Characteristic Analysis And Pipeline Leakage Detection Based On Fiber Optic Sensor Data

With the rapid development of national economy,refined oil has become an important and indispensable energy.so,in order to transport refined oil efficiently and safely,pipeline transportation has become the preferred transport mode because of its unique advantages.With the increase in automation,complexity and integration of refined oil pipeline networks,the safety of pipeline network is an urgent subject for people in present.As a new means,optical fiber detection has been developed rapidly in recent decades due to its strong anti-interference ability,small attenuation of signals during transmission,and sensitivity to external changes,it has been applied in many fields such as industry and military.Therefore,research on pipeline detection methods based on optical fiber sensing is of great significance.After lots of theoretical knowledge learning and literature reading reference,the work done in this thesis is as follows:First of all,the structure and transmission basis of fiber Bragg grating are analyzed.Based on the principle of fiber grating transmission,the sensing characteristics of fiber grating are studied and deduced.The fiber optic sensing principle is used as the theoretical support to complete construction of pipeline optical fiber detection system.The system includes a light source,a fiber grating,a spectrum analyzer,a fiber demodulator,etc,for completing fiber-based pipeline detection.Secondly,for the collected optical fiber sensing signals,which contains a large number of noise signals,the wavelet denoising method is used to denoise the signals.Based on the analysis of the factors affecting the denoising performance of the wavelet,respectively,from the decomposition scale,wavelet basis selection,threshold quantization and threshold function are explored and improved.The deficiencies of traditional threshold processing methods are analyzed and compared.The effectiveness and feasibility of the improved method are verified by experimental simulation.Thirdly,in the process of collecting the signal detection of the fiber grating,due to the nonlinear characteristics of the demodulation equipment and the disturbance of the environmental changes,the peak value fluctuation caused by the trapezoidal integral peak finding method is proposed.Perform peaking processing.The multi-scale entropy method is used to extract the feature of the detected signal.Then,the reliefF algorithm is used to select the different scale entropy values,and good results were obtained.Finally,in view of the fact that there are few pipeline anomaly data samples in actual experiments,a support vector machine method suitable for small sample identification is proposed to analyze the sensor data of pipeline fiber detection,and then determine the running state of the pipeline.Aiming at the selection of kernel function parameters in support vector machine algorithm,the particle swarm optimization algorithm is used to optimize the parameters of support vector machine,and the best result is obtained.Through experimental simulation,the optimized support vector machine classifier has a higher accuracy of classification.