Torque prediction in ultrasonic vibration assisted drilling for small and deep holes using AI

Excessive torque poses a significant risk of tool and workpiece damage during drilling, especially when drilling small, deep holes in challenging materials like stainless steel. Substantial friction, insufficient chip drainage, and chip clogging are the main contributors to torque during deep hole drilling. Accurate torque prediction is crucial to optimizing the drilling process and avoiding potential damage. However, torque data in deep hole drilling exhibit nonlinear and stochastic behavior, making accurate prediction challenging. To address this issue, we propose an innovative artificial intelligence approach known as a hybrid CNN–LSTM (convolutional neural network–long short-term memory) model for predicting torque during small and deep hole drilling in SUS-304 material using carbide bits, with a hole depth-to-diameter ratio of 8. Experimental results show that the proposed model achieves the highest prediction accuracy ( R 2 = 0.9509) and the lowest error values (MAE = 0.0452, mean squared error (MSE) = 0.0033, RMSE ≈ 0.0574), and is more robust to noise compared to conventional methods (support vector machine, artificial neural network, CNN, and LSTM). This approach offers significant potential for real-time monitoring, adaptive control, and optimization of deep drilling processes in advanced manufacturing.