Deformation Suppression Method for the CNC Machining Process of Parts Based on a Single Neuron PID

Abstract

Computer Numerical Control (CNC) machining plays a vital role in modern precision manufacturing but often suffers from part deformation due to thermal and mechanical stresses, compromising dimensional accuracy. Traditional CNC systems lack adaptive intelligence, operating with static parameters and failing to address real-time deformation risks. This study proposes an intelligent deformation suppression method using a lightweight single-neuron-based Proportional-Integral-Derivative (PID) neural model, termed NeuroPID-CNC, to predict and mitigate deformation during machining. The model was trained and tested on the CNC-DeformControl dataset containing machining parameters such as cutting speed, feed rate, depth of cut, tool temperature, and material type. Data preprocessing involved normalization and categorical encoding. The NeuroPID-CNC model, structured as a binary classifier with a single hidden neuron using a sigmoid activation function and Adam optimizer, was trained on 70% of the data and evaluated on the remaining 30%. It achieved 92% accuracy, 90% precision, 93% recall, 91.5% F1-score, and 0.84 MCC, outperforming conventional algorithms like SVM, RF, LR, and KNN. A real-time feedback loop further enables adaptive learning. The NeuroPID-CNC approach effectively predicts deformation risks and recommends real-time control actions, enhancing machining reliability and reducing material waste. This makes it a promising solution for smart, adaptive manufacturing environments.