Digital Twin-Driven Surface Roughness Prediction Based on Multi-sensor Fusion

Aiming at the problems in the current surface roughness prediction methods that interface with the normal machining of machine tools, poor real-time performance, inconvenient sensor installation and high cost, a multi-sensor surface roughness prediction method based on digital twin was proposed. Firstly, the digital twin model of intelligent workshop was established as the only data source for workshop monitoring and surface roughness prediction; secondly, vibration signal was preprocessed and combined with power, energy consumption and cutting parameters to construct joint multi feature vector, and feature fusion was performed by principal component analysis; finally, support vector machine was used to predict surface roughness. The results showed that the average relative error was 4.00% and the maximum error was 0.07 μm, which verified the effectiveness of the method.