ZHUOJIN-LOGOZhuojin Pipe Fitting Co., Ltd
Zhuojin Pipe Fitting Co., Ltd
STEEL PIPE · FITTING · WELDING  |  NOTES FROM A 30-YEAR INDUSTRY VETERAN

Gaussian Surface Heat Source Loading Methods in TIG Welding Numerical Simulation: A Practical Review

Why This Paper Matters on the Shop Floor

I have spent more than three decades in the pipe and fitting business, and I have seen more than one welding procedure developed by trial and error when a decent simulation could have saved weeks of shop time. The paper by Zhang Weizhi and Cheng Yanlin from the China Academy of Engineering Physics addresses a deceptively simple question: how exactly do you apply a Gaussian surface heat source in a finite element model of a TIG weld? On the surface, this looks like a purely academic exercise. In practice, the choice of loading method changes the predicted peak temperature, the predicted weld width and penetration, and the convergence behaviour of the solver. If your simulation says the weld pool is 4 mm deep and the real weld is 5 mm deep, you have not saved time at all. You have simply moved the problem from the computer to the weld test coupon.

The authors used the ANSYS finite element software to simulate a TIG bead-on-plate weld on a flat plate. They compared three ways of loading the Gaussian heat flux: applying heat flux density to surface elements, applying volumetric heat generation rate to the surface layer of elements, and applying heat generation rate to the nodes on the surface layer. The findings are not what most engineers would guess, and they have direct consequences for anyone who uses simulation to set welding parameters for thin-wall pipe, small-diameter fittings, or repair welds on live equipment.

The Three Loading Methods Compared

The table below summarises the key differences observed in the study.

Loading Method How It Is Applied Main Observation Convergence Behaviour Sensitivity to Mesh
Surface element heat flux density Heat flux applied to the faces of surface elements Face numbering of elements significantly affects results; heat flux coverage problems occur when element faces do not align with the heat source distribution Acceptable but unstable Very high; results change with element face orientation
Surface element volumetric heat generation rate Heat generation rate applied to the volume of the top layer of elements The thickness of the surface layer affects the peak temperature; convergence is poor Poor; solver struggles to converge Moderate; peak temperature depends on layer thickness
Surface node heat generation rate Heat generation rate applied to the nodes on the surface layer The thickness of the surface layer affects the peak temperature, but the calculated weld width and penetration do not change significantly with layer thickness Good; stable and fast Low for weld width and penetration; moderate for peak temperature

The first method, surface element heat flux density, is the most intuitive and the most commonly used in the literature. The paper shows that it is also the most dangerous. Because the heat flux is applied to the faces of the elements, the result depends on which face of which element is selected. If the mesh is unstructured, or if the element faces are not perfectly aligned with the Gaussian distribution, you get heat flux coverage gaps or overlaps. I have seen this in practice when a young engineer tried to simulate a TIG root pass on a 6 mm wall pipe and ended up with a predicted penetration that was 30% off because the element face numbering was inconsistent. The second method, volumetric heat generation rate on the surface layer, avoids the face numbering problem but introduces a new one: the thickness of the surface layer becomes a user-defined parameter that directly affects the peak temperature. If you choose a layer that is too thick, you dilute the heat and underestimate the peak temperature. If you choose a layer that is too thin, you concentrate the heat and may get a spurious spike. The convergence is also poor because the volumetric heat source creates a steep temperature gradient that the solver has trouble handling.

The third method, surface node heat generation rate, is the one I would recommend for most TIG simulations. The paper shows that the calculated weld width and penetration are essentially independent of the surface layer thickness, which is a huge practical advantage. The convergence is good, and the results are stable. The only downside is that the peak temperature still depends on the layer thickness, but for most welding process development purposes, what you care about is the fusion boundary geometry, not the exact peak temperature in the centre of the arc.

Practical Implications for Process Development

When I sit with a process engineer who wants to use simulation to set TIG parameters for a new alloy or a new joint configuration, I always ask one question: what are you going to do with the result? If the answer is "predict the weld width and penetration," then the surface node heat generation rate method is the right choice. If the answer is "predict the peak temperature to check for over-tempering or grain growth in the HAZ," then you need to be much more careful about the mesh and the loading method, and you should probably validate the model against a thermocouple measurement on a real weld coupon.

This paper also reminds me of a common mistake in the industry: using a simulation model that was validated for one material and one thickness, and then applying it to a completely different case. The Gaussian heat source parameters, the arc efficiency, and the loading method all interact. A model that works for carbon steel may not work for stainless steel or titanium, because the thermal conductivity and the melting point are different, and the heat input distribution changes. The authors do not go into this, but the lesson is implicit: if you change the material, revalidate the model.

Lessons and Recommendations

I have taken three concrete lessons from this paper. First, the loading method is not a cosmetic detail. It is a modelling decision that can change your predicted weld geometry by tens of percent. Second, the surface node heat generation rate method is the most robust for predicting weld width and penetration, and it should be the default choice for TIG bead-on-plate and root pass simulations. Third, if you use the surface element heat flux density method, you must carefully control the element face numbering and ensure that the mesh is structured and aligned with the heat source. In my experience, this is rarely worth the effort.

For the pipe and fitting industry, this has a direct application. We weld TIG root passes on everything from small-bore stainless steel instrument lines to large-diameter CRA clad pipe. The root pass geometry controls the quality of the subsequent fill passes, and the penetration is critical for avoiding lack of fusion or burn-through. A reliable simulation model, properly loaded, can help us optimise the root pass parameters without burning through a dozen test coupons. That is real money saved, and this paper tells you how to do it right.

The final word is this: simulation is a tool, not a substitute for welding trials. But if you are going to use it, use it with the right loading method, validate it against a real weld, and never trust a model that has not been verified against a thermocouple or a macro-section. That is the difference between a simulation that saves money and a simulation that costs you a failed weld test.