基于神经网络的宽带激光熔覆熔池特征参数预测  

Prediction on characteristics of molten pool in wide-band laser cladding based on neural network

作　　者：雷凯云 秦训鹏[1,2,3] 刘华明[1,2] 冉渊[1,3] LEI Kai-yun;QIN Xun-peng;LIU Hua-ming;NI Mao(Hubei Key Laboratory of Advanced Technology of Automobile Components,Wuhan University of Technology,Wuhan 430070,China;Hubei Collaborative Innovation Center for Automotive Components Technology,Wuhan University of Technology,Wuhan 430070,China;School of Automotive Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学现代汽车零部件技术湖北省重点实验室,武汉430070 [2]武汉理工大学汽车零部件技术湖北省协同创新中心,武汉430070 [3]武汉理工大学汽车工程学院,武汉430070

出　　处：《光电子．激光》2018年第11期1212-1220,共9页Journal of Optoelectronics·Laser

基　　金：湖北省技术创新专项重大项目(cxzd2017000281);武汉理工大学研究生优秀学位论文培育项目资助(2017-YS-045)资助项目

摘　　要：为了实现宽带激光熔覆熔池特征的准确预测,从而对激光熔覆工艺过程进行实时监测、评价及反馈控制。通过宽带激光熔覆全因素工艺试验采集熔池特征参数样本数据,采用遗传算法优化BP神经网络的初始权值和初始阈值,建立激光熔覆工艺参数(激光功率、粉末厚度、扫描速度)与熔池特征参数之间的BP神经网络预测模型。利用训练集数据对所建立的神经网络进行训练,形成输入与输出之间的映射关系,并利用测试集数据对网络进行测试。试验结果表明,宽带激光熔覆熔池特征参数神经网络预测模型具有很高的精度。该神经网络预测模型对激光熔覆过程监测及熔覆层质量控制具有重要意义。In order to realize the accurate prediction on characteristics of molten pool in wide-band laser cladding,and achieve the real-time detection, evaluation and closed-loop control of laser cladding, a full factorial design method is used to conduct the experiments, and the experimental results are chosen ran- domly as sample data for neural network. Genetic algorithm is utilized to optimize the initial weights and thresholds of back propagation （BP） neural network. The BP neural network prediction model is devel- oped to express the relationship between the process parameters （laser power, powder thickness, scan- ning speed） and the characteristics of molten pool. The training set of data obtained in experiments is used to train the neural network to establish the perfect mapping relation between input and output of network. The testing set of data is used to verify the performance of the trained network. Simulation re- sults indicate that the prediction model of characteristic parameters of molten pool in wide-band laser cladding has sufficient accuracy. The neural network prediction model is of great significance for the real- time monitoring of laser cladding process and the quality control of cladding layers.

关 键 词：宽带激光熔覆 神经网络 熔池特征 预测模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]