Research For Support Vector Machine And Its Application To The Optimal Design Of Iuduction Heating Equipment

Support vector machine (SVM) is a new method for machine learning. It is a new milestone in the field of intellectual computation after the artificial neural net method. SVM is based on strictly justified statistical studying theory. With this method, the data from sampling space are maped to higher dimensional characteristic space by the kernel functions, and then the nonlinear problem can be converted into linear divisible problem to get optimum relation, which is a great innovation in the methodology. SVM has rigorous mathematic foundation, and the training result just has relationship with Support Vectors. Thus SVM becomes the important tool for solving nonlinear problem. Therefore, it is of broadly concerned in the field of intellectual computation and broadly used in Pattern Classification and Regression. Based on good understanding of the SVM theory and algorithm, this paper discussed the optimal design problem of transverse flux induction heating (TFIH) devices for continuously moving strip.In order to analyze the reasons resulting in prematurity in the genetic algorithm (GA) running procedure, the population statistical characters included inter generations hamming distance etc, were dynamically extracted. The changing rules of these characters were analyzed and predicted by SVM. The results indicated the population statistical characters had inner relationship with prematurity. Because the population statistical characters had global regularity, local irregularity and universality, the optimization strategy was improved in the global trend and local performance. The convergence rate of the improved algorithm had been increased, and the prematurity could be effectively avoided.The optimal design problem of TFIH devices for continuously moving strip involves three-dimensional coupled field computation of eddy current, temperature and the design of global optimization. The mathematical simulation of the problem was done, and the eddy current field on the strip was given. In order to calculate temperature field caused by eddy current field, the heat source of every computation point on the continuously moving strip was equivelent simplified as two parts. One part was the current fixed heat source which was only related to geometric position of the computation point. The other part was the mean heat source which was calculated in period from the strip entering into the device to current computation time. The mean heat source was the equivelent sum of these two parts, and heat exchange procedure was also considered. The temperature field was calculated employing the above weighted heat source simply. The calculated time for numerical simulation of TFIH was greatly reduced and the computation results was also creditableFinally, the paper discussed the optimal design problem of TFIH devices for continuously moving strip. Firstly, the mathematical simulation of the problem was done by Ansoft software, and the result was trained as the sample of support vector regression (SVR) to gain large numbers of data for optimization design. Secondly, the SVR model was built according to the samples, and then the built model was used to predict mean temperature and mean relative error of temperature on the outlet of device. Finally it is founded for the math model of optimal design of TFIH device for continuously moving strip with three variables as frequency, current and coil size parameters. The optimal results of the math model were also abtained by improved GA and the SVR.