音波法输气管道泄漏检测系统实施与应用研究

发布时间：2019-03-16 21:36

【摘要】：近年来,国内油气管道行业发展势头迅猛,储运管道总长已达十多万公里,但随之也带来了更加频繁的管道安全事故。为了减轻和预防此类事故对人民生命财产安全造成的伤害,必须制定一套合理的油气管线泄漏检测与定位系统。由于目前已应用的管道泄漏检测系统多采用单工况(泄漏工况、正常运行工况)辨识算法,因此对运行平稳的管道应用效果较好,而对运营状态变化较大的管道适应性却较差。针对该技术难点本文以管道音波信号波动为检测输入量,进行了管道泄漏检测与定位系统的开发。首先,为了解决管道运行工况辨识难的问题,本文引入善于模态识别的人工神经网络作为管道多工况判定识别算法。为了达到最好的多工况识别效果,本文还以工况区分度高、实时计算简易和神经网络分辨性强为准则,对种类繁多的音波信号特征量(时域特征量、频域特征量和时频域联合分析特征量)进行了优选,同时结合优选结果对多种类型的网络应用效果进行了对比和分析,从而得出BP神经网络在泄漏判定准确率和抗干扰能力上相比其它几种神经网络具有更高适用性。其次,BP网络虽然使用广泛,具有诸多优越性能,但也存在一些如:训练收敛不稳定、易陷局部最优、样本依赖性强等缺陷。因此本文针对BP网络的泛化能力、训练收敛稳定性以及模态识别精度三个方面提出了优化算法。其中分别采用了贝叶斯归一化训练方法用来提高网络泛化性能;改进的自适应遗传算法用来提高网络收敛稳定性;模糊神经网络算法用来提高模态识别精度。通过多次试验验证可以得出,优化后的BP网络具有更好的网络性能以及工况辨识效果。最后,为了编制出计算高效且适用性更强的长输管线音波泄漏检测与定位系统,本文选择编程简便、外部接口较多且计算机系统适用性较强的Visual Basic6.0来编写系统主体,同时引入MATLAB,利用其工具箱函数高效编写工况判断核心算法,并将Access数据库嵌入VB从而实现了泄漏报警记录的实时保存和显示。总体来说,本文介绍了一种以管道音波信号为基础,优化后的泄漏多工况辨识BP神经网络为核心算法,采用NI-DAQmx、MATLAB、Visual Basic以及Access数据库混合编制出的管道泄漏检测与定位系统,并通过系统有效性实验验证了该软件系统在实验室内的良好应用效果。
[Abstract]:In recent years, the domestic oil and gas pipeline industry has a rapid development momentum, the total length of storage and transportation pipelines has reached more than 100, 000 kilometers, but also brought more frequent pipeline safety accidents. In order to reduce and prevent the damage caused by such accidents to the safety of people's life and property, a reasonable leak detection and location system for oil and gas pipelines must be established. Because most of the pipeline leak detection systems used at present use the identification algorithm of single working condition (leakage condition, normal operation condition), so the application effect of pipeline with stable operation is better, but the adaptability of pipeline with great change of operation state is poor. In view of the technical difficulties, the pipeline leak detection and location system is developed in this paper, in which the acoustic wave signal fluctuation is used as the input to detect the pipeline leakage. Firstly, in order to solve the problem of difficult identification of pipeline operating conditions, this paper introduces the artificial neural network (Ann), which is good at modal identification, as the identification algorithm of pipeline multi-working conditions. In order to achieve the best identification effect of multi-working conditions, this paper is based on the criteria of high differentiation of working conditions, simple real-time calculation and strong resolution of neural network, for a wide variety of acoustic signal characteristics (time-domain characteristics, etc. The frequency domain characteristic quantity and the time-frequency domain joint analysis feature quantity are optimized. At the same time, many kinds of network application effects are compared and analyzed by combining the optimization results. It is concluded that BP neural network has higher applicability than other neural networks in the accuracy of leak detection and anti-jamming ability. Secondly, although BP network is widely used and has many superior performance, it also has some defects such as unstable training convergence, easy to fall into local optimization, strong sample dependence and so on. Therefore, this paper proposes an optimization algorithm for the generalization ability of BP network, the stability of training convergence and the accuracy of modal identification. The Bayesian normalization training method is used to improve the generalization performance of the network, the improved adaptive genetic algorithm is used to improve the convergence stability of the network, and the fuzzy neural network algorithm is used to improve the accuracy of modal identification. Through many experiments, it can be concluded that the optimized BP network has better network performance and better working condition identification effect. Finally, in order to develop a sound leakage detection and location system for long-distance pipeline with high efficiency and applicability, this paper chooses Visual Basic6.0, which has simple programming, more external interfaces and stronger applicability of computer system, to program the main body of the system. At the same time, MATLAB, is introduced to use its toolbox function to write the core algorithm of working condition judgment efficiently, and the Access database is embedded into VB to realize the real-time storage and display of leak alarm records. In general, this paper introduces an optimized leakage multi-condition identification BP neural network algorithm based on the pipeline acoustic signal, and uses NI-DAQmx,MATLAB, as the core algorithm. A pipeline leak detection and location system based on Visual Basic and Access database is developed. The effectiveness of the system is verified by the experimental results in the laboratory.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP183;TE973.6

【参考文献】