An Intelligent Thermal Compensation System Using Edge Computing for Machine Tools.

Area of Science:

• Manufacturing Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Lathe-cutting tool machines are susceptible to thermal deformation during prolonged operations, causing displacement errors and product damage.
• Smart manufacturing offers solutions for real-time monitoring and control to mitigate these issues.

Purpose of the Study:

• To develop a predictive system for thermal displacement in machine tools using time-series thermal compensation.
• To apply edge computing technology for industrial implementation of the predictive system.

Main Methods:

• Collected time-series data on time, speed, torque, and temperature from machine tools.
• Trained and compared seven machine learning models (WNN, LSTNet, TPA-LSTM, XGBoost, BiLSTM, CNN, GA-LSTM) for temperature prediction and five-axis displacement forecasting.
• Optimized temperature prediction models and validated forecasting accuracy through experimental analysis.

Main Results:

• The GA-LSTM model demonstrated superior performance, achieving an average precision greater than 90%.
• GA-LSTM outperformed other leading models including LSTM, GRU, and XGBoost.
• Analysis considered both model precision and training time for practical application.

Conclusions:

• The GA-LSTM model is highly effective for thermal displacement prediction in machine tools.
• A system utilizing LSTM, GRU, and XGBoost models is recommended for thermal compensation.
• Edge devices like Raspberry Pi are suitable for implementing such thermal compensation systems in industrial settings.