大型管件JCOE成形智能化控制中的材料性能参数识别

Parameters identification of material properties in intelligent control of large diameter pipe with JCOE forming

大型管件JCOE成形智能化控制系统可根据板材性能差异自动预测最佳成形工艺参数,为此采用神经网络技术,利用LM(Levenberg-Marquarat)算法,设计了材料性能参数识别神经网络模型,用于实时识别板坯的材料性能参数。通过10种不同的材料,采用正交实验设计网络模型训练方案,经10组样本数据检验,收敛精度小于1‰,检验误差小于3%,能够满足工程应用的需要。

The large diameter pipe with JCOE forming intelligent control system can automatically predict the best process para-meters according to the differences of sheet material properties.So the neural network technology and LM（Levenberg-Marquarat） algorithm are used to design neural network model for real-time material parameters identification.The training program of net-work model is designed by orthogonal experiment and it is trained through 10 different materials.Test by 10 groups of sample data,the convergence accuracy is less than 1‰ and test error is less than 3% to meet the needs of engineering applications.