Abstract: Aiming at the complicated thermal error measurement and control problems caused by the superposition of multiple heat sources in FAMTs, a FAMT thermal error modeling method was proposed, the important parameters of the thermal error models were evaluated by using LSO optimized least squares support vector machine (LSO-LSSVM), and then the efficiency and accuracy of the thermal error prediction models were effectively improved. The partial correlation analysis was used to screen a large number of temperature sensor positions, and the temperature variables with large correlation were selected. According to the selected measured temperature data, the multiple linear regression, particle optimization optimized LSSVM, and LSO-LSSVM modeling methods were respectively used to conduct thermal error modeling, and the prediction capabilities of the respective thermal error models were compared and analyzed. The results show that the accuracy and robustness of the thermal error prediction models established by LSO-LSSVM were greatly improved. The thermal error compensation tests were also carried out on the main parts of the FAMTs. The test results show that the LSO-LSSVM modeling method reduces the errors of the specimens in the three directions of X, Y and Z by 35.3%, 32.2% and 43.9% respectively.

Key words: five-axis CNC machine tool(FAMT), thermal error modeling, lion swarm optimization(LSO) algorithm, point optimization, compensation test

摘要： 针对五轴数控机床多个发热源叠加导致的较为复杂的热误差测控难题，提出了一种五轴数控机床热误差建模方法，采用狮群优化算法优化最小二乘支持向量机（LSO-LSSVM）方法对热误差模型的重要参数进行求解，从而有效提高热误差预测模型的效率和精度。使用偏相关分析对大量温度传感器位置进行初步筛选，选取关联性较大的温度变量，根据选取的实测温度数据，分别采用多元线性回归、粒子群优化最小二乘支持向量机与LSO-LSSVM建模方法进行热误差建模，同时对各热误差模型的预测能力进行对比分析，结果表明：使用LSO-LSSVM建立的热误差预测模型的精度和鲁棒性都有很大的提高。对五轴数控机床主要部位实施热误差补偿测试，测试结果表明，采用LSO-LSSVM建模方法可使试件在X、Y、Z三个方向的误差分别减小35.3%、32.2%和43.9%。

关键词: 五轴数控机床, 热误差建模, 狮群优化算法, 测点优化, 补偿测试