基于优化BP神经网络的钢板测速修正方法

A Speed Correcting Method on Steel Plate Based on Improved BP Neural Network

摘要:

在采用激光多普勒仪测速的钢板长度测量过程中,针对由于钢板表面因素和测量环境所引起的测量速度数据失真问题,设计了基于L-M(Levenberg-Marquardt)优化BP神经网络的钢板测速数据处理模型；通过进行多普勒测速数据分析调整BP网络结构和参数,依据误差反向传播理论将测速数据自适应地进行非线性拟合,修正测速粗差,最后计算钢板准确长度；与最小二乘(LS)拟合法的对比实验结果表明,该方法能更加准确地进行失真数据修正,现场运行结果表明该方法可将钢板测长精度提高10%以上,满足钢板长度测量的精度要求。

Abstract:

Due to the speed data distortion caused by the surface factor and measuring environment of the steel plate in the process of the steel plate length measurement using Laser Doppler Velocimetry, a data processing model of steel plate velocity measurement based on BP neural network which optimized by L-M has been designed. Adjust The BP network structure and parameters through analyzing the Doppler velocity data. Then the velocity data are adaptive nonlinear fitting based on the error back propagation theory and so correct the gross errors. Finally calculate the steel plate length accurately. Compared with LS, the results of the experiment show that this method can correct the fuzzy data more accurately. Results of field operation indicate that the steel plate length measurement accuracy can be increased by more than 10%, and meet the accuracy requirement of the steel plate length measurement.