raises concern about a crack, seam, lap, or other elongated discontinuity; a nonlinear pattern may correspond to a pit, pore, inclusion, or clustered discontinuities. The shape is evidence to investigate, not a final material diagnosis. #### Geometry to record - **Length and width** Describe the visible footprint, without assuming it represents the full discontinuity. - **Depth** Assess how much material is interrupted and whether a sharp subsurface tip may exist. - **Orientation** Record relation to rolling direction, weld axis, principal stress direction, or component geometry. - **Branching** Note branches, intersections, and changing direction because these may indicate a crack system. - **Location** Record relation to weld toes, corners, fillets, risers, edges, or abrupt section changes. ### Length, width, depth, orientation, and branching Length and width describe only the visible footprint. Depth determines how much material has been interrupted and whether the feature has a sharp subsurface tip. A long, shallow scale mark may be less significant than a short, deep crack. Conversely, a long surface-breaking crack can provide a large path for corrosion or fatigue growth even when its opening is barely visible. An **open crack** has a visible opening and may collect dirt, oxide, penetrant, or magnetic particles. Its boundaries can often be traced by direct visual testing, although glare, roughness, paint, scale, and poor lighting may obscure part of the path. A **tight crack** has little or no visible opening. It may be invisible to ordinary visual inspection but produce a clear liquid-penetrant indication if the discontinuity reaches the surface and the surface is clean enough for penetrant entry. On ferromagnetic steel, magnetic-particle testing can reveal a tight surface or near-surface crack through magnetic-flux leakage, even where the opening is difficult to see. #### Magnetic-particle testing limit Magnetic-particle sensitivity depends on field direction. Use suitable magnetization directions, often more than one, when the expected discontinuity orientation is uncertain. Orientation matters because inspection response is directional. ASTM E709-21 covers dry and wet magnetic-particle techniques for detecting surface and near-surface discontinuities in ferromagnetic materials, including steel. Particles gather where magnetization produces a magnetic-flux leakage field, so a crack is most strongly indicated when it interrupts the magnetic field rather than running parallel to it. Examiners therefore use suitable magnetization directions, often more than one, when the expected crack orientation is uncertain. A single favorable-looking scan cannot prove that all orientations have been examined. Liquid penetrant testing has a different physical basis. The penetrant must enter a surface-breaking opening and later bleed back to form an indication. It does not reliably reveal a discontinuity that remains entirely below the surface, and rough or porous surfaces can produce excessive background. Magnetic-particle testing can respond to some near-surface features, but its response depends on steel’s ferromagnetism, field strength, field direction, particle application, and surface condition. Branching is particularly important. A single line may be a machining groove or a seam; a network with branches, changing direction, or intersecting segments is more consistent with a crack system, though shrinkage, oxide entrapment, or processing damage can produce similar patterns. The examiner should record the indication’s location, apparent length, orientation, branching, spacing, and relation to geometric stress raisers such as corners, fillets, weld toes, risers, or abrupt section changes. Depth is commonly established by follow-up examination rather than by visual shape alone. Cleaning, light grinding, local sectioning, replication, ultrasonic testing, radiography, or an approved combination may be required, depending on the component and specification. Grinding can remove a shallow surface imperfection, but it can also smear a tight discontinuity or erase evidence before its origin has been assessed. ### Why geometry influences detection and acceptance Geometry affects both mechanical risk and test sensitivity. A sharp, elongated feature concentrates stress at its ends and can propagate under fatigue or brittle fracture conditions. A rounded cavity distributes stress more gradually, although a deep pit, clustered porosity, or exposed inclusion may still be unacceptable because it removes load-bearing material or provides a corrosion site. A broad finish variation may have little structural effect but can conceal a smaller, sharper discontinuity. Inspection methods do not see “defects” in the abstract. Visual testing directly observes a visible discontinuity, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media, a distinction made by the American Society for Nondestructive Testing in 2024. Camera-based machine vision records changes in intensity, color, texture, or shape. The 2023 review *Review of Surface Defect Detection of Steel Products Based on Machine Vision* states that no unified standard exists for defining and classifying steel surface defects; a camera classifier may therefore label a feature “scratch,” “pit,” or “crack-like” without establishing metallurgical identity. Automated optical inspection, described by the University of Oulu in 2021 for flat steel products, is useful for consistent detection and categorization, but it still requires calibrated imaging, suitable lighting, and a standards-based decision rule. Acceptance is a separate step. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings, while ASTM A903/A903M addresses acceptance of relevant linear and nonlinear indications found by magnetic-particle or liquid-penetrant inspection. Neither standard makes every visible line a reject, nor does a favorable visual appearance override a relevant test indication. The governing product specification, casting class, weld quality level, examination extent, and repair provisions control the decision. For welds, ISO 17637:2016 covers visual testing of fusion-welded joints and can also apply before welding; ISO 17635:2025 links NDT method selection, technique, acceptance level, and testing extent to quality requirements, material, thickness, welding process, and other conditions. The correct sequence is therefore: describe the geometry, select a method that responds to that geometry, determine whether the indication is relevant, identify its likely source, and compare the verified condition with the applicable acceptance criterion. A line is not a verdict. Neither is a spot, stain, pit, or broad finish change. Formal visual testing depends on controlled examination conditions. | Visual-testing control | Why it matters | |---|---| | Illumination | Insufficient light can hide shallow undercut or fine cracking; glare can create false lines. | | Access | Restricted surfaces may require mirrors, cameras, or another examination method. | | Surface cleanliness | Scale, oil, paint, dirt, and corrosion products can conceal or imitate discontinuities. | | Viewing angle | A mark may appear or disappear as reflected light changes. | | Documentation | Location, dimensions, photographs, examiner, date, and acceptance basis make the result traceable. | ## Visual Testing: More Than Looking at the Part Visual testing (VT) is often reduced to a quick look at a steel surface, followed by a statement such as “no defects found.” That description is too loose for inspection work. ASM’s *Metals Handbook* identifies visual inspection as a nondestructive method for finding surface flaws, including corrosion, contamination, surface-finish problems, and visible surface discontinuities. ASNT makes a related distinction: formal visual testing directly observes visible discontinuities, whereas magnetic-particle and liquid-penetrant testing observe indications created by test media. The difference matters. A person may see a dark line, but only a defined examination can establish what was observed, where it occurred, how it was measured, and whether an applicable acceptance criterion rejects it. ISO 8785:1998 supplies vocabulary for surface imperfections, including terms, definitions, parameters, permissible imperfections, and measurement methods. That vocabulary helps prevent a visible mark from being treated automatically as a defect. A surface imperfection is a condition or departure from the intended surface; an indication is a response or observation that requires interpretation; a defect is an imperfection judged unacceptable by the governing specification. A failure mechanism, such as fatigue-crack growth or corrosion-assisted cracking, is a different conclusion again. Visual testing may identify evidence associated with a failure mechanism, but it does not establish subsurface damage merely because the surface appears unusual. #### Controlled visual-testing sequence 1. **Identify the examination area** Define the component, surface, access, and required coverage. 2. **Prepare the surface** Remove obstructions without erasing or altering the condition being assessed. 3. **Control viewing conditions** Use suitable illumination, distance, angle, and optical aids. 4. **Observe and measure** Record profile, texture, discontinuity location, dimensions, and orientation. 5. **Document the result** Preserve photographs, equipment details, examiner identity, and the acceptance basis. ### Direct observation and controlled viewing conditions Direct visual testing uses the examiner’s eyes, with or without specified optical aids, to inspect an accessible surface. It is still a controlled examination rather than casual naked-eye inspection. The procedure should identify the component, inspection area, surface condition, viewing distance and angle, lighting arrangement, equipment, examiner qualifications, recording method, and acceptance standard. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and can also be applied before welding, when joint preparation, fit-up, alignment, cleanliness, and edge condition are being checked. Lighting is not a minor convenience. Insufficient illumination can hide a shallow undercut or a fine surface-breaking crack; glare can mask a discontinuity or create a false line. The light should provide enough intensity and suitable direction to reveal changes in profile, texture, and reflectivity. Oblique lighting can help show a shallow depression, while more direct lighting may make staining or discoloration easier to assess. The examiner may need to change viewing angle because a mark visible from one direction can disappear when reflected light shifts across the surface. The part must be clean enough for the purpose of the examination. Scale, oil, welding spatter, paint, dirt, corrosion products, and handling residue can conceal the surface or imitate a discontinuity. Cleaning itself must not remove the condition being assessed. A weld may require removal of spatter before profile examination, but aggressive grinding can erase evidence of overlap, undercut, or a crack. If the specified procedure calls for magnification, a mirror, a borescope, or another viewing aid, those aids become part of the examination rather than optional additions. Magnification can improve recognition of a tight surface-breaking crack, but it does not turn an inaccessible or poorly prepared surface into a fully inspected one. Documentation gives the observation a traceable form. Useful records include the location relative to a datum or weld reference, orientation, length and width where measurable, photographs, lighting conditions, surface preparation, equipment identification, examiner, date, and disposition. A photograph alone is rarely enough: scale, location, and acceptance basis may be missing. ### What visual testing can reveal VT is well suited to conditions that reach or alter the visible surface. ASM’s examples include corrosion, contamination, surface-finish problems, and surface discontinuities. Corrosion may appear as uniform wastage, pitting, flaking scale, or localized attack. Visual examination can show staining and deposits, but it cannot by appearance alone determine the remaining wall thickness or confirm the full depth of a pit. Thickness measurement or another applicable method is needed for that question. Contamination includes oil, moisture, embedded debris, oxides, paint residue, and other substances that can interfere with welding, coating, bonding, or service. Surface-finish problems may include grinding marks, gouges, laps, roll marks, rough machining, excessive roughness, or an abrupt transition between surfaces. Their acceptability depends on the product specification and the function of the surface, not simply on whether they look untidy. On welds, visual testing can reveal an irregular profile, excessive reinforcement, insufficiently filled areas, poor alignment, crater damage, visible porosity, arc strikes, spatter, overlap, and undercut. It can also identify a crack that opens at the surface, provided the crack is visible under the examination conditions. Before welding, the same approach can expose poor fit-up, incorrect bevel geometry, a damaged edge, inadequate cleaning, or an unacceptable root gap. ISO 17637:2016 is therefore not limited to inspecting a finished weld bead. Obvious mechanical damage is another target: dents, bends, gouges, impact marks, torn edges, distortion, and damage to machined or coated areas. On steel castings, ASTM A802-19 addresses visual examination and surface-acceptance standards. For castings examined by magnetic-particle or liquid-penetrant methods, ASTM A903/A903M specifies acceptance criteria and requires evaluation of relevant linear and nonlinear indications. Those standards illustrate the essential point: observation and acceptance are connected, but they are not the same act. A visible line may be a scratch, a lap, a weld crack, a machining mark, or residue in a groove. The examiner records and evaluates the indication against the applicable requirement. If the requirement does not define the feature, the report should not silently invent a rejection limit. ### Limits imposed by access, surface condition, and human judgment Visual testing cannot reveal a defect that has no visible surface expression. It does not establish the presence or absence of a subsurface crack, internal porosity, lack of fusion hidden beneath a weld cap, or a buried inclusion. Magnetic-particle testing has a different physical basis: ASTM E709-21 covers dry and wet techniques for ferromagnetic materials, including steel, in which magnetization and particle accumulation at magnetic-flux leakage fields can disclose surface and near-surface discontinuities. Liquid penetrant testing likewise depends on a penetrant entering surface-opening flaws. Neither method is interchangeable with direct VT, and VT cannot substitute for either when the specification requires them. Access sets a hard boundary. A narrow crevice, the back side of a stiffener, an internal bore, or a surface hidden beneath insulation may need mirrors, remote cameras, borescopes, or a different examination method. Even then, the result is limited by image quality, focus, field of view, and the ability to position the equipment. A clean, brightly lit surface is not necessarily an inspected surface if the examiner could not reach or resolve it. Human judgment also introduces variation. Fatigue, unfamiliarity with the product, expectation bias, and inconsistent terminology can change how two examiners describe the same mark. Competence therefore includes more than eyesight. The examiner must understand the material, manufacturing process, likely discontinuities, procedure, measuring tools, and acceptance criteria. ISO 17635:2025 addresses selection of weld NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. That framework supports a measured decision: use VT where visible conditions are the target, and select another method when the required evidence lies below or within the surface. Machine vision can improve repeatability and coverage, especially on flat steel products; a 2021 University of Oulu thesis describes automated optical inspection for detecting and categorizing such defects. Yet a 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects. Cameras record image features. They do not remove the need for agreed terminology, validated procedures, competent interpretation, and a stated acceptance rule. ASTM A802-19 addresses observable casting-surface conditions, not every internal discontinuity. | Visual examination can reveal | Visual examination cannot establish alone | |---|---| | Surface-breaking cracks, pits, folds, laps, scabs, and open porosity | Depth of concealed discontinuities | | Adhered molding material, fins, weld repairs, and grinding damage | Internal shrinkage, inclusions, or buried porosity | | Distortion, mismatch, incomplete gate removal, and irregular finish | Reliable detection of tight, non-visible cracks | | Accessible surface contamination and coating problems | Volumetric soundness | ## Visual Examination of Steel Castings Under ASTM A802-19 ASTM A802-19 provides a practice for visual examination and surface-acceptance standards for steel castings. Strong evidence ASTM A802-19 is a standard practice for visual examination and surface-acceptance standards for steel castings. Its subject is the condition that can be observed at the casting surface, not every discontinuity that may exist below it. That boundary matters. A visible dark patch, raised area, cavity, crack-like line, or rough region is first an observed feature. It becomes an imperfection or discontinuity only when its character is identified; it becomes a rejectable defect only when it exceeds the acceptance requirements applied to that casting. The terminology should not be treated casually. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections, including ways to describe and measure them. A foundry may report a feature as a scab, inclusion, gas hole, cold shut, lap, veining, roughness, or adhering sand, but the name alone does not determine its disposition. Location, length, width, depth, spacing, orientation, and effect on the required surface condition may all matter. A failure mechanism is a separate question again: shrinkage, gas evolution, inadequate feeding, mold erosion, thermal contraction, or handling damage may explain how the feature formed, but ASTM A802-19 is concerned primarily with what is present and whether it meets the specified visual requirement. ![Close view of a steel casting surface with a crack-like line, cavity, molding residue, and grinding mark.](/images/uploads/6c1d2d28-e835-4611-bb68-5dfcea986594/wiki-inline-a-cleaned-steel-casting-surface-with-a-visible-crack-like-line-shallow-cavity-ad-1520x1920.jpg)[](/images/uploads/6c1d2d28-e835-4611-bb68-5dfcea986594/wiki-inline-a-cleaned-steel-casting-surface-with-a-visible-crack-like-line-shallow-cavity-ad-2027x2560.avif "Enlarge image — Close view of a steel casting surface with a crack-like line, cavity, molding residue, and grinding mark.")Visual examination records accessible surface conditions before acceptance is judged. ### The scope of visual surface acceptance Visual examination under ASTM A802-19 is a direct examination of accessible casting surfaces. It can reveal surface-breaking cracks, pits, folds, laps, scabs, porosity open to the surface, adhered molding material, fins, weld repairs, arc strikes, grinding damage, and irregular finish. It can also identify distortion, mismatch, incomplete removal of riser or gate metal, and conditions that make later examination unreliable. ASM’s *Metals Handbook* identifies visual inspection as a nondestructive method for detecting surface flaws such as corrosion, contamination, surface-finish problems, and surface discontinuities. That capability has limits. Visual examination does not establish the depth of a concealed discontinuity, reliably detect a tight crack that does not produce a visible contrast, or replace volumetric examination for internal shrinkage and inclusions. A smooth surface can conceal a subsurface condition. Conversely, a prominent mark can be harmless excess metal or molding texture rather than a crack. Acceptance therefore depends on the applicable casting specification, the purchaser’s requirements, the drawing, and the selected acceptance level. ASTM A802-19 supplies the practice and surface-acceptance framework; it does not give every casting the same universal rejection threshold. A valve body, pump casing, pressure-retaining component, and structural counterweight may have different permitted conditions, examination areas, or finishing requirements. The final decision must follow the governing purchase specification and its stated acceptance level. Visual acceptance also differs from acceptance under other nondestructive methods. ASNT distinguishes visual testing from ordinary naked-eye inspection because visual testing directly observes visible discontinuities, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media. Under ASTM E709-21, a ferromagnetic steel casting is magnetized and particles gather at magnetic-flux leakage fields caused by surface or near-surface discontinuities. A particle indication is not the same physical observation as a line seen directly on the metal. ASTM A903/A903M addresses surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. Those criteria must not be copied into a visual examination report unless the contract explicitly requires that method and criterion. A casting may pass visual examination yet fail penetrant testing because a tight surface-breaking crack draws penetrant and produces a relevant indication. The reverse can also occur: a rough or stained area may look objectionable while producing no relevant magnetic-particle or penetrant indication. ### Casting surface condition and preparation The examination starts with the surface actually presented to the inspector. Loose sand, scale, oil, paint, shot residue, rust, burn-on, moisture, and machining chips can mask a discontinuity or create a false visual feature. The required cleaning method depends on the casting and the specification. Brushing, washing, blasting, shot cleaning, grinding, or machining may expose the metal, but each can also alter it. Aggressive grinding can remove evidence of a shallow defect, smear metal across a crack mouth, or create new grinding checks. Blasting may close a narrow opening or obscure a shallow fold with peening and roughness. Preparation must also preserve the geometry relevant to acceptance. A fillet, rib intersection, threaded hole, boss, or internal passage may need a different viewing angle or supplemental lighting. Castings with complex geometry cannot be judged from one sweep around the outside. Corners and changes in section deserve attention because they can hide discontinuities, retain cleaning media, or produce reflections that resemble linear defects. Lighting is a measurement condition, not a cosmetic detail. The inspector needs enough illumination to see the specified surface, with shadows controlled and glare reduced. A bright reflection on a machined face can conceal a fine line; oblique light can make shallow relief visible but can also exaggerate harmless texture. The viewing distance, angle, surface contrast, and cleanliness should be suitable for the required examination. Where access is restricted, mirrors, borescopes, or cameras may extend the view, but the record should identify that indirect method and its limits. Temperature, surface wetness, and coating condition can affect what is visible. A wet surface may fill a shallow cavity and change contrast. A coating can hide the base metal unless the specification permits examination through it. If the surface cannot be adequately cleaned or viewed, the proper response is not to call it acceptable by default. The inaccessible or obscured area should be identified, and the responsible authority should determine whether additional preparation, a different examination method, or a repair is required. ### Recording and disposition of observed features A useful visual report records more than “pass” or “fail.” It should identify the casting, heat or lot when required, examination date, examiner, surface condition, lighting or viewing limitations, examined areas, and the location of each relevant feature. Location can be given by datum, drawing zone, casting feature, grid, or a marked photograph. Size and orientation should be recorded when the acceptance criteria depend on them. The report should separate observation from judgment. “Dark linear feature, 38 mm long, at the fillet near datum B” is an observation. “Rejectable crack” is a disposition that requires confirmation against the governing criterion. If the feature is uncertain, it may require cleaning, grinding for verification, dimensional measurement, or another NDT method. Magnetic-particle testing is appropriate for surface and near-surface discontinuities in ferromagnetic steel; liquid penetrant testing is suited to open-to-surface flaws on compatible, suitably prepared surfaces. Neither method automatically answers questions outside its physical range. Disposition may be acceptance, repair and re-examination, rejection, or referral for engineering review. A repair does not erase the original finding: the report should retain the feature’s location, repair method where specified, and results of the required re-examination. If visual inspection identifies a suspected crack, the casting should not be released merely because the line is short or difficult to photograph. The applicable specification, purchaser requirements, and acceptance level control the decision, while supplementary NDT supplies evidence when visual observation alone cannot characterize the feature. Visual testing of welds can be applied before, during, and after welding. | Inspection stage | Typical observations | |---|---| | Before welding | Joint preparation, fit-up, alignment, cleanliness, root gap, and edge condition | | During welding | Interpass cleaning, root-pass condition, distortion, arc strikes, and visible process problems | | After welding | Cracks, porosity, undercut, overlap, weld profile, crater areas, and dimensional conformity | | After repair | Excavation condition, repaired geometry, and required re-examination results | ## Visual Testing of Welds Under ISO 17637:2016 ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials. It also permits visual testing of joints before welding, which is important because a weld can be correctly deposited onto a wrongly prepared or badly fitted joint. Inspection is therefore not one event performed after cooling. It follows the joint through preparation, fit-up, welding, and final examination. The method is often reduced to “looking at the weld,” but that description is too narrow. ASNT distinguishes visual testing from an informal naked-eye check: visual testing directly observes visible discontinuities under defined examination conditions, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media. The distinction matters when a surface is clean and apparently sound but contains a tight crack that is not visible without penetrant or magnetic particles. ISO 17637:2016 does not turn every visible mark into a defect. ISO 8785:1998 provides terminology for surface imperfections, including definitions, parameters, permissible imperfections, and measurement methods. A surface imperfection is an observed condition. An indication is a signal or visible feature found by an examination method. It becomes a defect only when the applicable specification judges it unacceptable. A failure mechanism, such as fatigue-crack initiation, hydrogen-assisted cracking, or lack of load-bearing fusion, is a further engineering interpretation and should not be confused with the observation itself. ![Inspector checking the root gap and alignment of a prepared steel joint before welding.](/images/uploads/f3917438-ab50-49cb-8ab6-cd30177d8349/wiki-inline-a-welder-inspector-checking-a-prepared-steel-groove-joint-for-root-gap-alignment-1920x1288.jpg)[](/images/uploads/f3917438-ab50-49cb-8ab6-cd30177d8349/wiki-inline-a-welder-inspector-checking-a-prepared-steel-groove-joint-for-root-gap-alignment-1920x1288.avif "Enlarge image — Inspector checking the root gap and alignment of a prepared steel joint before welding.")Pre-weld inspection checks the joint condition before the arc is struck. ### Inspection before welding Pre-weld visual testing begins with the materials, joint faces, and preparation details. The examiner checks whether the specified parent materials and weld consumables have been identified, whether joint edges match the drawing or welding procedure, and whether surfaces are free from oil, paint, scale, moisture, loose rust, and other contamination that could affect fusion or obscure a discontinuity. Surface condition is not cosmetic. Contamination can produce porosity, unstable arc behaviour, or incomplete fusion, while excessive grinding can change the joint geometry. Joint preparation is examined for bevel angle, root face, root gap, groove cleanliness, and damage at the edges. These dimensions must be compared with the applicable drawing, welding procedure specification, or contract requirement rather than with a universal visual limit. A preparation that is acceptable for one process and thickness may be unsuitable for another. The examiner also looks for laminations, seams, gouges, and cracks extending from cut edges. A visible linear mark in the parent metal should be recorded and assessed; it should not be renamed a weld crack merely because it lies near the joint. Fit-up is the next concern. Parts should be aligned, supported, and restrained as specified, with the intended root opening and joint orientation maintained along the weld length. Excessive mismatch, angular distortion, an inconsistent gap, or abrupt transitions may produce dimensional deviation before an arc is struck. Tack welds require examination for cracking, poor fusion, excessive profile, and unsuitable termination. Defective tack welds can remain inside the joint or become initiation points during subsequent passes. Inspection at this stage also confirms access for welding and later examination. A joint may be geometrically acceptable on one side yet inaccessible for cleaning or visual observation on the other. That limitation belongs in the inspection record. It cannot be solved by declaring the unobserved surface acceptable. ### Inspection during and after welding During welding, visual testing provides process feedback rather than merely a final verdict. The examiner observes whether joint cleanliness is maintained, whether the arc is placed and manipulated as required, and whether each pass has an acceptable appearance before the next pass conceals it. Interpass cleaning is checked for slag, oxides, spatter, and trapped debris. The root pass is particularly significant: visible lack of root penetration, an irregular root bead, burn-through, or surface-breaking porosity can justify correction before additional metal is deposited. The welding sequence may reveal distortion, movement, or a widening root gap. Temporary attachments and arc strikes outside the joint are also examined. An arc strike can leave a locally hardened or cracked area, especially in susceptible steels, even when the mark is small. Spatter may be harmless residue in one application, but it can damage coatings, interfere with dimensional fit, or conceal a crack in another. Its significance comes from the specification and service conditions, not from its appearance alone. After welding, the joint is cleaned sufficiently for examination. Slag, spatter, temporary attachments, and surface contaminants must not hide the weld toes or face. The examiner observes the full accessible length, weld ends, starts and stops, crater areas, toe transitions, and adjacent parent metal. Visible cracks receive immediate attention because a crack is a planar discontinuity with a potentially severe effect, but visual detection does not establish its depth or full extent. A crack-like line may require removal, re-examination, or another NDT method. Surface-breaking porosity may appear as rounded cavities or pits. Undercut is a groove at the weld toe or edge of the parent metal, while overlap occurs when weld metal rolls onto the parent surface without proper fusion. Both are profile-related observations, but they are not interchangeable. A concave or convex weld face, excessive reinforcement, insufficient fill, an abrupt toe, or an irregular width may indicate poor profile. The applicable acceptance criterion determines whether the condition is permissible and how it is measured. Dimensional examination includes weld size, length, location, alignment, distortion, and deviation from the drawing. Visual testing can identify that a weld is misplaced or that a member has moved, but it does not by itself determine residual stress, internal lack of fusion, or metallurgical transformation. #### Weld profile terms Undercut A groove at the weld toe or edge of the parent metal. Overlap Weld metal rolls onto the parent surface without proper fusion. Porosity Rounded cavities or pits that may break through the weld surface. Arc strike A local mark outside the intended weld that may require assessment for cracking or hardening. Reinforcement Weld metal projecting above the parent surface; its acceptability depends on the applicable criterion. ### Weld-profile and surface-discontinuity observations A useful report describes what is seen, where it occurs, and how it was measured. “Bad weld” is not an examination result. A record might identify a linear surface indication at the weld toe, scattered surface-breaking pores in a specified region, undercut along a measured length, or an arc strike adjacent to the joint. Photographs, locations, dimensions, lighting conditions, and the reference requirement make the observation reproducible. Acceptance must remain tied to the governing product, fabrication, or project standard. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings, while ASTM A903/A903M addresses steel-casting surface criteria examined by magnetic-particle or liquid-penetrant methods and requires evaluation of relevant linear and nonlinear indications. Those casting standards should not be imported into a welded-joint assessment without authority from the applicable specification. Visual testing has a defined field of application. It can identify accessible surface conditions and profile deviations, but it cannot reliably reveal every tight surface crack, subsurface discontinuity, or internal volumetric flaw. ASTM E709-21 covers dry and wet magnetic-particle techniques for surface and near-surface discontinuities in ferromagnetic materials, including steel; magnetization produces leakage fields where particles accumulate. Liquid penetrant testing has a different physical basis and requires a suitable, open surface. Machine vision can improve repeatability on flat steel products, but a 2023 review states that no unified standard defines and classifies all steel surface defects. ISO 17635:2025 therefore places method selection, acceptance levels, and testing extent in relation to quality requirements, material, thickness, welding process, and other conditions. The visible mark is only the starting point. Magnetic-particle testing reveals magnetic-flux leakage, not a complete image of the discontinuity. | Feature | Magnetic-particle testing implication | |---|---| | Material | Requires ferromagnetic steel or another ferromagnetic material. | | Field direction | Discontinuities crossing the field generally produce stronger leakage responses. | | Detection range | Most sensitive to surface and near-surface discontinuities. | | Particle technique | Dry powder suits some rough or field conditions; wet particles provide a more uniform coating on suitable surfaces. | | Disposition | The particle pattern is an indication requiring relevance evaluation and comparison with acceptance criteria. | ## Magnetic-Particle Testing for Ferromagnetic Steel Magnetic-particle testing (MT) is a method for finding surface and near-surface discontinuities in ferromagnetic steel. It does not make every visible mark a defect, and it does not inspect the full volume of a component. The result is an *indication*: a visible or fluorescent accumulation of particles produced by the test. Whether that indication represents an unacceptable defect depends on its size, shape, location, orientation, and the acceptance criteria specified for the component. ASNT describes the physical sequence plainly: magnetize the ferromagnetic steel, apply ferromagnetic particles, and observe where they gather at magnetic-flux leakage fields (American Society for Nondestructive Testing, 2024). ASTM E709-21 covers dry and wet magnetic-particle techniques for detecting cracks and other surface or near-surface discontinuities in ferromagnetic materials, including steel. The method is therefore different from visual testing, which directly observes a visible condition, and from liquid-penetrant testing, which relies on a penetrant entering an open surface discontinuity. ![Schematic of magnetic flux leaking at a steel surface crack and attracting particles.](/images/uploads/5f9559b0-10e2-42c6-9dce-8f3ec4ffc047/wiki-inline-a-ferromagnetic-steel-bar-undergoing-magnetic-particle-testing-showing-magnetic-1920x1094.jpg)[](/images/uploads/5f9559b0-10e2-42c6-9dce-8f3ec4ffc047/wiki-inline-a-ferromagnetic-steel-bar-undergoing-magnetic-particle-testing-showing-magnetic-1920x1094.avif "Enlarge image — Schematic of magnetic flux leaking at a steel surface crack and attracting particles.")Magnetic particles gather where a discontinuity causes magnetic-flux leakage. ### Magnetization and magnetic-flux leakage A magnetizing field passes through the steel when current flows through the part, through a coil around it, or between an electromagnetic yoke’s poles. In an intact, uniformly magnetized region, magnetic flux travels mainly through the metal. A crack, lack of fusion, lap, seam, or other discontinuity interrupts that path. Some of the flux is forced out of the steel and back into it, producing a leakage field at or near the discontinuity. Ferromagnetic particles placed in this field are attracted to the leakage region. They form a line, cluster, or other pattern that outlines the source. The indication may be viewed under white light when visible particles are used, or under ultraviolet-A illumination when fluorescent particles are used. The test medium makes a magnetic disturbance visible; it does not expose the entire geometry of the discontinuity. The field direction is central to the result. A crack produces the strongest leakage when it lies approximately perpendicular to the magnetizing field. A discontinuity running parallel, or nearly parallel, to the field can produce weak leakage and a faint indication, even when the discontinuity is significant. This is why a single magnetizing direction cannot be assumed to find every crack. A longitudinal field is suited to discontinuities transverse to that field; a circular field produced by current passed through the component tends to reveal discontinuities running along the component. Inspectors often apply two substantially different field directions, commonly using separate shots or a multidirectional technique, when the procedure and component geometry permit it. Magnetization must also be sufficient for the material and the inspection task. Excessively weak magnetization may produce no useful leakage field. Excessive magnetization can create particle background, interfere with interpretation, or make particle removal difficult. Geometry matters as well. Sharp corners, keyways, changes in section, holes, weld toes, and contact points can distort the field and create nonrelevant indications. These patterns are not automatically cracks. They must be distinguished from relevant indications by their location, shape, repeatability, and response to changes in magnetization. Surface condition affects reliability. Oil, scale, paint, rust, weld spatter, and rough machining can prevent particles from moving freely or can create false particle accumulations. Cleaning and preparation are therefore part of the inspection, not cosmetic preliminaries. After testing, residual magnetism may need removal if it could affect machining, assembly, instrumentation, or service. MT applies to ferromagnetic steel, such as ferritic and many martensitic grades. AISI 410 stainless steel is normally ferromagnetic; annealed AISI 304 stainless steel is generally nonferromagnetic and is not a suitable material for ordinary MT. Alloy composition, heat treatment, cold work, and temperature can alter magnetic behavior, so a grade name alone does not replace verification of the actual component response. MT is not a general method for aluminum, copper, titanium, or nonferromagnetic austenitic stainless steel. ### Dry and wet particle techniques under ASTM E709-21 ASTM E709-21 describes both dry-particle and wet-particle approaches. In the dry technique, dry ferromagnetic powder is applied to the magnetized surface, commonly by a hand bulb, powder blower, or other controlled applicator. The particles move across the surface under the leakage field and collect at a discontinuity. Dry powder is useful on relatively rough surfaces and in field conditions, but wind, vibration, excess powder, and heavy surface texture can reduce sensitivity or obscure a fine pattern. Wet-particle testing suspends finely divided magnetic particles in a liquid vehicle. The suspension is applied by spray, flow, or immersion while the component is magnetized, or during magnetization according to the procedure. Wet particles generally provide a more uniform coating and are particularly effective for fine surface indications on smoother surfaces. Fluorescent wet particles are widely used where controlled dark-room or ultraviolet-light conditions are available. The liquid must wet the surface properly; contamination, poor concentration, settling, and inadequate agitation can change the result. The choice between dry and wet particles is not a simple ranking. Particle type, carrier, surface finish, lighting, component size, access, magnetizing equipment, and required sensitivity all matter. ASTM E709-21 is a practice and guide, not a universal acceptance specification. It addresses how the examination can be performed, while a product, construction, or contract standard determines what indications may be accepted. A technician must establish adequate field strength and confirm equipment performance using the checks required by the applicable procedure. The component is then magnetized in a direction that gives the expected discontinuity a useful crossing angle. Particle application, observation, removal of excess particles, and interpretation must occur in the correct sequence. If the field is removed before particles have formed an indication, or if excess particles are not cleared, the image can be misleading. ### Surface and near-surface detection limits MT is most sensitive to discontinuities that break the surface or lie very close beneath it. Surface-breaking cracks, seams, laps, and weld toe cracks commonly produce sharp, concentrated indications. A shallow subsurface discontinuity can also be detected because its leakage field reaches the surface. As depth increases, however, the leakage field weakens and spreads. The indication becomes less distinct, and small discontinuities may disappear into background particle patterns. #### Do not assign a fixed MT depth Detectability varies with discontinuity size, opening, orientation, depth, permeability, magnetizing current, field direction, surface roughness, particle properties, and viewing conditions. There is no single depth limit that applies to every steel component. Detectability changes with discontinuity size, opening, orientation, depth, steel permeability, magnetizing current, field direction, surface roughness, particle properties, and viewing conditions. A broad subsurface flaw may produce a detectable pattern while a smaller flaw at a more favorable depth does not. For that reason, claims such as “MT detects defects to a fixed depth” are misleading unless tied to a qualified procedure and reference standard. MT also cannot determine volumetric condition. It may miss internal porosity, slag inclusions, laminations, or lack of penetration that produces little surface leakage. Ultrasonic testing, radiographic testing, or another volumetric method may be needed when internal soundness is part of the requirement. MT is not a substitute for volumetric examination. Interpretation must separate terminology from disposition. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections, but an imperfection is not automatically an unacceptable defect. A particle pattern is an indication. It becomes a rejectable defect only when evaluated against the governing acceptance criteria. For steel castings, ASTM A903/A903M specifies surface-acceptance criteria for magnetic-particle or liquid-penetrant examination and requires evaluation of relevant linear and nonlinear indications. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings; visual acceptance and MT acceptance are related but not interchangeable. Weld inspection follows the same logic. ISO 17635:2025 links selection of NDT methods, techniques, acceptance levels, and testing extent to quality requirements, material, thickness, welding process, and other conditions. MT can reveal a weld toe crack that visual testing misses, but it cannot establish that the weld is free of internal discontinuities. The method answers a specific question: does magnetization create a particle indication associated with a surface or near-surface magnetic-flux leakage field? Acceptance standards answer the separate question: what may remain in service? Magnetic-particle and liquid-penetrant acceptance requires method-specific interpretation. | Method | Physical basis | Principal limitation | |---|---|---| | Magnetic-particle testing | Particles gather at magnetic-flux leakage fields in magnetized ferromagnetic steel. | Response weakens with depth and may be weak when discontinuity orientation is unfavorable. | | Liquid-penetrant testing | Penetrant enters a surface-breaking opening and developer draws it back to the surface. | Does not normally detect a fully subsurface discontinuity. | | Visual testing | Examiner directly observes an accessible visible condition. | Cannot reliably establish hidden or internal soundness. | ## Surface Acceptance by Magnetic Particle and Liquid Penetrant Inspection A visible mark is not automatically a rejectable defect, and a particle or penetrant indication is not automatically proof of a harmful discontinuity. The inspection method produces evidence; the applicable product standard then determines how that evidence is classified and accepted. This distinction matters particularly for steel castings, where casting geometry, surface condition, machining, weld repair, and local changes in permeability can all affect the indication pattern. For the casting-acceptance discussion here, ASTM A903/A903M is the controlling reference. It specifies surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. The applicable edition must be identified on the purchase specification, drawing, inspection procedure, or contract. ASTM A903/A903M should not be treated as a universal acceptance rule for every steel product, weld, forging, or fabricated component. ### Relevant linear and nonlinear indications The word *indication* describes the response produced by an examination. It does not, by itself, describe the actual discontinuity or its effect on service. A surface-breaking crack may produce a sharp linear indication, while a rounded cavity, isolated pore, or cluster of small discontinuities may produce a nonlinear indication. The indication’s appearance is important, but interpretation also depends on its location, orientation, continuity, contrast, and relationship to the test technique. A relevant indication is one judged to result from a discontinuity rather than from the component’s shape, surface roughness, magnetic condition, residual penetrant, or another examination artifact. Nonrelevant indications can arise at sharp corners, keyways, threads, abrupt section changes, boundaries between different materials, or areas where the test medium collects without a service-significant flaw. The examiner must therefore distinguish a response caused by geometry from one caused by a discontinuity. For magnetic-particle testing, orientation is especially important. A discontinuity is most readily shown when it lies approximately across the direction of the magnetic field, because the discontinuity interrupts the magnetic path and creates flux leakage. A crack parallel to the applied field may produce a weak or incomplete response. A second magnetization direction may be required when the procedure or acceptance standard demands coverage of differently oriented flaws. “Linear” and “nonlinear” are not casual descriptions of whether a mark looks long or round. ASTM A903/A903M provides the governing framework for evaluating these categories in steel castings. The examiner must use the definitions, examination conditions, recording rules, and acceptance requirements in the specified edition rather than applying a home-made length-to-width rule. No universal indication-size limit should be assumed. The permitted size, spacing, grouping, and location depend on the applicable ASTM A903/A903M edition and on any product-specific requirement invoked by the contract. Surface condition also affects the result. Scale, paint, oil, machining marks, weld spatter, and excessive roughness can mask a discontinuity or create a misleading response. Cleaning is not merely cosmetic: it allows particles or penetrant to interact with the surface in a controlled way. On a casting, a grinding mark may resemble a crack under visual examination, while a shallow open shrinkage feature may retain penetrant and generate a stronger indication. Neither should be accepted or rejected from appearance alone. ### ASTM A903/A903M and steel castings ASTM A903/A903M addresses surface examination of steel castings by magnetic-particle and liquid-penetrant methods. That scope makes it different from a method standard such as ASTM E709-21. ASTM E709-21 describes dry and wet magnetic-particle techniques for detecting cracks and other surface or near-surface discontinuities in ferromagnetic materials, including steel. It explains how the examination is performed. ASTM A903/A903M supplies casting-oriented acceptance criteria and evaluation requirements. The distinction is practical. ASTM E709-21 can establish that particles accumulate in a particular pattern under specified magnetization conditions. It does not, standing alone, establish that every such pattern is unacceptable in a casting. ASTM A903/A903M, the casting drawing, and the governing procurement specification determine what happens next. Magnetic-particle testing requires a ferromagnetic material. The steel casting is magnetized, and dry powder or a wet particle suspension is applied. Particles gather where magnetic flux leaks from the surface or near-surface region, forming an indication. The method can reveal some discontinuities below the surface, although response decreases with depth and depends on discontinuity size, orientation, magnetizing strength, and material condition. It is therefore not a general substitute for radiographic, ultrasonic, or visual examination. Liquid-penetrant testing has a different physical basis and can be applied to nonferromagnetic materials as well as steel. A penetrant must wet the surface and enter a discontinuity that is open to the surface. After excess penetrant is removed, developer draws trapped penetrant back toward the surface, creating a visible or fluorescent indication. A sealed subsurface discontinuity will not normally be detected, regardless of its size. Surface cleaning, dwell time, removal technique, developer condition, lighting, and penetrant system sensitivity all influence the result. These methods can produce different answers on the same casting. Magnetic particles may respond to a near-surface crack that is not open to the surface, whereas liquid penetrant may show a very fine open crack that has insufficient magnetic leakage for a clear particle pattern. Conversely, rough or porous cast surfaces can retain penetrant and create diffuse, difficult-to-interpret indications. Method selection must follow the material, surface state, discontinuity type, examination stage, and required acceptance level. #### Evaluate an indication before disposition 1. **Verify** Confirm that the response is genuine and not caused by geometry, roughness, residual medium, or another artifact. 2. **Classify** Record whether the relevant indication is linear, nonlinear, rounded, clustered, intermittent, or otherwise defined by the procedure. 3. **Measure** Establish dimensions, spacing, location, orientation, and other parameters required by the acceptance rule. 4. **Compare** Apply the specified edition, product requirement, drawing, or contract criterion. 5. **Disposition** Accept, repair and re-examine, reject, or refer for engineering review as authorized. ### Indication evaluation versus automatic rejection An indication begins a decision process; it does not end it. The examiner first verifies that the indication is genuine and relevant, then records its form and location, and finally compares it with the applicable acceptance criteria. Where permitted by the procedure, cleaning, light dressing, or re-examination can help determine whether a response comes from a removable surface condition or from a persistent discontinuity. Any repair and re-examination requirements must come from the governing specification, not from informal practice. Automatic rejection is therefore poor inspection practice unless the contract explicitly defines the detected condition as rejectable. A long-looking indication may be nonrelevant. A small indication may be unacceptable if it lies in a critical area or represents a crack. Clusters and aligned nonlinear indications may also be assessed differently from isolated rounded responses. ASTM A903/A903M must be consulted for the applicable category and acceptance rule; this article does not assign numerical size limits because those limits must come from the specified edition, product specification, drawing, or contract requirement. The report should preserve the chain from method to decision: examination method and technique, surface preparation, equipment or consumables, relevant indication location, classification as linear or nonlinear, disposition, and the acceptance document used. That record prevents a common error—reporting “MT failed” or “PT failed” without stating what indication was found and which requirement made it unacceptable. Visual examination remains useful before and after either method, but it is not interchangeable with them. ISO 8785:1998 provides vocabulary and parameters for surface imperfections, while ASTM A802/A802M addresses visual examination and surface-acceptance standards for steel castings. Visual testing observes visible discontinuities directly; magnetic-particle and liquid-penetrant testing observe responses produced by test media. The method, the material, and the acceptance standard must remain connected. Method selection should follow the physical inspection question and the governing quality requirement. | Inspection question | Suitable method or method group | |---|---| | Is an accessible surface profile or visible discontinuity acceptable? | Visual testing | | Is a surface or near-surface discontinuity present in ferromagnetic steel? | Magnetic-particle testing | | Does a nonporous surface contain an opening to which penetrant can gain access? | Liquid-penetrant testing | | Is internal weld soundness or volumetric condition required? | An appropriate volumetric method under the inspection plan | | Can a moving flat product be screened for repeatable optical features? | Machine vision or automated optical inspection, followed by standards-based disposition | ## Choosing NDT Methods for Welded Steel A mark on a weld is not automatically a defect, and a test indication is not automatically rejectable. A surface imperfection is a feature described by agreed terminology; an indication is the response produced by an inspection method; a defect is an imperfection judged unacceptable against a specified criterion. A failure mechanism is different again: cracking, hydrogen-assisted fracture, fatigue growth, corrosion, or brittle fracture describes how a component may lose function. Inspection planning must keep these categories separate. ISO 17635:2025 provides rules for selecting NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. Strong evidence ISO 17635:2025 provides the main framework for selecting nondestructive testing (NDT) methods, techniques, acceptance levels, and testing extent for welded joints. Its selection logic relates the examination plan to quality requirements, material, thickness, welding process, and other conditions. That makes method selection a design and manufacturing decision, not a final checklist applied after fabrication. ### Method selection under ISO 17635:2025 ISO 17635:2025 does not turn one NDT method into a universal answer. Each method responds to particular physical features and has blind spots. Visual testing can reveal an open crack, an undercut, an irregular weld profile, arc strikes, visible porosity, or poor alignment when the feature is exposed and accessible. It cannot reliably reveal a tight subsurface lack of fusion merely because the weld surface looks acceptable. Visual testing also means more than an unaided glance. ASNT distinguishes visual testing from ordinary naked-eye inspection: visual testing directly observes visible discontinuities, while magnetic-particle testing (MT) and liquid-penetrant testing (PT) observe indications produced by test media. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and may also be applied before welding. That pre-weld use matters where joint preparation, fit-up, cleanliness, and edge condition can affect the finished weld. MT is based on magnetism. The component is magnetized, and ferromagnetic particles gather where magnetic flux leaks from a discontinuity. ASTM E709-21 covers dry and wet magnetic-particle techniques for surface and near-surface discontinuities in ferromagnetic materials, including steel. MT therefore fits carbon and low-alloy steels that can be magnetized, but it is not a general solution for every alloy or every defect depth. Its response also depends on field direction: a discontinuity parallel to the magnetic field may produce a weak indication, so more than one magnetization direction may be required. PT relies on a penetrant entering a surface-breaking discontinuity and then being drawn out by a developer. It can be used on many nonporous materials, including steels, but it only addresses defects open to the surface. Paint, scale, oil, roughness, poor cleaning, or an unsuitable surface condition can mask or distort the result. PT and MT are therefore not interchangeable merely because both can reveal cracks. The wider ISO 17635:2025 decision may also require volumetric or other examinations when the concern is internal weld geometry or discontinuity content. The supplied standards establish the framework for choosing such methods, but they do not justify treating radiographic testing (RT) or ultrasonic testing (UT) as automatic substitutes for surface examinations. A sound plan states which defect population is being addressed, which method can physically respond to it, and which other method covers the remaining risk. Machine vision belongs in the same disciplined framework. A 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects and identifies camera-based optical inspection as a key component of machine-vision systems. A 2021 University of Oulu thesis describes automated optical inspection for detecting and categorizing defects on flat steel products. These systems can improve repeatability and record keeping, but an image classification is not, by itself, an acceptance decision. The defect vocabulary, lighting, calibration, resolution, surface condition, and governing acceptance standard still control what the result means. ### Material, thickness, process, and quality requirements Material is the first practical filter. MT requires ferromagnetic steel and can be affected by magnetic permeability, geometry, residual magnetism, and field strength. PT does not require magnetization, but it requires a sufficiently clean, nonporous surface and a discontinuity that reaches that surface. Visual testing requires line of sight, adequate illumination, access, and a surface condition that permits the relevant feature to be seen. Thickness changes the inspection problem. A thin weld may permit direct observation of a profile or surface-breaking crack while leaving little volume for an internal imperfection. A thick joint can conceal planar or volumetric discontinuities that surface methods cannot reach. The selection should therefore follow the likely defect geometry rather than the convenience of the test: surface methods for exposed or near-surface features, and an appropriate volumetric method when internal soundness is part of the requirement. The welding process changes the expected discontinuity population. Gas pores, slag-related features, lack of fusion, incomplete penetration, crater cracking, solidification cracking, hydrogen-assisted cracking, excessive reinforcement, and undercut do not have the same geometry or location. Joint design, welding position, heat input, shielding, consumable, preheat, interpass temperature, and restraint also affect risk. A process known to create a surface-breaking crack threat may justify VT followed by MT or PT; a process and joint configuration associated with internal lack of fusion may require a method capable of examining the weld volume. Quality requirements must be stated before test selection. ISO 17635:2025 links NDT choices to the required weld quality level and to the applicable acceptance levels. A high-consequence pressure boundary, fatigue-sensitive attachment, lifting component, and lightly loaded structural connection may not need identical examination strategies. That difference is not permission to lower a requirement informally. It means the design authority, product specification, welding standard, and inspection plan must identify the applicable quality and acceptance basis. Terminology also affects the decision. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections, including permissible imperfections and measurement methods. Without that vocabulary, an inspector may record “crack,” “line,” “mark,” or “indication” inconsistently, making comparison with the acceptance standard unreliable. ### Inspection extent and acceptance levels Inspection extent answers how much of the weld is examined: all weld length, selected portions, particular zones, or additional areas triggered by an initial finding. It is separate from method capability. A method capable of detecting a relevant indication does not prove that an unexamined length is sound. The plan should identify weld categories, locations, timing, method sequence, access restrictions, and escalation rules. VT is commonly used during fit-up, during welding where specified, and after completion. MT or PT may follow surface preparation and may be repeated after grinding or repair. If a relevant indication is found, the plan should define whether adjacent weld length, linked welds, or the full population receives additional examination. Acceptance criteria are applied to indications, not to vague visual impressions. ASTM A903/A903M, for example, specifies surface-acceptance criteria for steel castings examined by MT or PT and requires evaluation of relevant linear and nonlinear indications. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings. These casting standards should not be copied onto welded joints unless the governing specification actually invokes them. A weld can pass VT and still require MT, PT, RT, or UT because the methods address different defect populations. Conversely, a recorded MT indication may be nonrelevant, removable surface contamination, or a geometric response rather than a rejectable crack. The inspector must characterize, size, and assess the indication against the specified acceptance level. #### Coordinate complementary methods Define the failure concerns, material, weld process, thickness, access, examination extent, and acceptance criteria before selecting complementary NDT methods. The correct approach is coordinated inspection: define the failure concerns, identify the material and weld process, consider thickness and access, select complementary methods under ISO 17635:2025, set the examination extent, and name the acceptance criteria before testing begins. One test rarely answers every question. A defensible result comes from matching each method to what the weld can physically reveal. Machine vision separates optical detection from classification and acceptance. | Machine-vision stage | Output | |---|---| | Image acquisition | Pixels representing reflected or transmitted light from the surface | | Pre-processing | Corrected, normalized, or noise-reduced image | | Localization or segmentation | Candidate pixels, regions, or coordinates | | Feature extraction | Length, width, area, orientation, texture, contrast, or position | | Classification | A label such as scratch, pit, scale, inclusion, or stain | | Disposition | A standards-based decision or referral for follow-up examination | ## Machine Vision and Automated Optical Inspection of Steel Machine vision **Machine vision** A camera-based inspection system that acquires surface images and uses image-processing or classification rules to locate and categorize optical features. Machine vision gives a production line a repeatable way to record and assess steel surfaces, but it does not turn every recorded mark into a rejectable defect. A camera captures reflected or transmitted light; software then identifies patterns within the image. The result is an optical indication whose significance depends on the product specification, inspection stage, and acceptance rule. Automated optical inspection can detect and categorize defects on flat steel products. Limited evidence This distinction matters particularly for flat products. Strip, sheet, and plate can pass a camera at high speed while their broad surfaces contain scale, lubricant, roll marks, stains, scratches, pits, slivers, and other irregularities. The University of Oulu thesis (2021) describes automated optical inspection as a method for detecting and categorizing defects on flat steel products. The machine-vision review *Review of Surface Defect Detection of Steel Products Based on Machine Vision* (2023) identifies camera-based optical inspection as a key part of these systems, while also stating that no unified standard exists for defining and classifying steel surface defects. That absence of common terminology is not a minor documentation problem. A visible line may be called a scratch in one dataset, a rolled-in scale defect in another, and a surface crack in a third. Those labels imply different causes, dimensions, and acceptance decisions. ![Line-scan cameras and lights inspecting a moving flat steel strip.](/images/uploads/c640f4a7-a579-48c1-880e-a49ce4626df2/wiki-inline-a-line-scan-machine-vision-station-imaging-a-moving-flat-steel-strip-under-contr-1920x1094.jpg)[](/images/uploads/c640f4a7-a579-48c1-880e-a49ce4626df2/wiki-inline-a-line-scan-machine-vision-station-imaging-a-moving-flat-steel-strip-under-contr-1920x1094.avif "Enlarge image — Line-scan cameras and lights inspecting a moving flat steel strip.")Machine vision records optical features; acceptance still depends on defined criteria. ### Cameras, illumination, and image acquisition Image acquisition is the first technical stage, before segmentation, feature extraction, or classification. The system must form a usable image of a moving steel surface and preserve the visual contrast associated with the indication of interest. A camera with high pixel resolution cannot compensate for illumination that hides the relevant geometry. Line-scan cameras are common on continuous strip lines because they build an image one transverse line at a time as the product moves beneath the camera. Their timing must match line speed. If acquisition is too slow, the image stretches or gaps appear; if it is too fast, the same surface region may be sampled inefficiently. Area-scan cameras capture rectangular frames and may suit stationary plate, cut sections, or slower inspection zones, but a large moving surface can require several cameras and careful image stitching. Illumination determines what “visible” means. Diffuse lighting can reduce glare from polished or bright steel and reveal broad changes in reflectance. Low-angle, or dark-field, lighting makes shallow ridges, grooves, and raised particles cast stronger intensity changes. Coaxial lighting can help with relatively flat surfaces, while backlighting emphasizes silhouettes and through-thickness holes rather than surface texture. Infrared, ultraviolet, or multispectral arrangements may separate compounds or residues that appear similar under visible light, but they do not automatically establish the defect’s metallurgical cause. Steel is especially difficult to image because its reflectivity changes with finish, oxidation, roughness, curvature, and contamination. A scale patch can scatter light differently from bare metal; coolant can produce a bright streak; a shallow depression can disappear when illumination is aligned with the surface. Camera exposure, lens angle, polarization, working distance, vibration, and lens contamination therefore become part of the inspection result. Changes in surface scale or production speed can alter the apparent size and contrast of the same physical feature. Optical inspection also has a physical boundary. A camera records surface appearance, not magnetic-flux leakage or penetrant accumulation. ASTM E709-21 covers dry and wet magnetic-particle techniques for detecting surface and near-surface discontinuities in ferromagnetic materials, including steel. In that method, magnetization creates leakage fields and ferromagnetic particles gather at them. Liquid-penetrant testing similarly depends on a penetrant entering and later emerging from surface-breaking discontinuities. Machine vision may observe a crack opening, stain, or particle deposit, but it does not reproduce either test merely by producing a sharper image. The same principle applies to visual testing. ASNT distinguishes visual testing from ordinary naked-eye inspection because visual testing directly observes visible discontinuities under a defined method, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and can also be applied before welding. These standards do not make a camera image equivalent to every other nondestructive test. ### Detection versus classification An automated optical system normally performs several different operations. Image acquisition records pixels. Pre-processing may correct uneven illumination, suppress noise, or normalize brightness. Segmentation or localization identifies the pixels, regions, or coordinates that differ from the surrounding surface. Feature extraction converts those regions into measurable properties such as length, width, area, orientation, edge sharpness, texture, grey-level variation, or position across the strip. Classification then assigns a category, such as scratch, pit, scale, inclusion, or stain. Detection and classification answer different questions. Detection asks, “Where is an unusual image pattern?” Classification asks, “What label should be assigned to that pattern?” A system can locate a narrow indication accurately while assigning the wrong cause. It can also classify a large, obvious stain correctly but miss a faint crack because the image does not contain enough contrast. A reported detection result therefore should not be treated as proof that the system has identified a metallurgical failure mechanism. The distinction between an indication and a rejectable defect remains essential after classification. ISO 8785:1998 provides terms, definitions, and parameters for surface imperfections, including permissible imperfections and measurement methods. A machine-vision model may label a region “scratch,” but the relevant product standard may permit scratches below a specified depth, length, or density. Conversely, a small linear indication may require further examination if the applicable acceptance criterion treats it as significant. For steel castings, ASTM A802-19 addresses visual examination and surface-acceptance standards. ASTM A903/A903M specifies surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. Those criteria cannot be inferred from a generic image classifier. They require a defined examination method, an evaluation rule, and a product-specific decision. A useful machine-vision installation can still support these decisions. It may trigger a marker, record the location for grinding or sampling, measure the extent of a surface condition, or route suspect material to a qualified follow-up test. It should not silently replace the acceptance standard. For welds, ISO 17635:2025 sets rules for selecting nondestructive testing methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. The selected method follows the inspection question; a camera is not automatically the correct answer. ### Why datasets and terminology affect reported performance Performance figures from machine-vision studies are meaningful only when the test conditions and labels are clear. The 2023 review’s statement that no unified standard exists for defining and classifying steel surface defects explains why two systems can report different results while examining similar material. One may treat scale as a defect class; another may remove scale-related images during pre-processing. One may label every connected region separately; another may combine adjacent marks into one event. Class imbalance creates another problem. Rare defects can be the most important operationally, yet they contribute few images to a dataset. A model can appear successful by recognizing common stains and roll marks while failing to identify scarce cracks or slivers. Randomly dividing images can also place nearly identical frames from one coil in both training and test sets, making performance look better than it would on a different heat, finish, camera setup, or production campaign. Lighting, reflectivity, surface scale, and line speed change the data distribution. A model trained on matte hot-rolled strip may not transfer to bright cold-rolled sheet. A label assigned by one inspector may encode a visual impression rather than a verified cause. If the dataset does not record product grade, surface finish, thickness, camera arrangement, illumination, speed, and inspection stage, later comparisons lose useful context. For that reason, a credible evaluation should report the unit being scored: pixel, defect region, image, length of strip, or coil. It should separate localization from classification and state how borderline indications were handled. Most importantly, it should connect labels to a documented acceptance criterion rather than treating every anomaly as an automatic reject. Machine vision is a measurement and decision-support system. Its value depends as much on controlled acquisition and disciplined terminology as on the algorithm applied after the image is captured. #### Planning questions - **Visibility** Can the feature be seen or measured on the accessible surface? - **Surface opening** Does the feature break the surface? - **Magnetic response** Is it close enough to the surface to disturb a magnetic field in ferromagnetic steel? - **Geometry** Does curvature, access, coating, or a change in section prevent reliable coverage? - **Acceptance basis** Does the governing standard require this indication to be evaluated? ## Inspection Planning by Product Form and Manufacturing Stage Inspection planning starts with the product’s geometry and the manufacturing event that created its surface. A dark line on a rolled strip, a cavity on a casting, and a groove beside a weld may look similar in a photograph, but they do not pose the same inspection problem. Access, surface condition, magnetic response, lighting, coating, curvature, and the required acceptance standard all affect the result. The terminology must remain controlled. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections; it does not declare every imperfection unacceptable. An inspection method produces an observation or indication. The applicable product or fabrication standard then determines whether that indication is relevant, whether it exceeds an acceptance criterion, and whether it represents a rejectable defect. A failure mechanism is a further step: cracking, fatigue initiation, leakage, fracture, or loss of section may result from a defect, but the inspection result itself is not a failure mechanism. The physical question changes with each stage: - Can the feature be seen or measured on the accessible surface? - Does it break the surface? - Is it close enough to the surface to disturb a magnetic field? - Does the geometry prevent reliable coverage? - Does the governing standard require this indication to be evaluated? A method should be selected to answer that question, not because it is familiar or because a visible mark has already been labeled a “defect.” ### Flat products and continuous surfaces Flat products such as sheet, plate, strip, and other broad rolled surfaces present a relatively continuous inspection field. That favors optical inspection, provided the surface is accessible and the contrast between the feature and surrounding steel is sufficient. Direct visual testing can identify visible discontinuities, corrosion, contamination, surface-finish changes, stains, laps, or obvious mechanical damage. ASM’s *Metals Handbook* treats visual inspection as a nondestructive method for detecting surface flaws of this kind. Optical inspection may be performed by an inspector or by a camera-based system. Machine vision is particularly suited to continuous products because cameras can observe a moving surface repeatedly while software records position, length, width, contrast, and distribution. A 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects. That limitation matters: a camera can classify a pattern according to its training or programmed rules, but classification is not the same as acceptance. A University of Oulu thesis published in 2021 describes automated optical inspection for detecting and categorizing defects on flat steel products, yet the detected category still has to be mapped to the relevant product specification. Lighting and preparation are part of the inspection method, not secondary details. Glare can hide shallow depressions; scale can resemble a crack; oil can mask a surface opening; and a rough rolled finish can create false edges. The inspection plan should state viewing distance, illumination, scanning speed, surface cleanliness, camera position, and the minimum reportable size where those factors affect detection. Visual or optical inspection answers, “Is the feature visible?” It does not reliably answer whether a line is surface-breaking beneath scale, or whether a subsurface discontinuity lies below an apparently sound face. For ferromagnetic steel, magnetic-particle testing can address the second question when the discontinuity is at the surface or near it. ASTM E709-21 covers dry and wet magnetic-particle techniques for detecting cracks and other surface or near-surface discontinuities in ferromagnetic materials, including steel. Magnetization creates a magnetic-flux leakage field at a discontinuity, and particles accumulate at that field. The method therefore depends on suitable magnetization direction, particle visibility, surface condition, and access. It is not a replacement for optical inspection across every square metre of a moving strip, nor does it reveal every internal discontinuity. ### Steel castings and complex geometries Castings change the inspection problem through curvature, pockets, ribs, fillets, changes in section, parting lines, and difficult-to-reach surfaces. A visual examination may find a raised area, cold-shut-like line, open shrinkage cavity, scab, or machining damage, but shadows and irregular texture can obscure the feature. The inspection plan should divide the casting into accessible and restricted areas and should identify whether examination occurs in the as-cast, cleaned, ground, or machined condition. ASTM A802-19 is a standard practice for visual examination and surface-acceptance standards for steel castings. It addresses the visual assessment of casting surfaces, where the examiner must distinguish a permitted surface condition from an indication requiring further examination. A visually suspect area may then receive magnetic-particle or liquid-penetrant testing, depending on material, surface state, geometry, and the required sensitivity. Magnetic-particle testing is appropriate for ferromagnetic steel when the question concerns a surface or near-surface discontinuity. It is sensitive to cracks and other features that interrupt the magnetic field, but sensitivity depends on field orientation. A discontinuity parallel to the applied field may produce weak leakage and may be missed unless the component is magnetized in another direction. Complex geometry can also create non-relevant particle patterns at abrupt changes in section, keyways, edges, or magnetic poles. Liquid-penetrant testing asks a different question: does a discontinuity open to the surface and permit penetrant to enter? It can be applied to ferromagnetic or nonferromagnetic materials and is useful on machined areas or geometries where magnetic-particle testing is unsuitable. It cannot disclose a fully subsurface discontinuity, and excessive roughness, porosity, dirt, residual penetrant, or inadequate removal can obscure or imitate indications. Penetrant indications are produced by test media; they are not direct images of the full discontinuity. For steel castings, ASTM A903/A903M specifies surface-acceptance criteria for magnetic-particle or liquid-penetrant examination and requires evaluation of relevant linear and nonlinear indications. That wording is important. The examiner does not reject every particle accumulation or penetrant bleed-out automatically. The indication must be interpreted under the specified procedure, categorized by form and size, and compared with the acceptance requirement. The casting standard controls the decision. Machining also changes access and exposure. A discontinuity hidden beneath an allowance may become surface-breaking after material removal. Conversely, grinding can remove a shallow indication or smear metal across an opening. Inspection timing should therefore be linked to the manufacturing operation, with re-examination after significant repair or machining when the standard requires it. ### Welded components, repairs, and final examination Weld inspection must begin before welding. Joint preparation, groove angle, root face, alignment, cleanliness, tack welds, and fit-up can expose conditions that later become lack of fusion, incomplete penetration, cracking, or unacceptable weld profile. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and can also be applied to visual testing of joints before welding. Pre-weld examination is not administrative paperwork; it asks whether the joint is in a condition that permits the intended weld to be made. During welding, visual checks may identify arc strikes, excessive reinforcement, undercut, overlap, crater damage, distortion, spatter, and visible cracking. Interpass examination can reveal discontinuities before additional layers conceal them. For repairs, the excavation must be examined after removal of defective material and before deposition of replacement weld metal. If the repair area is not cleaned and inspected at that point, the process may simply bury the original problem. After welding, visual testing examines the finished surface, weld contour, transition to parent metal, end regions, and adjacent heat-affected areas. ASNT distinguishes visual testing from ordinary naked-eye inspection: visual testing directly observes visible discontinuities, while magnetic-particle and liquid-penetrant testing observe indications produced by test media. That distinction prevents a common error—calling all NDT “visual” because the final result is viewed by an examiner. Magnetic-particle testing can examine ferromagnetic welded components for surface and near-surface discontinuities, including cracks at weld toes, crater regions, and repaired areas. Liquid penetrant can examine surface-breaking flaws where magnetic-particle testing is not suitable or where the procedure specifies it. Neither method automatically establishes weld integrity throughout the thickness. Radiographic or ultrasonic methods may be needed when the physical question concerns internal lack of fusion, volumetric discontinuities, or planar flaws below the surface. ISO 17635:2025 provides rules for selecting weld NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. The final examination plan should therefore state the weld category, examination stage, coverage, method, technique, reporting threshold, and acceptance level. A visible line may be acceptable after evaluation, while a faint magnetic-particle indication may be rejectable. The difference comes from geometry, method response, and the governing criterion—not from appearance alone. A defensible evaluation separates observation, characterization, and disposition. | Step | Required action | |---|---| | Describe and preserve | Identify product form, material, stage, location, dimensions, orientation, and surface condition before altering the area. | | Select a confirmatory method | Choose visual, magnetic-particle, penetrant, optical, or another method based on suspected geometry and material. | | Characterize the result | Record method, technique, indication form, dimensions, location, orientation, and relevance. | | Compare with the governing requirement | Apply the product specification, drawing, weld code, casting standard, repair procedure, or service plan. | | Disposition and trace | Accept, reject, repair, re-examine, or refer for engineering review; preserve the complete record. | ## A Practical Workflow for Evaluating a Suspected Surface Defect A dark line, shallow groove, raised scale, or isolated pit is not automatically a crack. The visible feature is first a **surface imperfection** or an observation requiring investigation. It becomes a **detected indication** only after an examination method reveals or confirms it. It becomes a **defect** when the indication exceeds an applicable acceptance criterion. A failure mechanism is a separate conclusion about cause, such as fatigue, hydrogen-assisted cracking, lamellar tearing, corrosion, or solidification shrinkage. That sequence matters. ISO 8785:1998 provides terminology, definitions, parameters, permissible imperfections, and measurement methods for surface imperfections, but it does not make every imperfection unacceptable. Product form, material grade, manufacturing route, inspection stage, and the controlling specification determine the decision. ### Describe and preserve the condition Begin by identifying what is being examined. Record whether the item is plate, sheet, bar, tube, forging, casting, weld, or a machined component. Record the material designation and heat or batch identification where available: for example, ASTM A36 plate, ASTM A516/A516M Grade 70 pressure-vessel plate, EN 10025-2 S355JR structural steel, or a casting supplied to ASTM A216/A216M Grade WCB. The applicable standard may change with the form. A casting acceptance rule cannot automatically be applied to a rolled plate, and a weld acceptance level cannot automatically be applied to the parent metal. Establish the inspection stage. A feature found before welding may be treated differently from one found in a completed weld, after forming, after heat treatment, or during service. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and also permits visual testing before welding. For weld examination, ISO 17635:2025 links method selection, technique, acceptance level, and testing extent to quality requirements, material, thickness, welding process, and related conditions. Those conditions should be identified before selecting a test. Do not grind, wire-brush, polish, weld over, or otherwise alter the area before recording it. The action may remove the evidence needed to determine whether the mark was a crack, a laps, a handling scar, embedded scale, or a machining artifact. If cleaning is necessary, use a controlled method that removes oil, loose dirt, and corrosion products without changing the geometry. Note the cleaning method and, if practical, photograph the condition both before and after cleaning. Use a scale, dimensional reference, and consistent lighting in photographs. Mark the feature’s position from identifiable datum points, such as a plate edge, weld toe, casting datum, or component end. Record its orientation relative to the rolling direction, weld axis, principal stress direction, or casting geometry. Measure length, maximum visible width, depth where it can be measured without damage, spacing between parallel features, and the affected area. Record whether it is straight, branched, curved, intermittent, clustered, aligned with a weld, or associated with a pit or local loss of section. Appearance can guide the next step, but it does not establish identity. A sharp, continuous line with a dark opening may be surface-breaking. A smooth depression may be a rolled-in mark or local mechanical damage. A tight feature may be closed at the surface while continuing below it. Note oxidation, discoloration, raised lips, smeared metal, and any relation to scale or coating. The first technical question is therefore geometric: is the feature open to the surface, merely visible because of contrast, or potentially near-surface but closed? The second is material-dependent: is the steel ferromagnetic, and is the surface condition suitable for the intended method? ### Select a confirmatory method Visual testing is the starting point when the feature is visible and access, lighting, cleanliness, and surface condition are adequate. It is more than an unaided glance. ASNT distinguishes visual testing from ordinary naked-eye inspection because visual testing directly observes visible discontinuities under defined conditions, while magnetic-particle and liquid-penetrant examinations observe indications produced by test media. Magnification, lighting, gauges, mirrors, cameras, and documented procedures may all form part of visual testing. For welds, use the applicable visual-testing procedure and acceptance level rather than a generic shop inspection. ISO 17637:2016 addresses fusion-welded joints, including examination before welding. For steel castings, ASTM A802-19 addresses visual examination and surface-acceptance standards. A camera image can help locate and measure a feature, especially on flat products, but machine vision is not a universal classification authority. A 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects. Automated optical inspection, described in a 2021 University of Oulu thesis for detecting and categorizing defects on flat steel products, still requires a defined defect catalogue and acceptance rule. Magnetic-particle testing is appropriate when the component is ferromagnetic and the concern includes surface-breaking or near-surface discontinuities. ASTM E709-21 covers dry and wet techniques for ferromagnetic materials, including steel. The component is magnetized, and particles accumulate where magnetic-flux leakage indicates a discontinuity. The result is an indication, not a photograph of the actual crack. Particle buildup can be affected by field direction, field strength, surface roughness, coating, geometry, and adjacent magnetic fields. A single magnetization direction may miss a discontinuity oriented unfavorably to the field, so the written procedure may require multiple directions. Liquid-penetrant examination is useful for surface-breaking discontinuities in materials and geometries where the penetrant can enter the opening. It does not reliably reveal a fully subsurface feature. Surface preparation, dwell time, removal of excess penetrant, developer application, temperature, and lighting affect the indication. Porosity, roughness, smeared metal, and an open seam can produce confusing patterns. Penetrant testing also cannot determine depth simply from indication length or brightness. Select the method from the suspected geometry, not from convenience. Visual testing may establish a rolled-in mark or gouge. Magnetic-particle testing can test a ferromagnetic steel weld toe for a tight surface or near-surface discontinuity. Penetrant testing can confirm an open crack-like line on a nonmagnetic insert or a carefully prepared steel surface. If the result would affect structural disposition and the geometry remains uncertain, supplementary examination or sectioning may be required under the governing procedure. ### Compare the result with the governing requirement After examination, characterize what was found without overstating it. State the method, equipment or consumables, surface preparation, test direction, relevant indication dimensions, location, orientation, and whether the indication is linear, rounded, clustered, intermittent, or nonrelevant. “Linear indication” is an examination description. It is not, by itself, proof of a metallurgical crack. Then identify the controlling requirement. It may be the product specification, purchase or manufacturing specification, drawing, weld code, casting standard, repair procedure, or service-inspection plan. ASTM A903/A903M specifies surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. Its criteria are not automatically transferable to plate, forgings, or welds. ASTM A802-19 likewise addresses visual examination and acceptance standards for steel castings, not every steel product. Compare like with like: method against the method named by the requirement, indication type against the defined category, and measured size against the stated limit. Check whether the criterion applies to all surfaces or only specified zones, whether isolated and aligned indications are treated differently, and whether grinding or repair is permitted. Also verify the required testing extent; a satisfactory result at one location does not validate an unexamined length. Keep three decisions separate. **Diagnosis** asks what the visible feature might be. **Characterization** records what the selected examination actually demonstrates. **Disposition** decides accept, reject, repair, re-examine, or seek engineering review under the governing requirement. If the evidence supports only “linear magnetic-particle indication,” use that wording. Call it a crack only when the examination, metallurgical evidence, or an authorized engineering investigation supports that conclusion. ## Reporting, Traceability, and Acceptance Decisions An inspection report is not merely a statement that a surface “passed” or “failed.” It is the link between a physical observation, the method that produced it, and the requirement used to judge it. Without that link, a later reviewer cannot determine whether the reported feature was a surface imperfection, a test indication, an unacceptable defect, or evidence of a failure mechanism. ### What an inspection record should identify The record should first identify the item examined: component or product number, heat or cast number where applicable, drawing or weld identification, material grade, and relevant dimensions. A weld report should identify the joint, weld number, weld process if required, and inspection extent. A casting report should distinguish the casting from adjacent parts or repair welds. Traceability must remain possible after the component leaves the inspection area. Location needs equal precision. “Defect on the flange” is weak evidence. A useful description gives the face or side, datum or weld reference, distance from a known feature, orientation, and zone examined. A sketch, grid reference, coordinate, or marked-up drawing may be necessary for large plate, pressure-retaining components, and complex castings. If the observation concerns a weld, the report should state whether it lies in the weld metal, fusion boundary, heat-affected zone, or parent material when that can be established. Surface condition affects what can be seen and what a test can reveal. The report should therefore record whether the area was as-rolled, machined, ground, blasted, painted, coated, oxidized, wet, dirty, or otherwise obstructed. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and also permits visual testing before welding. That timing matters: a bevel can be acceptable before welding but conceal a preparation crack under later weld metal, while paint can prevent direct visual observation or interfere with penetrant testing. The method and technique must be stated, not implied. “NDT completed” has little technical value. The report should identify visual testing, liquid-penetrant testing, magnetic-particle testing, or machine-vision inspection, along with the applicable procedure and revision. For visual testing, record lighting or illumination where required, viewing distance, optical aids, surface preparation, and access. For magnetic-particle testing, identify wet or dry particles, visible or fluorescent media, magnetizing technique and direction, field strength or verification method where specified, and equipment identification. ASTM E709-21 covers dry and wet magnetic-particle techniques for surface and near-surface discontinuities in ferromagnetic materials, including steel; particles gather where magnetic-flux leakage marks a discontinuity. That physical response is not interchangeable with direct observation. For penetrant testing, the record should identify the penetrant system, cleaner and developer, dwell times, removal method, inspection lighting, and whether the system was visible-dye or fluorescent. For automated optical inspection, identify camera or sensor arrangement, resolution, lighting, software or model version, calibration status, and the product surface viewed. A 2021 University of Oulu thesis describes automated optical inspection for detecting and categorizing defects on flat steel products, but an automated label still requires a defined classification and acceptance rule. The examiner’s name or identification, qualification or authorization where required, inspection date and time, environmental conditions when material, and equipment or media batch should be recorded. Include calibration or function-check status. A report that names the instrument but not its status leaves uncertainty about the measurement. The same applies to a fluorescent magnetic-particle medium with no indication of its batch, concentration, or operating check when the procedure requires those controls. #### Report terminology Surface imperfection A feature or departure from ideal geometry, not automatically harmful or rejectable. Indication A response observed during a defined examination method. Relevant indication An indication judged to result from a discontinuity rather than an examination artifact. Defect A condition judged unacceptable by the specified requirement. Disposition The authorized decision to accept, reject, repair, re-examine, or seek engineering review. ### Indication terminology and photographic evidence Terminology should preserve the sequence from observation to decision. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections. An imperfection is a feature or departure from ideal geometry; it is not automatically harmful and is not automatically rejectable. An indication is the response produced or observed during a test. In magnetic-particle testing, for example, a particle accumulation is an indication. In penetrant testing, a bleed-out is an indication. In visual testing, the directly observed mark may be described without calling it a defect. A defect is an imperfection or discontinuity judged unacceptable by the specified requirement. The governing document may set limits for length, depth, spacing, area, alignment, or number of indications. ASTM A903/A903M, for example, specifies surface-acceptance criteria for steel castings examined by magnetic-particle or liquid-penetrant inspection and requires evaluation of relevant linear and nonlinear indications. ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings. The report should quote or identify the clause, table, drawing note, purchase specification, or weld quality level used. “Does not meet standard” is not a substitute for the actual criterion. Shape must be described carefully. Record whether an indication is linear, rounded, intermittent, clustered, aligned, branching, or diffuse, and state whether its classification is provisional. Give length, width, spacing, and orientation when measured; report depth only when a valid method has measured it. Do not convert a vague visual impression into a precise dimension. A photograph can support the record, but it cannot replace measurement or the acceptance clause. Photographs should show the whole component reference, the local area, and a close view with a scale, orientation marker, and indication identifier. Lighting and contrast should make the feature reproducible rather than dramatic. For fluorescent methods, retain the relevant image conditions and identify whether the photograph shows the actual indication, a cleaned surface, or a post-processing view. Digital images should preserve original files and metadata where practical. Machine-vision images also need the threshold or classification output, not only a cropped defect picture. The report must separate “indication observed” from “accepted” or “rejected.” ASNT distinguishes visual testing from ordinary naked-eye inspection: visual testing directly observes visible discontinuities, whereas magnetic-particle and liquid-penetrant testing observe indications produced by test media. A visible line and a magnetic indication may correspond to one crack, two different features, or a non-relevant response. The examiner records the response first; evaluation establishes relevance; the responsible acceptance decision follows the specified standard. ### Repair, re-examination, and unresolved findings Repair changes the evidence. Grinding can remove a surface-breaking indication, enlarge the examined area, alter local geometry, and leave scoring that resembles a new discontinuity. Weld excavation exposes fresh metal but may introduce heat-affected-zone cracking, undercut, porosity, or residual contamination. Blasting, machining, solvent cleaning, and coating can each change surface condition and test sensitivity. A repaired area therefore cannot be cleared by deleting the original report. The original indication, disposition, repair method, repaired location, and responsible authorization should remain traceable. Record the excavation dimensions when required, filler metal or repair welding details, preheat and post-weld treatment where applicable, and the inspection stage at which the repair was accepted. Re-examination must follow the applicable procedure and acceptance standard, not merely repeat the previous visual glance. A ground magnetic indication may require magnetic-particle testing in two field directions. A weld repair may require visual testing plus the radiographic, ultrasonic, magnetic-particle, or liquid-penetrant examinations specified for that joint. ISO 17635:2025 provides rules for selecting weld NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. The required extent can therefore change with the repair or with the governing weld category. #### Unresolved findings Record the reason for uncertainty, identify the inaccessible or inadequate area, and assign additional examination, engineering assessment, repair, or hold status rather than closing the report as acceptable. If an indication cannot be classified, measured, removed, or tested adequately, it is unresolved—not automatically acceptable and not automatically a confirmed defect. State the reason: inaccessible surface, inadequate cleaning, uncertain geometry, conflicting method results, or missing acceptance information. Assign an owner and disposition, such as additional examination, engineering assessment, repair, or hold. Closing the report with “no further action” when the evidence is incomplete weakens traceability and disguises an acceptance decision as an inspection result. ## Common Misinterpretations and Limits of Steel Surface Inspection Inspection results are often overstated because four different judgments are treated as interchangeable: a surface imperfection, a detected indication, a defect under a specified acceptance rule, and a failure mechanism. They are not the same. ISO 8785:1998 defines terms, definitions, and parameters for surface imperfections, but it does not turn every observed mark into a rejectable defect. A discontinuity becomes unacceptable only when the applicable product, casting, weld, or service specification defines its type, size, location, orientation, or severity as outside the permitted range. ### A visible mark is not automatically a crack A line on steel may be a crack, but it may also be a scratch, lap, seam, fold, machining mark, scale edge, coating fracture, grinding groove, or dirt trapped in a surface depression. Shape alone rarely settles the question. Width, continuity, sharpness, depth, branching, local deformation, relation to rolling or welding direction, and response to a second method all matter. Visual testing can directly observe a discontinuity that is visible under suitable lighting, magnification, surface preparation, and viewing access. That is more controlled than casually looking at the component, but it still does not identify every cause of the mark. The American Society for Nondestructive Testing distinguishes visual testing from ordinary naked-eye inspection because formal visual testing uses defined conditions and procedures, while magnetic-particle and liquid-penetrant testing observe indications produced by test media. Magnetic-flux leakage **Magnetic-flux leakage** Magnetic flux that exits and re-enters ferromagnetic material when a discontinuity interrupts the preferred magnetic path. A magnetic-particle indication is not a photograph of a crack. In accordance with ASTM E709-21, the component is magnetized and ferromagnetic particles collect where magnetic flux leaks from a surface or near-surface discontinuity. A particle pattern can therefore reveal a crack, lack of fusion, a seam, or another condition that disturbs the magnetic field. It can also be affected by field direction, particle concentration, surface roughness, geometry, residual magnetism, and inadequate magnetization. A rounded indication and a linear indication do not carry the same significance, which is why ASTM A903/A903M requires evaluation of relevant linear and nonlinear indications for steel castings examined by magnetic-particle or liquid-penetrant inspection. Liquid penetrant testing has a different physical basis. The penetrant must reach an opening at the surface, remain there for the required dwell period, and bleed back out to form an indication. It cannot reveal a completely subsurface void, an embedded inclusion, or a crack sealed by scale, paint, oil, oxide, weld spatter, or smeared metal. Cleaning may be essential, not cosmetic. A surface that looks acceptable may still prevent penetrant entry, while excessive background staining can obscure a relevant indication. The correct question is not “Does this line look like a crack?” It is “What indication was produced, by which method, under which conditions, and does the governing specification classify it as acceptable?” ASTM A802-19 addresses visual examination and surface-acceptance standards for steel castings; it does not replace ASTM E709-21 or a penetrant procedure. ISO 17637:2016 specifies visual testing of fusion-welded joints in metallic materials and can also apply before welding. Those documents serve different inspection purposes. ### A clean visual surface is not proof of internal soundness Visual inspection has a shallow information boundary: it reports what the inspector can observe at the accessible surface. It cannot establish the absence of internal porosity, inclusions, shrinkage cavities, laminations, lack of penetration, or subsurface cracking. Even a highly prepared surface may conceal a discontinuity below the inspection plane. The limitation is physical, not a matter of inspector confidence. Light does not pass through opaque steel in a way that permits ordinary visual examination to map its interior. Surface preparation can expose a defect that was previously hidden, but it cannot certify material below the exposed layer. A clean surface may indicate good finishing, effective cleaning, or an absence of visible surface-breaking conditions. It says little about internal soundness unless other examination methods support that conclusion. Magnetic-particle testing extends detection below the surface, but only within a limited near-surface range and only when the material is sufficiently ferromagnetic. ASTM E709-21 covers dry and wet magnetic-particle techniques for ferromagnetic materials, including steel. Austenitic [stainless steels](/categories/stainless-steels "stainless steels") such as many grades specified to ASTM A240/A240M are generally not suitable for conventional magnetic-particle examination because their response is not comparable to that of ferritic or martensitic steel. Even in ferromagnetic steel, detection depends on magnetization geometry. A discontinuity is most readily indicated when it lies approximately perpendicular to the magnetic flux; a crack aligned with the field may produce weak leakage and escape detection. Multi-directional magnetization or separate examinations may therefore be required. Geometry creates further blind spots. Sharp corners, threads, keyways, weld toes, changes in section, and inaccessible faces can create background indications or prevent adequate field strength. Demagnetization, field verification, particle application, lighting, and surface condition are procedural controls, not optional refinements. Acceptance remains specification-dependent. A casting standard may permit a small rounded indication while rejecting a shorter linear indication in the same region. A weld standard may impose different limits based on quality level, thickness, joint category, or service consequence. ISO 17635:2025 provides rules for selecting weld NDT methods, techniques, acceptance levels, and testing extent according to quality requirements, material, thickness, welding process, and other conditions. It does not make visual testing a substitute for radiographic, ultrasonic, magnetic-particle, or penetrant examination where those methods are required. ### A machine-vision classification is not a metallurgical diagnosis Machine vision can inspect more area, more quickly, and with greater repeatability than a person performing a basic visual survey. Automated optical inspection is used to detect and categorize defects on flat steel products, as described in a 2021 University of Oulu thesis. Cameras, illumination, image processing, and trained classification models can identify patterns such as scale patches, scratches, pits, roll marks, slivers, or stains. That classification is bounded by the imaging system and its definitions. A camera records reflected or emitted light from the visible surface; it does not directly measure crack depth, subsurface position, fracture toughness, residual stress, or metallurgical origin. A dark line may be assigned to a “crack” class because it resembles labelled examples, even when it is a groove or scale edge. Conversely, a tight crack may be missed because its contrast is low, its orientation is unfavorable, the lighting is wrong, or the training images contain too few comparable examples. The problem is not only algorithmic. A 2023 review, *Review of Surface Defect Detection of Steel Products Based on Machine Vision*, states that no unified standard exists for defining and classifying steel surface defects. One system’s “defect” may be another system’s “surface imperfection,” while labels may combine appearance, presumed cause, and acceptance status. That makes reported accuracy difficult to interpret unless the dataset, class definitions, resolution, lighting, false-negative rate, and decision threshold are stated. Machine vision should therefore be treated as an optical detection and sorting tool, not as a metallurgical diagnosis. Suspect images may require visual confirmation, surface preparation, magnetic-particle or penetrant examination, and, where internal soundness matters, an appropriate volumetric method. Final disposition must follow the governing specification. ISO 8785:1998 supplies terminology; ASTM A802-19 addresses visual acceptance of steel castings; ASTM A903/A903M addresses acceptance criteria for casting indications; ISO 17637:2016 concerns visual testing of fusion welds; ASTM E709-21 defines magnetic-particle practice; and ISO 17635:2025 supports method selection for weld NDT. None of these standards permits a clean image, a particle pattern, or a model label to replace the acceptance rule that applies to the steel product and its service. ## References 1. \[1\] Review of Surface Defect Detection of Steel Products Based on Machine Vision. Review article, 2023. [](/wiki/reading-a-datasheet/steel-surface-defects-and-inspection-methods#wiki-cite-ref-1) Review of Surface Defect Detection of Steel Products Based on Machine Vision, 2023 **At a glance** Core distinction A surface mark is not automatically a defect Key inspection responses Visual observation, penetrant indication, magnetic-particle indication, and image feature Final disposition Determined by the applicable specification and acceptance criteria [Back to Reading a Datasheet](/wiki/reading-a-datasheet "Reading a Datasheet — Steel Wiki")