Blind Source Separation Of Vibration Signals Of Ac Motor Speed Control System

Blind source separation (BSS) is a multi-dimensional signal processing method, is the process of recovering each individual component of source signals which is under the situation of unknowing source signals and mixed models. Blind source separation has become a hot issue on modern signal processing field and very important theoretical significance and wide application in voice processing, image processing, seismic exploration, biomedical, radar and other fields of economic data analysis.In this paper, the AC motor variable speed system vibration signal as the research object, introduces the fast fixed-point independent component analysis (FastICA) method and information maximization (Infomax) method.First, this paper introduces the source of the problem about the blind source separation, research status and main applications, describes the basic principles, mathematical models and classification methods of blind source separation, discussed in detail the basic characteristics of blind source separation and performance evaluation criteria of blind source separation method.Second, the paper focused on the FastICA algorithm. FastICA algorithm is based on the principle of non-Gaussian to maximize, use of fixed-point (Fixed-point) iteration theory to find the maximum value of non-Gaussian, and then using Newton's iterative algorithm to batch processing for a large number samples of observed variables, each time separated a independent component from observation signals. Therefore, FastICA is a fast algorithm in the independent component analysis algorithm. Then study the information maximization (Infomax) algorithm, which use a great principle of information transmission, namely Infomax (information maximization) principle, propose blind source separation algorithm. Concrete can be described as:While the mutual information of the network input and output to maximum, the correlation of the output between the various components become to minimum. Infomax algorithm is a new self-organizing learning algorithm based on the information maximization criterion. Infomax algorithm is a product of combination of information theory and neural networks.Finally, I conducted a specific application study of FastICA algorithm and Infomax algorithm. This is first time applied them to simulation experiment of vibration signals of AC motor speed control system, the experimental results show that both methods can be achieved extraction of the main source of vibration signals in mixed-signal. Then, I detailed description of the motor vibration causes and analyzed the data of experiment.