Study On Signal Analysis Methods In The State Analysis Of Non-Stationary Mechanical And Electrical System

In the fault diagnosis and trend prediction of non-stationary mechanical and electrical system, state analysis is a necessary process. But in various state analysis methods, signal analysis method, which determines the accuracy and validity of state analysis, is a restrict factor. This paper based on the study of signal analysis method of non-stationary mechanical and electrical system. First of all, it was explained that the essentiality of vibration signal analysis in the state analysis. Existing signal analysis methods, suitable for the analysis of non-stationary vibration signal, and their limitation in principle were expounded. Hilbert-Huang Transform (HHT), based on Empirical Mode Decomposition (EMD), was introduced. It's superiority was analyzed. And it was carried out through program. During the process of programming, some detail realizations of theories were researched audaciously. Accordingly, the algorithm has better self-adapting than before. Secondly, the concept of end effects in EMD was introduced, and methods suitable for restrain it were explored in various approaches. The result indicated that, if the speed and space are considered, mirror extension is the best, polynomial fitting extension is the second, parallel line extension is the next, the method of series extension using neural network have perfect effect, but also have disadvantage as, lower speed and possess more memory space. It was discussed that how to select proper method to restrain the end effects when the signal was different. Then, through the analytic of a simulated typical signal, the validity of EMD was confirmed. The conclusion shows that HHT is better than other methods. HHT has a lot of advantages, such as better resolution capability, better self-adapting, better information integrity, clearer physical meaning, more compact manner, and easier to analyze accurately. Finally, A system named "signal analysis integrated system in state analysis of non-stationary mechanical and electrical system" was designed both in hardware and software. The system can obtain the online and offline signal of monitored mechanical and electrical system, and also can analyze the signal by several methods. Fast Fourier Transform, Short Time Fourier Transform, Wavelet Transform, and Hilbert-Huang Transform, etc. are all included. The interface of new analysis methods is reserved, which provide precondition to implement opening system development.