Research On Train Safety Testing Technique Based On Audio Analysis

Railway transportation plays the leading role in the transportation industry.Once there is any fault with a train,it may seriously affect its normal operation,and if the fault cannot be dealt in time,it will cause an incalculable loss.Therefore,it is of great significance to test train faults.Aiming at the abundant sound source information during train traveling,this paper presents a train fault testing method based on audio analysis.As the environment is quite complicated during its travelling period,it is difficult to get complete fault samples in the process of actual diagnosis train fault diagnosis.Therefore,the fault recognition method based on supervised learning is difficult to be applied in the real situation.At present,the common methods for train fault diagnosis include image diagnosis,acceleration diagnosis and displacement diagnosis,among which there are shortcomings.The audio signal has some advantages such as convenient acquisition and non-contact measurement.Employing the unsupervised K-means clustering algorithm,the above method applied to train fault diagnosis better.A preliminary study has been presented,which is regarding the technique of train fault diagnosis based on audio analysis.The main works are as follows:1.The propagation properties of sound waves are introduced and analyzed,which provides the basis for the audio analysis of the train.2.Unsupervised learning and supervised learning algorithms have been compared,and eventually the unsupervised learning algorithm is used according to actual requirements.3.The wavelet packet algorithm is used to analyze the train audio signals and at the same time the eigenvector has been achieved accordingly.4.An unsupervised K-means clustering algorithm is proposed,and the feature vectors are clustered.5.The train fault audio signal diagnosis system has been designed,which is proven to be effective to identify the normal samples and abnormal samples in the field experiments.6.The algorithm has been verified to be effective further through the real train operation data.The unsupervised testing technique based on audio analysis can identify the train fault online,and ensure the safety and stable operation of trains,which is proven to be accurate,reliable and of great values for practice.