A Study And Application On Volterra Series Theory Utilized In Signal Feature Extraction Of The Non-stationary And Nonlinear Mechanical Failure

Although effective handling methods have been established in the field of signal feature extraction of mechanical failure at present,the majorit y of the methods onl y aim at stationary signals.Thus,onl y a small amount of research findings are acquired,which concern the non-stationary signals or even non-stationary and nonlinear signals.Solutions to ph ysical modeling of the mechanic al s ystems involved in the nonlinear s ystem have been regarded as an important research field b y man y domestic and international scholars.A new method concerning the nonlinear s ystem identification has been widel y applied in every walk of life in recent years,which is based on kernel of the volterra series,and also obtains great progresses and achievements.This paper based on volterra series of the nonlinear system,solution of the kernel function and it's representing method has put forward the estimat ion method of low order kernel of the volterra series.The low order kernel's estimation algorithm,the form of convergence and truncated form have been researched respectivel y in this study,and all the research findings have been experienced computer simu lations one b y one.The estimation method of low order kernel of the volterra series is applied to the bearing fault diagnosis so as to identify breakdown of the bearings and it's effectiveness has been verified in practical application.The research conte nts of the paper and accomplishments obtained have been shown as follows:1)The thesis presents the overseas and domestic research status of nonlinear s ystems identification and fault detection and diagnosis of mechanical fault,and summarizes the commonest research methods utilized in fault diagnosis of the nonlinear s ystems,and then compares the methods mentioned in this paper.Finall y,volterra series diagnostic is elicited,and innovation points of the research is presented.2)Volterra series theory is e licited b y introduction of the nonlinear systems identification methods,and theoretical foundations of the nonlinear volterra series are also discussed in this dissertation.The properties and manifestation of time domain and frequency domain of volterra series are also introduced thoroughl y.3)Calculating method concerning the kernels of volterra series,the least squares method,the differential equation and chaos method are mentioned in this paper.Through comparison to the above three methods,the defi ciencies appearing in modeling have been summarized and then the solving method of the kernels of volterra series based on multiple puls e excitation method is proposed.4)The optimal solving model of the low order kernels' volterra series is established b y stud ying the convergence problem of volterra series bas ed on multiple-pulse excitation,and the merits of this method are found,such as the uncomplicated solving progress and explicit ph ysical meaning.It's also shown that the estimation difference of vo lterra series is connected with amplitude of the pulse excitation.Thus,the relationship between the amplitude and kernels of series has been researched.5)The stud y on the truncated form of volterra series finds that different truncated forms of volterra s eries affect the solution of the function of series' kernels obviously,which can lead to estimated results appearing non-uniqueness.6)The viewpoints of this paper have been verified b y experiments.Datum of the three different t ypes of rolling bear are collected in the first place,and then datum are solved b y multiple-pulse excitation method in the second place.According to identification of the bearing failure under different loading conditions and depths,the feasibilit y and strong points of multiple-pulse excitation can be obtained.