关键词: 连铸漏钢预报, 支持向量机, 粒子群算法, 模拟退火算法

Abstract: In order to overcome the problems that the neural network model was difficult to obtain a high accurate breakout prediction for continuous casting under the conditions of small sample training data, a breakout prediction for continuous casting was proposed based on SA-PSO algorithm to optimize the parameters of SVM. Firstly, the PSO algorithm was introduced into the training processes of SVM, increasing the optimization speeds of breakout prediction model by using the advantages of less parameters and fast optimization speeds. Secondly, SA algorithm was used to evaluate the new positions of updated particles, and to determine whether the new positions were accepted, which could avoid the PSO algorithm steped into the local extremum in optimization processes. Finally, the breakout prediction for continuous casting was tested by the history data of continuous casting. The results show that the proposed algorithm may make the breakout prediction accuracy reach 98.8%.

Key words: breakout prediction for continuous casting, support vector machine (SVM), particle swarm optimization(PSO) algorithm, simulated annealing(SA) algorithm.