Post weld heat treatment simulation for thick welds and local heating bands
A thick weld needs stress relief, and the furnace slot or the heating plan is the bottleneck. Simulating the soak, the heated band and the gradients shows how much residual stress is left and whether the cycle adds new stress, before the first heating pad is wired.
Post weld heat treatment simulation predicts how much residual stress a stress-relief cycle leaves in a weld, how hot each part of the joint gets, and whether heating and cooling add new stress. We build that model around your existing thermal-structural solver, so a soak, heating band or cooling rate is tested before the furnace is booked or the pads are wired.
The last seam of a thick-walled vessel is welded and the specification calls for post weld heat treatment. The vessel will not fit the furnace, so the plan is a local cycle: heating pads on a band round the seam, insulation, a controller. How wide must the band be, how fast may it heat, how steep may its edge gradient be, and how much stress is left for the fracture and fatigue assessment that follows? A band too narrow or a ramp too fast puts new stress into the shell; one wider than needed costs time for stress already low enough.
What does post weld heat treatment do to residual stress?
As a weld cools, the metal contracts against the colder metal around it and the stress rises to the yield strength. TWI describes the next step: reheating lowers the yield strength, so the residual stress redistributes by creep, and what remains is bounded by the proof strength at the soak temperature. Tempering of hard heat-affected zones happens in the same range, so one cycle does two jobs.
The model carries both mechanisms. During heating the stress follows the falling yield strength down. During the soak, creep turns elastic strain into inelastic strain at constant total strain: for a Norton law, dεcr/dt = Aσne−Q/RT, and modulus E, dσ/dt = −EAσne−Q/RT, which integrates to σ(t) = [σ01−n + (n−1)EA e−Q/RT t]−1/(n−1). Relaxation is fast at first, then slows, and temperature sits in the exponent, so the hold temperature moves the result strongly.
- Heating ramp. The heating rate is limited. A thick section heats from the outside in, and the temperature difference across it turns into stress. Too fast, and a restrained part can distort or crack on the way up.
- Yield strength falls with temperature. Residual stress cannot stay above the yield strength of the hot metal, so it is pushed down as the part heats. The dashed line is that limit: the stress follows it down and then leaves it.
- Soak and creep relaxation. Held at temperature, creep turns elastic strain into inelastic strain and the stress keeps falling, quickly at first and then slowly. Time and temperature trade against each other, and hard heat-affected zones temper in the same hold.
- Controlled cooling. Cooling is limited too. The yield strength recovers as the part cools, so the stress is locked in at the level it reached, and a steep temperature difference during cooling can put new stress in.
- Stress that remains. What is left after cooling is lower than what welding left, but not zero. It is the input a fracture or fatigue assessment needs, and the quantity the simulation reports at the locations you choose.
- As-welded stress. Weld shrinkage held back by the cold metal around it leaves tensile stress that is often close to the yield strength at room temperature. The model starts from this state.
Neither mechanism takes the stress to zero. In the pipeline butt weld of the SINTAP residual stress paper (30 mm wall, a Norton-type creep law, three hours at 550 °C, controlled cooling), the stress after PWHT peaked near 10% of the weld-metal yield strength, inside the levels the paper quotes from BS 7910 of that date: 15% of yield transverse, 30% longitudinal. A fracture assessment takes that fraction as input. Fatigue is different: TWI notes that PWHT has little beneficial effect on fatigue performance unless the stresses are mostly compressive, so credit earned in a fracture assessment does not carry across to a fatigue assessment.
Why are thick sections and local heating bands hard to get right?
Codes tie the need for PWHT and the soak time to thickness, and TWI gives one hour per 25 mm as a general rule, so a thick shell means a long cycle and often a wait for a furnace. A thick wall also heats unevenly, and at soak temperature steel is weak: TWI warns that a component may be unable to support its own weight, so supports and sag belong in the model.
A local cycle adds geometry. TWI's note on local PWHT separates a soak band around the weld, a wider heated band under the elements and a gradient control band under the insulation, with gradients along the axis and through the thickness. The edge of the heated band is generally required to reach at least half the soak temperature. A region held back by cooler metal and heated by ΔT more carries a thermal stress of order EαΔT, and because yield is low at soak temperature, a modest difference exceeds it. Uneven or rapid heating can therefore leave new residual stress on cooling, as TWI states.
- Weld and heat-affected zone. The seam and the metal beside it hold the residual stress that welding left. The model starts from this stress field, taken from a weld simulation or from a measured or code profile.
- Heated band. Heating elements on the outer surface cover the heated band, and insulation over them cuts the loss of heat. Element zones can be controlled separately, so the heating can be shaped to the weld.
- Soak band. The weld, its heat-affected zones and some parent metal are held within the specified temperature and time. This is where stress relaxes by yield-strength drop and creep.
- Axial gradient. Temperature falls from the soak band to ambient under the insulation. A steep fall or a fast ramp makes hot metal push against cold metal, and the stress can exceed the hot yield strength and leave new residual stress. A common check is that the edge of the heated band reaches at least half of the soak temperature.
- Through-thickness gradient. Heat is lost from the inner surface, so the inside is cooler than the outside and the soak band is narrower there. The larger the diameter, the greater the loss, which is one reason band widths are tied to thickness.
- Control thermocouples. Thermocouples at the weld centre line and at the edges of the soak band and the heated band feed the controller and the record. The same positions are probes in the model, so a recorded cycle can be set beside the simulated plan.
How is PWHT specified today, and where does that break?
Construction codes set when PWHT is required, by material and thickness, and give temperature, soak and rate limits. TWI tabulates the PWHT temperatures of EN 13445, ASME VIII and PD 5500, and the local band rules of ASME VIII, ASME B31.3 and EN 13445. Such rules are written to be checked, not to show the stress you get. Four things go wrong beyond them.
- The rule stands in for the stress. A band width or half the soak temperature is met on paper while the stress at the weld toe that a fracture assessment needs is never computed.
- Odd geometry. A nozzle or a repair beside a support has no standard band. TWI's paper on avoiding PWHT describes a boiler stub repair where heating a band round the header was cumbersome, and notes that ASME allows a patch width to vary if the gradients are not harmful, with FEA able to compare the varied patch with the code band.
- A single flat number. A fraction of yield cannot tell a clean cycle from one whose gradients added stress.
- Waiver cases. The scope of BS 7910:2019 lists justifying waiver of post-weld heat-treatment in thick-walled steel structures. TWI adds that PD 5500:2003 permits the route subject to the agreement of the concerned parties, and that it cannot justify a general relaxation of the code criteria. The stress put in decides the result.
| Question | What the model carries | What it reports | Decision it supports |
|---|---|---|---|
| Furnace cycle for a thick shell | Furnace control, heating through the thickness, supports at temperature | Temperature difference across the section, stress after cooling, sag | Heating and cooling rates, thermocouple positions, supports |
| Local band on a circumferential seam | Heater zones, insulation, inner-surface loss, welding stress as the start state | Axial and through-thickness gradients, stress at the weld and heat-affected zone | Band, heater and insulation widths, ramp rates |
| Nozzle, stub or repair with no standard band | Real geometry and restraint | Stress from a varied band beside the code band | Whether to propose the varied band to the authority |
| Waiving PWHT by fracture assessment | As-welded and PWHT stress at the postulated flaw | Stress distribution at the flaw | The stress input the parties agree to assess with |
| Changing a qualified cycle | Relaxation by yield drop and creep | Stress left against time and temperature | Whether a change is worth requalification |
How is post weld heat treatment simulation done properly?
The chain has four links. First, the starting state: PWHT acts on the stress welding left, so the model starts from a weld simulation of the same joint, or a measured or code profile when the weld exists. The table of contents of API 579-1/ASME FFS-1 (December 2021) lists, in Annex 9D on residual stresses in a fitness-for-service evaluation, residual stress distributions from welding simulation, and post weld heat treatment among the modifying factors.
Second, the thermal model: heater zones, insulation, controller thermocouples and surface losses, so gradients are computed, not assumed. Third, the stress model: temperature-dependent yield, modulus and expansion and a creep law calibrated for relaxation, run through heating, soak and cooling. TWI's paper asks for elastic-plastic FEA that allows relaxation during PWHT and includes the initial welding stresses, validated by measuring the stress after a patch PWHT. Fourth, the outputs: gradient history against your limits, stress along the lines your assessment uses, and sag at temperature.
Verification: mesh and time-step convergence, an energy balance, and the creep law reproducing its own relaxation test. Validation: thermocouple records from an earlier cycle and a residual stress measurement on the same joint, against tolerances set before the run. Uncertainty: creep constants, starting stress, insulation, thermocouple contact and band position, swept so the result is a range. The numbers are estimates that depend on material data and calibration.
A stress estimate is not a PWHT approval
Where does AI help post weld heat treatment, and where does it not?
- Plan screening. Runs planned over band width, insulation, ramp rate, thickness and diameter train a surrogate that returns gradients and the stress left at the weld for any plan inside its range; full runs are kept for the plans that decide. See surrogate models for CFD and FEA.
- Calibration. Relaxation constants are fitted to your test data by an inverse method that returns a range, which goes into the uncertainty record.
- Record checking. A script compares thermocouple records with the plan and flags deviations outside the tolerances you set.
It does not replace material data. A surrogate cannot predict a cycle outside its training range, inherits every error of the thermal-structural model, and never decides whether a stress is acceptable. The metallurgy stays with testing: TWI warns that exceeding the original tempering temperature risks strength below the specified minimum. See AI surrogate models you can sign off and the thermal stress article.
From one heating plan to a PWHT capability you keep
The return comes when your engineers can put the next weld, band layout or cycle change through the same chain without rebuilding anything.
- Physics model. CFD and FEA set up once, as templates: geometry clean-up, mesh, solver settings and post-processing that every later run reuses.
- Validation. Mesh and time-step studies, and comparison with tests or full-scale data, so the model's uncertainty is known before it is trusted.
- Automated runs. Scripted runs over the parameters and operating conditions that matter, on your hardware or in the cloud.
- AI surrogate. A machine-learning model trained on those runs answers in seconds and says how sure it is; low-confidence questions go back to the solver.
- Decision tool. Optimisation, trade-off charts or operating guidance, in a form your engineers and operators use every day.
- Field data. Test, trial and monitoring data flow back to recalibrate the model, so it improves with every project instead of being rebuilt.
- A cycle file holding soak, rates, band, heater and insulation widths, thermocouple positions and supports.
- A thermal template in your existing thermal-structural solver, Abaqus, Ansys or LS-DYNA, with heater zones, insulation and controller thermocouples built in.
- A stress template that starts from the weld residual stress field, runs the cycle with temperature-dependent properties and a relaxation law, and writes the stress along your assessment lines.
- A validation record against a recorded cycle and a residual stress measurement on your own joint, with tolerances set before the run.
- A surrogate with a stated range in thickness, diameter, material and band type, and a gate that sends anything outside it to the full model.
- Handover and training, so your team runs and extends it. See our training and software development services, and the welding simulation hub.
Questions engineers ask about post weld heat treatment
What is the purpose of post weld heat treatment?
How long is the soak time for PWHT?
Does PWHT remove all residual stress?
Does PWHT improve fatigue life?
Can simulation replace PWHT or a heat treatment procedure?
If a heating plan or a PWHT waiver is waiting on a stress number, send us the joint drawing with thickness and diameter, the material and its PWHT requirement, the planned cycle with band and heater layout, and any thermocouple record or residual stress measurement you hold. We reply with how we would build the model, the gradients and stress lines it would report, and what the validation record should contain. Start with a conversation, or read the stress relief capability page.
