Research On Modelling Of Materials Processing Based Support Vector Regression Method

Automation is a trend of materials processing technical development. However, hot work processes involve the temperature field, flow field, organizational field, stress field, deformation field, melting, solidification and other complicated change, which make it hard to obtain accurate mathematical models of hot work processes using classical modeling methods. Obtaining knowledge models of hot working processes is helpful for understanding the processes, obtaining and even sublimating the experiential knowledge of human intelligence and realizing the intelligent control of these processes. Therefore, it is of great significance to obtain knowledge models of hot processing of metal.In recent years, it is becoming the focus of scientific research personnel to get knowledge models of hot work processes with the fuzzy set method, the neural network method, the rough set methods and their combination. Many interesting results have been acquired, however, they could not completely satisfy the practical needs. Further modeling techniques are necessary. Support vector regression modeling method based on structural risk minimization principle, has strong generalization ability and overcomes"dimension disaster","over learning"and other problems of past modeling methods. For complex process modeling, it has a good applicability. Therefore, this paper introduced support vector regression modeling method, made necessary formalization of related problems and provided relevant solutions in each step of SVR modeling for hot work processes. Eventually, each step was modularized, and MPSVR knowledge modeling software system for hot work process was developed.Based on HUNTER sand casting production line for Piston Ring of Huizhong Automotive Axle Works, modeling experiments on obtaining knowledge models of green sand casting process using MPSVR software system are carried out. For green sand casting process, SVR method was compared with grey correlation degree analysis (GCDA) method. The experiment results indicate that knowledge model of green sand casting process using MPSVR method possesses good generalization capability as well as high comprehensibility and compared with the GCDA method, MPSVR method is more intuitive, reliable and precise, which makes up for deficiencies in past traditional methods. MPSVR modeling method can effectively obtain the process models of complex processes with a great number of uncertainties such as green sand casting processes.In this paper, welding process based on visual sensing as background, MPSVR modeling method was applied to aluminum alloy pulse GTAW welding dynamic process modeling, proposed three different models input form for welding process modeling considering different welding materials and welding methods having different heat accumulation. For the random welding experiments, we compared features of the three kinds of model input and summarized application scope of them. The experiment results indicate that, in the condition of selecting reasonable models input form, knowledge model of aluminum alloy pulse GTAW welding dynamic process using MPSVR method possesses good generalization capability as well as high comprehensibility, which laid the foundation of achieving intelligent control of welding process.