A ball screw all-round error compensation technology based on novel hybrid deep learning for CNC machine tool

Abstract

Considering the detrimental impact of thermal phenomena on the geometric precision of machine tools, a machine tool ball screw’s omni-directional error model is created using the LSTM neural network algorithm. Subsequently, the machine tool ball screw's omni-directional error compensation module is devised by combining the core functions of the Huazhong numerical control system with the visual programming environment of QT and the numerical computation capability of Matlab. To enhance the practicality and accuracy of the compensation model, this study has employed the Whale Optimization Algorithm (WOA) to optimize the parameters of the LSTM model. This has resulted in an improvement in the model's generalization ability and prediction performance, making it more effective. During the experimental validation phase, the Z-axis error of the machine tool was practically operated and analyzed using the compensation method. Results manifestly show that, after employing the compensation method, the peak amplitude of the Z-axis error fluctuations have been notably curtailed to ±0.006 mm – a considerable reduction compared to the initial error bandwidth of ±0.0145 mm. These empirical findings substantiate the efficacy of the proposed compensation strategy in substantially boosting the machining precision of products, thus furnishing a substantial and instructive benchmark for future inquiries into CNC machine tool error compensation technologies.