基于人工神经网络的施釉机器人釉料厚度沉积率建模与预测

Glazing Thickness Deposition Rate Modeling and Prediction for Robotic Glazing Based on Artificial Neural Network

为进行陶瓷产品的施釉机器人离线编程作业的轨迹自动规划,实现釉料厚度的精确性,提高釉面质量,提出基于人工神经网络拟合釉料厚度沉积率模型的方法。以釉料厚度试验数据为基础,采用贝叶斯归一化算法和LM优化算法进行拟合,通过试验结果对比分析表明,两种拟合模型与试验数据基本吻合,验证了模型的正确性和有效性。在此基础上进行模型的筛选,选择执行效率高、精度高、抗噪能力强的贝叶斯归一化算法拟合模型。提出的方法符合工程实际,有助于提高施釉机器人釉料厚度的控制精度,为陶瓷施釉自动轨迹规划的软件编程和仿真实现提供了模型依据与方法指导。

A key problem of the off-line trajectory planning for robotic glazing is to determine the glazing deposition rate model.For automatic trajectory planning of robot offline programming tasks about ceramic products,ensuring accuracy of glazing thickness and improving the quality of products,the model of glazing thickness deposition rate is developed based on artificial neural network.Based on the experimental data of glazing thickness,the glazing deposition rate model is fitted by using the Bayes normalization algorithm and LM optimization algorithm respectively.The result shows that all the two models have the high precision.However,compared with LM optimization algorithm,the Bayes normalization algorithm converges faster and more accurate.So Bayes normalization algorithm is better than LM optimization algorithm in modeling glazing deposition rate.The method increases the control accuracy of glazing thickness.The paper provides a specific theoretical and methodological support for the realization of process planning and simulation system in ceramic glazing manufacturing.It will make the future developed system meet the actual processing requirement.