应用激光, 2018, 38 (4): 649, 网络出版: 2018-10-06  

基于反向传播神经网络的激光弱化残余厚度预测

Residual Thickness Prediction of Laser Weakening Based on Back-propagation Neural Network
作者单位
江苏大学 汽车与交通工程学院, 江苏 镇江212013
摘要
激光弱化加工是目前的新兴技术, 弱化后残余厚度的大小是其关键问题。将残余厚度与各项激光加工工艺参数结合起来, 建立相应的预测模型。首先对汽车仪表板常用材料PC(聚碳酸酯)硬塑进行弱化加工, 沿弱化孔中心线将试件剖开并通过影像测量系统测得残余厚度值。建立激光脉冲宽度、离焦量和加工速率这三个工艺参数与残余厚度之间的BP(反向传播)神经网络预测模型, 使用大量试验数据训练网络, 并使用试验样本中的部分数据检验所建网络。最终得到了最大误差率不超过3%, 收敛速率及预测准确性高的预测模型。使用该模型, 可以精确地预测残余厚度的大小, 缩短了激光弱化加工作业的准备时间。
Abstract
Laser weakening processing is the current emerging technology. The residual thickness after weakening is the key issue. The residual thickness and the laser processing parameters were combined to establish a corresponding prediction model. First of all, the material PC (polycarbonate) hard plastic common used in automobile dashboard was weakened. The specimen was cut along the weakened hole center line and the residual thickness value was measured by an video measuring system. The BP(back-propagation)neural network prediction model was established between the three process parameters of laser pulse width, defocusing and processing rate and the residual thickness. Use a large amount of test data to train the network and use the data in the test sample to test the network. Finally, a prediction model with maximum error rate less than 3%, high convergence rate and high prediction accuracy is obtained. Using this model, the size of the residual thickness can be accurately predicted and the preparation time for laser weakening process can be reduced.

丁华, 殷潇. 基于反向传播神经网络的激光弱化残余厚度预测[J]. 应用激光, 2018, 38(4): 649. Ding Hua, Yin Xiao. Residual Thickness Prediction of Laser Weakening Based on Back-propagation Neural Network[J]. APPLIED LASER, 2018, 38(4): 649.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!