Abstract:  It was difficult to accurately detect the weak signal of rotating machinery system in strong noise environment, therefore, a method of self-adaptive multistable stochastic resonance (MSR) and adaptive particle swarm optimization (APSO) was proposed. Firstly, self-adaptive MSR was chosen as the basic detection method, and twice sampling (TS) method was introduced in when numerical solving the output signal which solved the problem of stochastic resonance's poor adaptability to high frequency signal. Then, the output signal to noise ratio (SNR) was selected as the fitness function and particle swarm optimization (PSO) was used to optimize structural parameters of multistable system. The inertial weight coefficient was self-adaptively adjusted according to the global optimum position and the modification changed particle swarm optimization into adaptive particle swarm optimization. Finally, the simulation test of sinusoidal small signal with high intensity Gaussian noise was set up, on this basis the weak malfunction diagnosis experiment of mechanical system was completed by using this method. The result of the research show that this method can accurately highlight the malfunction frequency component at 161.1Hz, at the same time, the frequency doubling component which describes the actual system running state can be obtained. This method has the accurate weak signal detection ability.

Key words:  rotating machinery system;fault diagnosis;weak signal detection; multistable stochastic resonance (MSR) ; particle swarm optimization (PSO); twice sampling (TS)