Fault Diagnosis Based On Incomplete Gear Acoustic Signal

Rotating machinery is a very important mechanical equipment in daily life.It covers a wide variety of fields has lots of species.As an important component in rotating machinery,gear plays a key role.However,due to the special function and position of the gears,many faults often occur.In order to avoid the hidden danger of the gear of production,fault diagnosis technology is needed to diagnose the working condition of the gear.But it is difficult to use vibration sensors to collect vibration signals directly under some working conditions.So,this paper study acoustic signal and proposes a fault diagnosis method and support vector machine model for the special case of incomplete acoustic signal.The research of the incomplete acoustic signal is focused on the fact that the acoustic signal of the faulty gear is a nonlinear and non-stationary signal,and in the case of incomplete situation,the blind source separation correlation algorithm cannot be used to separate the source signal directly.So,variational mode decomposition(VMD)is selected to decompose the collected signals,and the components obtained by VMD and the collected signals are used to form a complete signal model.When the signal set is a positive definite blind source separation model,conventional blind source separation algorithms,such as Fast independent component analysis(Fast ICA),can be used to separate the source signals.Finally,feature extraction is carried out for the components with less noise and more information separated by Fast ICA.The sample data of the rubbing fault belongs to a small sample,so a support vector machine(SVM)is used as a classification model for pattern recognition.For small sample data,the SVM classification model with the radial basis function as the kernel function has higher accuracy,and the selection of the penalty factor C and the parameters of the kernel function σ in the kernel function is optimized by the PSO optimization algorithm.Determine the optimal parameters to improve the accuracy of fault diagnosis and identification.Finally,the operation experiment of normal gear and broken gear is designed.The acoustic data of the gear operation under the two working conditions are collected and analyzed.The experiment data are analyzed by using the VMD-Fast ICA algorithm and SVM classification model,and the effectiveness of the method is verified.