基于填充函数算法的工业产品小波网络质量模型

A Wavelet Neural Network Model for Industrial Product Quality Based on Filled Function Algorithm

为避免反传学习(BP)算法易于落入局部极小点,该文提出一种基于新填充函数的小波神经网络全局优化学习算法,用来解决连铸连轧过程的产品质量建模问题.该过程很复杂,影响其产品性能的因素很多,物理模型难以建立.该文以小波神经网络为模型,建立连铸连轧产品质量与其化学成分和轧制参数之间的复杂非线性模型.该模型用来对板材产品的断裂延伸率、屈服强度等质量性能指标进行预测.数值实验表明:所建立的模型拟合与校验命中率较高,能够较好地预测产品的物理性能.

To avoid local minimum solutions in the back propagation learning, a new global optimization algorithm based on filled function is proposed to model the product quality of continuous casting furnace and hot rolling mill. The industrial process is very complicated and the number of parameters which determine the final properties can be quite large. It is extremely difficult to develop a physical model for predicting various properties like elongation and yield and tensile strengths. In the present work, a wavelet neural network has been employed to develop a quantitative method for estimating the elongation and yield and tensile strengths as a function of steel composition and rolling parameters. Experimental studies demonstrate that the predicted mechanical proper- ties have a good agreement with the measured data by using the developed wavelet network model.