A Study Of The Method Of Slowly Varying Small Faults Detection

Since 1970 s,electronics and information technology have been widely used in industrial production.Advanced automated machines have gradually replaced manual operations,which enormously elevates the production efficiency.However,automated production lines produce inevitable faults when it comes to continuous long-term production operations.Lots of slowly varying small faults are difficult to be detected at the beginning of the production operation period and eventually cause major damage after these faults worsen.Supported by the Sichuan Science and Technology Plan "Intelligent Maintenance and Remote Monitoring System for Lifecycle Management of Automated Flexible Production Line"(No.2017GZ0060),this thesis studies the early detection of slowly varying small faults in the production line equipment.Combining with the time-frequency characteristics of slowly varying small faults,a new fault detection method is proposed based on traditional fault detection methods.The main research contents are as follows:First,empirical mode decomposition(EMD)is combined with wavelet threshold denoising(WTD).Traditional EMD can effectively extract the characteristic components from input signals,but the background noise in the signal of the actual production line will cause the modal aliasing.The false mode components will increase with the increase of decomposition levels-which will affect the signal processing effect.In order to reduce false modal components and modal aliasing,the input signal is filtered by WTD,and further decomposed by the EMD.In order to overcome the shortcomings of traditional threshold function in WTD,an improved wavelet threshold function is designed to improve the denoising effect,which effectively reduces the false modal component and modal aliasing during empirical mode decomposition.Secondly,in order to enhance the output response of slowly varying small faults,stochastic resonance(SR)is selected to process the signal.The system parameters need to be optimized to achieve the optimal stochastic resonance state.In this thesis,fruit fly optimization algorithm(FOA)is used to optimize the parameters of SR system globally and adaptively.Because of defects in the fixed step size of traditional FOA,a nonlinear decreasing double-step optimization method is proposed to improve its global search ability.The best SR state of the system can be achieved with more accurate calculated system parameters.Finally,after analyzing the time-frequency characteristics of slowly varying small faults,a fault detection method is proposed by combining the signal processing methods mentioned above.In this method,the original signal is first denoised by WTD,EMD is then applied on the filtered output result to get a group of intrinsic mode functions(IMF).Next,the signal is reconstructed and processed by the SR to make the slowly varying small fault signal stand out and finally achieve the purpose of detecting the actual fault of the actual signal in the production line.The slowly varying small fault detection method proposed in this paper has certain practical value for the fault detection of the body-inwhite welding production line.