Research On Adaptive Control Of Fractional Stochastic Resonance And Application Of Bearing Fault Detection

As for the problem that fractional-order stochastic resonance involves multi-parameter joint tuning,which is not friendly to manual debugging,this paper proposes to use particle swarm optimization to adaptively optimize multiple parameters of a fractional-order bistable system,making the weak target signal masked by noise effectively enhanced in system’s output.Because bearing vibration signals are complex signals,the traditional single-signal evaluation parameters have certain defects in the application of such signals.This subject analyzes and investigates the types of bearing failures and multiple signal evaluation indicators.BP neural network is used for training,and multiple evaluation indexes are integrated into a custom comprehensive index(CCI).Finally,the CCI is used as the evaluation parameter for the optimization of the particle swarm optimization algorithm,and the fractional stochastic resonance system is used to adaptively process the bearing vibration signal,which can effectively reduce the noise energy while enhancing the target signal.In the course of the research,the following tasks have been done:(1)Improve the ORA algorithm to make it approach the integer-order approximation of the fractional-order bistable system that handles the bearing fault signal,and perform modular encapsulation to facilitate subsequent calls;(2)Explore the influence mechanism of each parameter involved in the generation of stochastic resonance in the fractional-order bistable system based on the principle of single variable,and select and reduce the number of parameters that need to be optimized according to the summarized law;(3)Optimize the particle swarm algorithm based on the Levy flight strategy and set up the corresponding stochastic resonance program framework under the fractional-order bistable system to achieve adaptive parameter adjustment of the system’s input signal to generate fractional-order stochastic resonance;(4)Use BP neural network to train multiple signal evaluation indicators to obtain custom comprehensive index for the output signals of fractional-order stochastic resonance.