Research Article | Open Access
Thongchai Arunchai, Kawin Sonthipermpoon, Phisut Apichayakul, Kreangsak Tamee, "Resistance Spot Welding Optimization Based on Artificial Neural Network", International Journal of Manufacturing Engineering, vol. 2014, Article ID 154784, 6 pages, 2014. https://doi.org/10.1155/2014/154784
Resistance Spot Welding Optimization Based on Artificial Neural Network
Resistance Spot Welding (RSW) is processed by using aluminum alloy used in the automotive industry. The difficulty of RSW parameter setting leads to inconsistent quality between welds. The important RSW parameters are the welding current, electrode force, and welding time. An additional RSW parameter, that is, the electrical resistance of the aluminum alloy, which varies depending on the thickness of the material, is considered to be a necessary parameter. The parameters applied to the RSW process, with aluminum alloy, are sensitive to exact measurement. Parameter prediction by the use of an artificial neural network (ANN) as a tool in finding the parameter optimization was investigated. The ANN was designed and tested for predictive weld quality by using the input and output data in parameters and tensile shear strength of the aluminum alloy, respectively. The results of the tensile shear strength testing and the estimated parameter optimization are applied to the RSW process. The achieved results of the tensile shear strength output were mean squared error (MSE) and accuracy equal to 0.054 and 95%, respectively. This indicates that that the application of the ANN in welding machine control is highly successful in setting the welding parameters.
In automotive production, each automobile has approximately 7,000 to 12,000 spot welds. The welds are done using the Resistance Spot Welding (RSW) process which is done by a computer controlled robotic welder. The use of RSW on lightweight aluminum alloy is increasing [1, 2]. RSW is a rapid joining technique extensively used to join thin shell assemblies in automotive manufacturing operations. It is an important process to ensure strong structural car bodies using lightweight materials to save both energy and natural resources . The aluminum alloy material is a low density material with significant mechanical properties which is expected to be extensively used in the future to partially replace steel which is currently the primary production material in automobiles . The 6061-T6 aluminum alloy is of light weight and has significant mechanical properties which are of interest in this research.
The weld quality of the RSW process has been a significant problem for the automotive industry. Manual calculation of welding parameters, operator experience, and technician expertise in adjusting the parameter settings have not been consistently accurate or correct. The calculations have previously been unable to be confirmed against optimal parameters . The parameter settings of each welding machine have been difficult because there are many sensitive factors. This has led to the high cost of requiring many specimens of the construction material to be tested to achieve adequate experimental results to derive optimal parameter values. Each car body factory has more than 200 welding machines. In order to get standard weld quality overall has led to high costs of adjusting the parameter settings for each different model of welding machine, for different materials (e.g., thickness), for replacement electrodes, and so on [6–8]. Therefore, it is important to understand the parameter relationships in the RSW process, quality improvement measures, assessment and efficiency prediction, and appropriate parameter optimization.
The problems mentioned have motivated this study, formally titled “parameter optimization for resistance spot welding of 6061-T6 aluminum alloy based on artificial neural network.” The estimation of RSW parameters using ANN will be more efficiently and correctly optimized. The ANN proved to be effective in resolving both linear and nonlinear functions required for adjusting the RSW parameter settings of computer controlled robotic welders, especially in the automotive industry.
2. RSW Process
The objective of the RSW process is to generate heat rapidly in the joints of the material being welded while minimizing conduction of heat to cooler adjacent material. This heat generation can be expressed by where is the heat energy in joules, is the current in amperes, is the resistance in ohms, and is the time in seconds . The series of resistances contributed by the secondary circuit resistance welding machine are shown in Figure 1.
From Figure 1, the resistance in the RSW process depends on the total resistance which can be expressed by where is the total resistance, is the electrode resistance, is the specimen electrode contact resistance, is the specimen resistance, and is the specimen’s contact resistance.
The resistance transformation of the material work piece affects the melting temperature in the spot weld in the RSW process and the weld ability of the material [11, 12]. Optimal electrode force is needed for the explosive solutions during material smelting [13, 14], while the intensity of the electric welding current is an important factor that affects heat generation leading to smelting. When the electric intensity is high the spot weld will also have intensive heat. This will affect the nugget size and strength of the weld joint . The nugget size and weld joint strength depend on the resistance and welding time, so the welding duration influences the mechanical properties of the weld joints [16, 17]. Melting occurs as a result of the relationship between welding current, welding time, and electrode force, as shown in schematic diagram RSW cycle Figure 2, where is welding current, is welding time, and is electrode force.
Researchers have studied the relationship of the RSW parameters and confirm that the welding process parameters have a great influence on the weld quality [18, 19]. In addition, the resistance of the material due to the thickness of the work piece is also important, as can be seen in greater nugget growth in thicker pieces than thin pieces. In different materials, a satisfactory nugget will be generated in high resistance materials or materials with lower thermal conductivity . Therefore, this study adds the resistance as a factor.
3. Material and Methods
This research studied the RSW process under actual conditions in a car body factory which used gun welds controlled by a MFDC Rexroth Bosch welder. Aluminum alloy specimens (6061-T6) of both 1 and 2 mm thickness were welded at three resistance levels. Figure 3 shows the three resistance levels of the high, medium, and low thicknesses of 2-2, 2-1, and 1-1 mm material, respectively.
This research used 6 mm copper alloy electrode tips. The experimental design included full factorial analysis based on low, medium, and high thickness of material, with all parameter settings varied at each level of thickness, giving runs. The parameter settings are welding current, electrode force, welding time, and resistance of 20, 28, and 36 kA, 2, 4, and 6 kN, 100, 150, and 200 ms and 2-2, 2-1, and 1-1 mm, respectively. Subsequently the welding parameters were input according to the experimental design and welding tests were done. The finished weld joints and the maximum tensile shear strength were tested and recorded. The tests were done using the HOUNSFIELD 50 kN at maximum load as shown in Figure 4.
In this research, 75% of the 243 experimental results of the ANN were used to train the robotic welders. 25% of the experimental results were selected randomly for testing, for comparison between the predicted shear strength and the experimental shear strength results. The ANN models were configured to establish the relationship between the welding parameters and the shear strength. The ANN model is a multilayered feed-forward algorithm, as shown in Figure 5. In that algorithm, there are three layers: the input layer, the output layer, and the hidden layer between the input and output layers. In the input layer, the number of the input neurons is set to the level of welding parameters in each run order, and the output layer is the shear strength.
The ANN was designed and tested for estimating the shear strength by using the welding parameters as input data and the shear strength as output data, using the multilayered feed-forward algorithm, and the ANN was trained by the backpropagation algorithm. The final transfer function was a sigmoidal function and was a pure linear function.
4. Results and Discussion
This work studied the application of the ANN to calculate optimized parameters for controlling welding robots. Table 3 shows the tensile shear strength achieved from actual application of the experimental results as calculated by the ANN; this variable is the calculation of the mean squared error (MSE) of the shear strength. The MSE is equal to 0.05. According to these results, it can be seen that the ANN determined appropriate parameter optimization with satisfactory results.
The shear strength is obtained when applying the experimental results from the ANN. Applying the ANN calculations results in good parameter optimization as shown in Figure 6. ANN is therefore seen as an appropriate method optimization for finding an answer that is both linear and nonlinear.
The ANN model is composed of three different layers: input layer, hidden layer, and output layer. The ANN was used to calculate the shear strength at the output layer and the RSW parameters at the input layer while training and using the backpropagation algorithm. 75% of the 243 data values from the full factorial experimental results were used for training and 25% of the data were used for testing the accuracy of the ANN’s calculated values. The results show that hidden nodes, learning rate, MSE goal, and maximum learning epoch are 30, 0.001, 0.04, and 3,000,000, respectively.
When tested, 20 samples were found to have a mean squared error (MSE) of 0.047896 kN. This indicates that the ANN model is capable of predicting the shear strength for adjusting the RSW parameters under test with a small sample which are shown in Table 4 and Figure 7, respectively.
From this study, it is suggested that the RSW operator can use a small sample set appropriate as input into the ANN model which would be sufficient for setting the optimization parameters. However, the accuracy of the parameters may be affected by other factors outside the scope of the ANN, such as the particular welding machine model, the surface condition of the specimen material, the cooling effect of the environment, and electrode wear.
The full factorial experimental results and ANN were developed and successfully tested in an auto body industry plant in Thailand. The results were MSE and accuracy equal to 0.054 and 95%, respectively. Calculating a reliable estimation of shear strength enables parameter settings to achieve high weld quality and reduces both the setting time process and the sample testing. Future work should test the accuracy of other algorithms used to calculate these parameters to enable comparison of the accuracy of other algorithms.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
The authors would like to express their profound gratitude to the UK Engineering and Supply Co. Ltd. and Thonburi Automotive Assembly Plant Co., Ltd., Thailand, for allowing use of their premises and providing research equipment to support the project.
- H. Huh and W. Kang, “Electrothermal analysis of electric resistance spot welding processes by a 3-D finite element method,” Journal of Materials Processing Technology, vol. 63, no. 1–3, pp. 672–677, 1997.
- W. Li, S. Cheng, S. J. Hu, and J. Shriver, “Statistical investigation on resistance spot welding quality using a two-state, sliding-level experiment,” Journal of Manufacturing Science and Engineering, Transactions of the ASME, vol. 123, no. 3, pp. 513–520, 2001.
- J. A. Khan, L. Xu, Y.-J. Chao, and K. Broach, “Numerical simulation of resistance spot welding process,” Numerical Heat Transfer Part A: Applications, vol. 37, no. 5, pp. 425–446, 2000.
- J. Zhu, L. Li, and Z. Liu, “CO2 and diode laser welding of AZ31 magnesium alloy,” Applied Surface Science, vol. 247, no. 1–4, pp. 300–306, 2005.
- H.-L. Lin, T. Chou, and C.-P. Chou, “Optimation of resistance spot welding process using Taguchi method and a neural network,” Experimental Techniques, vol. 31, no. 5, pp. 30–36, 2007.
- K. D. Weiss, “Paint and coatings: a mature industry in transition,” Progress in Polymer Science, vol. 22, no. 2, pp. 203–245, 1997.
- X. Deng, W. Chen, and G. Shi, “Three-dimensional finite element analysis of the mechanical behavior of spot welds,” Finite Elements in Analysis and Design, vol. 35, no. 1, pp. 17–39, 2000.
- S.-H. Lin, J. Pan, T. Tyan, and P. Prasad, “A general failure criterion for spot welds under combined loading conditions,” International Journal of Solids and Structures, vol. 40, no. 21, pp. 5539–5564, 2003.
- N. T. Williams and J. D. Parker, “Review of resistance spot welding of steel sheets: part 2—factors influencing electrode life,” International Materials Reviews, vol. 49, no. 2, pp. 77–108, 2004.
- Y. Luo, J. Liu, H. Xu, C. Xiong, and L. Liu, “Regression modeling and process analysis of resistance spot welding on galvanized steel sheet,” Materials and Design, vol. 30, no. 7, pp. 2547–2555, 2009.
- P. H. Thornton, A. R. Krause, and R. G. Davies, “Contact resistances in spot welding,” Welding Journal, vol. 75, no. 12, pp. 402.s–412.s, 1996.
- A. G. Livshits, “Universal quality assurance method for resistance spot welding based on dynamic resistance,” Welding Journal, vol. 76, no. 9, pp. 383–390, 1997.
- R. B. Hirsch, “Making resistance spot welding safer,” Welding Journal, vol. 86, no. 2, pp. 32–37, 2007.
- J. Senkara, H. Zhang, and S. J. Hu, “Expulsion prediction in resistance spot welding,” Welding Journal, vol. 83, no. 4, p. 123-S, 2004.
- S. Satonaka, K. Kaieda, and S. Okamoto, “Prediction of tensile-shear strength of spot welds based on fracture modes,” Welding in the World, vol. 48, no. 5-6, pp. 39–45, 2004.
- S. Aslanlar, A. Ogur, U. Ozsarac, and E. Ilhan, “Welding time effect on mechanical properties of automotive sheets in electrical resistance spot welding,” Materials and Design, vol. 29, no. 7, pp. 1427–1431, 2008.
- F. Hayat, “The effects of the welding current on heat input, nugget geometry, and the mechanical and fractural properties of resistance spot welding on Mg/Al dissimilar materials,” Materials and Design, vol. 32, no. 4, pp. 2476–2484, 2011.
- T. Kim, H. Park, and S. Rhee, “Optimization of welding parameters for resistance spot welding of TRIP steel with response surface methodology,” International Journal of Production Research, vol. 43, no. 21, pp. 4643–4657, 2005.
- R. S. Florea, D. J. Bammann, A. Yeldell, K. N. Solanki, and Y. Hammi, “Welding parameters influence on fatigue life and microstructure in resistance spot welding of 6061-T6 aluminum alloy,” Materials and Design, vol. 45, pp. 456–465, 2013.
- C. Tsai and J. Papritan, “Modeling of resistance spot weld nugget growth,” Welding Journal, vol. 71, no. 2, pp. 47–54, 1992.
- KAISER ALUMINUM, “Trent wood Works—WA 99215 CERTIFIED Serial Number 4315467,” Article ID 4315467, TEST REPORT, December 2013.
Copyright © 2014 Thongchai Arunchai et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.