Abstract: Most of the existing modular design methods focus on the optimization iteration of modular index.In order to improve the information interaction between components,a product modular partition method based on ensemble learning algorithm is proposed.Firstly,the product connection and function design structure matrix are constructed.Secondly,the information transfer and information loop between components are taken as the construction criteria of weak classifiers.In the last,the strong classifier model is constructed by combining a large number of weak classifier models,and the product module partition results are given by the strong classifier.A ZY3200 hydraulic support case is used to demonstrate the proposed module partition method.Compared with direct clustering and genetic algorithm,the modularity index I1 of ensemble learning algorithm is 0.8438,which is closer to the design structure matrix diagonal.Sensitivity analysis shows that the Q value is stable at 0.5562when the number of loop components is greater than 3,which further verifies that the proposed method has the intelligence and robustness of machine learning.