Health Estimation And Prognostics For Complex Electromechanical System Based On Semi-quantitative Information

Complex electromechanical systems are one of the crucial applications in manufacturing industry,which give a fundamental direction for national economy and development.With the development of modern industrial technology,the safety operations and costs of complex electromechanical system are mainly affected by the reasonable health management.Recently,the studies about health management of complex electromechanical systems are attracted more attention.Health estimation and prognostics are crucial for the health management of complex electromechanical systems.However,Due to the features containing strong coupling and non-linear complexity in complex electromechanical system,to date there has not been a complete and mature theory.Thus,this research has essential significance in both academia and engineering practice.To reflect the current health states and tendency of the system status in the future,this paper focus on the health estimation and prognostics for complex electromechanical systems.A certain type of aero-engine is taken as a research object.Moreover,a semiquantitative information approach,which contains both qualitative knowledge and quantitative information,is employed to construct the health estimation and prediction model for complex electromechanical systems.This approach can effectively solve the limitation of sample scarcity,which improves the accuracy of health estimation and prediction.It provides an effective basis for decision making of health management in complex electromechanical systems.Four main aspects of this research are presented as follow:1.To analyze the failure mechanism of aero-engine,this chapter presents a method for health feature extraction in complex electromechanical systems.Based on the relationship between the failure symptom and the change of parameters,the feature quantities are determined and the influence of the change for characteristic quantity on a system from the aspect of mechanical dynamics is carried out.A simulation is tested to obtain the adequate valid feature data.Moreover,infinite irrelevance method is taken in features selection to reduce the calculation of the model and the structure can be optimized.Finally,an idea of feature extraction for the health in complex electromechanical systems in a certain aero-engine is summarized to verify the effectiveness.2.To solve the problem of health states estimation in complex electromechanical systems,this chapter proposed a health states estimation model based on Belief Rule Base(BRB).Evidential Reasoning(ER)is introduced as the inference engine to realize the reasoning of knowledge,which can fuse more abundant uncertain information and provide a more realistic way of knowledge expressing.Differential Evolution(DE)is used to update the parameters,where the limitation of lacking of subjective knowledge in input parameters can be addressed.An experiment of health estimation for critical components in a certain type of aero-engine is carried out.The results are verified the effectiveness of proposed BRB model in describing the health state of system.Compared with other models,the estimated health states follow the true value very well,and the accurate results are obtained.3.To solve the problem of healthy state prediction of the complex electromechanical system,a model of health state prediction of a complex electromechanical system based on double-layer BRB is proposed in this chapter.The double-layer BRB model contains two layers of model that achieve different functions.The first layer of BRB defined as BRB_layer1 is used to establish the dynamic change of the time series of features,which considers both historical and current information.The second layer of BRB defined as BRB_layer2 is employed to combine the features for predicting the health state of the complex electromechanical system.Due to the initial parameters are given by experts,the training may have boundedness and not be appropriate for engineering practice.The projection covariance matrix adaption evolution strategy(P-CMA-ES)is chosen as the optimization algorithm to train the initial parameters.Finally,the vibration fault experiment of a certain type of aero-engine is analyzed.Compared with other models,the double-layer BRB model has higher prediction accuracy and can predict the health state very well.4.To overcome the low reliability of the observed feature data caused by certain degradation of the sensor and noise disturbance in the actual working conditions,this chapter proposed a health prediction model for complex electromechanical systems with the consideration of measurement errors.Considering the disturbance and errors,which are random and unpredictable,these factors are often impacted on the data collection as the input of BRB rules due to the low tracking ability of the sensor.When these errors are entered as input BRB attributes,the information expression ability of features is affected.The measurement error fused in the double-layer BRB prediction model is calculated via proximity,whereas the accuracy of the health states is further improved.Finally,the vibration fault experiment of a certain type of aero-engine is further analyzed.This approach is formulated more consistent with the actual practical engineering and the effectiveness and accuracy are verified.