Aero-engine Gas Path Monitoring Technology Based On Electrostatic Induction

Electrostatic monitoring technology is one of the tools with prediction ability that can be used to effectively resolve the problem of real time and on-line monitoring of the heat components of aero-engine. It can improve PHM capabilities of aero-engine, and promote the introduction of advanced maintenance strategy and the implementation of advanced maintenance methods, and increase the proportion of the conditioned based maintenance in the maintenance support, improving the safety while decreasing the maintenance cost. Electrostatic monitoring technology is to monitor the electrostatic charge in exhaust gas of aero-engine via electrostatic sensor, and make performance and state prediction of gas path competent by means of signal processing algorithms and intelligent decision model. Firstly, the formation mechanism of charged particles, the composition and influencing factors of charge level are analyzed, and then the key technologies for electrostatic monitoring are researched in the thesis as follows:(1) Sensor technologyAn electrostatic sensor is designed to meet the requirement of gas path monitoring under high temperature. The sensitivity distribution of electrostatic sensor probe is hard to obtain by Analytical method and experimental method. Aiming at this problem, based on point charge principle the mathematic model of electrostatic field generated by a club-shaped probe is established. Numerical solution method based on finite element is proposed to obtain sensitivity distribution of the sensor probe. The distribution function of sensitivity is obtained by data fitting method, then the effect of structure parameters and material characteristics on the sensing fielding are explored detailedly; Aiming at problems of weak electrostatic induction signal detection, the noise sources of detection circuit are discussed and the relevant solutions are proposed, a preamplifier embedded in sensor is designed. The temporal frequency characteristic of the sensor output signal and influence factors are investigated theoretically and experimentally.(2) Noise reduction method for electrostatic monitoring signalThe de-noise effect is not good when wavelet threshold filtering is applied to process the electrostatic inducing signal, Based on comprehensive consideration of energy distribution and noise composition, a de-noise method combining with the median filter method and Birge-Massart threshold wavelet method is researched in this thesis and the better filter effect is obtained by experiment. Moreover, A de-noise method for electrostatic monitoring signal based on Independent Component Analysis is researched. A construction method of reference noise signal based on Empirical Mode Decomposition is proposed to construct reference noise signal similar with overall noise in original signal and solve the underdetermined problem while de-noising by Independent Component Analysis method. The experimental result of simulated signal and measured singal shows the two methods can increase the signal-to-noise ratio effectively.(3) Character extraction and abnormal particles distinguishing methodAccording to the difference of frequency spectrum of induced by single particle and multi-particles, a pretreatment method based on frequency spectrum characteristics is applied to resample signal, then extract energy distribution characteristic in relative frequency range via wavelet decomposition method.A knowledge acquirement model based on rough sets theory and neural network is proposed in this thesis. Based on the importance of attribute and the consistency of decision table, SOM neural network is employed to discretize continuous data.In order to improve neural network's generalization ability. A structure auto-adaptive Neural Network model based on genetic algorithm is proposed to optimize structure parameter of Neural Network, then Neural Network matching the training requirement is employed to generate more samples containing foreknowledge, which are used to extract rules. The model provides an effective method for distinguishing abnormal particle.Lastly, an experimental platform is set up to simulate aero-engine gas path environment, and experimentize study on the electrostatic monitoring technology, electrostatic sensor, and rule extraction model via the experimental platform. The relationship between change law of characteristic parameters and varied conditions is obtained by typical simulated work condition experiment and provides a reference for distinguishing the abnormal particles.