Detection Of Non-Contact Mechanical Seal Film Thickness Based On Acoustic Emission

Non-contact mechanical seal is widely used in fluid machinery equipment, and is also one of the key equipment of nuclear power plant. The performance and the operating condition of mechanical seal often affect the whole mechanical equipment directly. In recent years, the increasingly serious environmental problem has put forward a huge challenge to the research of non-contact mechanical seal. The ideal operating condition of the non-contact mechanical seal is to achieve the optimization of the thickness of the opening seal face. At this time, the seal face can not only get good lubrication, but also can achieve zero leakage. The research object of this paper is the liquid film lubrication non-contact mechanical seal. The research purpose is to realize the condition monitoring of the seal face film thickness effectively. The main research contents of this paper are as follows:When the contact state of non-contact mechanical seal face changes, it will produce acoustic emission wave, thus the acoustic emission signals of the seal face can reflect the contact state of seal face well. Acoustic emission technology can realize the dynamic nondestructive testing of mechanical seal, and it has the characteristics of high sensitivity and wide applicable scope. In this paper, the acoustic emission technology was used to build the test bench and collect the acoustic emission signal. The acoustic emission signal analysis and processing method was formulated based on the characteristics of the acoustic emission signal.The collected acoustic emission signal contains a lot of background noise, so it is difficult to reflect the true state of mechanical seal face. To solve this problem, this paper proposed a filtering technology of acoustic emission signal based on Gaussian sum particle filter (GSPF). GSPF improved the problems of particle filter theoretically, and the simulation experiments illustrated GSPF had the advantages of low process noise and high estimation accuracy. Applying GSPF to filter acoustic emission signal, the result showed that after filtered by GSPF, the acoustic emission signal can reflect the characteristics of real signal effectively.This paper extracted features of the filtered acoustic emission signal from three aspects of time domain, frequency domain and time-frequency domain, including:variance, root mean square, peak, skewness, waveform factor, center frequency, frequency variance and mean square frequency, effective value, standard deviation, sub band energy ratio. The data showed that multiple groups of features of acoustic emission signals had obvious differences under different states, and had good consistency in the same states. On the one hand, it showed that the effect of the used filtering technology in this paper is prominent. On the other hand, the extracted features could improve the recognition efficiency of the state of seal face film thickness.For the BP network existed some problems, the Adaptive Genetic Algorithm (AGA) was used in this paper to optimize the BP network. AGA adopts adaptive crossover rate and mutation rate to improve the convergence of the algorithm, and prevents the phenomenon of "premature convergence" by pruning similar individuals and adding new individuals. AGA has the better optimization ability and approximation ability. So this paper used AGA to optimize the initial weights, threshold and network structure of BP network at the same time. Applying AGA-BP network in recognition of film thickness obtained high recognition rate. The contrast experiment showed that the performance of AGA-BP network is better than that of basic BP network in the aspects of training rate and recognition rate.The research of this paper mainly realized the pattern recognition of film thickness of the non-contact mechanical seal, and it provided the theoretical basis and technical support for further in-depth analysis of non-contact mechanical seal.