Research On Key Technology Of Fault Prognostic And Health Management For Complex Equipment

In recent years, the security and reliability of aerospace, ship and other complexequipment systems have become the focus of attention. Due to its complex structure, it willcause a huge loss in the event of failure. Therefore, it is urgent to improve the reliability,repairability and security of complex equipment systems. However, current fault diagnosisresearches mainly focus on the research of system condition evaluation and fault diagnosis,concern about the "current" running status, while the research on system failure predictionand health management is quite less. The traditional "posterior repair" and "plannedmaintenance" approaches are quite insufficient in response to rapidly changing newsituations."On-condition repair" and "predicting maintenance" will prevent the fault in thebud, and become the development direction of system maintenance in the future.As a result, the research about fault prognostic and health management (PHM) hasbecome a focus of scholars at home and abroad. PHM technology represents thetransformation of fault diagnosis technology from the traditional diagnosis based on sensorto forecast based on intelligent systems. However, the current domestic research in thisrespect is only in the beginning, and there are less research results. Aiming at the key theoryand technologies involved in PHM technologies of complex equipments, this paper carriesout the research, which mainly includes:Firstly, the PHM service model research on complex equipments. Aiming at thecurrent problems such as the lack of effective organization and management of faultdiagnosis resources, and the low resource sharing rate, a service-oriented PHM systemarchitecture is proposed, through the analysis of the characteristics of complex equipmentsand the shortage of traditional fault diagnosis models. The system service mode andcharacteristics are studied and key technology and main function modules needed in theconstruction of the system are analyzed, providing a new thoughting for research on publicPHM system service platform in complex equipment field.Secondly, Research on data dimension reduction and fault diagnosis technology. Faultdiagnosis is the core of PHM. In view of the features of complex equipments such ashigh-dimensional, nonlinear, and the difficulty to identify early faults, etc, a method basedon the nonlinear manifold learning and mixed Hidden Markov Model (Hidden MarkovModel, HMM) for fault diagnosis is put forward. In order to fully reflect the true state ofsystem, system state can be divided by mixed HMM model into the normal state, intermittent fault, intermediate state and failure state. Through the nonlinear manifoldlearning algorithm (the Locality preserving the projection, LPP), the original highdimensional failure data is mapped to low dimensional space, and the inner manifoldfeatures are extracted from data as the characteristic vector; Then the hybrid HMM is usedas a classifier to recognize each state. Through the simulation analysis, the proposedmethod is compared with other methods, and the result shows that the LPP-HMM methodcan effectively identify the early fault features and has a higher fault recognition rate.Thirdly, the state of time series prediction technology research. Fault prediction is thekey of PHM aiming at the gradual changing fault types, a state prediction method based onthe Auto-Regressive Moving Average Model (ARMA) and Artificial Neural Network(ANN) is put forward. Using the advantage of ARMA model in capturing the linear part ofthe time series and the good performance of ANN in processing nonlinear time series, ahybrid forecasting model is established. The proposed method uses ARMA model to modelthe linear part of the time series, and uses the remainder to build a neural network model,which finishes nonlinear modeling part. Based on the hybrid model, the paper takes theremote sensing voltage data of a certain type of satellite as an example, to forecast thetelemetry parameter using static model and dynamic model respectively. The experimentalresults show that the dynamic model has higher prediction accuracy. Then based on thehybrid model, the paper takes the remote sensing pressure data as an example for conditionmonitoring, which shows that this method can effectively reduce the failure rate of falsealarm.Fourthly, the research on fault knowledge modeling and service. Fault knowledgemanagement is the foundation of PHM. Aiming at the serious problems such as the lowreuse rate of fault diagnosis knowledge resources and the serious knowledge fault, themodeling technology and service based on domain ontology is proposed. First of all, twotypes of fault knowledge classification methods, the static and dynamic are put forward,and the reasoning knowledge ontology and multi-domain knowledge ontology are built;Then, based on the reasoning knowledge ontology, the fault knowledge reasoning isrealized, Based on domain ontology, the distributed knowledge resource index, semanticannotation, semantic retrieval and knowledge management are realized, and the diagnosisstatic knowledge of product design, test stage is effectively acquired; Through theservitization encapsulation and semantic annotation of the class knowledge resources, therelative curing diagnosis and prediction model, the dynamic knowledge retrieval and invoking service are realized, which effectively improves the use efficiency and reuse rateof diagnostic knowledge resources.Fifthly, the PHM service system development. Based on the above research content,according to the characteristics of spacecraft in orbit, the spacecraft ground PHM systemresearch and development are carried out. Fault prediction, fault diagnosis, and faultknowledge management and service are integrated in one, which provides an effectivetechnical support for improving the operational security of spacecraft in orbit and groundtest experiment capability.This paper makes systematic analysis and research on the service mode and keytechnologies of complex equipment PHM system. By combining the characteristics ofspacecraft, a corresponding PHM prototype system is developed, which verifies thefeasibility and effectiveness of the proposed method in this paper. The research in this paperis of great significance to improve the complex equipment diagnosis and maintenancecapability, and low the operational cost.