Application Of Artificial Intelligent Algorithm In GTAW

As the welding technology is widely applied to more and more areas, welding quality gets a high degree of attention, while the penetration control for obtaining stable welding quality is essential, and penetration control is essential to stable welding quality, therefore accurate prediction of penetration and welding parameters determined to become an important research topic in the field of welding. Penetration forecasting model, which can map the penetration parameter by the positive characteristics of the pool, provides the basis for real-time control of the penetration. Welding expert system can help to improve work efficiency, economic benefits, and promote the development of enterprises by reducing the time required to determine parameters.The artificial neural network is an important branch of artificial intelligence, with a unique self-organizing, self-learning, fast processing, highly fault-tolerance and nonlinear function approximation ability. As welding process is so complex, ANN is used to establish the penetration predictive soft sensor model for its advantages. Then globalbest adaptive mutation particle swarm optimization (GBAMPSO) was proposed to improve the limitations of BPNN, such as slow convergence speed of learning algorithm, local minimization, and over-fitting phenomenon. And the predictive model based on GBAMPSO-BPNN was established.ANN gets a high attention once again in expert system design for its technological superiority. It is applied to knowledge acquisition part, which is conductive managing and updating the database. The expert system designed in this paper with independent modules is consists process parameter selection, design, welding equipment selection and network self-learning sections. Visual Basic6.0development tools, object-oriented programming method are used to achieve the establishment of all the modules. Matlab is considered to train and test ANN model because of its powerful computing capabilities, interacting with the database is realized by ODBC data source interface, process parameter optimization and database expansion are achieved.Simulation results show that the GBAMPSO-BPNN model is better than BPNN model in penetration prediction. The predictive model based on GBAMPSO-BPNN is capable of obtaining desire destination. The test results show that the ES designed in this paper is stable, and it can give reliable results, provide useful knowledge for expert system design and development.