Fault Diagnosis And Leasing Pricing For Intelligent Equipment Via Cloud Manufacturing Platform

The manufacturing industry is the very foundation for the establishment and rejuvenation of a nation.The development,progress and innovation of the industry have always been a critical research subject in various areas.With the emergence of the new generation information technology(New IT)represented by cloud computing,big data,industrial internet and mobile internet,and the new generation artificial intelligence(New AI)represented by deep learning,deep reinforcement learning and swarm intelligence,the intelligent manufacturing technology has become the core driver for the current upgrade and transformation of the manufacturing industry.Multiple new manufacturing modes and business formats have derived from intelligent manufacturing,which defines the modern manufacturing industry.The user-and demand-oriented cloud manufacturing is one of the advanced modes among them.In this mode,the cloud manufacturing platform is an important bond connecting the intelligent equipment manufacturers and the clients,as well as a major impetus boosting the manufacturing industry’s transformation to the service industry.This research focuses on some of the problems lying on the way of the manufacturing industry’s transformation based on cloud manufacturing platform.They are:(1)problem in the reliability of the fault diagnosis model caused by insufficient and unevenly distributed training data;(2)due to complex operational conditions,the training data and actual test data are unevenly distributed,resulting in narrow applicability of the diagnosis model;(3)the redundancy and waste of manufacturing resources caused by maldistribution;(4)failure to provide efficient leasing service due to lack of reasonable pricing strategy.Above those problems,this paper proposes two service modes based on the cloud manufacturing platform,including fault diagnosis and service leasing.The research methods and novelty are described as follows:(1)A new fault diagnosis method,i.e.Digital-twin-assisted Fault Diagnosis method using Deep transfer learning(DFDD),is proposed in order to solve the problem of low-reliability and poor applicability diagnosis models.At first,the potential problems that are not considered at design time can be discovered through front running the ultra-high-fidelity model in the virtual space,while a Deep Neural Network(DNN)based diagnosis model will be fully trained.In the second phase,the previously trained diagnosis model can be migrated from the virtual space to physical space using Deep Transfer Learning(DTL)for real-time monitoring and predictive maintenance.This ensures the accuracy of the diagnosis as well as avoids wasting time and knowledge.(2)The DFDD is further improved by using an Improved Salp Swarm Algorithm(ISSA),to solve the problems of relying on artificial experience and further improve the diagnosis performance.DTL can classify and predict the equipment status among different distributed data,but the initialization parameters of the fault diagnosis mode are provided through random initialization or empirical methods,which is inefficient and difficult to ensure optimal parameters.In this paper,ISSA is used to adaptively optimize the network initialization parameters,including the number of hidden layer nodes and sparsity penalty coefficients.(3)The intelligent equipment leasing mode is proposed and the pricing model is established in order to share manufacturing resources among the whole industry.In cloud manufacturing,intelligent equipment is leased to clients in the form of services.Meanwhile,preventive maintenance is performed by real-time monitoring and fault diagnosis for extending the service life of equipment and keeping the equipment running.The leasing mode can effectively solve the problem of uneven distribution of resources and realize more efficient utilization of manufacturing resources.From the perspective of supply and demand Stackelberg game,a bi-level programming model for service leasing in cloud manufacturing platform is established.Bi-level programming problem is a leader-follower hierarchical complex decision making problem,and is NP-hard.The Pareto optimal solution of the proposed model is obtained by Nondominated Sorting Genetic Alogorithm II(NSGA-II),which maximizes the utility of both the service providers and the service demanders.(4)The proposed methods are tested and demonstrated by several case studies.The proposed fault diagnosis methods(DFDD and ISSA-DFDD)are tested by a case of an automobile manufacturer in Guangdong Province.The results show that the proposed methods can predict and judge the health status of intelligent equipment accurately.The results also show that the ISSA-DFDD method outperforms the DFDD method in accuracy and robustness.Both the leasing mode and the pricing model are tested by a case of a tool manufacturer.Using the proposed mode and model can maximize the utility between service providers and service demanders.The results show that the leasing mode and pricing model are feasible and reasonable in practice.