【摘要】：管道運輸業(yè)的發(fā)展極大促進了化工工業(yè)和油氣資源行業(yè)的發(fā)展,管道運輸正逐漸成為油氣資源運輸?shù)氖走x方式。與此同時,油氣輸送管線鋪設逐年增加,服役期也在不斷增長,由于施工缺陷、沖刷腐蝕、人為破壞、設備老化以及地理和氣候環(huán)境變化等原因,會頻繁導致油氣運輸管道介質泄漏事故的發(fā)生,從而造成了大量的資源浪費和財物損失。管道泄漏診斷技術是一項綜合了多領域多學科知識的復雜性技術,特別是間歇性輸送管道,管道工況極其復雜,并且受多種因素制約,并且到目前為止,還沒有一種管道泄漏診斷方法是通用的,每一種方法都有它自身的適應性。因此,實現(xiàn)間歇性輸送管道泄漏的可靠診斷具有重要的理論和實際應用價值。本文在時域統(tǒng)計特征分析和信息熵理論研究的基礎上,提出了一種聲波信號的特征提取方法。通過對聲波信號統(tǒng)計特性的研究,提取了聲波信號時域統(tǒng)計特征,描述了管道發(fā)生泄漏時,在頻率分布直方圖中表現(xiàn)出的“長拖尾”現(xiàn)象；對于實際運行管道頻繁受到的輸送工藝突發(fā)和持續(xù)的干擾,本文深入探究了信息熵理論在間歇性輸送管道泄漏診斷中的應用,提出了聲波信號頻域功率譜熵和聯(lián)合時-頻域小波包空間特征譜熵的特征提取方法,完善了聲波信號的特征描述；由于實際運行管道很難獲取大量泄漏樣本,本文管道泄漏診斷模型采用了具有單值分類特性的支持向量數(shù)據(jù)描述模型,通過對正常樣本的訓練,避免了泄漏形式多樣化對模型泛化能力的影響。現(xiàn)場測試表明,本文所提出的間歇性管道泄漏診斷方法能夠有效實現(xiàn)間歇性輸送管道泄漏的快速、準確診斷,并且具有較高的可靠性、穩(wěn)定性和工況適應性,為解決間歇性輸送管道的泄漏診斷提供了一種有效、可靠的方法。
[Abstract]:The development of pipeline transportation industry has greatly promoted the development of chemical industry and oil and gas resources industry. Pipeline transportation is gradually becoming the preferred mode of oil and gas resources transportation. At the same time, the number of oil and gas pipelines is increasing year by year, and the duration of service is also increasing. Due to construction defects, erosion, artificial destruction, aging of equipment and changes in geographical and climatic environment, Oil and gas transportation pipeline medium leakage accidents occur frequently, resulting in a large number of waste of resources and property losses. Pipeline leak diagnosis technology is a complex technology which integrates multi-domain and multi-disciplinary knowledge, especially intermittent transportation pipeline. The pipeline working condition is extremely complex, and is restricted by many factors, and up to now, No pipeline leak diagnosis method is universal, and each method has its own adaptability. Therefore, the reliable diagnosis of intermittent pipeline leakage has important theoretical and practical application value. On the basis of time domain statistical feature analysis and information entropy theory, a method for feature extraction of acoustic signals is proposed in this paper. By studying the statistical characteristics of acoustic signals, the time-domain statistical characteristics of acoustic signals are extracted, and the phenomenon of "long trailing" in the frequency distribution histogram of pipeline leakage is described. For the frequent sudden and continuous interference of transportation technology in actual running pipeline, this paper probes into the application of information entropy theory in the diagnosis of intermittent pipeline leakage. The feature extraction method of frequency domain power spectrum entropy and combined time-frequency wavelet packet spatial characteristic spectrum entropy of acoustic signal is proposed to improve the feature description of acoustic signal. Because it is very difficult to obtain a large number of leakage samples in actual pipeline operation, the pipeline leak diagnosis model in this paper adopts the support vector data description model with the characteristics of single value classification, and through the training of normal samples, The influence of leakage form diversification on the generalization ability of the model is avoided. Field tests show that the method proposed in this paper can effectively diagnose intermittent pipeline leakage quickly and accurately, and has high reliability, stability and adaptability. It provides an effective and reliable method to solve the leakage diagnosis of intermittent pipeline.
【學位授予單位】：北京化工大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TE973.6

【參考文獻】

相關期刊論文 前2條

1 唐秀家,顏大椿;基于神經(jīng)網(wǎng)絡的管道泄漏檢測方法及儀器[J];北京大學學報(自然科學版);1997年03期

2 陶洛文,方崇智,肖德云;以辨識為基礎的長輸管線故障定位[J];清華大學學報(自然科學版);1986年02期

相關碩士學位論文 前1條

1 朱愛華;卡爾曼濾波和序貫概率比檢驗在管道泄漏監(jiān)測中的應用[D];天津大學;2006年

，

本文編號：2346295