Research On Rolling Bearing Health Management Model And Predictive Maintenance Optimization Strategy Including Remaining Service Lif

Under the background of comprehensively promoting the implementation of the manufacturing power strategy,the state has included the production and operation and quality management for intelligent manufacturing in the ”14th five-year plan” of the Ministry of management science and the key project funding areas of the Ministry of Management Science in 2022.《Made in China 2025》also proposes to take ”quality first” as the development concept.Moreover,with the increasing precision and complexity of modern equipment,in practical application,it is particularly critical to obtain the health status of complex components with high failure rate and difficult to disassemble.The multiple consequences of component failure or degradation can also be included in the operation reliability measurement.It is necessary to take overall consideration in the operation and maintenance decision-making to meet the actual operation decisionmaking needs,but because of this,the maintenance cost of equipment accounts for a large proportion of the enterprise expenditure.As the core supporting technology for health management and maintenance decision-making,mechanical equipment condition monitoring and remaining useful life(RUL)prediction need to be developed,and predictive maintenance technology has not been well realized at present.It is necessary for researchers to constantly try various new and effective methods to better solve this problem and speed up the mature process of its implementation methods and technology application.Based on this,facing the major national needs,this study explores new solutions and approaches for predictive maintenance of key components of major equipment from the perspective of health status monitoring and obtaining RUL,and deeply studies and discusses the health management and scientific maintenance methods of key components of major equipment based on RUL.The research work of this paper is as follows: The rolling bearing,a multi frequency parameter fault component,is taken as the research object.First,combining the basic concept of failure prediction and health management(PHM)and the actual needs of engineering applications,based on the analysis of the basic connotation of machinery and equipment health management,the object of condition monitoring and RUL prediction is further abstracted.Based on the characteristics of small samples of degradation process data,the methods including degradation feature extraction,construction of degradation feature set.The general process model of three sub problems of state prediction in order to find the causal relationship between health state and operation and maintenance strategy;The implementation path of health management objectives is further abstracted,and a general process model is proposed,which includes three sub problems: fault consequence analysis,maintenance cost modeling and maintenance decision-making.Second,through the analysis of the application status and problems of existing research technologies,the research is carried out for the three sub-problems of condition monitoring and remaining service life prediction technologies.To solve the problem of insufficient ability of data features to characterize degradation information in equipment condition monitoring data,the degradation mechanism of key components of mechanical equipment is analyzed and a three-stage degradation feature indicator set construction method is proposed,and an integrated robust prediction method is proposed to predict bearing RUL based on the construction of bearing degradation indicator set: fast correlation-based approximate Markov blanket filtering(FCBF-AMB)and maximum information coefficient(MIC)selection method to construct a subset of bearing degradation indicators.These selected degradation indicators are then fed into a long short-term memory(LSTM)neural network prediction model enhanced by Ada Boost algorithm.The method is also validated using the bearing degradation dataset from Xi’an Jiaotong University,and after comparing with other typical feature extraction methods,the prediction accuracy of the proposed method in this study is improved by 1.8%-14.87%.Thirdly,based on the previous study,in order to reveal the health status of rolling bearings more comprehensively,quantify the uncertainty of equipment health status,and better serve the health management and decision making.The problem of efficient feature extraction and RUL prediction in the operational degradation process of rolling bearings is further investigated in depth.Through data reduction and key feature mining analysis,a feature vector based on joint features in the time-frequency domain is proposed,which can describe the bearing degradation process more comprehensively.In order to maintain effective information without increasing the scale of the neural network,a joint feature compression calculation method based on degradation metrics is proposed to determine the input data set.A temporal convolutional network is combined with a quantile regression algorithm to predict the conditional distribution of predicted values based on kernel density estimation(KDE)of the probability density at any moment,so as to obtain more accurate prediction intervals and probability density curves and quantify the uncertainty of the bearing operating state.Applying the proposed method to the experimental data of bearings in Xi’an Jiaotong University and comparing it with the prediction results of other typical prediction models,the proposed method in this study can obtain higher prediction interval coverage and narrower average prediction interval width.Fourth,on the basis of obtaining the operation status of complex mechanical equipment,health management and maintenance decisions need to be made for it.It is proposed that the Markov decision process is used for maintenance scheme selection on the basis of obtaining the degradation state of the equipment.It is proposed that the multiple consequences of component failure are incorporated into the maintenance cost analysis in the maintenance decision-making process,so that the maintenance scheme can take into account the operational reliability and maintenance economy of the complex system.The complex system of subway train is also analyzed and studied as an example,and the failure rate model based on the memory factor is established by combining the dynamic decay law of subway train,and the failure risk of subway train components is quantitatively evaluated,and the risk penalty cost model is constructed.Fifth,to address the conflict of interests between operators who aim to improve operational reliability and maintainers who aim to reduce maintenance cost in the process of actual formulation of operations and maintenance strategy.We propose a rolling bearing preventive maintenance decision method based on a dynamic noncooperative game model,and use particle swarm optimization algorithm to solve the equilibrium problem and achieve a balance of interests among the decision participants and realizing cost-effective maintenance.Case analysis shows that the dynamic non-cooperative game approach can better guide the maintenance decision,and the preferred bidder has more advantages.Through the above research,we can achieve high-efficiency and high-precision control of equipment health status,and provide scientific guidance for equipment operation management and maintenance of industrial enterprises,so as to make the response measures in advance,ensure the sustainable operation of equipment and industrial line,and promote the sustainable development of enterprises.The conclusion of the above research can be used as an effective solution to the problem of parametric failure and degradation of key components of complex mechanical system.The proposed policy recommendations are applicable to industrial enterprises that need to ensure the stable and sustainable development of operation and make business decisions quickly,so as to improve the production safety and operation reliability of enterprises and systematically reduce the operation cost of enterprises.