Prediction of weld formation in 5083 aluminum alloy by twin-wire CMT welding based on deep learning
- 30 Downloads
Based on a large amount of experimental data from twin-wire CMT welding of 5083 aluminum alloy, deep neural network technology was adopted to analyze the welding process parameters and the weld dimensions, and a precise prediction model for the weld formation parameters was established. The results show that the key parameters influencing the prediction accuracy of the twin-wire CMT deep neural network model are the number of hidden layer neurons, the number of network training iterations, and the learning rate of the deep network. For a single factor, regardless of the weld width, weld penetration or weld reinforcement, the predicted value curve changes smoothly and without distortion from the measured value curve. The accuracy of the complex nonlinear model can be evaluated by linear regression analyses of the predicted data and the measured data. In addition, the deep neural network has the obvious advantages of high efficiency and precision due to its strong multi-dimensional nonlinear fitting abilities in the quantitative analysis of the arc welding system from the input welding parameters to the output weld dimensions. This model can provide data support and scientific reference for the process designs for 5083 aluminum alloy twin-wire CMT welding or additive manufacturing and the determination of the numerically calculated heat source size. Also, this model provides innovative ideas for the application of deep learning technology in the welding field.
KeywordsTwin-wire CMT welding 5083 aluminum alloy Deep learning Neural network Formation prediction
This research was supported by the National Natural Science Foundation of China (No. 51674056), the Frontier and Applied Basic Research Projects of Chongqing (No. cstc2018jcyjAX0108), the Opening Project of Materials Corrosion and Protection Key Laboratory of Sichuan Province (No. 2016CL15), and the Science and Technology Planning Project of Guangdong Province (2017A010106007, 2017A070701026, 2017GDASCX-0113, 2018GDASCX-0803).
- 2.Ahsan MRU, Cheepu M, Kim TH, Jeong C, Park YD (2017) Mechanisms of weld pool flow and slag formation location in cold metal transfer (CMT) gas metal arc welding (GMAW). Weld World 61(2):1–11Google Scholar
- 7.Heaton J (2017) Ian goodfellow, yoshua bengio, and aaron courville: deep learning. Genet Program Evolvable Mach 19(1–2):1–3Google Scholar
- 12.Montes-Atenas G, Seguel F, Valencia A, Bhatti SM, Khan MS, Soto I, Yoma NB (2016) Predicting bubble size and bubble rate data in water and in froth flotation-like slurry from computational fluid dynamics (CFD) by applying deep neural networks (DNN). Int Commun Heat Mass Transfer 76:197–201CrossRefGoogle Scholar
- 14.Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R (2014) Dropout: a simple way to prevent neural networks from overfitting. J Mach Learn Res 15(1):1929–1958Google Scholar
- 15.Bengio Y (2012) Practical recommendations for gradient-based training of deep architectures, neural networks: tricks of the trade. In: Montavon G, Orr GB, Müller KR (eds) Lecture notes in computer science. Springer, Berlin, pp 437–478Google Scholar
- 16.Nielsen MA (2015) Neural networks and deep learning. Determination Press http://neuralnetworksanddeeplearning.com