A Performance Prediction Method for a High-Precision Servo Valve Supported by Digital Twin Assembly-Commissioning

The manufacturing of a high-precision servo valve belongs to multi-variety, small-batch, and customized production modes. In the process of assembly and commissioning, various characteristic parameters are critical indicators to measure product performance. To meet the performance requirements of a high-precision servo valve, the traditional method usually relies on the test bench and manual experience for continuous trial and error commissioning, which significantly prolongs the whole assembly-commissioning cycle. Therefore, this paper proposed a performance prediction method for a high-precision servo valve supported by digital twin assembly-commissioning. Firstly, the cloud-edge computing network is deployed in the digital twin assembly-commissioning system to improve the efficiency and flexibility of data processing. Secondly, the method workflow of performance prediction is described. In order to improve the accuracy of measurement data, a data correction method based on model simulation and gross error processing is proposed. Aiming at the problem of high input dimension of the prediction model, a key assembly feature parameters (KAFPs) selection method, based on information entropy (IE), is proposed and given interpretability. Additionally, to avoid the poor prediction accuracy caused by small sample data, a performance prediction method based on TrAdaboost was utilized. Finally, the hysteresis characteristic commissioning of a high-precision servo valve is taken as an example to verify the application. The results indicate that the proposed method would enable accurate performance prediction and fast iteration of commissioning decisions.