Research On Early Fault Diagnosis And Maintenance Strategy Of Equipment

The degree of automation,precision and complexity of equipment used in modern industrial production process is rapidly improving.On the one hand,it greatly improves the production efficiency,saves human resources,improves product quality and service level,but on the other hand,it also puts forward higher requirements for the control of equipment health.Once there is a significant functional failure in the operation of the equipment,it will lead to unexpected shutdown,high maintenance costs,or damage the production system,resulting in personal injury and safety accidents.Therefore,how to avoid sudden,serious and uncontrollable functional failure of production equipment is a problem worthy of study.In order to avoid the serious loss caused by the fault developing to the uncontrollable stage,it is a research direction worth considering to intervene and control the fault developing process from the early stage.Early failure of maintenance costs low,less maintenance difficulty,the less time required,can be used to maintain the long time,than in the middle-late stage of fault,maintenance and control of the early maintenance has a better economy,higher fault tolerance and are more likely to optimize allocation of maintenance resources and the maintenance time,so the early fault diagnosis and maintenance can effectively solve the sudden failure of unplanned downtime and production system damage and other problems.Due to the early fault feature was not significant,the signal is submerged by noise,judge fault reason and lack of obvious signs,so on the reasoning fault diagnosis and fault reasons there is more to solve the problem,at the same time also need to consider a maintenance policy in order to make full use of the equipment from early failures,to develop to a long period of time before failure,by optimizing the maintenance cycle to maximize the production value of equipment.Therefore,based on the manufacturing equipment as the research object,from the early fault diagnosis,failure cause analysis and repair and maintenance of failure of three parts,aimed at establishing scientific eventually through preventive maintenance strategy of equipment for effective maintenance and management,control the early development of small fault process,make the equipment can maintain a longterm stable operation,maximize the value of production.The main contents of this paper are as follows:(1)Facing the fact that there is a lack of label information in fault diagnosis and it is difficult to classify,a fault diagnosis model based on state estimation is studied,which can effectively identify small faults in early stage,lay the foundation for the follow-up preventive maintenance.In feature extraction,the noise interference in the original signal is eliminated by combining the de-noising self-encoder and the undercomplete self-encoder,based on the long-term and short-term neural networks and the encoder-decoder model,the predictive sequence data of the device state are output,and the problem of information dilution is solved by introducing attention mechanism,finally,a fault diagnosis method based on residual sequence is presented and verified on the IMS data set provided by NASA,and the capability of the proposed method for early fault diagnosis is illustrated.(2)Aiming at the problems of lack of information in the early stage of fault,heavy workload,long time consuming and strong blindness in fault cause analysis,a fault cause reasoning algorithm based on Bayesian network is proposed.The directional relationship in the fault cause relationship network is described by the way of conditional probability,and the possible fault cause set and its confidence are completed by the weighted transfer method of confidence Finally,the automatic analysis of fault causes based on fuzzy input is realized,which can effectively help to locate the fault causes.In addition,the fault cause relation network based on Bayesian network is also the standardization of reasoning process,which greatly reduces the dependence on the experience and knowledge level of relevant personnel,and makes the analysis and experience of the same fault type quantitative,transferable and iterative.(3)Based on the control of fault development process,a dynamic multi-objective maintenance strategy is studied.On the one hand,the setting of multi-objective weight coefficient can meet different management objectives.On the other hand,the dynamic output of maintenance cycle can better match the uncertain requirements of enterprise production site.Finally,through preventive maintenance,the reliability of equipment can be restored,and the process from minor fault to significant fault can be delayed,and the life cycle of equipment can be extended.In this process,based on the importance of each equipment to the system reliability,the value of preventive maintenance is described,and then the optimal preventive maintenance cycle is obtained.By combining with the availability model,the input of multi-objective maintenance model can be obtained,and the dynamic maintenance plan can be made.