Reading Welding Symbols and Fabrication Drawings

A surprising share of fabrication errors is not born at the welding machine. It is born at the desk where the drawing is read. A fillet run on the wrong face, a thickness figure read as "z" when it says "a", a bracket in an intermittent weld callout taken as the gap when it means the pitch — each of these can burn tons of surplus filler material, pull needless distortion into the assembly and, worst of all, leave a joint that cannot carry its design load. Welding symbols are an extraordinarily dense form of compression: they carry the joint geometry from the designer's head to the fabricator, the welder and the inspector in a single graphic. This article unpacks that language at the syntax level: the logic of the arrow line and the reference line, the single most critical positional difference between ISO and AWS, the sizing rules, how complete and partial penetration are shown, the supplementary symbols, how a testing requirement appears on the drawing, and finally the bridge that runs from the fabrication drawing to the WPS and the inspection plan.

Standard references carry their edition year throughout: ISO 2553:2019 (5th edition) and AWS A2.4:2020 (8th edition). That is not pedantry; it is necessity. At least four of the points below are valid only in the current editions and have changed since the previous ones.

1. Anatomy of a welding symbol: arrow, reference line and tail

Every welding symbol has two mandatory components: the arrow line and the reference line. To these you add the elementary symbol that defines the joint geometry. The tail is generally an optional component; draw it only when it must carry information the symbol language cannot express. One exception matters. ISO 2553:2019, Clause 4.2 and Figure 1 treat the tail as a mandatory component of the basic weld symbol, the one that says only "this joint shall be welded" and carries no geometric detail. In that case the symbol consists of arrow line, reference line and tail, and the detail is read from the reference in the tail.

The arrow line physically points at the joint. Where the arrow tip lands is the joint line to be welded; the direction the arrow comes from defines the face known as the "arrow side". The arrow line may be broken — particularly on single-side prepared joints (single-bevel, single-J), where the break identifies the member to be prepared. That break is not arbitrary. It answers the question "which plate gets the preparation?" and it drives the preparation cost directly.

As a rule, draw the reference line horizontal, parallel to the bottom edge of the drawing. Where the layout does not allow it, ISO 2553 permits a vertical reference line when unavoidable; AWS A2.4:2020, by contrast, requires the reference line to be parallel to the bottom edge of the drawing at all times. The difference looks minor, but a detail drawn with a vertical reference line out of ISO habit will be rejected at drawing approval on an AWS-governed project. The elementary symbol, the dimensions and the supplementary marks sit above or below this line. The arrow-side end of the reference line joins the arrow line; the other end opens into a tail where required.

Multiple reference lines: a single reference line establishes the side logic; more than one reference line shows not side but sequence of operations. The line closest to the arrow is the first operation, the ones beyond it follow in order. A typical example: V-groove weld on the first line, back gouging on the second, the far-side run on the third, UT on the fourth. AWS A2.4:2020 defines this explicitly, and for the inspector it reads directly as the order of work. It is also the only mechanism that makes the back weld / backing weld distinction of §7 readable from the drawing.

The tail carries what the symbol language cannot. In ISO practice that is typically the ISO 4063 process number (111 covered electrode, 135 MAG solid wire, 136 flux-cored wire, 121 submerged arc, 141 TIG), the quality level where required (ISO 5817 B/C/D for steel, nickel and titanium; ISO 10042 B/C/D for aluminium and its alloys), the welding position (ISO 6947) or the filler material classification. Commit the position abbreviations to memory once: PA flat, PB horizontal-vertical fillet, PC horizontal, PF vertical up, PG vertical down. Tail content is not written in random order. In ISO practice the items follow a fixed sequence, separated by forward slashes — the typical string is process / quality (acceptance) level / position / filler material, for example 135 / ISO 5817-B / PF / ISO 14341-A G 42 4 M21 3Si1. Keeping the order and the separator consistent across the drawing is what makes the tail machine-readable.

In AWS practice the tail usually carries the WPS number, the governing specification reference, the backing material, or a pointer of the "see detail on sheet X" type. Where no extra information is needed, no tail is drawn; in AWS practice the absence of a tail means "no specific specification reference has been given". If the note "TYP" (typical) appears in or beside the tail, that symbol applies to every similar joint on the drawing — one of the costliest details to miss.

2. Arrow side and other side: where ISO and AWS part company

This is the single most critical difference in the symbol language, and it is exactly where teams mixing the two standards on one project go wrong.

In the AWS A2.4:2020 system the side logic is built on a single reference line. The rule is unambiguous: a symbol written BELOW the reference line defines the arrow-side weld, and a symbol written ABOVE it defines the other-side weld. There is no dashed (identification) line. Close off one common misreading straight away: AWS allows more than one reference line, but those lines show sequence of operations, not side (§1); the line closest to the arrow is the first operation.

ISO 2553:2019 defines two systems. The System A / System B duality entered the standard with the 2013 edition and is retained in the current ISO 2553:2019 (5th edition, superseding ISO 2553:2013).

System A (dual-line system): a continuous reference line is used together with a parallel dashed identification line. The dashed line may sit below or above the continuous line. The rule depends not on position but on line type: a symbol on the continuous line is an arrow-side weld, a symbol on the dashed line is an other-side weld. On symmetrical double-sided welds (double-V, double-bevel, double fillet) the dashed line may be omitted entirely. This simplification is not new; it has applied since the 1992 edition of ISO 2553 (the European adoption EN 22553) and has been kept in later revisions.
System B (single-line system): there is no dashed line; below is arrow side, above is other side. System B originates in Pacific Rim (JIS) practice. For side indication it agrees with AWS, but it is not identical to AWS in dimension notation and in some supplementary symbol details.

Here is the trap. In System A, if the dashed line sits below and the symbol is on the dashed line, the symbol appears "below" yet defines the other-side weld. A fabricator reading with AWS reflexes will run the weld on the wrong face. Checking only "is there a dashed line or not" is not enough in the field, because System B has no dashed line either.

That is why declaring the system on the drawing is not advice but a requirement of the standard. ISO 2553:2019, Clause 4.3 requires that System A and System B shall not be mixed and that drawings shall clearly indicate which system is used, together with the unit of measurement in accordance with ISO 129-1. If the title block carries no such declaration, ask before work starts. The first item of the checklist in §12 — "which standard, which system, which unit" — is the field expression of exactly this requirement.

One more detail: where the concept of side means something and where it does not. For resistance spot and seam welds, side has no meaning; the symbol is centered so that it straddles the reference line. Arc spot welds carry no side meaning either; again the symbol is centered on the reference line. Side placement is meaningful in only two cases: for projection welds the arrow side indicates the member carrying the projection, and for plug and slot welds it indicates the member in which the hole or slot is made. Lumping arc spot welds in with plug welds is a common shortcut, and it is wrong. In an arc spot weld no hole is made; penetration is achieved by melting through the upper sheet with the arc. In a plug weld a pre-made hole is filled. They are defined by separate symbols.

Finally an update note. Within the resistance family, flash welding (FW) and upset welding (UW) are separate processes and cannot be lumped together as "upset/flash butt". More importantly, AWS A2.4:2020 has removed the symbols for both processes from the standard altogether. In today's symbol language, flash and upset welding are defined not by a symbol but by a tail note or a WPS reference. Watch the wording, too: the "butt" in flash butt welding belongs to the resistance welding family and must not be confused with the square butt (I) weld described in §3.

3. Elementary symbols

Elementary symbols are the joint geometry itself. The ones you meet most often:

Square butt (I weld): two parallel lines. No preparation; on thin sections, full penetration is achieved through the root gap.
Single-V butt: the classic "V". Its double-sided form becomes a double-V (X) joint, with opposing Vs on both sides of the reference line.
Single-bevel: a symbol with one inclined face; only one member is prepared. The arrow line is broken towards that member. Its double-sided form is the double-bevel (K) joint.
Single-U with root face (Y) and single-bevel with root face (half-Y): a V with a vertical root face beneath it. The root face dimension is written beside the symbol. AWS A2.4:2020 has no separate "Y" elementary symbol; the same geometry is given by adding root face and depth of preparation dimensions to the V symbol.
Steep-flanked single-V: a narrow-angle V preparation with near-vertical flanks. It reduces the fill volume but demands particular care in procedure qualification because of root access and the risk of lack of side wall fusion.
Single-U and single-J: radiused preparations that reduce fill volume on thick sections. As thickness increases (typically above roughly 20 mm for U, and double-U on heavier sections) the advantage in fill volume and distortion becomes clear. The crossover point depends on process, position, groove angle and degree of mechanization, and should be confirmed by calculation.
Flare-bevel and flare-V: joints where at least one flank is curved. They are standard where a pipe or hollow section meets a plate, where two pipes meet, or where reinforcing bars are joined. Because of the curved flank, the effective throat is calculated from the weld size rather than from the geometry. AWS A2.4:2020 has notably expanded the depiction and sizing of flare grooves; on structures framed with hollow sections, failing to recognize this symbol leads straight to a capacity error.
Fillet: a right-triangle symbol. The vertical leg of the symbol always faces left — a rule quietly violated on many drawings but mandated by both standards.
Plug and slot welds: a rectangular symbol. In AWS A2.4 notation the hole diameter (or slot width) goes to the left, the depth of filling inside the symbol, the pitch to the right, and the number in parentheses; the countersink angle is written above or below the symbol. With AWS A2.4:2020 the Ø sign has been removed from the plug weld symbol — the number on the left is read directly as the diameter. ISO 2553 places these items differently; do not carry one notation into the other.
Spot and seam welds: the two standards diverge here. ISO 2553: spot is a circle, seam is a rectangle crossed by a centre line. AWS A2.4: spot is a circle, seam is that circle combined with a rectangle. Whether the weld is resistance or arc spot is settled by the process in the tail.
Stud welding: it has its own elementary symbol; sizing gives stud diameter, number and pitch. This is how shear studs are shown on composite steel decks.
Root penetration / melt-through: requires the root to melt visibly through to the far face on a single-sided weld, with the resulting root reinforcement on that face dimensioned. It is a critical item on single-sided full-penetration joints made without backing.
Consumable insert: a ring or strip placed in the root that melts in the first pass and becomes part of the weld. Used in pipework for repeatable root quality; its class is given in the tail or in a note.
Spacer: a separating element placed inside the groove; it fixes the root gap on double-sided preparations in heavy sections.
Surfacing and edge welds: for wear or build-up overlays and for plate edge joints.

The groove angle (α), root gap (b) and root face (c) dimensions written beside these symbols complete the preparation geometry. ISO 9692-1 gives the recommended dimensions for these preparation forms; where the drawing shows only a symbol and no dimensions, the fabricator is expected to fall back on ISO 9692-1 or on the WPS.

Showing complete and partial penetration (CJP / PJP)

This is the item with the most direct effect on the structural calculation, and most symbol guides skip it.

On the ISO 2553:2019 side the rule is hidden in a footnote (Table 1, footnote b): butt welds are full penetration unless otherwise stated by a dimension on the weld symbol or by reference to other information, such as a WPS. In other words, an undimensioned butt weld symbol means full penetration under ISO, and partial penetration can only be defined by writing the penetration depth dimension (Clause 5.4.1). For a fabricator reading with the reflex "no dimension means free choice", that is a surprise in exactly the wrong direction.

On the AWS A2.4:2020 side the distinction is drawn openly. CJP (complete joint penetration) is usually given by a "CJP" note in the tail with no elementary symbol drawn; the choice of joint detail is left to the fabricator. PJP (partial joint penetration) is defined by two numbers: the groove/weld size (S) to the left of the symbol and the effective throat (E) in parentheses immediately beside it. The notation 12 (10) means "depth of preparation 12 mm, effective throat to be used in calculation 10 mm". Confuse those two numbers and the joint is fabricated straight into a capacity shortfall.

Under both standards the penetration claim has to be proven: partial or deep penetration values may not appear on a drawing unless they are supported by a procedure qualification record (WPQR).

4. Sizing: cross section on the left, length on the right

The golden rule of sizing is simple: the cross-sectional size goes to the LEFT of the symbol, the length information to the RIGHT.

For fillets the cross-sectional size can be given in two different ways, and confusing the two is the most common error in the field. Keep all three dimensions straight:

a — design throat thickness: the perpendicular distance from the corner of the fillet triangle to the hypotenuse. This is the dimension the structural calculation rests on.
z — leg length: the side of the fillet triangle. This is the dimension the welder checks easily, by eye and with a gauge.
s — penetration throat / depth of penetration: ISO 2553:2019, Clause 3.19 defines it as the "deep penetration throat thickness" for deep-penetration fillet welds; but the same letter is also used under Clause 5.4.1 for the penetration depth / design throat of partial (and full) penetration butt welds. Treating s as nothing more than an "automatic fillet" dimension means you cannot size a partial-penetration butt weld at all. In both uses the s value shall be proven by procedure qualification (WPQR).

For a 45° equal-leg fillet the relationship is clean: z = a × √2 ≈ 1.41 a and a = z / √2 ≈ 0.7 z. So on a joint calling for a6, z is roughly 8.5 mm. Reading a drawing that says "a6" as "z6" leaves the throat about 30% short; doing the reverse means surplus metal, excessive heat input and distortion. It is worth stressing that this relationship holds only for the 45° equal-leg fillet; the moment the leg ratio changes, the conversion fails.

The unequal-leg fillet is a separate item for exactly that reason. In ISO notation the two legs are given individually as z1 × z2 (for example z6 × z10); AWS notation likewise writes the two leg sizes separately. Under both standards, if the symbol placement does not make it certain which leg belongs to which member, a detail section shall be drawn. Just as common in the field as a/z confusion is applying the right dimensions to the wrong member: a 6 × 10 fillet rotated through 90° gives a completely different section modulus.

On the length side, a continuous weld carries only the weld length (for example "150"). How welds with no length given are to be interpreted differs by standard; that rule is covered in §6, together with the all-around symbol.

5. Intermittent fillets: the trap inside the parentheses

The way intermittent fillet welds are shown carries the second major interpretation difference between ISO and AWS.

The typical expression in the main text of ISO 2553:2019 (Clause 5.3.2) looks like this: z6 4 × 50 (100) — that is, four weld elements of 50 mm length with a 6 mm leg, and (100) giving the gap. In this clause the value in parentheses is the clear distance (e) between two adjacent weld elements. The pitch (centre-to-centre distance) is therefore 50 + 100 = 150 mm. This clause carries no A/B suffix, so it applies to System A and System B alike.

Even so, saying "in ISO the parentheses always mean the gap" would be too absolute: the standard's informative Annex C (alternative methods for identifying intermittent butt and fillet welds) defines other notations as well. If an alternative method is used, it shall be clearly stated on the drawing; if it is not stated, the notation of the main text (5.3.2) governs.

Under AWS A2.4:2020 the second number to the right of the symbol is the pitch, that is, the centre-to-centre distance. The expression "6 50-150" means a 50 mm weld at a 150 mm pitch; the gap is 100 mm.

The same numerical values produce different geometries under the two standards. On drawings with intermittent welds the standard reference is therefore vital, and where there is any doubt, cross-check against the total weld length.

Where the intermittent weld is double sided, there are two arrangements: chain and staggered. The staggered arrangement is shown by drawing the fillet symbols on the two sides of the reference line offset relative to each other. But offsetting the symbols alone does not define the geometry: without the offset dimension defined in ISO 2553:2019, Clause 3.12, the position of the runs on the two sides relative to each other stays undefined and every shop applies its own interpretation. Looking for the offset dimension on a staggered arrangement belongs in the inspector's routine checks.

The staggered arrangement reduces distortion, but it needs care on joints under fatigue loading because of the notch effect; where dynamic loads are present, avoiding intermittent welds altogether is often the right engineering decision.

6. All-around welds and the field weld flag

Two special marks sit at the corner where the arrow line meets the reference line:

Circle (weld all around): states that the weld runs completely around the joint. It is indispensable on pipe-to-flange, pipe-to-plate and hollow section-to-plate joints. Where there is no circle, the continuity assumption changes by standard: in AWS A2.4 notation, a weld with no length given is taken as continuous between abrupt changes in direction; in ISO 2553 notation, a weld with no length given is taken as continuous over the entire length of the joint, and the requirement to run around the perimeter is defined separately by the circle. If the weld is required to turn the corners and run right around, the circle is mandatory under both standards. The crater and start-stop problems seen at the corner turns of hollow sections usually trace back to this distinction being skipped. On complex three-dimensional geometry, question at design stage whether "all around" is actually achievable — that is, accessible.
Flag (field weld): the small flag raised at the junction of the arrow line and the reference line shows that the weld is to be made on the erection site, not in the shop. Its direction and position are defined: in AWS A2.4:2020 the flag is drawn at the junction of the arrow line and the reference line, perpendicular to the reference line and pointing upwards, with its open end towards the tail. In ISO 2553 the flag likewise sits at that junction, above the reference line. Flags drawn the wrong way round are common in the field and cause arguments during revision tracking; check the direction as well when you check the drawing. This mark is more than location information. A field weld means a different position (often PF or PC), wind and moisture protection, preheat logistics, a different WPS and usually a heavier testing regime. In cost estimating, count flagged welds as a separate item.

7. Surface shape, finishing and supplementary symbols

Lines added on top of the elementary symbol define the finished surface of the weld:

Flat contour — a straight line: AWS A2.4:2020 has split this symbol in two. The flat contour symbol is now used only for fillet welds, while flush (level with the base metal) applies to groove, plug and slot welds. ISO 2553 has a single flat contour symbol. AWS A2.4:2020 also made the contour achievable "by any method" — by welding technique or by mechanical means — so the flat contour symbol on its own is not an instruction to grind; where a method is required, state it with the finishing letter. In practice, bringing the surface flush with the base metal usually does require grinding, and that adds cost directly.
Convex — a curved line: excess weld metal is accepted; its size is limited by the quality level.
Concave: improves fatigue behavior, particularly on fillets.
Toe blended: the weld toe blends smoothly into the base metal; required on critical details under fatigue loading. What is meant here is the line where the weld meets the base metal, not the leg of the fillet.

Finishing method letters (per AWS A2.4): G grinding, M machining, C chipping, H hammering, R rolling, P planishing, U method unspecified. This letter set is AWS notation; ISO has no letter set of comparable breadth, and surface finishing requirements are generally left to a note, the tail or the WPS. The combination "flat + G" means "this weld shall be ground flush with the base metal", and it also matters for NDT: a ground surface makes surface testing easier and at the same time prepares a scanning surface for UT.

Other important supplementary items:

Back weld (back run): the completing run made from the far face after the main weld, usually following back gouging and cleaning. Where the back gouging requirement is not stated on the drawing, define it in the WPS.
Backing weld: the run made on the far side as root support before the main weld. The two concepts are opposites in sequence and must not be confused. Critical warning: in AWS A2.4 the symbol for these two welds is the same; the distinction cannot be read from the shape of the symbol. Which one is meant can only be established from a tail note ("back weld" / "backing weld") or from the sequence of operations given by multiple reference lines. If neither piece of evidence appears on the drawing, the requirement is undefined — raise a question before you weld.

How backing is shown differs by standard — not knowing this difference leaves a permanent notch in the root:

ISO 2553: the letter M inside the backing symbol means permanent backing (it stays in the structure); MR means removable backing (it is taken out after welding).
AWS A2.4: the letter R inside the backing symbol means backing that shall be removed after welding; where there is no letter inside the symbol, the backing is permanent. The backing material is stated in the tail.

A reader looking at an AWS drawing with ISO reflexes may dismiss the "R" as meaningless and assume the backing is permanent — which means leaving in place backing that should have been removed, and a permanent notch in the root with the fatigue and corrosion risk that follows. The reverse is just as expensive: grinding out backing that never had to be removed costs both labor and section. The backing decision governs not only the assembly sequence but the lifetime behavior of the joint.

A note on notation: the letter M appears in this article in two different meanings, not because of anything inherent in the letters but because two standards are being set side by side — the finishing letters (G/M/C/H/R/P/U) are AWS notation; the M/MR backing letters are ISO notation. Position settles the distinction absolutely: the backing letter goes INSIDE the backing symbol (the rectangle), while the finishing letter goes BESIDE the surface contour symbol.

8. Showing the NDT requirement on the drawing

A non-destructive testing requirement reaches the drawing by two routes.

On the AWS side the situation is direct: AWS A2.4:2020 standardizes not only welding and brazing symbols but nondestructive examination symbols, and that clause was rewritten and expanded in the 2020 edition. Using the same arrow/reference line logic, the method is written as a letter abbreviation: RT radiographic, UT ultrasonic, MT magnetic particle, PT penetrant, VT visual, ET electromagnetic (eddy current is a sub-branch of this method), LT leak, AET acoustic emission, NRT neutron radiographic, PRT proof. A method written below the reference line means examination from the arrow side, above it means from the other side; where examination is required from both sides, write it in both positions. The examination length is given to the right of the symbol. The number of examinations to be made at random locations is given in parentheses together with the examination symbol — for example (3); the tail is not used for this, it is reserved for the specification or procedure reference. The same circle serves for all-around examination and the same flag for field examination. The welding symbol and the examination symbol can be combined on one reference line, so that "6 mm fillet + MT full length" is expressed in a single graphic. Where the sequence of welding and examination must be shown (for example weld → grind → UT), use the multiple reference lines described in §1.

On the ISO side, ISO 2553 does not define a separate NDT symbol set as AWS does. Testing information is given through a tail note, a weld number and a weld map. In the ISO world the testing requirement is generally defined through: a weld map linked to a weld ID on the drawing, the quality level given in the title block (ISO 5817, or ISO 10042 for aluminium), and the testing percentage tables of the application standard. For metallic materials, ISO 17635 gives the general rules for method selection and testing levels; in steel structures the extent of testing comes from the execution class (EXC1–EXC4) tables of EN 1090-2:2018. For pressure equipment the relevant design code (EN 13445 or ASME BPVC, for example) sets the testing rate.

In practice the right approach is this: give every weld on the drawing a unique number, and map that number in the weld map to geometry, WPS, welder, test method, extent and acceptance criteria. The symbol language answers "what shall be welded"; the weld map answers "by whom, how, and on what evidence".

9. ISO 2553:2019 and AWS A2.4:2020 side by side

In summary, the clear differences between the two worlds:

Topic	ISO 2553:2019	AWS A2.4:2020
Side indication	System A: continuous/dashed line distinction; System B: below/above	Below = arrow side, above = other side
Dashed identification line	Present in System A (may be omitted on symmetrical welds)	None
Multiple reference lines	May be used	Show sequence of operations; the line nearest the arrow is the first
Reference line direction	Horizontal; may be vertical where unavoidable	Always horizontal (parallel to the bottom edge of the drawing)
Fillet cross section	System A: a (throat) or z (leg), labelled with the letter. System B: dimension notation differs (closer to AWS) — do not interpret a fillet size without the system declared in the title block	Usually the leg, with no letter; where the effective throat is needed it is given in parentheses as (E)
Unequal-leg fillet	z1 × z2	The two leg sizes are written separately; a detail section is mandatory if the orientation is ambiguous
Penetration	An undimensioned butt weld is full penetration (Table 1, footnote b); partial penetration is defined only by a penetration depth dimension	CJP by tail note / undimensioned; PJP with size (S) on the left and effective throat (E) in parentheses
Intermittent welds	Main text (5.3.2): the value in parentheses = the gap (e); Annex C defines alternative methods	Second value = pitch (centre to centre)
Backing	M = permanent, MR = removable (inside the backing symbol)	R = removable; no letter means permanent
Surface finishing	No broad letter set; via note/tail/WPS	Letter set G, M, C, H, R, P, U
Tail content	ISO 4063 process number, quality level, position, filler material (separated by slashes)	WPS/specification reference, backing material
NDT symbols	No separate symbol set; via weld map and application standard	Defined within the same standard (expanded in the 2020 edition)

Rule shared by both: the vertical leg of the fillet symbol faces left under either standard.

These differences are not academic. In Türkiye (Turkey) it is entirely normal for one fabricator to run a European project under EN 1090 and an American project under AWS D1.1 in the same week. Two symbol languages circulating in one shop turn the title block check into a mandatory procedural step. The symbol family also has neighboring standards; the symbolic representation of adhesive bonded, folded and pressed joints on drawings, for example, is defined by ISO 15785, and that standard must not be confused with the welding symbols themselves. On the numbering side, one fact is enough: EN 22553 has been withdrawn; the current European number for ISO 2553 is EN ISO 2553:2019.

10. The bridge from the fabrication drawing to the WPS

Reading the drawing correctly is not enough on its own; what has been read has to turn into a welding procedure. That conversion runs through the following mappings:

Elementary symbol + preparation dimensions → the joint design in the WPS. The choice of V, X, K or U drives the pass plan and the filler material consumption directly.
Penetration requirement (CJP/PJP, the s dimension) → root technique and pass sequence in the WPS. A partial penetration value cannot reach the work order unless it is proven by a WPQR.
Material thickness → the thickness range of validity of the WPS. A joint that falls outside the thickness range of a WPQR under ISO 15614-1 requires a new qualification.
Process number in the tail → the WPS process. You cannot weld with process 135 under a procedure approved for 111.
Assembly orientation of the part → welding position (ISO 6947). The orientation on the drawing decides whether the weld is PF (vertical up) or PA (flat), and that in turn changes the limits on heat input, weave and travel speed.
Backing indication (M/MR under ISO, R or no letter under AWS) and the back weld → root technique in the WPS. On a single-sided full-penetration joint with no backing, a TIG root or a gas-shielded root technique is required; where a consumable insert is specified, its class and size go into the WPS.
Surface finishing symbol → the post-weld operation step. Grinding time does not show up on the drawing, but it must always appear in the work order.

11. From the WPS to the inspection plan (ITP)

The inspection plan is the line-by-line counterpart of the weld numbers on the drawing. A sound ITP line contains: weld number, joint type and size, WPS number, welder identity (with the ISO 9606-1 qualification certificate), preheat and interpass temperature requirements, post-heating where required, PWHT requirement and record — with hardness testing where required, visual testing (ISO 17637), the extent of surface NDT (MT/PT), the extent of volumetric NDT (UT/RT), acceptance criteria, the delay time before testing, and the record form number.

Acceptance criteria are not read from a single standard. ISO 5817 (and ISO 10042 for aluminium and its alloys) is a quality level standard; the method-based acceptance levels come from separate standards, and ISO 17635 establishes that mapping: ISO 10675-1 for RT, ISO 11666 for UT, ISO 23277 for PT, ISO 23278 for MT; for visual testing, ISO 17637 together with ISO 5817 / ISO 10042. An ITP line that says only "ISO 5817 B" leaves the accept/reject decision in a UT or RT report without a basis, and it turns into an audit finding. The correct wording is this: visual: ISO 5817 (or ISO 10042) level; volumetric and surface NDT: the relevant acceptance level standard via ISO 17635.

The direction of severity of the quality levels must also be clearly understood: in ISO 5817, B is the most stringent, C intermediate and D the most permissive. Level B is required on highly fatigue-loaded and critical connections. The difference is not cosmetic; on the same weld, the acceptable depth of continuous undercut is typically of the order of h ≤ 0.2t (max. 1 mm) at D, h ≤ 0.1t (max. 0.5 mm) at C and h ≤ 0.05t (max. 0.5 mm) at B — values differ between editions, so verify them one to one against the ISO 5817:2023 table before writing them into an ITP. Excess weld metal height and porosity ratios likewise vary appreciably with level. Where the drawing states no quality level, a fabricator applying ISO 3834 as its quality management framework has to clear this up before contract.

The delay time cannot be given as a single figure; it is applied because of the risk of delayed (hydrogen-induced) cracking, and it comes from the relevant code according to material class, weld size and heat input.

EN 1090-2:2018 (Clause 12.4.2.2 and Table 23) grades the delay time not as one value but on three variables: steel grade + weld size (throat thickness a) + heat input. Roughly: for S235–S420, cooling to ambient temperature / 8 / 16 / 40 hours; for S460 and above, 24 / 40 / 48 hours. For S460 and above, for example, 24 hours applies for a ≤ 6 mm or for 6 < a ≤ 12 mm with ≤ 3 kJ/mm, 40 hours for 6 < a ≤ 12 mm with > 3 kJ/mm, and 48 hours for a > 12 mm. The often-quoted "16 hours" is one cell of the S235–S420 grade (6 < a ≤ 12 mm with > 3 kJ/mm, or a > 12 mm with ≤ 3 kJ/mm) and cannot be used on its own as a general rule.
EN 1011-2 and the BCSA/SCI guides give delay time recommendations graded by thickness; those grades are a separate source from the EN 1090-2 table and must not be mixed with it.
On the AWS D1.1/D1.1M side, the 48-hour delay is defined not for a broad class such as "quenched and tempered steels" but by name for ASTM A514, ASTM A517 and ASTM A709 Gr. HPS 100/100W: the acceptance criteria are based on NDT carried out at least 48 hours after the weld is completed. For other grades, cooling of the weld to ambient temperature is sufficient. The D1.1 edition governing the project (D1.1/D1.1M:2020, for example) and the clause number must be written into the ITP.

The correct practice is to read the time from the code the project is governed by and from the relevant table of that code, and to write it into the ITP line together with the edition year and the clause number.

One last link in the chain of responsibility: the introduction to ISO 2553:2019 foresees that welding drawings are prepared and checked by competent welding coordination personnel appointed in accordance with ISO 14731. In other words, the institutional answer to "who drew the symbol, who approved it" is the welding coordinator, and that role must be named as such in the ITP and in the quality plan.

12. The order of checks when reading a drawing

The sequence an experienced inspector follows when looking at a drawing runs roughly as follows:

Title block: which standard, which system (ISO 2553 System A or B), which revision, which scale, which unit (declared in accordance with ISO 129-1?).
Material list and material group (ISO/TR 15608 group); the first sign of a preheat requirement.
Material group + thickness → preheat and PWHT requirement; a hardness check plan where required.
Side logic of the welding symbol: is there a dashed line, and if so, which line carries the symbol.
Cross-sectional size: a or z, equal leg or not (is there a z1 × z2), is the unit mm.
Penetration: complete or partial; is there a dimension under ISO, is there a CJP note or an S and (E) notation under AWS.
Length and continuity: continuous or intermittent, do the parentheses mean gap or pitch, and if staggered, is the offset given.
All-around circle and field weld flag (is the flag drawn the right way round).
Supplementary symbols: backing (ISO M/MR — AWS R or no letter), back weld or backing weld, surface finishing.
Tail: process, quality level, position, filler material, WPS reference.
Weld number and its match in the weld map; is the acceptance level standard (the ISO 17635 mapping) stated.
Accessibility check: can the weld the symbol describes actually be made, can the torch get in, can the test probe reach the surface.

The last item is the one most often skipped and it produces the most expensive surprises. A symbol can be flawless on paper; if the welder's hand cannot get in there on site, that symbol is void.

Related standards

ISO 2553:2019 — Welding and allied processes: symbolic representation on drawings — Welded joints (System A and System B; 5th edition, superseding ISO 2553:2013)
EN ISO 2553:2019 — the current European number for ISO 2553:2019 (EN 22553 has been withdrawn)
AWS A2.4:2020 — Standard Symbols for Welding, Brazing, and Nondestructive Examination (8th edition; supersedes AWS A2.4:2012)
ISO 129-1 — General principles for dimensioning and tolerancing on technical drawings (system and unit declaration)
ISO 4063 — Nomenclature of processes and reference numbers for welding and allied processes
ISO 9692-1 — Joint preparation for fusion welding of steels
ISO 6947 — Welding positions: definitions
ISO/TR 15608 — Grouping systems for metallic materials for welding purposes
ISO 5817:2023 — Quality levels for imperfections in fusion-welded joints in steel, nickel, titanium and their alloys (4th edition; beam welding excluded)
ISO 10042 — Quality levels for imperfections in arc-welded joints in aluminium and its alloys
ISO 17635 — Non-destructive testing of welds: general rules for metallic materials (quality level ↔ acceptance level mapping)
ISO 10675-1 — RT acceptance levels (steel, nickel, titanium and their alloys)
ISO 11666 — UT acceptance levels
ISO 23277 — PT acceptance levels
ISO 23278 — MT acceptance levels
ISO 17637 — Visual testing of fusion-welded joints
ISO 15609-1 / ISO 15614-1 — Welding procedure specification and procedure qualification (arc welding)
ISO 9606-1 — Qualification testing of welders (steels)
ISO 14731 — Welding coordination: tasks and responsibilities
ISO 3834 — Quality requirements for fusion welding
EN 1090-2:2018 — Execution of steel structures; execution classes (EXC), extent of testing and NDT delay times (Clause 12.4.2.2, Table 23)
EN 1011-2 — Recommendations for arc welding of steels (preheat, heat input, delay time guidance)
AWS D1.1/D1.1M:2020 — Structural Welding Code — Steel
ISO 15785 — Technical drawings: symbolic representation of adhesive, fold and pressed joints (neighboring symbol family)

From the field

On a pipe support bracket, the drawing called out a double-sided fillet as "a5". Out of habit, the shop read that as the leg size and welded a 5 mm leg. The throat came out at about 3.5 mm — roughly 30% below the required value. Under visual testing the welds looked flawless: no porosity, no undercut. The error surfaced only during dimensional checking with a throat gauge, by which point every one of the brackets — more than 60 — had to be cut off and rewelded.

Three lessons came out of it. First, visual testing must answer more than "is there a discontinuity?"; it must also answer "is the size right?", which makes the fillet gauge mandatory equipment. Second, never skim past the letters "a" and "z" on a drawing; where there is any doubt, write both values into the work order (for example "a5 = z7"). Third, going into series production without a first article check multiplies a single reading error across hundreds of parts. That shop now holds a 15-minute "symbol reading" meeting before fabrication starts on every new drawing. Against the cost of that rework, those 15 minutes are nothing.

Take it to the field: To keep the welding standards, the procedure logic and the field check steps in this article in your pocket — completely offline and free — take a look at the Doawise Welding Eng Guide app.

At DoaWise we set up welding engineering and inspection services as a single traceable chain, from the first moment a drawing is read to the moment the final test report is signed. From symbol interpretation to WPS preparation, from the weld map to the inspection plan and the acceptance criteria, every step is tied to a document — producing results that are workable in the field, recordable and auditable.