Research On The Blockage Recognition Of Buried Drainage Pipes Based On Acoustic Detection Methods

With the continuous construction of China’s infrastructure,the process of urbanization is taking the fast train of China’s high-speed rail into a stage of rapid development.However,in this rising period,there is still a phenomenon of uneven development.The water supply and drainage pipeline serves as an infrastructure,just like the blood vessels of the city,ensuring the stable operation of the city.In recent years,with the increase of service years,the problem of pipeline blockage has gradually emerged.Often because of a heavy rain,the whole city has entered the‘seeing the sea’mode,affecting people’s normal travel and production and life in an orderly manner,while the pipeline is blocked.It will also slowly accumulate as a safety hazard,causing damage to people’s lives and property.Therefore,timely detection of pipeline blockage and control of the hazard hazard caused by the blockage of pipelines is of great significance for saving water resources,ensuring urban water use,and promoting the sustained and healthy development of China’s economy^[1].At present,domestic research on pipelines mostly focuses on pipeline defects,that is,pipelines produce cracks and dents,and research on pipeline clogging is less carried out.Pipeline blockage is an operating condition that is very likely to cause accidents during pipeline operation.Because of its existence and development time,there is a certain hysteresis and incompatibility to the detection of blockage.Therefore,it is mainly used as a passive method for leak detection.In this paper,the acoustic detection method is used to study the pipeline blockage condition,and the pipeline blockage condition is judged by analyzing the echo signal in the pipeline.Compared with the water supply pipeline,the drainage pipeline has less full-filled state,and the internal sewage of the drainage pipeline is more complicated,often with foreign matter doping,and the water flow density cannot be determined.Therefore,the signal analysis of the drainage pipeline is also complicated.In this paper,the echo signal is obtained by the acoustic detection method,and the feature set is used to select the feature set.Finally,the semi-supervised learning method is combined with the actual engineering data to judge the pipeline operating conditions.The main research work of the thesis is as follows:（1）Using the data obtained by acoustic active detection,from the acoustic signal mechanism,other acoustic parameters such as particle velocity and other acoustic parameters are indirectly obtained through the measurement of sound pressure,and the acoustic signal characteristics are analyzed,and the acoustic signals in the pipeline are encountered in the blockage and The acoustic changes produced by the pipe components of the tee are extracted from the acoustic characteristics of the pipe.（2）Using feature selection method to select feature set,effectively select feature subset,solve the problem of optimal feature selection in pattern recognition,avoid feature redundancy as much as possible,reduce feature set dimension,and save computing resources.Improve work efficiency in real-world applications.The acoustic signal is analyzed by signal processing method,and the features are extracted based on the components of different frequency bands decomposed to form a high-dimensional feature set.The feature set is sorted by the generalized Fisher method,and the mutual information between each feature is calculated.Then,the features are added to the feature subsets in turn until the classification correctness rate shows a downward trend,and the optimal feature subset is obtained.（3）In the actual engineering application,the amount of data collected is getting larger and larger.With the manpower and experience,all the data is marked and the workload is gradually impossible.The semi-supervised method can greatly alleviate this phenomenon,which guarantees The correct guidance of the classification process for a small number of labeled samples reduces the need for a large number of labeled samples.S4VM can realize two classifications.This paper classifies various working conditions of pipelines.It needs to improve S4VM algorithm to realize multi-classification identification.The basic method of expanding method is dismantling.This study adopts one-versus-rest method.The S4VM method is extended to achieve the needs of this article.