Research On Sensor Fault Diagnosis Method And Its Application In Mechanical Vibration Monitoring

In the condition monitoring of mechanical equipment,the judgment of the operating status of equipment is usually based on the results of the frequency spectrum analysis,and the acquisition of the spectrogram depends on the time domain data of the vibration signal.Sensors are used as the front end of vibration monitoring signal acquisition,and their working status directly affects the accuracy and reliability of time domain data.For this reason,this paper studies the sensor fault diagnosis algorithm to solve the problem of low accuracy and long time-consuming of the existing methods.Specific work includes:First,aiming at the shortcomings of existing single sensor fault signal feature extraction methods that do not fully consider the local characteristics of the fault signal,the local ternary pattern in image processing is introduced into the processing of one-dimensional signals.Combined with sliding filtering,a feature extraction method based on multi-scale one-dimensional local ternary pattern is proposed.Using Gaussian white noise to carry out fault simulation experiment,it is found that proposed method is more accurate compared with the method based on wavelet packet transform and time domain characteristics.Secondly,in view of the problem that the Fast ICA-based sensor fault detection algorithm is easy to fall into local optimum,an ICA sensor fault detection method based on the improved grey wolf optimization is proposed.This paper proposes a nonlinear convergence factor and introduces the chaotic map into the grey wolf population initialization to improve the global search ability.Numerical analysis experiments show that the improved grey wolf optimization has better optimization ability than the basic grey wolf optimization and particle swarm optimization.Then,a system based on Lora is designed for mechanical vibration signal collection.Compared with traditional wireless transmission methods,Lora can better balance cost,safety,transmission distance,power consumption and networking convenience.The system consists of two parts: the upper computer and the lower computer.The hardware design of the lower computer includes accelerometer,analog-to-digital conversion module,power management module,core processor,and wireless transmission module.The signal collected by the lower computer is sent to the Lora gateway through the Lora module,and then transmitted from the Lora gateway to the upper computer for data analysis.Finally,install a vibration acquisition module on the vibration test bench to obtain acceleration data.By adjusting the tightness of the bolt connection between the wireless vibration acquisition module and the test bench,the sensor fault data under the loose bolt is collected.Combining the mathematical models of different sensor fault types,various types of sensor fault data are obtained,and the application verification of the aforementioned single sensor and multi-dimensional sensor fault diagnosis methods is carried out.The results show that the fault recognition accuracy of the proposed method is better than the existing methods.