Failure mechanism · AISI 347 / UNS S34700

Knife-Line Attack in AISI 347 Welds: Why “Sensitization-Immune” Steel Still Fails at the Fusion Line

Niobium stabilization ends weld decay. It does not end every form of intergranular corrosion. A band roughly 0.1–1 mm wide, pressed flat against the weld, survives one thermal cycle and is destroyed by the next — and standard acceptance testing will happily pass the material that contains it.

Jiangsu Liangyi · Engineering Department Published 28 July 2026 Last reviewed 28 July 2026 13 min read Applies to 347 · 347H · 348 · 321 · 1.4550

Quick answer

Knife-line attack is intergranular corrosion confined to a band roughly 0.1–1.0 mm wide immediately adjacent to the fusion line of a weld in stabilized austenitic stainless steel such as AISI 347 or AISI 321. It requires two thermal cycles: welding first dissolves niobium carbide above about 1250 °C (2280 °F), then a later exposure to 425–815 °C (800–1500 °F) allows chromium carbide to precipitate in that narrow supersaturated band.

The remedy is a post-weld stabilizing anneal at 870–900 °C (1600–1650 °F) for 1–4 hours, air cooled. Detection requires a welded coupon tested to ASTM A262 Practice C; a mill test certificate cannot predict it.

Key takeaways

Peak temperature versus distance from the fusion line in an AISI 347 weld Peak welding temperature falls from about 1400 degrees Celsius at the fusion line to under 300 degrees twelve millimetres away. The knife-line attack band sits from 0 to about 0.5 millimetres where niobium carbide dissolved. The weld decay band for unstabilized 304 sits between roughly 2 and 7 millimetres, inside the shaded 425 to 815 degree sensitization window. Sensitization window 425–815 °C NbC dissolves above 1250 °C WELD Weld decay — grade 304 band 2–7 mm wide Knife-line attack 0.1–1.0 mm · grade 347 / 321 1400 1250 815 425 100 Peak temperature, °C 0 2 4 6 8 10 12 Distance from fusion line, mm NbC fully dissolved — sensitizes only if reheated Cr₂₃C₆ forms in unstabilized grades
Peak temperature versus distance from the fusion line in an AISI 347 weld Same data as the desktop chart, drawn for a narrow screen. Peak temperature falls from about 1400 degrees Celsius at the fusion line to under 300 degrees twelve millimetres away. The knife-line attack band occupies the first half millimetre. The weld decay band for unstabilized 304 lies between about 2 and 6 millimetres inside the 425 to 815 degree sensitization window. Sensitization 425–815 °C NbC dissolves >1250 °C weld Weld decay grade 304 · 2–7 mm Knife-line attack 0.1–1.0 mm 1400 1250 815 425 100 Peak temp, °C 0 2 4 6 8 10 12 Distance from fusion line, mm NbC dissolved — sensitizes if reheated Cr₂₃C₆ forms in unstabilized grades

Scroll the chart sideways to see the full width →

Figure 1. The same weld, two different failure modes, two different places. Weld decay needs the metal to sit in the 425–815 °C window. Knife-line attack needs the metal to shoot straight past that window to above 1250 °C, then come back to it later. Distances are typical for arc welding on 15–25 mm section; actual widths scale with heat input.
Affected band
0.1–1.0 mmAdjacent to the fusion line
Trigger, step 1
>1250 °C2280 °F — NbC dissolves
Trigger, step 2
425–815 °C800–1500 °F — Cr₂₃C₆ forms
Remedy
870–900 °C1600–1650 °F, 1–4 h, air cool

The claim, and its one exception #

AISI 347 is specified in place of AISI 304 because niobium prevents chromium carbide from forming, so the heat-affected zone does not sensitize and the assembly can enter service as-welded. That claim is correct everywhere on the weld except a band 0.1–1 mm wide against the fusion line.

Every datasheet for AISI 347 (UNS S34700) forged parts says the same thing: niobium ties up the carbon, chromium stays in solution, the heat-affected zone never sensitizes, and no post-weld heat treatment is needed. All of that is true, and it is why 347 appears in refinery, nuclear and API specifications instead of 304.

The exception is narrow — literally. In a band you could cover with a pencil line, the stabilization can be undone during welding and then exploited by a later heating cycle. The result is intergranular corrosion in a strip so thin that visual inspection reads it as a scratch, and so close to the weld that people blame the filler metal. Metallurgists have called it knife-line attack since the 1950s, and it remains the most misunderstood limitation of stabilized austenitic stainless steel.

What knife-line attack actually is #

Knife-line attack (KLA) is intergranular corrosion confined to a band approximately 0.1–1.0 mm wide immediately adjacent to the fusion line of a weld in a stabilized austenitic stainless steel — AISI 347 and 348 (niobium-stabilized) or AISI 321 (titanium-stabilized). It is produced by a two-step thermal history. Neither step alone causes it.

Definition

Step 1: welding dissolves niobium carbide above about 1250 °C and leaves a supersaturated band.

Step 2: a later reheating into 425–815 °C allows chromium carbide to precipitate and deplete chromium at grain boundaries in that same band. Neither step alone causes KLA.

Nothing on a mill test certificate predicts knife-line attack, because it is created after the forging leaves the mill — during fabrication. The material is fully compliant: chemistry in order, Nb at or above 10×C, solution annealed, mechanical properties passed.

The two-step thermal history #

Step 1 dissolves niobium carbide above about 1250 °C (2280 °F) and leaves a supersaturated band. Step 2 reheats that band into 425–815 °C (800–1500 °F), where chromium carbide precipitates and strips chromium from the grain boundaries.

Step 1 — on heating

The stabilizer is dissolved

Metal within roughly a millimetre of the fusion boundary reaches 1250–1400 °C. Niobium carbide begins dissolving above about 1200 °C and is substantially back in solution above 1250 °C. Cooling from a normal arc weld is fast enough that neither NbC nor Cr₂₃C₆ has time to form. The band solidifies as a supersaturated solid solution — high in dissolved carbon, with the niobium sitting uselessly in the matrix.

Result: unstabilized, but not yet sensitized
Step 2 — on reheating

Chromium carbide wins the race

The same band is later taken back to 425–815 °C by the next weld pass, a repair, a post-weld heat treatment on an attached component, or the service temperature itself. Chromium carbide now precipitates on the grain boundaries, because at these temperatures Cr₂₃C₆ nucleates far faster than NbC. Chromium is stripped from the boundary regions below roughly 10.5 % and the band becomes a continuous corrosion path.

Result: a sensitized strip 0.1–1.0 mm wide

Why kinetics beat thermodynamics here #

Niobium carbide is thermodynamically favoured over chromium carbide, but at 650–750 °C it precipitates far too slowly to compete. Chromium carbide forms first simply because niobium diffuses slowly in austenite and NbC needs roughly 870–950 °C to precipitate at a useful rate.

The usual explanation for stabilization — niobium bonds carbon more strongly than chromium does — is a statement about thermodynamics, and it is correct. Knife-line attack is a statement about kinetics, and it points the other way. At 650–750 °C, the nose of the chromium-carbide precipitation curve, Cr₂₃C₆ can nucleate and grow in minutes, while NbC is too sluggish to compete.

The practical consequence: stabilization is not a property the steel carries permanently. It is a property of the carbide distribution, and that distribution can be erased by heat and must be deliberately restored.

Knife-line attack vs weld decay #

They are different defects. Knife-line attack is a 0.1–1 mm band touching the fusion line in stabilized grades after two thermal cycles. Weld decay is a 2–6 mm band located 2–7 mm away in unstabilized grades after one thermal cycle.

Table 1 — Differentiating the two intergranular mechanisms on a stainless weld
CharacteristicKnife-line attackWeld decay
Grades affectedStabilized only — 347, 348, 321Unstabilized — 304, 316, 309, 310
LocationTouching the fusion line2–7 mm away, in the outer HAZ
Band width0.1–1.0 mm2–6 mm typically
Peak temperature reachedAbove 1250 °C (2280 °F)425–815 °C (800–1500 °F)
Thermal cycles requiredTwo — dissolve, then reheatOne
Present as-welded?No — needs the second cycleYes
Fixed by low carbon (L grade)?Partly — reduces severityYes, the standard remedy
Fixed by stabilizing anneal?Yes — 870–900 °CNo, needs full solution anneal

Where it actually bites — and where it does not #

Knife-line attack only progresses in strongly oxidizing electrolytes. Hot nitric acid and mixed acids are high risk. Steam, feedwater, flue gas and most hydrocarbon streams are low risk, and no stabilizing anneal is normally specified for them.

Table 2 — Service environments ranked by knife-line attack risk
EnvironmentRiskComment
Hot concentrated nitric acidHighThe classic KLA environment. Strongly oxidizing conditions drive rapid dissolution of chromium-depleted boundaries.
Mixed acids, nitric–hydrofluoricHighCommon in pickling lines and reprocessing plant. A stabilizing anneal is normally mandatory.
Urea and nitrate process streamsModerate–highDepends strongly on oxygen content and temperature. Qualify with a Practice C test.
Polythionic acid during shutdownModerate347 resists PTA-SCC because it does not sensitize — but a knife-line band is sensitized, so the immunity is locally lost.
Hot caustic, high-purity waterLowNot aggressive to depleted boundaries at normal operating chemistry.
Steam, feedwater, flue gasLowBand is metallurgically present but electrochemically inert. No anneal specified.
Hydrocarbon process, most refinery dutyLowStandard practice is as-welded. Reserve the anneal for oxidizing-acid exposure.

The refinery nuance worth knowing

Grade 347 is specified for reformers and hydrotreaters precisely because it cannot sensitize, and therefore cannot suffer polythionic acid stress corrosion cracking during shutdown. A knife-line band is the one place on the component where that logic no longer holds. It is a small exposure — a fraction of a millimetre — but on a component whose entire justification is PTA-SCC immunity, some owners choose to close it on the most critical welds. That is a risk-tolerance decision, not a metallurgical requirement.

Four fabrication sequences that create it #

Step 1 happens on essentially every arc weld in 347. What separates a healthy weld from a knife-line failure is whether step 2 follows — through multipass welding, a dissimilar-joint PWHT, a repair, or the service temperature itself.

How to detect it before it detects you #

Test a welded coupon that has carried the full fabrication thermal history, not a mill coupon. ASTM A262 Practice C is the practice most likely to reveal knife-line attack. Practice E is a faster pass/fail screen. Practice A is an etch classification only and is not an acceptance test.

Table 3 — Test methods, and what each one is actually good for
MethodMedium and principleSuitability for KLA
ASTM A262 Practice C (Huey)Boiling 65 % nitric acid, five 48-hour periods, mass-loss measurementBest. The oxidizing medium mirrors real KLA service, and the multi-period format exposes the narrow band clearly.
ASTM A262 Practice E (Strauss)Boiling copper–copper sulfate–16 % sulfuric acid, followed by a bend testGood, fast pass/fail screen. A knife-line band shows as a fine continuous fissure parallel to the weld after bending.
ASTM A262 Practice B (Streicher)Ferric sulfate–50 % sulfuric acid, mass lossUseful quantitative screen; less specific to the narrow band than Practice C.
ASTM A262 Practice AOxalic acid electrolytic etch, with the resulting structure classified by microscopyScreening only. A dual or ditch structure justifies further testing; not an acceptance test on its own.
ASTM A262 Practice FCopper–copper sulfate–50 % sulfuric acidWritten for molybdenum-bearing grades. Not applicable to 347, which contains no molybdenum.
DL-EPRDouble-loop electrochemical potentiokinetic reactivationNon-destructive on small areas and quantitative, but needs careful surface preparation and skilled interpretation.
MetallographyCross-section, etch, 200–500×Confirms the mechanism and measures band width once attack is suspected. Not a screening method.

Two details that catch people out

Practice D was withdrawn. ASTM removed Practice D from A262 years ago. Specifications that still call for “A262 Practice D” cite a withdrawn method and will generate a technical query at order review.

These tests are susceptibility screens, not service simulations. A pass tells you the microstructure is sound; it does not predict a corrosion rate in your actual process stream.

Whichever practice is chosen, the coupon must carry the full simulated thermal history: the production weld procedure, the actual number of passes, and any PWHT the assembly will see. Testing an as-welded coupon when the real component gets a 700 °C stress relief afterwards proves nothing.

The remedy: a stabilizing anneal, not a solution anneal #

Apply a post-weld stabilizing anneal at 870–900 °C (1600–1650 °F), held 1–4 hours by section thickness and air cooled. This range is above the stability field of chromium carbide and inside the range where niobium carbide precipitates at a useful rate.

Table 4 — Post-weld heat treatment options for grade 347
TreatmentParametersNotes
Stabilizing anneal (preferred)870–900 °C (1600–1650 °F) · 1–4 h by section thickness · air coolRe-precipitates NbC. Air cooling is preferred — the point is to allow carbide formation, not suppress it. Minimal distortion.
Full solution anneal1010–1065 °C (1850–1950 °F) · water quench within 3 minAlso effective, and standard for supplied forgings. Rarely practical on a welded assembly due to distortion and quench access on large fabrications.
No treatmentAs-weldedCorrect for the great majority of applications. See Table 2 before deciding.

Practical cautions

A specification trap on heavy 347 forgings

ASTM A182/A182M limits products made to that specification to a maximum weight of 4,540 kg (10,000 lb). Buyers often write “ASTM A182 Gr. F347” on forgings that finish heavier than this. For austenitic forgings above the A182 weight limit, the correct companion specification is ASTM A965/A965M. Getting this right at the RFQ stage avoids a certification dispute at final inspection.

Decision matrix: do you need a stabilizing anneal? #

Specify the anneal for hot oxidizing acid service, for dissimilar joints that receive a 690–720 °C PWHT, and where an owner requires full PTA-SCC immunity. Skip it for steam, hydrocarbon and ambient non-oxidizing service.

Table 5 — Specify a post-weld stabilizing anneal when the answer is yes
ConditionAnneal required?
Service in hot oxidizing acid (nitric, mixed acid)Yes — specify without exception
Dissimilar joint to low-alloy steel requiring PWHT at 690–720 °CYes
Continuous service between 425 and 815 °C in a corrosive mediumYes, if the medium is oxidizing
Nuclear or safety-critical component requiring PTA-SCC immunity across 100 % of the surfaceOwner’s call — commonly yes
Multipass welds in hydrocarbon or steam serviceNo
Ambient or low-temperature service, non-oxidizing mediumNo

Wording you can put straight into a purchase order #

Most knife-line disputes are contractual before they are metallurgical. This clause closes the gaps by requiring the full weld thermal history at quote stage, the stabilizing anneal before delivery, and a welded corrosion test coupon rather than a generic mill coupon.

Suggested PO clause — grade 347 welded assemblies, oxidizing service
MATERIAL: AISI 347 / UNS S34700 to ASTM A182 Gr. F347 (or ASTM
A965 Gr. F347 if the forging exceeds the 4540 kg A182 weight
limit), supplied solution annealed 1010-1065 C and water quenched.

WELD PROCEDURE: ER347 / E347 matching filler. Maximum interpass
temperature 150 C. Full weld thermal history, including number of
passes and any PWHT of attached components, to be submitted for
review prior to production welding.

POST-WELD HEAT TREATMENT: Stabilizing anneal 885 C +/- 15 C,
minimum 2 hours at temperature (or 1 hour per 25 mm of thickness,
whichever is greater), air cool. All welding and repairs to be
complete before the anneal. Furnace survey and calibrated part
thermocouple records required.

CORROSION QUALIFICATION: One production-equivalent weld coupon
carrying the complete simulated thermal history shall be tested
to ASTM A262 Practice C. Acceptance: corrosion rate not exceeding
the project limit in any of the five test periods, with no
increasing trend between periods. Practice E bend test may be
substituted with purchaser approval.

DOCUMENTATION: EN 10204 Type 3.2 certificate covering chemistry
(including Nb/C ratio), mechanical properties, heat treatment
records and corrosion test results.

Getting this specified correctly

If you are working through a specification and want a second opinion on whether a stabilizing anneal is justified for your service conditions, send us the operating temperature, medium and weld sequence. ASTM A262 Practice C and Practice E testing can be arranged on coupons carrying your simulated weld thermal history rather than a generic mill sample. Technical enquiries are normally answered within one working day.

Grade chemistry, mechanical and elevated-temperature data, 347 versus 347H, size ranges and the applicable ASTM and ASME standards are all set out on our AISI 347 and 347H forged parts specification.

Ask our engineering team

Inquiry Email: sales@jnmtforgedparts.com
Phone / WhatsApp: +86-13585067993
Website: www.jnmtforgedparts.com
Address: Chengchang Industry Park, Jiangyin City, Jiangsu Province, China

Frequently asked questions #

What is knife-line attack in AISI 347?

Knife-line attack is intergranular corrosion confined to a band roughly 0.1–1.0 mm wide immediately adjacent to the fusion line of a weld in stabilized austenitic stainless steel such as AISI 347 or 321. It occurs when weld heat dissolves niobium carbide above about 1250 °C and a later thermal cycle in the 425–815 °C range allows chromium carbide to precipitate in that narrow supersaturated band.

Is knife-line attack the same as weld decay?

No. Weld decay is a band 2–6 mm wide located several millimetres from the fusion line in unstabilized grades such as 304 and 316, after a single thermal cycle. Knife-line attack is a band 0.1–1 mm wide pressed directly against the fusion line, occurs only in stabilized grades such as 347 and 321, and requires two thermal cycles. See Table 1 for a full comparison.

Does AISI 347 suffer knife-line attack in the as-welded condition?

Generally no. A single-pass weld that is never reheated leaves the near-fusion band supersaturated in carbon but free of chromium carbide. Knife-line attack needs a second heating cycle into roughly 425–815 °C: a following weld pass, a repair, a post-weld heat treatment on an attached low-alloy component, or high-temperature service.

Is AISI 321 more susceptible to knife-line attack than AISI 347?

In practice yes. Titanium oxidizes and forms nitrides readily in the arc, so some stabilizing element is lost from the weld pool before it can do any work. Niobium is not lost the same way. Both grades share the same mechanism, but 347 starts each thermal cycle with more of its stabilizer intact — which is why 347, not 321, is the default for as-welded high-temperature service.

What stabilizing anneal removes knife-line attack in 347?

A stabilizing anneal at 870–900 °C (1600–1650 °F), held 1–4 hours depending on section thickness and followed by air cooling. Niobium carbide precipitates readily at this temperature while chromium carbide is not stable, so carbon dissolved during welding is re-fixed as NbC throughout the affected band. A full solution anneal at 1010–1065 °C with water quench also works but is usually impractical on a welded assembly.

Will specifying a low-carbon 347 prevent knife-line attack?

It reduces severity rather than eliminating the mechanism. Less dissolved carbon means less chromium carbide can precipitate during step 2. But 347H deliberately raises carbon to 0.04–0.10 % for creep strength, so you cannot simply drop the carbon if the design is creep-limited. The stabilizing anneal is the correct control.

Which ASTM A262 practice detects knife-line attack?

Practice C, the boiling 65 % nitric acid test, is the most reliable. Practice E, the copper–copper sulfate–sulfuric acid bend test, is a faster screen that usually shows the band as a fine fissure line parallel to the weld. Practice A is an etch screening test only and should never be used as an acceptance criterion. Practice F is written for molybdenum-bearing grades and does not apply to 347.

Do I need to worry about this in a refinery or steam application?

Usually not. Knife-line attack needs a strongly oxidizing electrolyte to progress. In steam, hydrocarbon, feedwater and most refinery process streams the band is metallurgically present but electrochemically inactive, and no stabilizing anneal is specified. Table 2 sets out the full boundary.

Glossary #

Table 6 — Terms and abbreviations used in this article
TermMeaning
KLAKnife-line attack
SensitizationPrecipitation of chromium carbide at grain boundaries between 425 and 815 °C, locally depleting chromium below passivity level
Weld decayIntergranular corrosion in a 2–6 mm band in the outer HAZ of unstabilized grades after one thermal cycle
Stabilizing annealHeat treatment at 870–900 °C followed by air cooling, to re-precipitate NbC or TiC
Solution annealHeat treatment at 1010–1065 °C followed by water quench, to dissolve carbides and hold them in solution
NbC / Cr₂₃C₆Niobium carbide (the stabilizing precipitate) and chromium carbide (the damaging precipitate)
HAZHeat-affected zone
PWHTPost-weld heat treatment
PTA-SCCPolythionic acid stress corrosion cracking
DL-EPRDouble-loop electrochemical potentiokinetic reactivation

Standards and references #

  1. ASTM A262-15(2021), Standard Practices for Detecting Susceptibility to Intergranular Attack in Austenitic Stainless Steels. ASTM International. store.astm.org
  2. ASTM International, Committee A01 on Steel, Stainless Steel and Related Alloys — background on A262 revision and Subcommittee A01.14. astm.org
  3. ASTM A182/A182M, Standard Specification for Forged or Rolled Alloy and Stainless Steel Pipe Flanges, Forged Fittings, and Valves and Parts for High-Temperature Service — Grade F347, including the 4,540 kg weight limit. astm.org
  4. ASTM A965/A965M, Standard Specification for Steel Forgings, Austenitic, for Pressure and High Temperature Parts — companion specification for austenitic forgings above the A182 weight limit.
  5. ASTM A336/A336M, Standard Specification for Steel Forgings, Alloy, for Pressure and High-Temperature Parts.
  6. AWS A5.4 / A5.9 — filler metal specifications covering E347 and ER347 matching consumables.
  7. ASME Boiler and Pressure Vessel Code, Section VIII and Section IX — construction and welding qualification requirements.

Standards are cited by designation for identification only. Always work from the current edition issued by the relevant standards body.

About this article

Written and technically reviewed by the Engineering Department of Jiangsu Liangyi Co., Limited, an ISO 9001:2015 certified open-die forging manufacturer established in 1997 in Jiangyin, Jiangsu, China. We produce forgings from 30 kg to 30,000 kg across a full range of stainless, alloy and carbon steel grades, supplying customers in more than 50 countries. Third-party inspection by any customer-nominated body is welcome.

Published 28 July 2026. Last technically reviewed 28 July 2026. Corrections and technical challenges are welcome at sales@jnmtforgedparts.com.

Cite this article Jiangsu Liangyi Co., Limited Engineering Department (2026). Knife-Line Attack in AISI 347 Welds: Why “Sensitization-Immune” Steel Still Fails at the Fusion Line. https://www.jnmtforgedparts.com/blog/knife-line-attack-in-aisi-347-welds.html