Methods And Applications Of Improved Two-dimensional Complementary Stochastic Resonance In Weak Signal Detection

The fault diagnosis of the mechanical equipment is of great significance for the security of the people's life,reducing the unnecessary economic loss,avoiding the stagnation of the progress of the social production and the natural environment.A series of means that are used to determine whether there is the fault or there is a malfunction of the trend,the measures are very necessary in the industrial field.We should collect the mechanical sound or current signals to detect the fault,and do the signal processing analysis.But,there will be a lot of running noises and environmental noises from other mechanical parts in the signals which are collected from the sensors.The noise component can make the signal processing become very difficult.Weak signal detection is of great significant for suppressing the noise and enhancing the weak signal and improving the effect of fault diagnosis.In the view of the signal processing,traditional one-dimensional stochastic resonance(1DSR)can be regarded as a specific nonlinear filter,and it is able to enhance the weak periodic signal in the nonlinear system by a certain amount of noise.The traditional linear filter does not have this specific filtering de-noising mechanism.So,1DSR method is beneficial for the signal extraction,especially when the signal bandwidth is overlapped with the noise bandwidth.However,the nature of 1DSR is similar to a low-pass filter,and the de-noising effect can be further improved.This study investigates a kind of noise-enhanced weak signal detection method,termed as two-dimensional stochastic resonance,with its applications in fault diagnosis.The paper summarizes and analyzes the discrete stochastic resonance as a weak signal detection algorithm;on this basis,a novel ensemble average two-dimensional stochastic resonance(E2DSR)algorithm is proposed to do weak signal detection.E2DSR method contains two-dimensional stochastic resonance and ensemble average method.The signal-noise ratio(SNR)of E2DSR output is higher than 1DSR,and the fault characteristic frequency(FCF)is outstanding in the power spectrum.Both the simulations and the actual bearing fault signals confirm the advantages of the proposed E2DSR algorithm,it has a better performance of de-noising,and remarkable ability of eliminating noise.To further learn about the two-dimensional stochastic resonance method,this paper puts forward a new two-dimensional complementary stochastic resonance(2DCSR)method.To start with,the bearing fault signal which is collected from the sensors is bandpass-filtered and then demodulated.The demodulated signal is split into two sub-signals and the sub-signals are sent to the two input channels of the 2DCSR model.The weighted power spectral kurtosis(WPSK)of the output is used as the criterion to adaptively guide parameters tuning in the system.Then one output of 2DCSR is used to enhance the other output signal,this can be used to identify the bearing fault types.Simulations,bearing data experiments and the value of WPSK confirm that it has a better performance of 2DCSR method.In summary,this paper investigates the noise-enhanced weak signal detection techniques and their applications in the bearing fault diagnosis based on two-dimensional stochastic resonance method.The two new methods have distinct merits including good performance,high efficiency,easy to implement,etc.,as compared with the traditional way.Meanwhile,the actual fault signal experiments also verify the superiority and practicability of the improved two-dimensional complementary stochastic resonance method.