Research And Application Of Fault Diagnosis And Classification Based On The Stability Margin

With the development of science and technology and the continuous advancement of the "Made in China 2025" strategy,the scale of modern industrial systems is increasing,and the degree of automation is higher.Correspondingly,people have more requirements for safety in operation and the quality of production,thus attracting more attention to fault diagnosis technology.At the same time,data storage technology has developed rapidly in recent years.More and more process data can be saved in industrial systems,which also makes data-driven fault diagnosis technology more and more applied to industry systems and achieved good application results.Furthermore,a rising demand for industrial system health management and maintenance requires fault classification technology,an important branch of fault diagnosis technology,to play an important role in them.This paper focuses on the realization and application of fault diagnosis and classification technology based on stability margin.First of all,stability margin is regarded as an indicator to quantify the stability of the system,and the definition of the model-based stability margin is given in this paper.Next step is to find the relationship between the stability margin and the system stable kernel description(SKR)/stable image description(SIR).Accordingly,another description form of the stability margin is given in this paper.Secondly,it is difficult to describe the system and its stability margin with an accurate mathematical model due to the increasing complexity of modern industrial systems.This paper mainly studies data-driven stability margins of the system,which to an extent,can be translated into solving the data-driven SKR of the controller and the data-driven SIR of the controlled object.This paper explains the SKR estimation method when the controller information is known and when it is unknown,and illustrates how to realize data-driven SIR using the least square method.In addition,the realization of real-time stability margin is changed into that of real-time data-driven SIR,and the real-time estimation of SIR comes true via improved recursive least square method.Thus,real-time estimation of system stability margin is achieved.Thirdly,in response to the needs of real-time fault classification of industrial systems,this paper takes the real-time estimated stability margin as an important indicator of fault classification.Moreover,the system input and output data within a period of time are introduced to assist the stability margin for fault classification.In the process of designing the classifier,the LM-BP neural network multi-classifier is finally adopted for classifying multiple faults after the effects of classification multi-classifiers are compared.Finally,the double closed loop DC motor speed control system is used for the application and verification of the above research content,showing that the data-driven stability margin estimation method proposed in this paper possesses high accuracy in estimation,and the designed multi-classifier has a good effect on the classification of five faults in the DC motor.