Modeling of thermal errors for dual-spindle turning-milling compound machine tools based on hybrid networks

Area of Science:

• Mechanical Engineering
• Manufacturing Technology
• Artificial Intelligence in Manufacturing

Background:

• Thermal error is a critical factor limiting machining accuracy in high-precision CNC machine tools.
• Dual-spindle turning and milling machine tools present unique challenges for thermal error compensation due to complex thermal behaviors.

Purpose of the Study:

• To develop an intelligent and highly accurate method for predicting thermal error in dual-spindle turning-milling machine tools.
• To enhance the modeling efficacy of a hybrid neural network through advanced optimization techniques.

Main Methods:

• Collected temperature field and thermal error data from a 1200MSY turning-milling double-spindle composite machine.
• Employed K-means clustering-grey association coupling algorithm for selecting temperature-sensitive measurement points.
• Optimized a CNN-BiLSTM hybrid network using the SCSSA algorithm for improved thermal error modeling.

Main Results:

• The SCSSA algorithm demonstrated superior convergence and global search capabilities over GWO and WOA.
• The proposed model reduced Root Mean Square Error (RMSE) by 38.77% and Mean Absolute Error (MAE) by 47.06% compared to benchmark models.
• Prediction errors were maintained within ±4.6 μm (speed conditions) and ±5 μm (spindle types).

Conclusions:

• The SCSSA-optimized CNN-BiLSTM network provides an effective intelligent prediction scheme for thermal error in dual-spindle machine tools.
• The developed method offers high engineering practicality and significantly improves machining accuracy.
• This research contributes a robust solution for mitigating thermal errors in advanced manufacturing processes.