Research On Recognition Of Steel Structure Vibration Patterns Of Railway Canopy Under The Influence Of Train Arriving And Leaving

Structure early warning is the core function of the steel structure health monitoring system of railway platform canopy.Early warning effect is an important index to determine the quality of monitoring system.Through the actual vibration monitoring,it is found that the early warning threshold is frequently triggered due to the influence of the train arriving and leaving the station.The number of false alarms in the system is too large,which leads to the unsatisfactory warning effect.To solve this problem,this paper proposes a multi threshold early warning method based on pattern recognition.With wavelet analysis and support vector machine as the core,the vibration pattern recognition of rain shed steel structure affected by train in and out of the station is studied.Combined with multi threshold early warning,the optimization of the early warning effect is verified by an example.The specific work is as follows:(1)This paper studies the vibration monitoring and early warning strategy of canopy steel structure.The existing problems of early warning strategy are analyzed with the actual data.Furthermore,a multi threshold early warning method based on pattern recognition is proposed.According to the different vibration characteristics of the steel structure of the canopy caused by the train arriving and leaving the station,the vibration modes of the steel structure are divided into four categories: arriving,leaving,passing through the station at a constant speed and without the train.By setting different warning thresholds for vibration data in different modes,the number of false alarms can be reduced.(2)In order to realize the vibration pattern recognition,it is necessary to extract the characteristics of vibration signals.Based on wavelet analysis,the feature extraction method for the vibration signal of canopy steel structure is studied.It mainly includes wavelet threshold noise-reduced and wavelet packet of energy feature extraction.According to the characteristics of the vibration signal of the canopy steel structure,the db3 wavelet basis function is selected and the three-layer wavelet decomposition is carried out.The hard threshold function is used to noise-reduced the signal.Then,db3 wavelet packet is also used to decompose the noise-reduced signal in four layers.Extract the decomposition coefficient,calculate the node energy,normalize the energy data and sequence the gray code as the feature vector of the signal.It lays a foundation for pattern recognition.(3)Based on the support vector machine(SVM),the vibration pattern of recognition algorithm of rain shed steel structure is studied.The v-SVM multi classifier is used to build the model,and the method of model training and testing is established.In the aspect of model parameter optimization,aiming at the shortcomings of PSO,an improved PSO is proposed.The simulation results show the superiority of the improved PSO in optimizing the parameters of SVM.Combined with the method of signal feature extraction,the realization scheme of pattern recognition is established.It provides classification basis for multi threshold early warning method.Based on FBG acceleration sensor,the vibration data of canopy steel structure is obtained.According to the signal feature extraction and pattern recognition methods studied in this paper,an example of multi threshold early warning method is analyzed.The optimization of early warning effect is verified.