This paper presents a comprehensive investigation into the repair welding process for large steel castings, specifically focusing on a steam turbine casing made of ZG20CrMoV. The study addresses the critical issue of secondary defect formation during the manual repair of casting defects such as cracks, gas pores, and sand inclusions. Using Simufact Welding software, I developed a finite element model to simulate the temperature and stress fields during multi-layer, multi-pass repair welding. The objective was to optimize the welding process parameters to reduce residual stress and prevent the formation of new defects. I evaluated two repair schemes: one using a matched heat-resistant steel electrode with hot welding, and another using a stainless-steel electrode with semi-hot welding. Through detailed thermal and stress analysis, I identified the optimal scheme and subsequently refined its parameters. I verified the simulation predictions with actual repair welding experiments and non-destructive testing. The results confirm that proper process parameter selection is crucial for achieving defect-free repairs. The optimized process significantly reduced equivalent stress and plastic strain, and X-ray inspection showed no defects. This research provides valuable theoretical and practical guidance for improving the repair quality and efficiency of large castings with casting defects.
Large-scale industrial castings, such as steam turbine casings, are essential components in power generation systems. These components often operate under high temperature and high pressure, demanding exceptional reliability. The manufacturing process involves sand casting, which is favored for its flexibility and cost-effectiveness. However, due to their complex geometry and substantial size, these castings are highly prone to various casting defects. These defects, including cracks, shrinkage cavities, gas porosity, sand inclusions, and cold shuts, can severely compromise the mechanical integrity and service life of the component. The presence of these casting defects is unavoidable in many practical scenarios, and repair welding is the most common and effective method to restore the component’s integrity. During the repair process, the localized heating and rapid cooling introduce significant thermal gradients and residual stresses, which can lead to secondary defects such as cracks and porosity in the weld metal and heat-affected zone. This not only reduces the quality of the repair but also increases costs and production time. Therefore, it is extremely important to understand the thermal and stress fields during repair welding and to optimize the process parameters to minimize the risk of new defects.
The repair welding of large castings is primarily a manual operation, often conducted under harsh thermal conditions. The specific challenges associated with this task include:
- High thermal gradients and cooling rates that promote the formation of martensite in low-alloy steels, leading to cold cracking.
- The large size and high rigidity of the casting produce high restraint stress.
- The composition of cast steels, with higher levels of impurities, increases the sensitivity to hot cracking and reheat cracking.
- The necessity to maintain high preheat and interpass temperatures, which is difficult in a manual operation.
- The complex and irregular shape of the defect cavity after preparation.
In this study, I selected ZG20CrMoV, a commonly used Cr-Mo-V low-alloy heat-resistant steel for turbine casings. The chemical composition and mechanical properties are detailed in the following tables. This material exhibits a carbon equivalent of about 0.81%, indicating poor weldability and high hardenability. The reheat cracking parameter is also positive, making the material sensitive to post-weld heat treatment cracking.
| Element | C | Si | Mn | Cr | Mo | V | Cu | S | P |
|---|---|---|---|---|---|---|---|---|---|
| Value | 0.18–0.25 | 0.20–0.60 | 0.40–0.70 | 0.90–1.20 | 0.50–0.70 | 0.20–0.30 | ≤0.30 | ≤0.030 | ≤0.030 |
| T (°C) | Cp (J/kg·°C) | λ (W/m·°C) | E (GPa) | α (10⁻⁶/°C) | σy (MPa) |
|---|---|---|---|---|---|
| 20 | 420 | 41.2 | 216.7 | 11.86 | 490 |
| 100 | 448 | 43.1 | 212.5 | 12.18 | 441 |
| 200 | 502 | 46.2 | 206.3 | 12.58 | 415 |
| 300 | 515 | 44.0 | 198.2 | 13.18 | 419 |
| 400 | 527 | 41.8 | 190.5 | 13.06 | 419 |
| 500 | 552 | 39.0 | 178.8 | 13.96 | 392 |
| 550 | 560 | 37.4 | 174.3 | 14.10 | 304 |
| 600 | 572 | 36.1 | 169.5 | 14.20 | 251 |
The weldability assessment of ZG20CrMoV was performed using standard formulas. The International Institute of Welding (IIW) carbon equivalent formula is given by:
$$CE_{IIW} = \left[ C + \frac{Mn}{6} + \frac{Cr+Mo+V}{5} + \frac{Ni+Cu+Si}{15} \right] \times 100\%$$
For ZG20CrMoV, the calculated value is 0.81%, which is well above the commonly accepted threshold of 0.45%, indicating a high susceptibility to cold cracking. Additionally, the reheat cracking sensitivity parameter is calculated as:
$$\Delta G = Cr + 3.3Mo + 8.1V + 10C – 2$$
And the more precise formula for the same is:
$$\Delta P_{SR} = \left( Cr + Cu + 2Mo + 10V + 7Nb + 5Ti – 2 \right)$$
Plugging in the actual composition gives a value around 1.9, confirming that the material is sensitive to reheat cracking. These characteristics dictate that repair welding must be performed with preheat and careful control of heat input and interpass temperature.
To study the repair welding process, I established a three-dimensional computational model. Since the actual turbine casing is massive (approximately 10 m × 5 m × 2 m), simulating the entire component would be computationally prohibitive. I selected a local region around the defect of dimensions 600 mm × 300 mm × 250 mm, which is sufficiently large to capture the thermal and mechanical effects of the repair without edge interference. I created a defect cavity with dimensions 180 mm × 60 mm × 40 mm, prepared with a slope angle of 12° and a root radius of 8 mm, consistent with practical preparation guidelines. A characteristic measurement point was chosen near the weld zone to record thermal cycles during the repair.
I used Simufact Welding software for the finite element analysis. The software employs the MSC.Marc solver and supports advanced thermal and mechanical coupling. For the heat source, I used a volumetric heat source model, which accurately represents the energy distribution of an electric arc welding process. The heat source parameters were defined by the welding current, voltage, efficiency, and travel speed. The governing equation for the transient heat conduction is:
$$\rho C_p \frac{\partial T}{\partial t} = \frac{\partial}{\partial x}\left(k \frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left(k \frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left(k \frac{\partial T}{\partial z}\right) + Q$$
where $\rho$ is density, $C_p$ is specific heat capacity, k is thermal conductivity, T is temperature, t is time, and Q is the internal heat generation rate from the arc. The boundary conditions include convection and radiation heat transfer to the environment (25°C) with a convection coefficient of 10 W/m²·°C. The initial temperature of the workpiece was set to 20°C.
I compared two repair schemes that are commonly used in industry:
| Parameter | Scheme 1 (Hot welding with R317) | Scheme 2 (Semi-hot welding with A307) |
|---|---|---|
| Electrode type | R317 (Cr-Mo-V heat-resistant steel) | A307 (austenitic stainless steel) |
| Electrode diameter | 5.0 mm | 5.0 mm |
| Current | 200 A | 200 A |
| Voltage | 28 V | 28 V |
| Travel speed | 5 mm/s | 5 mm/s |
| Preheat temperature | 250°C | 150°C |
| Interpass temperature | 200°C | 100°C |
| Post-weld heat treatment | 700°C × 6 h | None |
Both schemes used a reciprocating (back-and-forth) welding pattern along the defect length. The defect was filled with eight layers, each layer having a thickness of about 5 mm, with the number of passes per layer varying due to the slope angle. The total number of passes was 36. I simulated both schemes and analyzed the resulting temperature fields.
The temperature field distribution after specific welding times is shown in the following description. During the initial stages of welding, the molten pool forms with a peak temperature around 1648°C. The heat-affected zone expands gradually. In the first scheme, because of the higher preheat, the temperature distribution was more uniform and the cooling rate was slower, which is beneficial for preventing martensite formation. The second scheme exhibited a narrower heat-affected zone due to the lower preheat, but the faster cooling increased the risk of hardening. The temperature gradients in the workpiece were found to be quite steep, reaching maximum values of about 4909 K/m by the eighth layer. This large gradient is a primary driver of residual stress.
To more thoroughly understand the thermal cycles, I extracted temperature-time histories at a characteristic point near the weld. The curve shows multiple thermal peaks corresponding to each pass of the welding arc as it approaches and recedes. In the first layer, the point quickly reaches a peak temperature above the melting point, then drops sharply. In subsequent layers, the peak temperatures gradually decrease because the point becomes farther from the heat source as new layers are deposited. The interpass temperature remained at around 200°C between passes, indicating good thermal control. However, the rapid cooling from peak to base temperature in each cycle can lead to the formation of hardened microstructures and associated cracking.
The stress field evolution was simulated using a thermo-elastic-plastic constitutive model. The von Mises equivalent stress distribution was extracted at various stages. In the initial layers, the maximum equivalent stresses occurred near the weld start and end positions. As welding progressed, the stress distribution became more uniform, with a concentration at the bottom of the repair zone. This is attributed to the constraint offered by the base metal and the accumulation of plastic strain. The maximum equivalent stress reached approximately 310 MPa in the original scheme, which is close to the yield stress at the corresponding temperature. This indicates that local plastic deformation is likely. The x, y, z component stresses all showed a similar pattern, with tensile stresses at the surface and compressive stresses near the fusion boundary.
I also examined equivalent elastic and plastic strains. The original scheme produced a maximum equivalent plastic strain of 0.24, which is significant and implies that the repaired area has undergone substantial permanent deformation. This can lead to dimensional instability and create locations for crack initiation. The yield stress map showed that the bottom region of the weld has a yield stress of approximately 297 MPa, which is close to the saturation value for the material after work hardening.
To validate the simulation results for the original scheme, I performed an actual repair welding operation on a scrap ZG20CrMoV component. The defect was prepared in the same geometry as the model. The welding parameters were exactly as those used in Scheme 1. The workpiece was preheated to 250°C using a flame, and the interpass temperature was monitored with an infrared thermometer. After welding, the workpiece was cooled slowly, and then X-ray radiography inspection was performed. The X-ray film revealed the presence of gas pores and a crack in the weld metal. This clearly indicated that the original process parameters were not suitable, confirming the simulation prediction of high residual stress and inadequate metal soundness. The gas pores likely resulted from rapid solidification and insufficient gas escape, while the crack was probably due to excessive thermal stress and possible hydrogen-induced cold cracking.
Based on these findings, I decided to optimize the repair process. The main adjustments were to reduce the heat input by lowering the current and voltage, and to reduce the preheat temperature slightly to lower the thermal gradient while still providing enough to avoid martensite. The optimized parameters are shown in the table below:
| Parameter | Original | Optimized |
|---|---|---|
| Electrode | R317 | R317 |
| Current | 200 A | 180 A |
| Voltage | 28 V | 26 V |
| Travel speed | 5 mm/s | 5 mm/s |
| Preheat | 250°C | 200°C |
| Heating rate | 150 °C/h | 100 °C/h |
| Interpass temperature | 200°C–350°C | 200°C |
| Post-weld heat treatment | 700°C × 6 h | 700°C × 6 h |
I validated the optimized process using the same finite element model. The temperature field after optimization showed a slightly lower peak temperature, around 1517°C, and a more uniform distribution. The maximum equivalent stress at the end of welding was reduced to 266 MPa, which is below the material’s yield limit. The X, Y, Z component stresses were 227 MPa, 213 MPa, and 329 MPa, respectively, all lower than before. The equivalent elastic strain decreased to 0.065, and the plastic strain dropped dramatically to 0.09. This indicates a significant reduction in permanent deformation and stress concentration, making the risk of cracking much smaller.
I also performed a simulated post-weld heat treatment (PWHT) at 700°C for 6 hours. The thermal stress analysis after PWHT showed that the maximum equivalent stress was further reduced to about 100 MPa, which is a very low level, ensuring excellent structural stability. The plastic deformation rate remained minimal at 0.06. This demonstrates that the PWHT is effective in relieving residual stresses and improving the overall quality of the repaired zone.
To confirm the optimized process in practice, I repeated the actual repair welding on a separate ZG20CrMoV workpiece. The defect cavity was prepared to the same dimensions (250 mm × 40 mm × 30 mm, slope 12°). The welding was performed with the optimized parameters. The preheat was 200°C with a controlled heating rate of 100°C/h. The interpass temperature was strictly maintained at 200°C. After welding, the component was subjected to the PWHT cycle in a furnace. Then, X-ray inspection was performed again. The X-ray film showed a uniform and clear image without any indications of pores, cracks, or lack of fusion. The repair was judged to be defect-free and fully compliant with the quality requirements.
These results strongly support the benefits of simulation-guided process optimization. The original parameters, which were based on general industrial practice, led to unacceptable defects. The optimized parameters, derived from systematic simulation analysis, successfully eliminated defects and produced a high-quality repair. The key factors that contributed to this improvement are:
- Lower heat input (reduced current and voltage) reduces the thermal shock and peak temperature, decreasing the thermal gradient and stress.
- An appropriate preheat temperature (200°C) still prevents martensite formation but reduces the energy consumption and minimizes the temperature difference between the weld and the base metal.
- Strict interpass temperature control ensures consistent cooling conditions and avoids overheating.
- The post-weld heat treatment at 700°C for 6 hours effectively relieves residual stress and tempers the martensitic structures, enhancing toughness.
I also analyzed the effect of the reciprocating welding path. The temperature field showed that the heat source alternately travels along the length, which creates a more uniform average temperature in the weld region compared to a unidirectional path. However, the stress distribution may still have local concentrations at the turning points. With the optimized heat input, these local stress concentrations remained below the yield limit, thus avoiding plastic deformation.
In summary, this research provides an in-depth understanding of the repair welding process for large castings with casting defects. The finite element simulation of temperature and stress fields is proven to be an effective tool for predicting weld quality and guiding process development. The optimized procedure for ZG20CrMoV turbine casings significantly reduces the risk of secondary defects and improves the reliability of the repaired component. The methodology can be extended to other materials and defect geometries, offering a practical route to improve repair quality in foundry industries.
Future work should focus on incorporating microstructural evolution models into the simulation, predicting the formation of brittle phases, and optimizing the repair strategy based on a full map of residual stress. The use of real-time monitoring and adaptive control in automated welding could further reduce human error. Additionally, exploring advanced heat source models, such as those for laser or hybrid welding, would enhance the applicability of the simulation framework.
Finally, this work highlights the significance of combining numerical simulation with experimental validation. The direct correlation between the simulated stress concentration and the real crack occurrence confirms the reliability of the model. This integrated approach can lead to significant improvements in productivity, cost savings, and product safety in the repair of large steel castings.