Research And Implementation Of PHM Model In Intelligent Control System

In the past decades,AI,IoT and other technologies have developed rapidly,which has set off tremendous changes in many fields,and has also promoted intelligence in various fields.As the main body to undertake informational tasks,the structure and scale of various types of electrical equipments and information systems have become larger and more complex,and the requirements for stability and reliability become higher and higher,so the traditional intelligent control system needs to be updated to adapt to the complex equipment system environment.Prognostics and Health Management(PHM)technology has been put forward as a new solution,and it's just right for improving intelligent control system.However,there are very few researches and applications on PHM models in intelligent control systems,and PHM technology itself has many problems that need to be solved urgently.The main work of this paper is as follows:1)Proposing a predictive maintenance framework based on PHM modelCompared with other PHM applications of fixed object scenes,intelligent control system contains various devices or systems,in order to empower intelligent control system better,this paper starts from the system composition structure,and divides hierarchically,the PHM predictive maintenance framework for intelligent control system is established,according to the characteristics of performance degradation;2)Proposing an health status assessing method based on improved Mahal distanceCompared with the traditional Mars distance,the improved method proposed in this paper is to reduce the amplification of some nuances by using the Principal Component Analysis method,and to highlight the process of performance degradation by using normal parameter values instead of sample averages;3)Classify the remaining useful life method based on data driveThe application of RUL prediction method on traditional single-scene is narrow,and according to the characteristics of controlling objects,the data-driven method is divided into three categories:health assessment,data fusion and degradation parameter sequence,which can contain most of the existing applicable scenarios and provide architectural support for the deployment application of PHM model;4)Proposing three residual applicable life prediction methodsFor scenarios where there are no parameters that can directly symptom the degradation process,the method applies least squares method to fit the linear model,and reverses to obtain the remaining useful life;For scenarios where parameters contain partial degradation information but are not sufficient to fully symptom degradation,a RUL prediction method based on PCA-ELM is proposed,the original feature data is fused and form a new data set,and the ELM prediction model is trained.;The RUL direct prediction method based on LSTM network is proposed.This method is suitable for scenarios with direct parameters that can characteristic degradation performance.