Research on classification method of abnormal vibration of pipeline based on SVM
Zhicheng Zhong,Qichen Gao,Fudong Zhang
DOI: https://doi.org/10.1117/12.2624213
2022-03-27
Abstract:In the process of pipeline transportation of liquid, due to the influence of corrosion, environmental effects and third-party damage on the pipeline, abnormal conditions such as leakage accidents often occur in the pipeline. In the actual pipeline operation, pipeline abnormality detection and location has become one of the effective means to ensure the safe operation of pipelines. Therefore, the abnormal vibration classification research of pipelines is of great importance whether in practical engineering applications or in the field of academic research. Due to the rapid development of sensor technology and signal processing and analysis technology, more and more data information can be monitored and stored during pipeline transportation. How to extract useful information that can be used to judge abnormal pipeline vibration from a large amount of monitoring data is an urgent problem that needs to be solved by pipeline abnormal vibration monitoring and classification methods. Distributed optical fiber disturbance sensing system has been widely used in many fields such as perimeter security, oil and gas pipelines, communication optical cable monitoring, etc., due to its high sensitivity, simple structure, no need for external power supply, and long detection distance. The -OTDR system has been gradually applied to oil and gas pipeline monitoring and other fields because of its excellent anti-electromagnetic interference, good transmission characteristics, extremely high resolution and sensitivity, and wide detection range. However, the actual data is often interfered by random noise, which seriously affects the monitoring and judgment of the pipeline state. Therefore, in order to suppress noise interference and improve the classification of pipeline vibration, a pipeline vibration classification method based on SVM algorithm is proposed. By extracting the characteristics of the signal's average value, variance, mean square error and signal power in the time domain and frequency domain, a classifier based on the SVM algorithm is established to judge the vibration and recognize the vibration pattern. According to the characteristics of the sensor signal, the classifier I is used to determine the presence or absence of a vibration signal, and the vibration signal is further identified by the classifier to identify the pattern of the vibration event. The experimental results show that the correct vibration discrimination rate is above 95%, the false alarm rate and the false alarm rate are below 5%; the correct pattern recognition rate is above 95%.