Weld heat input and CWI acceptance for structural steel

Weld heat input, and why the WPS puts a band on it

Weld heat input is the amount of energy you put into a unit length of weld, normally written in kilojoules per inch. The arc delivers power, the torch moves down the joint, and the slower you move the more energy lands in each inch. That energy does not just melt the filler. It heats the base metal next to the weld, and how fast that heated metal cools afterward is what decides the strength and toughness you end up with.

Cooling rate is the whole game. Weld cold and fast and the heat-affected zone next to the bead cools quick, hardens, and turns brittle, which is where cracks start. Weld hot and slow and the same zone cools so gradually that it loses strength and toughness, which is exactly what you cannot afford on a connection that has to absorb a seismic event. The right answer is a window, not a target, and that window is written into the welding procedure specification.

On a datacenter or mission-critical job this is not paperwork. The structure carries millions of dollars of equipment and has to ride out an earthquake without the moment connections fracturing. The heat-input band on the WPS is the link between the procedure qualification done in the shop and the weld an ironworker actually lays on the deck at two in the morning. The Certified Welding Inspector's job is to confirm that the weld in front of them was made inside that band, and to have the record to prove it.

How do you calculate weld heat input?

Heat input in kilojoules per inch equals arc voltage times welding current times 60, divided by travel speed in inches per minute, divided by 1000. The 60 converts amp-volts (watts) per minute into the same time base as travel speed, and the 1000 takes joules to kilojoules. For metric work the result in kilojoules per millimeter is the same arithmetic with travel speed in millimeters per minute, or divide the per-inch figure by 25.4.

There is a distinction the codes treat differently, and you need to know which one your job runs on. What the formula above gives is strictly arc energy, the energy leaving the power source. The energy that actually reaches the steel is arc energy times a thermal efficiency factor that depends on the process. Submerged arc (SAW) is the most efficient, commonly taken near 1.0. Stick (SMAW), gas metal arc (GMAW), and flux-cored (FCAW) run around 0.8. Gas tungsten arc (GTAW) and plasma are lower, often near 0.6. Verify the exact factor your code and procedure use, because it changes the number you compare to the band.

Here is the trap. AWS and ASME practice has historically called the arc-energy formula result the heat input and applied no efficiency factor, while the European EN ISO 1011 approach multiplies by efficiency to get a lower true heat input. So a SMAW weld at 40 kJ/in of arc energy is 40 kJ/in under AWS bookkeeping and about 32 kJ/in under the European method. When a WPS states a heat-input range, confirm which convention it is written in before you accept or reject a pass against it.

E

Arc voltage in volts, read at the arc, not the open-circuit voltage on the dial

I

Welding current in amps, the actual current carried during the pass

S

Travel speed in inches per minute along the joint, the input crews most often miss

η (eta)

Thermal efficiency factor by process: SAW near 1.0, SMAW/GMAW/FCAW near 0.8, GTAW near 0.6

Field example: holding the band on a CJP column splice

Take an FCAW fill pass at 27 V and 280 A moving 12 in/min. Heat input is 27 times 280 times 60, divided by 12, divided by 1000, which is 37.8 kJ/in. If the WPS band is 25 to 40 kJ/in, that pass is good with a little room left.

Now the welder slows to chase a wide gap and travel drops to 10 in/min at the same volts and amps. The number jumps to 45.4 kJ/in, over the top of the band, and the toughness the PQR proved is no longer guaranteed. Travel speed is the lever that moves heat input the fastest, and it is the one nobody is metering in real time. Volts and amps sit on a display. Travel speed lives in the welder's hands.

Run it the other direction for the floor. To stay at or below 40 kJ/in at 27 V and 280 A you need travel of at least 11.3 in/min, and to stay at or above 25 kJ/in you cannot run faster than about 18 in/min. That spread is the procedure window expressed as travel speed, and it is the most useful way to hand the band to a welder. Give them the travel range that keeps volts and amps inside the band, not an abstract kilojoule number they cannot see.

Why do heat input bands matter?

Heat input bands matter because both ends of the range produce a defective weld in different ways, and neither one shows on the surface. Too low and the weld and its heat-affected zone cool too fast. A fast quench in steel forms hard, brittle microstructure right beside the bead, and with any hydrogen present that hard zone cracks, sometimes days after the weld looks finished. Too high and the slow cool coarsens the grain, and the weld and the HAZ lose toughness and a measure of strength, which is the failure mode you least want on a connection asked to yield in an earthquake.

The low limit on the WPS is really a preheat-and-cooling-rate control. It exists so the joint does not quench into a crack. The high limit is a toughness control. It exists so the connection still has Charpy energy when it is cold and loaded hard. On ordinary statically loaded steel the high limit is often generous or absent. On demand-critical seismic welds it gets tight, because the whole point of the connection is ductility, and ductility is the first thing excess heat takes away.

The number people forget is that interpass temperature and heat input push the same direction. Pile passes in fast on a hot joint and you have effectively raised heat input even if each pass metered fine. The band on the WPS and the interpass maximum work together. Blow through one and you have usually blown through the other.

The WPS and the PQR behind it

The welding procedure specification is the recipe for the joint. It names the process, the filler metal classification and diameter, the position, the joint detail, the preheat and interpass temperatures, and the ranges for amperage, voltage, and travel speed, which together set the heat-input band. A welder should be able to stand at the joint, read the WPS, and reproduce a weld that meets the code. If the parameters on the machine fall outside the WPS, the weld is not qualified, full stop.

Behind most project-specific WPSs is a procedure qualification record. The PQR is the evidence: a test weld made to the proposed procedure, then cut up and tested with transverse tension and guided bend coupons, and Charpy impact specimens where toughness is required. For a groove weld, AWS D1.1 has historically called for two tension and four bend tests from the qualification plate. The PQR records the actual parameters and the test results, and the WPS ranges are bounded by what the PQR proved. You cannot write a WPS band wider than the qualification supports.

AWS D1.1 also allows prequalified WPSs for common processes, materials, and joint details that meet the code's prescriptive rules, with no PQR test required. That covers a lot of routine work. It does not cover the demand-critical and seismic welds on a mission-critical job, where the engineer of record and the seismic supplement usually drive project-specific qualification with toughness testing. When in doubt, the joint follows the more demanding of the contract documents, the code, and the EOR's notes. The adopted edition and project specification control.

Why control preheat and interpass temperature?

Preheat slows the cooling rate so the heat-affected zone does not quench hard and crack, and it drives off surface moisture before the arc can turn it into hydrogen. The minimum preheat climbs with material thickness and with the carbon equivalent of the steel, because thicker and richer steel pulls heat away faster and is more prone to a hard HAZ. AWS D1.1 publishes prequalified minimum preheat and interpass temperatures in a table organized by steel category and thickness, with a lower requirement when low-hydrogen processes are used. The code also offers a hydrogen-control method in an annex for cases the table does not cover.

Interpass temperature is the practical heat-input control on a multipass weld, and it is the one inspectors actually catch in the field. It is the temperature of the joint just before the next pass starts. There is a minimum, usually the same as the preheat, so the joint never drops back into the quench-crack range between passes. There is also a maximum, and that maximum is what keeps a hot joint from cooking the toughness out of the steel as the passes stack up. Exceed the interpass maximum and you have raised the effective heat input no matter what each pass metered.

Measure it, do not eyeball it. A calibrated surface thermometer or a temperature-indicating crayon at the right distance from the joint, checked before each pass, is how it gets verified. The rookie move is to keep welding a hot joint to stay on schedule and let interpass climb past the limit. The fix is to wait, or to weld a different joint while that one cools. Heat is cheaper to add than to take back.

Filler metal and hydrogen

Hydrogen plus a hard heat-affected zone plus restraint is the recipe for delayed cracking, and it is the reason low-hydrogen consumables exist. The hydrogen comes from moisture, and the most common source is a low-hydrogen stick electrode that sat out in humid air. The cracks it causes are the worst kind: they appear hours to days after the weld cools, often 24 to 72 hours later, after the inspector has already walked the joint and the crew has moved on.

Low-hydrogen electrodes like E7018 are hygroscopic. They pull moisture out of the air, and once they are wet the coating reintroduces hydrogen into every weld. The handling rules exist for exactly this. After the hermetically sealed can is opened, the electrodes go into a holding oven held around 250 to 300 degrees F. There is a maximum atmospheric exposure time before they must be rebaked, commonly stated as a few hours and tightened to about one hour for demand-critical welds under the AISC seismic provisions. Rebaking is a high-temperature bake, on the order of 500 to 800 degrees F for a couple of hours per the consumable and the code table, not a quick warm-up. Verify the values against the manufacturer's data and the adopted edition.

A welder pulling rod from a back pocket that has been open since the morning break is not running a low-hydrogen weld anymore, whatever the box says. On demand-critical work this is a hold point worth standing on. Cold cracks from wet rod do not fail the visual. They fail the building.

The per-pass ledger

The only way to prove a heat-input band was held is to record the parameters pass by pass, and the only way that record is worth anything is if it is taken at the joint while the weld is being made, not reconstructed from memory after. For each pass you want volts, amps, travel speed, the computed heat input, and the interpass temperature you measured before you struck the arc. That is the ledger that turns a WPS into evidence.

Travel speed is the input that makes or breaks the ledger, because it is the one nobody is reading off a meter. Volts and amps are on the machine. Travel speed has to be timed, an inch count against a clock, or backed out from deposition. A heat-input record that lists volts and amps but assumes a travel speed is not a record, it is a guess wearing a uniform. The discipline is to capture the real travel and let the kilojoule number fall out of it.

This is the work the HeatBandDC tool is built to take off the clipboard. It holds the WPS band for each joint, computes heat input from the parameters as they are entered, flags the pass that drifts over or under the band before the next one goes in, and keeps the per-pass ledger tied to the joint on the weld map. The CWI signs against a record that was right the first time instead of a stack of forms reconciled at the end of the shift.

Demand-critical and seismic welds (AWS D1.8)

Demand-critical welds are the ones the seismic system relies on to yield and absorb energy without fracturing, and they live under the AWS D1.8 seismic supplement to D1.1, working alongside AISC 341. The supplement tightens the rules that ordinary D1.1 leaves open, and heat input and toughness are the center of it. On a moment frame, the complete-joint-penetration welds of the beam flanges to the column are the classic demand-critical welds, and they are inspected and documented to the higher bar.

Toughness is specified, not assumed. D1.8 requires filler metals for demand-critical welds to meet a minimum Charpy V-notch toughness, commonly stated as 20 ft-lbs at 0 degrees F by the standard A5 classification test, with a higher requirement, on the order of 40 ft-lbs at a low test temperature tied to the lowest anticipated service temperature, where the structure runs cold. Verify the exact values against the adopted edition and the project, because they are edition and application specific.

Heat input gets bracketed at both ends for these welds. The supplement drives filler metals to be evaluated across a high-and-low heat-input envelope, so the consumable carrying the connection holds its toughness across the procedure window, not just at one comfortable setting. Consumables tested to that envelope carry a supplemental designator. Atmospheric exposure on low-hydrogen rod is cut tighter, and changes to preheat, interpass, and heat input that D1.1 would shrug at require requalification when the joint is demand-critical. The lesson for the field is that the heat-input band on a seismic WPS is narrow on purpose, and there is no good reason to wander outside it.

What fails a CWI visual inspection?

A crack fails, every time, any size, no exceptions. Beyond that, the visual acceptance criteria in AWS D1.1 cover undercut, porosity, weld profile, undersize, incomplete fusion, unfilled craters, and overlap, with limits that depend on whether the member is statically or cyclically loaded and whether the weld runs transverse to tensile stress. The criteria live in the inspection clause and its acceptance table, which the 2020 and later editions renumbered, so cite them by topic and confirm the table against the edition on the job.

The inspector works the joint in three passes. Before welding, the CWI checks the fit-up, the root opening, the joint preparation and cleanliness, the WPS at the station, the consumable and its handling, and the preheat. During welding, they confirm the parameters fall inside the WPS band, watch interpass temperature, and check interpass cleaning between passes. After welding, they do the visual against the acceptance criteria and call for the required nondestructive testing. The before-welding check is the one that prevents the most rework, and it is the one schedule pressure tries hardest to skip.

The tells a seasoned CWI reads fast: undercut as a shadow line along the toe, scattered or piping porosity, a cap that is too convex or has cold overlap rolling onto unfused base metal, an arc strike off the joint that left a hard spot, a crater crack at a stop. Undercut and porosity are limited by depth, length, and orientation to stress, not banned outright, but a crack or incomplete fusion is a reject. The visual is the cheapest inspection there is and it catches most of what is wrong. The volumetric testing is for what the eye cannot see.

Nondestructive testing: UT, MT, PT, RT

Visual testing is on every weld. Beyond it, the method is chosen for what you need to find and where it hides. Ultrasonic testing is the workhorse for the buried flaws in thick complete-joint-penetration welds, which is most of the demand-critical structural work, and phased-array UT has moved into D1.1 as an accepted variant. Magnetic particle testing finds surface and near-surface cracks in steel and is the common back-up to the visual on fillets, repairs, and back-gouged roots. Liquid penetrant testing catches surface-breaking flaws where MT does not suit. Radiography produces a volumetric film image but is less common on heavy structural sections than UT.

The CJP groove weld in a moment connection is the case that matters most here. It gets a 100 percent visual, and it gets UT for the interior, because a lack-of-fusion plane or a buried crack in that weld is invisible to the eye and fatal to the connection. The acceptance criteria for UT are defined in the code by indication rating against length and depth, and the frequency and extent of testing are set by the contract documents and the EOR, often 100 percent on demand-critical welds and a sampling rate elsewhere.

Match the method to the flaw. UT and RT see into the volume; MT and PT see the surface. Run a surface method on a buried flaw and it passes a weld that UT would reject. The CWI's call is which method the code and the spec require for that joint, then confirming the technician and the procedure are qualified for it.

When is a weld rejected, and how is it repaired?

A weld is rejected when it fails the visual criteria or the required NDT, and the disposition then runs through the engineer of record's repair procedure, not the welder's judgment. A crack is removed entirely, confirmed gone, and rewelded to a qualified procedure. Undercut and underfill are usually repaired by welding to a profile. Porosity, slag, and incomplete fusion are gouged or ground out to sound metal and rewelded. The repair cavity gets the same preheat, the same WPS, and the same inspection as the original, and the repaired area is re-tested by the same NDT that found the flaw.

Repairs are not free and the code does not treat them as routine. There are limits on how much base metal can be removed and on how many times a given area can be reworked before the engineer has to evaluate the member, because each cut-out-and-reweld cycle pours more heat into the same steel and stacks more residual stress and distortion into the connection. A joint that has been chased three times is telling you something about the fit-up or the procedure, not just the welder.

The blunt rule on a demand-critical weld: you do not grind a crack to make it disappear from the visual and walk away. Cracks propagate. Removing the surface evidence without removing the crack leaves the connection compromised and the next inspector, or the next earthquake, to find it. Document the reject, the repair WPS, and the re-inspection so the record shows the flaw was removed, not hidden.

Welder qualification and continuity

A qualified WPS proves the procedure makes a sound weld. Welder qualification proves the person can run that procedure in the position and on the material the job calls for. The two are separate, and a CWI checks both. A welder qualifies by making a test weld that passes the required tests, which earns them a range of positions, thicknesses, and processes they are then approved to weld.

Qualification lapses if the welder stops using the process. AWS D1.1 treats a welder's qualification as continuing as long as they have used the process within a defined period, commonly six months, after which the qualification for that process expires and has to be re-established. On a long job this catches people who moved to other work and came back. The inspector verifies the continuity record before accepting their welds, not after.

On the deck the practical check is the welder's stamp on the joint and a current qualification record on file that covers the process, position, and thickness of that weld. A weld with no traceable welder is a weld nobody owns. Tie the stamp to the joint on the weld map so the record connects the person, the procedure, and the location.

Distortion and shrinkage control

Weld metal shrinks as it cools, and that shrinkage pulls the steel. Left uncontrolled it warps members, closes or opens joints out of tolerance, and locks residual stress into the connection that adds to the service load. Heat input drives it directly. More heat per inch means more shrinkage and more distortion, which is one more reason the high end of the band is not a free parameter.

The field controls are sequence and balance. Weld in a sequence that lets the structure shrink toward where you want it, balance passes across the neutral axis of a joint so one side does not pull harder than the other, and use back-step or skip welding to spread the heat instead of dumping it in one place. Tack and fixture to hold position, and account for the shrinkage in the fit-up rather than fighting it after the fact. Preheat helps here too, because a more uniform temperature reduces the gradient that drives warping.

The number people miss is cumulative. Each pass adds a little shrinkage, and on a heavy multipass joint the total pull is real. Check the joint and the member for distortion as the weld builds, not just when it is capped, because by then the steel has already moved and the only fix left is heat straightening or rework.

The CWI report and the weld map

The CWI report is the document of record that the welds were made and inspected to the code and the contract, and the weld map is what makes it usable. The weld map is a marked-up set of drawings or a model that identifies every weld by a unique mark, ties each one to its joint detail and WPS, and tracks its inspection status. Without it, an inspection report is a pile of findings with no way to say which weld is which.

A complete record for a joint links the weld mark on the map to the WPS, the welder's stamp and continuity, the per-pass heat-input and interpass ledger, the visual result, the NDT report and its acceptance, and the disposition of any repair. That chain is what the engineer of record, the building official as the authority having jurisdiction, and the special inspector all read at turnover, and it is what defends the structure if anyone ever asks how a connection was made.

This guide sits inside the broader datacenter concrete and steel QA program. The structural steel weld map and CWI package are one workstream in that larger turnover, alongside the anchor bolt and base plate grout QA that ties the steel to the foundation. Keep the weld documentation aligned with the overall QA overview so the steel package closes out clean with the rest of the structure.

What to document

Document enough that a stranger could reconstruct how the weld was made and why it was accepted, joint by joint. The heat-input ledger is the heart of it, but the surrounding record is what gives the kilojoule numbers meaning: which WPS, which welder, what preheat, which consumable and lot, and what the inspection found.

If the heat input drifted on a pass and the joint was reworked, that belongs in the record too, with the repair WPS and the re-inspection result. The honest record that shows a flagged pass, a repair, and a passing re-test is worth more than a too-clean record that shows nothing ever went wrong on a hundred-foot run of multipass welding.

Common mistakes

• Running heat input out of the WPS band, usually by slowing travel speed to chase a gap, with nobody metering travel.
• Letting interpass temperature climb past the maximum to keep welding a hot joint on schedule.
• Welding with low-hydrogen rod that sat out past its exposure limit and was never rebaked.
• No WPS at the joint, or parameters on the machine that fall outside the WPS ranges.
• Grinding a crack out of the visual without removing the crack, then accepting the weld.
• Recording volts and amps but assuming travel speed, so the heat-input number is a guess.
• Treating demand-critical seismic welds like ordinary statically loaded welds and ignoring the tighter D1.8 limits.
• Accepting a CJP weld on the visual alone when the code and spec require UT.

Field checklist

Standards and references

AWS D1.1, Structural Welding Code for steel, is the central document. It defines the WPS and PQR, the prequalified procedures, the prequalified preheat and interpass tables, the visual acceptance criteria, and the NDT requirements and acceptance levels. The clause and table numbers were reorganized in the 2020 and later editions, so refer to requirements by topic and confirm the exact citation against the adopted edition rather than carrying an old table number onto a new job.

AWS D1.8 is the seismic supplement to D1.1, and it governs demand-critical and seismic welds together with AISC 341, the seismic provisions for structural steel buildings. AISC 360 is the base specification for structural steel design and references the welding codes for fabrication and erection. Base metals are specified by ASTM, commonly A992 for wide-flange shapes, A572 and A36 for plate and shapes, and A6 for general rolled-shape requirements, with the carbon equivalent and toughness of the steel feeding the preheat and consumable decisions.

The engineer of record sets the demand-critical designations, the toughness requirements, and the inspection extent on the contract documents, and the authority having jurisdiction enforces the adopted building code and its special-inspection requirements. Where the contract is stricter than the code, the contract controls. Confirm the adopted editions and any local amendments before citing a specific clause on a submittal.

Units, terms, and conversions

Heat input is reported in kilojoules per inch on AWS jobs and kilojoules per millimeter on metric work, and one kilojoule per inch is about 0.039 kJ/mm, so divide a per-inch figure by 25.4 to get per-mm. The same weld reads as a different number depending on whether the value is arc energy or true heat input with the efficiency factor applied, so always note which convention the WPS uses.

Travel speed is inches per minute or millimeters per minute, voltage is volts at the arc, and current is amps. Toughness is reported as Charpy V-notch energy in foot-pounds or joules at a stated test temperature, where 20 ft-lbs is about 27 J. Preheat and interpass temperatures run in degrees F on most US jobs and degrees C on metric drawings.

Heat input

Energy delivered per unit length of weld, kJ/in or kJ/mm, controlling cooling rate

Arc energy

The formula result before the process efficiency factor, what AWS practice calls heat input

WPS

Welding procedure specification, the recipe bounding process, filler, preheat, interpass, and parameters

PQR

Procedure qualification record, the tested evidence the WPS is built on

HAZ

Heat-affected zone, the base metal beside the weld whose properties the cooling rate changes

Interpass temperature

Joint temperature just before the next pass, the practical heat-input control on multipass welds

CVN

Charpy V-notch impact toughness at a stated temperature, in foot-pounds or joules

Demand-critical weld

A weld the seismic system relies on to yield without fracture, under AWS D1.8 and AISC 341

CJP

Complete joint penetration, a full-thickness groove weld, the usual UT-inspected case

Carbon equivalent (CE)

A weighted measure of alloy content that drives preheat and crack susceptibility

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is a special inspection on a construction project?
• Cast-in vs post-installed anchors: which should I use?
• What is the difference between a snug-tight and a pretensioned joint?
• What is structural steel erection?
• How much does a high-density AI rack weigh?