Research On Operational Reliability Assessment Of Mechanical Equipments Considering Performance Degradation

Moving from being a big manufacturer to a strong manufacturer,domestic mechanical equipment has increased greatly in every aspect.However,compare to international advanced level,the technology of high-end mechanical equipment is large far behind.The reliability of mechanical equipment is a key question,which restricts the development of high-end mechanical equipment.Operational reliability assessment is a necessary method for reliability quantitative analysis of mechanical equipment,and its main task is to measure the performance of mechanical equipment during operation,evaluate the mechanical equipment reliability condition in real time,and ensure the health operation of equipment.The traditional reliability assessment methods based on probability and statistic require a great deal of failure samples.These methods are not suit for reliability assessment of the specified mechanical equipment or in real time analytics.Therefore,this paper will investigate operational reliability assessment methods from performance degradation.The contains can be drawn as follows:(1)The characteristics of performance degradation of mechanical equipment are analyzed in this paper.A hidden Markov model(HMM)is employed to describe the hidden relation between the degradation process and the operational condition of mechanical equipment.Considering the traditional HMM cannot precisely depict the degradation process of mechanical equipment,a novel continue-time HMM method is developed for operational reliability assessment in this paper.In this method,the transition intensities are obtained by continue-time transform,and the reliability is calculated by solving the ChapmanKolmogorov differential equations.The validation case of slide guide illustrate that the proposed method can continuously assess the operational reliability of slide guide using discrete degradation data,and has higher precision than that of traditional HMM.The effectiveness and usefulness of the proposed method are further validated by a practical application of a milling machine.(2)Degradation data acquisition is a hard problem for mechanical equipment.Therefore,this paper proposed an operational reliability assessment method based on empirical mode decomposition(EMD)and sample entropy using accessible vibration signals,which can indirectly reflect the reliability condition of equipment.In this method,the vibration signals are decomposed by EMD algorithm,the sample entropies of the intrinsic mode functions are taken as degradation features,the reliability are calculated by Markov model,the relations between degradation features and reliability are mapped using a support vector regression model,and then the operational reliability of mechanical equipment is assessed by the proposed model.The results show that the proposed method can effectively extract the degradation features from vibration signals,and directly assess reliability from vibration signals.In addition,the practical application result of a milling machine shows the method is able to assess the operation reliability.(3)Aiming at the problem of current operational reliability assessment methods based on huge amount of signal processing technologies,a novel operational reliability assessment method using deep leaning theory is proposed in this paper,which only employs a few signal processing technologies and expert expertise to mine the underlying information of operational reliability.The deep degradation features are extracted from frequency domain vibration signals via an unsupervised stacked auto-encoder(SAE)model.Considering the characteristics of degradation processing of mechanical equipment,a Cri indicator,which consists of a correlation indicator and a monotonicity indicator,is employed to evaluate the degradation features.A novel method based on the domain distance of condition features is developed to dynamically assess the operational reliability of mechanical equipment.A practical application of operational reliability assessment of bearings is given in this paper,and the results show that the proposed method is able to adaptively extract the degradation features from raw signals without amount of complex signal processing technologies and expert expertise,and the reliability degrees calculated by the proposed method well describe the degradation process of bearings.(4)The change of operational reliability of mechanical equipment is an expression of complex performance degradation process,which under the multi-fact coupling influence.A simple SAE model,which extracts the deep features on one aspect,cannot depict exactly the degradation process of mechanical equipment.Therefore,a deep learning based multimodel ensemble feature extraction method is proposed for operational reliability assessment of mechanical equipment in this paper.Several SAE models with different hyper parameters(i.e.the number of hidden layers,the number of units in each hidden layer)are established to mine the deep degradation features in different ways.The mixed degradation features are clustered using K-means algorithm,and the features,which have higher correlation with cluster centers are selected.The selected features are utilized to calculate the operational reliability degrees of mechanical equipment by the method proposed in(3).This method is employed to evaluate the operational reliability of bearings.The bearing evaluation results show that the proposed method can independently and mutually extract deep features using several SAE models with different hyper parameters.The degradation features extracted by the ensemble method can more overall reflect the degradation properties of mechanical equipment.Moreover,the ensemble model can not only extract better features than simple SAE model,but also possess excellent generalization.