Nick Bagshaw1, Andrew Ezeilo1, Liwu Wei1, Frederick W. Brust 2 and J. C. Kennedy 2
1 TWI, UK
2 Battelle, USA.
Paper presented at ICCAS 2005 23-26 August 2005, Busan, Korea.
Copyright in this paper is owned jointly by TWI Ltd and Battelle Memorial Institute, first published in 2005. Copyright in the typesetting of this paper for this proceedings is owned by ICCAS.
Virtual Fabrication Technology (VFT TM ) is an emerging computer software package for prediction and control of distortion in welded structures. The remedial action to rectify distortion from welding is extremely expensive and time consuming. Despite efforts made in the industry to reduce distortion there are inherent difficulties in estimating the welding conditions, welding sequencing and the configuration of stiffeners. Dedicated weld simulation tools such as VFT TM can provide a good scientific basis for predicting welding residual stresses and distortion.
VFT TM has been demonstrated to predict distortion in ship panels. In analysing relatively large structures such as ship panels, it is important to consider the time to obtain a solution as well as the accuracy of the predicted distortion. Solution times up to 20 times faster than conventional finite element software packages have been achieved using VFT TM , which takes advantage of optimised solution controls and advanced material modelling capabilities.
In this paper examples will be presented that confirm the suitability of VFT TM to adequately simulate the distortions produced by welding in shipbuilding applications. In addition, the efficiency of VFT TM will be considered in the context of production schedules in the shipbuilding industry. Finally, the benefits that can be achieved by using VFT TM to improve the welding sequence and improve stiffener configurations to control distortion will be presented.
Lightweight structures are being increasingly used in recent years in both U. S. shipyards and European shipyards. From 1990 to 2000, the production ratio of thin steel (6-mm or less) to plate structures for certain specific vessels has risen to over 90% by weight (e.g. Huang et al  ). Severe distortions have been observed in the building of large thin ship panels. Moreover, distortion problems with thick welded fabrication can be even more difficult to correct because mechanical straightening is often not practical. In addition, fabrication residual stresses often reduce both fatigue life and resistance to fracture during attack. To understand the distortion mechanism and propose the proper method to control the distortion, shipbuilders need a distortion prediction tool which is able to be used to assist the building of the ship panels.
Many researchers [see for instance Masubuchi  , Goldak  , Ueda and Yuan  , or Tsai et al  for instance] have tried to develop modelling methodologies to simulate the welding process. However, the modelling techniques that have been developed have often been complex, less accurate, or too labour intensive to be applied routinely in an industrial setting. Although significant progress has been made in the finite-element modelling of welding processes in recent years, many of the modelling techniques are still far short of being used successfully for the control of residual stress and distortion in actual large structures. Reference  provides an updated summary of weld modelling approaches and applications (particularly Chapters 7 and 8).
Since 1996, Battelle, Caterpillar and the Welding Institute have been working together to develop an industrial-use methodology and user-friendly software for predicting weld residual stress and distortion in large and complicated structures. References [6-8] and the many references sited therein document much of this work. A large amount of manpower and equipment was invested for model development, program coding, and validation. Finally, a Virtual Fabrication Technology (VFT) and aweld modelling computer tool was developed and is now used routinely in the Caterpillar design process worldwide. VFT is a state-of-the-art fabrication simulation technology that allows rapid solutions for large, complex metallic structures containing both single-pass and multi-pass welds and allows the user to consider or input all critical variables. It can be used in product design stages to help the weld design and in the manufacture stage to determine the optimal weld processes to minimize welding-induced distortion.
Computational weld modelling
In the fabrication of large and complex structures in heavy industries, welding-induced distortion is a major concern. It has been realized that some methods (Reference  , Chapters 7 and 8) developed for controlling welding-induced distortion in thin structures are not very effective for controlling welding-induced distortion in thick structures. Therefore, many distortion control techniques, weld design optimization, pre-straining, welding sequence and pre-cambering, were developed by repeated experiment and experience. This makes welding operations very costly and time-consuming. Moreover, there is ongoing research which considers the use of special consumables, which results in 'designer' phase transformations to control residual stresses and distortions  .
Most computational weld models which are available commercially are mathematics and physics modules; (i) thermal module and (ii) a structural model. The following is a description of the VFT  code but other codes are similar.
There are two main analysis modules ( Figure 1); the thermal model  and the structural model  , that make up the weld process simulation models. The thermal model (CTSP) was developed based on superposition of complicated closed form analytical expressions and developed heat source theories. It can be used to obtain extremely rapid thermal solutions for complicated (and arbitrary) weld geometry's compared with traditional numerical thermal solutions. The structural model (UMAT) was developed based on ABAQUS commercial finite element codes by implementing a special materials module, which includes a constitutive law that permits stress relief due to weld melting/re-melting effects, strain hardening effects, large deformation mechanisms, rapid weld metal deposition features, etc. It is noted that our experience clearly suggests that uncoupled thermal/structural solutions for weld problems are accurate in all weld models. After the thermal analysis is performed, the temperature versus time histories is then written to a file in a format that can be automatically read by UMAT-WELD. The temperatures only need to be calculated near the current and prior weld locations. These specially written utility routines automate the weld modelling and make the simulation seamless.
Fig. 1. Overview of the VFT modelling process
The overall VFT simulation process flow is shown in Figs 1 and 2. The first step, which is not mandatory, is to develop a solid model of the part or structure that is to be welded. This could, for instance, be a Pro/E solid model that is passed on from the design department. By importing this solid model to a meshing generating tool, such as I-DEAS, FEMAP, or CUBIT, a finite element model can be developed. Note that CUBIT was specially developed for weld simulation by Sandia National Lab. The next step is to transform the finite element model to VFT-GUI to define weld parameters, weld passes, and welding sequences, then output thermal input file. Three kinds of thermal analysis are shown in Fig.1. Choosing which kind of thermal analysis is based on the user's analysis intention and the structures size or total weld length. Finally, the UMAT routine is a weld based constitutive routine which accounts for many features of weld modelling such as melting/re-melting, history annihilation, and etc., which are not properly accounted for in commercial packages. The simulation results are output as a residual stress and distortion. Solution speed with VFT is very important since several (often five to ten) analyses are required to optimize the structure to control the residual distortions to be within the desired tolerance. Please see references [6-10] and references sited therein for more details.
Fig. 2. Weld modelling process
Weld modelling examples
Some examples are provided in the following pages which illustrate the use of the VFT weld modelling tool.
Thin plate butt joint and importance of fabrication history
Two plates with dimensions, 450 mm long, 900 mm wide, and 5 mm thick, were tacked together from the back site with FCAW. After tack welding, the specimen was significantly distorted as shown in Fig.3 due to thin nature of the plate. The tack-weld induced distortions were measured with a specially designed device, and then mapped to the weld model with a VFT mapping code. Fig.3b shows the mapped pre-deformed distortion shape. A pre-deformed shape can have a great impact on such a thin, weak structure due to a weak resistance to buckling distortion. If buckling happens, the pre-deformed shape will lead to the final welded distortion direction. Therefore, if the simulation doesn't include the pre-weld deformation, incorrect results will be obtained.
Fig. 3. Distortion pattern in plates after tacking and before welding
Figure 4 shows a final-distortion comparison between experiment and prediction. The distortion in the weld area increases from 2.5 mm to 7mm. The overall distortion shape (W shape) is similar to the pre-deformed shape. A good agreement was achieved between the prediction and the experiment. If the initial distortions are included, the final distortion pattern is an inverted 'Vee shape', and even distorts in the wrong direction (see Reference  . The actual specimen became a W shape after welding. Therefore, it is important to include the pre-deformation in the weld model for predicting distortion accurately.
Fig. 4. Comparison of prediction versus measurements
Submarine hull manufacture
The finite element model in Fig.5 shows a 2540 mm diameter and 2286 mm long cylinder welded to a 101 mm thick ring with the GMAW welding process. Both cylinder and rings are made of high strength steel. The cylinder is originally a flat plate, edge prepared according to specified weld joint type, and then rolled, and welded longitudinally with a multi-pass single bevel seam weld. The cylinder was then welded to the end ring with a double side half-inch Tee fillet weld. This might be considered to be part of a submarine hull and ring stiffener weld problem.
Before fabricating this simple structure we used VFT TM to study the weld sequences. Three sequences were analyzed. The first sequence is to tack the cylinder, perform the multi-pass seam weld, and then weld the cylinder to the end ring with the outside fillet, and inside fillet. It was found that the cylinder was significantly distorted in the radial direction as shown in Fig.6a with a maximum value of about 15 mm. The roundness of the cylinder cannot meet the design requirement. The correction must be introduced before welding the cylinder to the end ring.
Fig. 5. Submarine cylinder and ring portion
The second sequence is to tack the cylinder, then tack the cylinder to the end ring from outside and inside Tee fillet locations, perform seam weld, and weld outside the Tee fillet and inside the Tee fillet. It was found that the radial distortion was significantly reduced as shown in Fig.6b. But the roundness of the cylinder was still not within the design requirement.
Fig. 6. Distortion patterns from three different VFT defined weld sequences
The third sequence has the same tacking procedures as the sequence two. But at first the inside Tee fillet was welded, then outside Tee fillet, and finally the seam weld. It was found that the radial distortion is reduced less than1mm as marked in Fig.6c and meet the design requirement. The reason that the third sequence produces less distortion is that after the Tee fillet weld the cylinder is stiffer so that the seam weld induced distortion is smaller. Sequencing is clearly the preferred procedure for this case, although fixture design may help. Many additional analyses were performed, including residual stress control. These are not shown here.
Ship panel construction
The structure illustrated in Figure 7 is part of the hull of a US Navy destroyer, fabricated by a US ship fabricator. This very large section incorporates numerous tee fillet welds, both vertical and horizontal. The key point illustrated here is the manpower effort required to perform this analysis. A solution to the distortion control of a panel such as this would mainly be sequencing. This analysis was performed quickly.
Three VFT TM experts (one from Battelle and two from Caterpillar) performed this analysis as a team, with direction from one other leader. From the time they received engineering drawings to the completion of an analysis took one week (not including computer time). The VFT TM graphical user interface (GUI) for shell models, developed since this analysis was performed three years ago, would reduce the required time. For the purposes of illustration, the following man-hours were required.
- Two key persons worked on this project.
- Number of welds modeled = 154.
- Model generation: person 1, 8 hours and person 2, 24 hours.
- Total shell elements 30,000.
- Set up thermal analysis, person 1, 8 hours.
- Set up stress analysis, person 2, 8 hours.
- Computer running time (three days - depends on computer - SUN Workstation here). This would be much faster today on an Itanium chip machine.
- Results analysis and summary: person 1, 16 hours and person 2, 20 hours.
- Total human hours: person 1, 32 hours; person 2, 52 hours.
- Modelling procedures: lump-pass modelling procedures.
Since the structure had single pass welds, a relatively simple shell model was used and only one demonstration analysis was performed. Typically, additional analyses using different sequences, pre-cambering, or other distortion control technique, would be required, but once the baseline model is built, setting up additional analyses using the VFT TM GUI is trivial.
An illustration of the distortion control procedures for typical ship-type fabrications are found in Figure 7. As shown in the diagram incorporated in Figure 7, they were joined using double sided welds where the weld was first finished along one side and then the other side was completed. After the predictions were made, predicted results were compared with service welded panels (qualitatively). The distortion predictions match qualitatively with those found after actual welding. The actual distortions were not measured precisely at the time but the 'bow' distortion was predicted to be about 50 mm. More importantly, it is clear that weld parameters (e.g. heat input, torch speed, material) strongly affect the final distortion.
Fig. 7. Ship hull weld model (size = 12 x 24 meter)
The 12 x 24 m hull assemblage of Figure 7 was modelled for illustration using a typical welding sequence applied to 6 mm plate.
The resulting distortions of the full structure illustrate the issues of concern, showing by comparison with the initial (un-deformed) shape that the hull assemblage bows up along the long length. An assemblage like that in Figure 7, for example, will require five to ten separate analyses with varying parameters before the optimum sequence for distortion control can be realized. Again, weld sequencing should be the preferred control method.
Bi-metallic nuclear pipe weld and crack growth study
As a result of recent unexpected cracking in the hot leg bi-metallic welds in some pressurized water reactors (PWR), the US Nuclear Regulatory Commission (NRC) initiated a study. This example illustrates how the fabrication modelling system may be used not only to model the full weld fabrication process, but to add service loads for an in depth study of the problem.
Figure 8 shows schematically the system of interest. The nozzle from the A508 pressure vessel is welded to a stainless steel pipe using an inconel buttering process (on the A508) first, followed by the inconel weld (theferritic steel is also clad with a stainless steel). The pipes are large diameter (760 mm) and thick (63 mm). The issue of concern was whether circumferential cracks (which can lead to catastrophic coolant loss) can grow faster than axial cracks (which can be tolerated temporarily) and whether similar problems might exist in the numerous other worldwide PWR nuclear plants. The problem was further complicated by the fact that the weld of concern had already undergone weld repairs.
Fig. 8. Bi-metallic pipe weld residual stresses
The entire weld process was first modelled. The analysis consisted of (1) preheat and modelling buttering of the nozzle, (2) post weld heat treat simulation of the buttered nozzle, (3) preheat and welding of the nozzle to stainless steel pipe, (4) grind out of a portion of the weld (5) repair weld analysis (6) hydro-test analysis (7) service load heat up (8) service load application, (9) introduction of cracks and determination of stress intensity factors (using a finite element alternating technique), and (10) SCC crack growth analysis. It was critical to include the entire history of stress, plastic and creep strain throughout the analysis since service stresses depend on this history. The final residual stress state at operating temperature and loads is shown in Figure 8.
The scale ranges from 50 ksi (red) to -50 ksi (purple). The axial stresses are clearly lower than hoop stresses, and the large hoop stresses are mainly confined to the inconel weld. Stress intensity factors were then calculated forabout 100 elliptical cracks inserted axially and circumferentially into this stress state (using FEAM, (FRAC@ALT,  )) and subsequent three-dimensional SCC analysis was performed (see Reference  ). The results indicated that axial cracks should grow and break through the pipe wall about twice as fast as circumferential cracks. This prediction correlated with crack experience to date. Moreover, the axial cracks should be confined to the inconel butter and weld zone. Circumferential cracks may grow to be long (circumferentially) before breaking through the pipe wall. This suggests that loss of coolant due to a double-ended pipe break is not likely. More details can be found in  . This clearly illustrates a case where the use of a model can be used to quantify and correct a service cracking problem. Many more examples of this are included in the references sited here and references therein.
Pre-straining technique for distortion control.
Pre-straining is another technique for control welding-induced distortion in heavy industries. With pre-straining, some plates are pre-bent, plastically, before welding. This is in contrast to pre-cambering, where the bending during the weld process is elastic. This may have applications in many large ship fabrication areas where thick plate must be welded and distortions controlled, such as bulkheads. Before assembling welded structure, plates are bent into permanent shape based on predictions using VFT. After welding, welding-induced distortion makes the bent shape become straight. To determine the pre-bent shape and magnitudes, a large amount of experiments needed to be conducted (in the past) before weld modelling software was available.
In Fig.9 and Fig.10, the pre-straining technique is described to mitigate distortion resulting from the J-groove circumferential fillet weld between a very long cylinder to a plate. Before welding, the plate was bent, as shown in Fig.9 and 10a. The end edge of the plate was pre-bent to about 7 mm away from middle straight part of the plate in the cylinder axle direction, as marked in Fig.10a. It is important to note that the pre-bent plate has 3 nearly straight sections with two bend radii. After welding, the pre-bent plate becomes almost flat in Fig.10b.
Fig. 9. Pre-straining technique for distortion mitigation (dimensions in mm)
Fig. 10. Shop floor practice of pre-straining
It should be pointed out that the plate was only bent in one dimension. This is because the plate is rectangular and the J-groove circumferential fillet weld is close to the two edges of the short side of the rectangular. If the plate is square and the center of circumferential weld is around the center of the plate, it is necessary to bend the plate in two dimensions. This pre-straining technique is easy to use and very effective. The use of weld modelling software makes the design of this type of procedure practical for any type of large structure.
Here the use of weld modelling software (VFT TM ) is illustrated via several examples. Caterpillar Inc., has implemented the VFT system throughout the design process for the worldwide business units. This saves money by controlling distortions 'before build'. The worldwide ship building industry could save both time and money by using weld modelling software as well.
- Huang, T.D., Dong, P., DeCan, L., and Harwig, D., 'Residual Stress and Distortions in Lightweight Ship Panel Structures,' TRW Technical Review Journal, Spring/Summer, 2003.
- Masubuchi, K., 'Analysis of Welded Structures,' Pergamon Press, Oxford, 1980, pp. 235, 331.
- Goldak, J., 'Progress and Pacing Trends in Computational Weld Mechanics,' Proceedings of International Conference on Computer Engineering and Science, 1998.
- Ueda, Y. and Yuan, M. G., 'Prediction of Residual Stress in Butt Welded Plates Using Inherent Strains', Journal of Engineering Materials and Technology, 1993, 115, pp. 417-423.
- Tsai, C. L., Cheng, W. T., and Lee, T., 'Modelling Strategy for Control of Welding-Induced Distortion. Modelling of Casting' Welding and Advanced Solidification Processes VII, The Minerals, Metals and Materials Society, 1995, pp. 335-345.
- Feng, Z. (editor), 'Welding Residual Stress and Distortion', Woodhead Publishing Co., Cambridge, U.K., to appear 2005.
- Brust, F. W., Yang, Y. Y., Ezeilo, A., and McPherson, N., 'Weld Modelling of Thin Structures with VFT', Proceedings of ASME Pressure Vessel and Piping Conference, San Diego, CA, July, 2004, in Residual Stress, Fracture, and Stress Corrosion Cracking, Principal Editor, Y. Y. Wang, 2004
- Yang, Y. P., Brust, F. W., and Kennedy, J. C., 'Lump-Pass Welding Simulation Technology Development for Shipbuilding Applications,' Proceeding of ASME Pressure Vessels and Piping Conference, 4-8 August, 2002, Vancouver, British Columbia, Canada, PVP Volume 434, princ. ed. F. W. Brust.
- User Manual for VFT - Virtual Fabrication and Weld Modelling Software by Battelle Memorial Institute, TWI, and Caterpillar Inc., February 2005.
- Cao, Z., Dong, P., and Brust, F. W., 'A Highly Efficient Heat-Flow Solution Procedure', Proceedings of ICES'98, October 7 - 9, Atlanta, 1998.
- Brust, F. W., Dong, P., and Zhang, J., 'A Constitutive Model for Welding Process Simulation Using Finite Element Methods,' Advances in Computational Engineering Science. Eds. S.N. Atluri, and G. Yagawa, pp. 51-56.
- Brust, F. W., Scott, P.M, and Yang, Y.P., 'Weld Residual Stress and Crack Growth in Bimetallic Pipe Welds,' Proceeding of 17th International Conference on Structural Mechanics in Reactor Technology (SMIRT 17), Prague, August 17-22, 2003.
- FRAC@ALT © (FRacture Analysis Code via ALTernating method), Version 2.0, January, 2003, Battelle Memorial Institute.