Issue No. 251〔Technical Papers〕

Numerical Simulation of Welding Deformation produced in Compressor Impeller

Author

Shinji KOBAYASHI*

Esao YAMADA**

Tetsu GO**

Shigetaka OKANO***

Masahito MOCHIZUKI***

Kenyu KIMURA****

Akiyoshi ANDO****

*

Production Process Innovation Division (Formerly Graduate School, Osaka University)

**

Production Process Innovation Division

***

Graduate School, Osaka University

****

Elliott Ebara Turbomachinery Corporation

Among many problems with welding is welding distortion. Since welding distortion significantly affects the performance and reliability of products, it has become increasingly required in recent years to appropriately foresee, control, and reduce distortion. One of the approaches to this issue is numerical analysis. As significant improvements have been made to welding dis- tortion analysis technology, its application has been increasing every year. So far, however, numerical analysis has only been applied to a limited number of large scale and/or complicated structures. Under such circumstances, with a compressor impel-ler having complicated structure as a target, distortion analysis and experiment was performed. The analysis and experiment results have revealed that it is possible to evaluate welding distortion occurring to an impeller with high accuracy by using numerical analysis with the help of the construction of a heat source model and taking into account phase transformation.

Keywords: Compressor, Impeller, Welding distortion, Numerical simulation, Distortion measurement, Phase transformation, Martensitic stainless steel, Displacement of discharge width, Displacement of cover height

1. Introduction

In fabricating machines and structures, welding plays an important part as a basic manufacturing technology. However, welding can be accompanied by defects such as poor welds and cracks, unevenness in strength and toughness ascribable to a heat-affected zone, mismatch in the shape of, for example, a bead toe, and residual stress and distortion, which may result in the decreased strength of the structure and other problems. These phenomena must be appropriately predicted evaluated, and controlled to build sound welded joints/structures.

Among these considerations, welding distortion is perceived as a critical problem because it sometimes decreases the strength of structures as well as affects assembly accuracy in fabricating machines/structures and spoil the appearance of products. Today, welding distortion is cramped before the welding process to restrict the occurrence of distortion or by mechanically or thermally correcting the distortion after the welding process. However, incorporating such steps into the welding process interferes with enhancement in the efficiency of the production process and increases pro-duction costs; under the circumstances, it has become increasingly required in recent years to streamline such steps by appropriately predicting controlling, and reducing welding distortion.

In recent years significant improvement has been made in numerical simulation technology for welding distortion and residual stress1), 2), enabling more accurate prediction and evaluation of welding distortion. This improvement has largely resulted from the introduction of modeling technologies that take into consideration the material behavior associated with the thermal cycles of phase transformation3) and recovery4)/recrystallization as well as heat source characteristics5) based on welding arc physics. So far, however, numerical simulation has only been applied to limited numbers of actual machines and similar complicated/large structures.

In this study, we performed numerical simulation of welding distortion occurring in a compressor impeller with a complicated shape and conducted experimental measurements to compare and verify the simulation and measurement results as well as examine the occurrence characteristics of welding distortion.

2. Numerical simulation of welding distortion occurring in a compressor impeller and measurement experiment

2.1 Compressor impeller and the welding conditions for it

The target compressor impeller consisted of a hub, a cover, and blades as shown in Figure 1. It was fabricated by welding a machined structural component−a combination of the cover and blades−to the hub. The impeller had 13 evenly-spaced blades. With both sides tack welded, these blades were manually fillet welded (with a welding heat input of 3.7 kW and at a welding speed of 3 mm/s) with 26 passes in total. The welding sequence was as shown in Figure 2. In this measurement experiment, we allowed each weld to completely cool down to room temperature before moving on to the next weld in order to observe whether distortion occurred in each welding pass. The material under test was martensitic stainless steel.

Fig. 1 General view of the compressor impeller

Fig. 2 Welding sequence

2.2 Numerical simulation

Figure 3 shows a finite element model of the compressor impeller. As the boundary condition in heat transfer analysis, heat transmission and the thermal radiation that obeyed the Stefan-Boltzmann law was considered, and as the boundary condition in thermo-elastic-plastic analysis, the model is restricted only the rigid-body movement and rotation. Figure 4 shows examples of material properties used for the numerical analysis. In order to take into consideration the phase transformation behavior associated with the thermal cycle, yield stress and the temperature dependence of the linear expansion coefficient (thermal strain curve) as shown in the figure were used in numerical analysis. A heat-source model was modeled by reference to the temperature history and angular distortion at the welds of T-shaped fillet welded joints observed with the same welding heat input as for the compressor impeller. In this analysis, heat source length was 6 mm, and thermal efficiency was 0.6.

Fig. 3 Finite element model of the compressor impeller

Fig. 4 Material properties used in numerical analysis in consideration of phase transformation

2.3 Procedure for evaluating welding distortion

In the numerical analysis and measurement experiment, three types of distortions were evaluated as shown in Figure 5 radial-direction distortion in the hub and cover (indicated as Dx in the figure), displacement of the cover height (indicated as H in the figure), and displacement of the discharge width (indicated as C in the figure), with each distortion point indicated in the figure. Distortions were evaluated after each of the 26 welding passes to record the displacement history.

Fig. 5 Measurement targets and their positions

3. Results of the numerical simulation of welding distortion occurring in the compressor impeller and experimental measurements

3.1 Comparison between the results of the numerical simulation on the occurrence characteristics of welding distortion in the compressor impeller and experimental measurements

The results of the numerical simulation and measurement experiment show that the radial-direction displacement D4 of the central opening width of the hub is almost 0, that the radial-direction displacements D1 and D3 of the central opening width of the cover are rather large, and that the radial-direction displacement D2 in the cover rim is somewhat small: no distortion is substantially large. This paper reviews in detail the displacement of the cover height H and the displacement of the discharge width C, which are larger than the displacements above. In the experimental measurements, assuming the state right after tack welding to be the initial state, we defined the changed amount from the initial state as distortion. In the numerical simulation, we used an integrated model that does not take tack welding into consideration was used.
Figures 6 and 7 show the displacement after each of the 26 welding passes revealed by the results of the numerical simulation and experimental measurements. At any position, distortion increases in the negative direction as the number of welding passes increases, ending up with the cover falling toward the hub and the discharge width being narrower after finishing all 26 passes. In addition, the results of both the numerical simulation and experimental measurements show similar quantitative tendencies of distortion occurrences, which has proven that the accuracy of the numerical simulation is excellent.

Fig. 6 Simulated and measured displacement H

Fig. 7 Simulated and measured displacement C

3.2 Consideration of the occurrence characteristics of welding distortion in a compressor impeller

In any type of distortion, the amount was relatively large when the area near the appropriate measurement position was welded, which is particularly notable for the displacement C of discharge width. In other words, when the two welding lines to the sides of a measurement position were welded, it was obvious that the discharge width becomes substantially narrower, which probably resulted from the angular distortions caused by the welding. On the other hand, when the welding lines symmetrical to the two welding lines mentioned above were welded, it was also observed that the displacement of discharge width tended to be slightly larger, revealing that the entire structure exhibited a distortion occurrence behavior ascribable to the complicated shape. The displacement of the cover height also exhibited a similar tendency while it was observed that in some cases, the cover was distorted to a relatively large extent when welding was performed away from the measurement position. This suggests that the displacement of the cover height was caused by welding-direction shrinkage as well as, unlike the displacements of the discharge width, the balance of the entire structure with a complicated shape.

As described above, through an accurate numerical simulation and detailed experimental measurements, the occurrence characteristics of welding distortion in the compressor impeller were examined to reveal that the following particularly large distortions occur: a distortion observed as a displacement of the cover rim and a distortion observed as a decreased discharge width between the hub and cover. In the future, the design and construction methods for reducing distortion based on this numerical simulation should be examined.

4. Conclusion

In this study, welding distortion occurring in a compressor impeller was calculated with numerical analysis and measured in experiment. And the accuracy of the simulation is verified based on a comparison between the simulation and experiment results as well as examining the occurrence characteristics of welding distortion. As a result, the following conclusions were reached:

（1）

The welding distortions in the compressor impeller obtained through the numerical simulation performed in this study substantially agree with the experimental measurement results quantitatively or otherwise, revealing that the simulation is accurate.

（2）

The review result of the welding distortions occurring in the compressor impeller has revealed that while radial-direction distortions are not so large, the following distortions are relatively large: a distortion observed as a fall of the cover rim in the hub (displacement of cover height) and a distortion observed as a decreased discharge width between the hub and the cover of each blade (displacement of discharge width).

（3）

The displacement of discharge width is large when two blades to the sides of the discharge width are welded, which is deemed to be caused by angular distortion.

（4）

The displacement of the cover height is deemed to be caused by shrinkage along the welding lines; in some cases, a large displacement is observed even when a blade away from the measurement position is welded. This suggests that this distortion can be also caused by the balance of the entire structure with a complicated shape.

5. Acknowledgments

We would like to thank Professor Hidekazu Murakawa at the Joining and Welding Research Institute, Osaka University for giving kind consideration to the implementation of our numerical simulation. We are grateful to the following organizations.

References

1) D. Deng, H. Murakawa and W. Liang: Numerical Simulation of Welding Distortion in Large Structures, Comput. Methods Appl. Mech. Eng., 196 (2007), 4613-4627.
2) M. Shibahara, K. Ikushima, S. Itoh and T. Masaoka: Com-putational Method for Transient Welding Deformation and Stress for Large Scale Structure Based on Dynamic Explicit FEM, Quar. J. JWS, 29-1 (2011), 1-9. (in Japanese)
3) Y. Mikami, M. Mochizuki and M. Toyoda: Angular Distortion of Fillet Welded T Joint using Low Transformation Temper-ature Welding Wire, Sci. Technol. Weld. Joining, 14-2 (2009),97-105.
4) R. Ihara, S. Okano, T. Hashimoto, Y. Mikami and M. Mochizuki:Visualization of Machining and Welding Residual Stress Variation by Numerical Simulation in Austenitic Stainless Steel, Proc. Visual-JW 2012, 1 (2012), 313-314. 5) S. Okano, M. Tanaka and M. Mochizuki: Arc Physics based Heat Source Modeling for Numerical Simulation of Weld Re- sidual stress and Distortion, Sci. Technol. Weld. Joining, 16-3 (2011), 209-214.

Reprinted with permission from the Welding structure Symposium 2014 Proceedings.

Recommended articles

Inquiry about Ebara Engineering Review