Research Of Source Noise Identification Based On The Modern Spectrum Estimation

The purpose of noise control is that according to the requirement and possibility, we control noises in an acceptable scope with the most economic method. For controlling noise effectively, we always control noise at source place. So it is important to find out the source noise accurately. In fact, we often meet multi-source noise in our lives. So we need to know specific weight of each source and then control them. We can call it"source noise diagnosis". In the noise control, it is an extremely important step to diagnose the source noise, and then we can put our effort into controlling the main one. For another, source noise diagnosis is very complicated; many noises exist in the same region or the same machine, and noises interfere each other. The purpose of source noise diagnosis is to find out the place of the main source noise,power distributing and frequency characters etc. with appropriate methods. After discriminating the main source noise, we still have to calculate the specific weight of every source noise。It is also important because we can choose specific methods to control noises.The method of noise identification can divide into two categories: the first method is a normal measurement and analytical method such as expunction, overlay, near field measurement, surface vibration technique and so on. The other method is sound signal processing method such as correlation function, frequency analysis, and surface strength method. This method developed according to modern signal processing. In different researching stage we choose different method to identify noise according to the characters of signals. Because of high request towards the measurement environment and many disadvantages exist in all methods described above, the advance of modern spectrum estimation can be shown in the text.Based on the introduction of some noise identification methods, we make use of modern spectrum estimation to identify noises. Then we use curve similarity degree and gray system relative degree to calculate the specific weight of each noise in the total noise. We mainly identify two source noises. First, we constructed a suitable model for every signal . Then we calculated the parameters and power spectrum of every signal. Last, we used curve similarity degree and curve relative degree to calculate the weight of every signal in the total signal by MATLAB. We found that the method of curve relative degree was better than curve similarity degree.