| Prognostics and health management is a kind of prediction and management method originated in aerospace industry.It can transform data and health status into degradation information of the system,so as to improve engineers’ cognition of the system.At present,it has been widely applied in heavy industry fields such as energy,automobile and nuclear industry.The working environment of marine transmission system is more severe than that of land equipment.Therefore,it is necessary to timely grasp the health status of critical components of the marine transmission system and formulate low-cost early maintenance strategies to reduce the cost caused by failure.In this paper,a new system health state prognostics and maintenance strategy is proposed for critical components of the marine transmission system in four aspects: feature extraction and choosing,system degradation state prediction,degradation stage evaluation and maintenance strategy.The main work of this paper is as follows:(1)Aiming at the problem of inaccurate prediction results caused by non-monotonic feature data containing noise in critical components of the marine transmission system,a multi-dimensional feature data processing method is proposed.Firstly,a method of extracting feature data based on accumulation is presented.The method obtains data preprocessing from the sensor based on multi-resolution signal decomposition,and then the classical features and inverse trigonometric features in the preprocessed data are extracted.The feature quality after accumulation is evaluated from three aspects of monotonicity,correlation and trendability,and the available data is selected to reduce the number of feature data.Then,a data choosing method based on compatibiliy is proposed.This method defines the compatibiliy function of the model and pre-selected features,and further reduces the data dimension through the compatibiliy function to get the features suitable for prediction.An example of bearing in the marine transmission system is used to verify the two methods proposed.(2)Aiming at the problem of inaccurate prediction caused by data uncertainty in critical components of the marine transmission system,a system degradation state prediction method based on grey interval particle filter is proposed.Firstly,the cognitive uncertainty is quantified and the grey interval theory is used to suppress the cognitive uncertainty in the parameters.The gray model is established to predict the interval boundary of prior data.Then,the kernel smoothing method is used to suppress the uncertainty propagation in the iteration process,which avoids the particle degradation problem while keeping the information of the generated parameter particles complete.Finally,according to the degradation state and the current observation data,the remaining service life is predicted and the assessment criteria is given.The degradation state and residual life of the system are predicted by taking the crack growth model of marine gear as an example,and the robustness and applicability of the proposed method are discussed.(3)Aiming at the problem that the degradation state evaluation of critical components of the marine transmission system is not accurate due to the lack of prior knowledge,a multi-dimensional degradation state dynamic evaluation strategy based on clustering is proposed.Firstly,this paper proposes a clustering algorithm for unlabeled data,which is fuzzy clustering method based on improved glowworm swarm algorithm.This method can deal with the high computational cost caused by improper clustering initial center points.Then,a dynamic evaluation strategy for system degradation stage is proposed.This method can update the failure time of the system according to the newly added observation data and update the degradation state of the system,so as to predict the remaining life of the system.Taking the crack growth model of the marine propulsion system as an example,the remaining life of the system and the failure time of the system are predicted and the effectiveness of the method is evaluated.(4)Aiming at the problem of increasing maintenance cost caused by complex relationship between coupling components of the marine transmission system,a predictive maintenance strategy of multi critical components based on dependency relationship is proposed.Firstly,a grouping model of multi-component system is proposed.The influence of system structure on grouping policy is considered in this grouping model.Secondly,the economic model of grouping model is established.To solve the Non-deterministic Polynomial-complete problem caused by the optimization of maintenance grouping strategy for multi-component systems,a random assignment scheme based on genetic algorithm is proposed to find the optimal grouping.This strategy selects the best maintenance strategy for each component or subsystem,ensuring the lowest expected depreciation cost of the system.The effectiveness of the maintenance strategy is illustrated by taking the multi-component complex system and the whole critical components of the marine transmission system as examples,and the influence of system structure,maintenance period and initial parameters on the maintenance strategy is discussed. |
