应用变量优选的PLSR分析直线进给轴热扭曲行为
Exploiting PLSR with Variable Optimization Selection in Thermal Distortion Behavior Analysis of Linear Feed Drive Axis

为了探索直线电机驱动的高速直线进给轴热扭曲变形的影响因素，在试验的基础上，给出应用向前变量智能自筛选的偏最小二乘线性回归模型（Partial least squares regression，PLSR）分析影响进给轴热扭曲行为关联因素的分析方法。通过在自构建的进给轴试验平台，建立进给轴扭曲变形的测试系统，给出直线进给轴在发热过程和强冷却作用过程的热扭曲变形采样与进给轴温度动态采集方案。应用周期大变异的遗传算法为偏最小二乘回归参数的自检验方法，给出分析方法的具体实现步骤。通过实验和回归识别计算，分析了进给轴的温度分布及其对热扭曲行为的影响规律。结果表明，给出的变量自筛选偏最小二乘线性回归分析方法，可有效的筛选复相关的温度测点变量，并保持较高的回归识别精度，给出的方法与全变量PLSR和向后变量筛选的Bootstrap方法进行了比较，进一步表明了给出的回归分析方法的优越性。
In order to explore the thermal effects of thermal distortion of high speed linear feed axis driven by linear motor, on the basis of experiment, an analytical method by a partial least squares linear regression model(PLSR)with forward variable intelligent self selection was proposed and the impact of the effect factors of thermal distortion behavior on linear feed drive axis was analyzed. According to the established feed axis distortion measurement system on the feed axis test platform, it gives the acquisition scheme of the heat distortion of the linear axis and dynamic temperature in heating process and cooling process. Using the cycle mutation genetic algorithm, the paper provides a self inspection method for acquiring the partial least squares regression parameter and gives the specific steps of the analysis method. With the experiments and regression recognition calculation, the feed axis temperature distribution and the rule of the thermal distortion behavior were analyzed. The results show that, the analysis method with variable self screening partial least squares linear, which can effectively filter the multiple correlation of the variables of temperature measurement points and maintains high regression identification accuracy. Compared with all variable PLSR and backward variables screening Bootstrap method, the regression analysis method can further demonstrates the superiority