| The natural gas pipeline has played an increasingly important role inthe development of the national economy. Unlike crude oil pipeline, onceleakage of the natural gas pipeline happened, it will cause greater harm aswell as more serious economic losses. Among the existing pipeline leakdetection techniques, diagnose methods based on neural network modelare commonly used. In terms of model-based leakage detection technique,it has a high generalization and rich samples requirement. however, theseare difficult to be satisfied.From the aspect of piezoelectric sensor signal amplification, thispaper analyzed and compared the voltage and charge amplifiercharacteristics of the piezoelectric sensor. A piezoelectric signaltransmitter based on the charge classified amplification, a dedicated RTUfor pipeline leak monitoring based on the ARM processor, and pipelineleak monitoring system structure are also designed.From the following two aspects, signal processing speed andde-noising effects, this paper analyze and compared three signals de-noising methods, they are moving average filter, de-noising withwavelet decomposition and empirical mode decomposition. thede-noising method with wavelet decomposition and reconstruction foronline real-time monitoring signal is chosed. On this premise, leakagedetection based on voiceprint comparison is proposed. In addition, basedon the different signals collected at the scene, this paper described thefeature extraction process of leak signal in detail using the voiceprintmethod. It also analyzed and compared the saliency of the leakage signalas well as leakage diagnosed reliability. At last, an algorithm dynamiclibrary based on Qt is designed.The practical application results show that: the natural gas pipelineleak detection based on voiceprint makes leakage diagnosis more reliable.in addition, it eliminates false alarm and omission alarm of the system,providing a new and effective technique for natural gas pipeline leakmonitoring. |
PDF Full Text Request