Abstract
Thermal error is the key factor affecting the machining accuracy of high-precision CNC machine tools. In order to solve the problem of thermal error of dual-spindle turning and milling machine tools, this study proposes an optimized CNN-BiLSTM Hybrid Network based on the SCSSA. Firstly, the thermal error experiment was carried out on the 1200MSY turning-milling double-spindle composite machine, and the temperature field data and thermal error data corresponding to the two spindles were collected. The K-means clustering-grey association coupling algorithm was used to screen the temperature-sensitive measurement points. Subsequently, the hyperparameters of the CNN-BiLSTM network were optimized with SCSSA to improve its modeling efficacy. The experimental findings indicate that: (1) SCSSA exhibits a superior convergence rate and enhanced global search capability compared to GWO and WOA algorithms during parameter optimization; (2) the proposed model achieves reductions in RMSE and MAE by 38.77% and 47.06%, respectively, when compared to benchmark models; (3) the prediction error variations are maintained within ± 4.6 μm under multiple speed conditions and ± 5 μm across various spindle types. This study provides an intelligent prediction scheme with high engineering practicality for thermal error prediction of dual-spindle turning-milling compound machine tools.