Privacy-preserving for assembly deviation prediction in a machine learning model of hydraulic equipment under value chain collaboration