Reseach On Weak Signal Detection Methods Of Planetary Gear Box Based On Stochastic Resonance And Chaos Theory

With the rapid development of industry, a lot of functions of the mechanical equipment are getting stronger, but the structures are become more complex. For fault diagnosis and state degradation assessment of mechanical equipment, many researches are widely appreciated by scholars, and scholars proposed many methods to detect whether the fault exists or not. Especially the gear box, which is a critical variable speed transmission part of mechanical equipment, diagnosing and evaluating it in time in order to find potential failure to repair can avoid a greater loss. However, due to the complicated working environment of gear box, the signal of fault will be affected by noise. Especially at the early stage of the fault, the signal of fault is too weak, which is submerged in strong background noise, so the detection is hard to conduct.In this paper, two methods of weak signal detection are studied, namely, stochastic resonance and chaos theory. Their principles of detecting weak signal are studied, then some methods for improving the detection effect are proposed, for example, combining cascade or correlation detection with stochastic resonance can improve stochastic resonance detection results effectively. For the disadvantage of stochastic resonance, the genetic algorithm is proposed to optimize system parameters, it can not only determine the parameter, but also can improve the output signal-noise ratio(SNR). For chaotic system, choosing a suitable chaotic model to build an effective chaotic system, and taking full use of the special mechanism of this system to detect whether the characteristic signal exists or not by observing the change of the phase trajectory from the phase trajectory diagram.This paper constructs a planet gear box test platform and implants fault into planetary gear and bearing artificially. Then collects the vibration data by acceleration sensor when gear box running, and sends these data into stochastic resonance system and chaotic system respectively to realize the fault diagnosis at the early stage and confirming the effectiveness of these improvement measures by comparing the spectrum diagram and values of output SNR before and after improvement. Moreover, genetic algorithm is used to optimize the two structural parameters for single-element and double-elements respectively. Not only confirms the effectiveness of this optimization algorithm for stochastic resonance, but also concludes that the output SNR of double-elements optimization is bigger than single-element optimization, so the genetic algorithm is better than the general optimization algorithm. Finally, inputting the normal and failure data of the bearing into chaotic system, and comparing the change of phase diagram to judge whether there is characteristic frequency in the input signal to confirm the application of chaos theory in fault diagnosis.