《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 嵌入式技術(shù) > 設(shè)計(jì)應(yīng)用 > 基于改進(jìn)的證據(jù)更新工業(yè)過程故障診斷研究
基于改進(jìn)的證據(jù)更新工業(yè)過程故障診斷研究
2019年電子技術(shù)應(yīng)用第11期
朱玉華,曲萍萍
沈陽(yáng)工業(yè)大學(xué) 化工過程自動(dòng)化學(xué)院,遼寧 遼陽(yáng)111003
摘要: 工業(yè)生產(chǎn)過程的故障成因頗為復(fù)雜,一種故障的故障特征可能有多種表現(xiàn)形式,而多種故障又有可能表現(xiàn)出一種故障特征。因此單模型、單因素的故障診斷方法已顯其不足。提出了改進(jìn)的證據(jù)更新的動(dòng)態(tài)故障診斷算法,并結(jié)合人工智能方法應(yīng)用到硝酸生產(chǎn)過程故障診斷系統(tǒng)中。該方法通過對(duì)模糊神經(jīng)網(wǎng)絡(luò)的描述來確定故障診斷的辨識(shí)框架,應(yīng)用新型的模糊推理方法生成診斷證據(jù),診斷證據(jù)再基于改進(jìn)的證據(jù)更新規(guī)則來實(shí)現(xiàn)證據(jù)的動(dòng)態(tài)更新,根據(jù)結(jié)果來進(jìn)行故障決策,從而解決了故障模式多樣性、故障診斷動(dòng)態(tài)性以及故障特征不確定性的問題。經(jīng)實(shí)例驗(yàn)證,該方法的應(yīng)用可提高故障診斷確診率。
中圖分類號(hào): TP277
文獻(xiàn)標(biāo)識(shí)碼: A
DOI:10.16157/j.issn.0258-7998.190536
中文引用格式: 朱玉華,曲萍萍. 基于改進(jìn)的證據(jù)更新工業(yè)過程故障診斷研究[J].電子技術(shù)應(yīng)用,2019,45(11):87-90,95.
英文引用格式: Zhu Yuhua,Qu Pingping. Research on industrial process fault diagnosis based on improved evidence updating[J]. Application of Electronic Technique,2019,45(11):87-90,95.
Research on industrial process fault diagnosis based on improved evidence updating
Zhu Yuhua,Qu Pingping
School of Chemical Process Automation,Shenyang University of Technology,Liaoyang 111003,China
Abstract: The causes of failures in industrial production processes are quite complex. The fault characteristics of a fault may be multiple, and multiple faults may exhibit the same fault characteristics. Therefore, the single-factor, single-model fault diagnosis method has been insufficient. This paper proposes an improved evidence-updated dynamic fault diagnosis algorithm, and combines it with the artificial intelligence method to apply to the industrial production process fault diagnosis system. The method determines the identification framework of fault diagnosis by describing the fuzzy neural network, applies the new fuzzy reasoning method to generate diagnostic evidence, and then based on the improved evidence update rule to realize the dynamic update of the diagnostic evidence before and after the acquisition, which will be updated dynamically. As a result, fault decision is made to solve the fault pattern diversity, fault diagnosis dynamics and uncertainty of fault characteristics. The example analysis proves that the method achieves the purpose of improving the diagnosis rate of fault diagnosis.
Key words : fault characteristics;fault diagnosis;diagnostic evidence;evidence update

0 引言

    工業(yè)生產(chǎn)過程運(yùn)行中,由于設(shè)備或人為等因素常常會(huì)出現(xiàn)一些故障,會(huì)嚴(yán)重影響產(chǎn)品質(zhì)量甚至對(duì)人身安全造成危害。傳統(tǒng)的更新方法的缺點(diǎn)在于過度依賴于當(dāng)前證據(jù)的作用,而忽略了歷史證據(jù)的作用。本文通過對(duì)原始證據(jù)更新規(guī)則進(jìn)行改進(jìn),提出了改進(jìn)的證據(jù)更新規(guī)則的動(dòng)態(tài)故障診斷算法[1-5],應(yīng)用到硝酸生產(chǎn)裝置故障診斷系統(tǒng)中,確定自動(dòng)生產(chǎn)裝置故障的辨識(shí)框架并生成診斷證據(jù),診斷證據(jù)進(jìn)行動(dòng)態(tài)的實(shí)時(shí)更新,將更新融合后的證據(jù)進(jìn)行故障決策[6],并與目前廣泛使用的原始證據(jù)理論和原始單獨(dú)模糊推理進(jìn)行分析對(duì)比,探討改進(jìn)的證據(jù)更新規(guī)則的動(dòng)態(tài)故障診斷算法的優(yōu)勢(shì)。本文利用證據(jù)更新,對(duì)歷史證據(jù)進(jìn)行實(shí)時(shí)更正,從而得到更全面可靠的決策。該方法在智能性、實(shí)時(shí)性和精確性方面對(duì)工業(yè)生產(chǎn)裝置的故障診斷效果都得到了有效的提高。

1 故障分析

    以稀硝酸生產(chǎn)過程系統(tǒng)為例,稀硝酸生產(chǎn)過程系統(tǒng)非常復(fù)雜,通過對(duì)邏輯控制關(guān)系和系統(tǒng)分析得到7個(gè)已知故障和5個(gè)故障特征參數(shù),從而進(jìn)行故障特征參數(shù)和故障模式分析。稀硝酸生產(chǎn)過程系統(tǒng)故障特征參數(shù)及故障模式如表1所示。jsj1-b1.gif

    故障診斷數(shù)據(jù)是采集系統(tǒng)的氣氨流量值、氨空比值、工藝水流量值、入出口壓力值,按照故障模式對(duì)稀硝酸生產(chǎn)過程系統(tǒng)進(jìn)行分析。共提取可檢測(cè)信號(hào)98項(xiàng),根據(jù)提取的故障數(shù)據(jù)找出故障。

2 辨識(shí)框架的生成

jsj1-gs1.gif

jsj1-b2.gif

jsj1-gs2-4.gif

3 改進(jìn)的診斷證據(jù)更新過程 

3.1 更新規(guī)則的建立

jsj1-gs5.gif

jsj1-gs6-7.gif

    改進(jìn)的全部更新過程如圖1所示。

jsj1-t1.gif

    改進(jìn)的證據(jù)更新規(guī)則具有實(shí)時(shí)性和證據(jù)的可信性,更新結(jié)果既與當(dāng)前證據(jù)有關(guān),也與歷史證據(jù)有關(guān),利用當(dāng)前證據(jù)來更新先前所獲取的故障診斷證據(jù),完成了歷史證據(jù)與當(dāng)前證據(jù)的信息整合,解決了模糊理論的缺陷問題,完全適用于動(dòng)態(tài)的故障診斷情況。

3.2 故障決策 

    故障診斷的關(guān)鍵一步是故障決策[7-10]。首先需要Pignistic變換,是在更新后k+1時(shí)刻進(jìn)行變換,故障決策是根據(jù)基本置信度轉(zhuǎn)換成Pignistic函數(shù)來進(jìn)行的[11-16]。首先采集故障特征的在線監(jiān)測(cè)值x1,x2,…,xn,通過改進(jìn)的證據(jù)更新規(guī)則來獲取更新后的診斷證據(jù),再經(jīng)過Pignistic變換得出概率值。

    Pignistic概率函數(shù)為:

jsj1-gs8.gif

4 故障診斷實(shí)驗(yàn)分析 

    動(dòng)態(tài)故障診斷分析以生產(chǎn)稀硝酸為例,構(gòu)造出模糊規(guī)則總數(shù)為1 247個(gè),生產(chǎn)稀硝酸的故障特征參數(shù)模糊語(yǔ)言集如下:

    U1=(A1,1,A1,2,A1,3,A1,4,A1,5,A1,6,A1,7),j=7

    U2=(A2,1,A2,2,A2,3,A2,4,A2,5,A2,6,A2,7),j=7

    U3=(A3,1,A3,2,A3,3,A3,4,A3,5,A3,6),j=6

    U4=(A4,1,A4,2,A4,3,A4,4,A4,5,A4,6,A4,7,A4,8),j=8

    U5=(A5,1,A5,2,A5,3,A5,4,A5,5,A5,6,A5,7,A5,8),j=8

    通過采集稀硝酸生產(chǎn)過程系統(tǒng)故障特征參數(shù)的在線監(jiān)測(cè)值,分析并計(jì)算在新型模糊推理規(guī)則下故障特征參數(shù)的實(shí)時(shí)監(jiān)測(cè)值所屬的模糊語(yǔ)言的項(xiàng)置信度。利用0時(shí)刻的稀硝酸生產(chǎn)過程系統(tǒng)故障特征參數(shù)在線監(jiān)測(cè)值,確定一條前項(xiàng)模糊規(guī)則,再經(jīng)過新型模糊推理獲得此刻的故障診斷證據(jù)。稀硝酸生產(chǎn)過程系統(tǒng)故障特征參數(shù)的在線監(jiān)測(cè)值及被選中的語(yǔ)言項(xiàng)歸一化置信度如表3所示。

jsj1-b3.gif

    因?yàn)橛?種故障特征參數(shù),則共有32種組合,即JNR=32條規(guī)則,從而得到特征參數(shù)監(jiān)測(cè)值所選模糊規(guī)則前項(xiàng)如表4所示(部分?jǐn)?shù)據(jù))。特征參數(shù)監(jiān)測(cè)值所選模糊規(guī)則后項(xiàng)如表5所示(部分?jǐn)?shù)據(jù))。根據(jù)改進(jìn)的證據(jù)更新規(guī)則得出Pignistic概率值和k=1時(shí)刻的診斷證據(jù)如表6和表7所示。

jsj1-b4.gif

jsj1-b5.gif

jsj1-b6.gif

jsj1-b7.gif

    最后根據(jù)故障決策準(zhǔn)則做出故障診斷,在K=1時(shí)刻故障Y7發(fā)生,經(jīng)過實(shí)驗(yàn)診斷分析結(jié)果與設(shè)定的故障模式一致。對(duì)診斷結(jié)果的進(jìn)行對(duì)比分析,根據(jù)稀硝酸生產(chǎn)過程系統(tǒng)狀態(tài)值采樣給出900個(gè)采樣數(shù)據(jù)點(diǎn),每個(gè)故障有100個(gè)采樣點(diǎn)經(jīng)過模糊神經(jīng)網(wǎng)絡(luò)訓(xùn)練,根據(jù)稀硝酸生產(chǎn)過程系統(tǒng)的已知在線診斷獲得245個(gè)測(cè)試數(shù)據(jù),經(jīng)過原始證據(jù)理論、模糊推理和改進(jìn)型證據(jù)更新規(guī)則3種方法對(duì)比,測(cè)試結(jié)果如表8所示。

jsj1-b8.gif

    從故障診斷結(jié)果的對(duì)比分析可以看出,改進(jìn)的證據(jù)更新規(guī)則在工業(yè)生產(chǎn)動(dòng)態(tài)故障診斷領(lǐng)域中具有更好的應(yīng)用優(yōu)勢(shì),并發(fā)揮了重要作用。

5 結(jié)論

    本文提出的改進(jìn)的證據(jù)更新規(guī)則的動(dòng)態(tài)故障診斷算法在工業(yè)領(lǐng)域解決了動(dòng)態(tài)的故障診斷的問題。新型的模糊推理和原始的模糊推理相比較,其推理結(jié)果更加精確,確保了證據(jù)更新的實(shí)時(shí)性,進(jìn)而確保了診斷結(jié)果的實(shí)時(shí)性。通過檢測(cè)系統(tǒng)的實(shí)時(shí)運(yùn)行狀態(tài)可以全面反映實(shí)際工況,給出各部件的維修建議,達(dá)到故障診斷的目的,體現(xiàn)了人工智能故障診斷在工業(yè)生產(chǎn)中的重要作用。利用證據(jù)更新對(duì)歷史證據(jù)進(jìn)行實(shí)時(shí)更正,從而得到更全面可靠的決策,來提高產(chǎn)品質(zhì)量和經(jīng)濟(jì)效益,在人身安全方面也有了更好的保障。

參考文獻(xiàn)

[1] KULASEKERE E C,PREMARATNE K,DEWASURENDRA D A,et al.Conditioning and updating evidence[J].International Journal of Approximate Reasoning,2004,36(1):75-108.

[2] DEMPSTER A P.Upper and lower probabilities induced by a multivalued mapping[J].The Annals of Mathematical Statistics,1967,38(2):325-339.

[3] 宋曉靜.基于證據(jù)理論的工業(yè)報(bào)警器設(shè)計(jì)方法研究[D].杭州:杭州電子科技大學(xué),2012.

[4] 史健.基于證據(jù)理論的動(dòng)態(tài)融合方法研究[D].杭州:杭州電子科技大學(xué),2012.

[5] 王玉成.沖突證據(jù)的相似性度量方法及其在信息融合故障診斷中的應(yīng)用[D].杭州:杭州電子科技大學(xué),2011.

[6] 張鎮(zhèn).基于證據(jù)推理與更新規(guī)則的動(dòng)態(tài)融合方法及其應(yīng)用研究[D].杭州:杭州電子科技大學(xué),2016.

[7] 徐曉濱,張鎮(zhèn),李世寶,等.基于診斷證據(jù)靜態(tài)融合與動(dòng)態(tài)更新的故障診斷方法[J].自動(dòng)化學(xué)報(bào),2016,42(1):107-108.

[8] 董煒,陳衛(wèi)征,徐曉濱,等.基于模糊區(qū)間優(yōu)化的模糊推理故障診斷方法[J].北京工業(yè)大學(xué)學(xué)報(bào),2012,38(12):1905-1912.

[9] 吳祎,周強(qiáng),吳文軍,等.動(dòng)態(tài)辨識(shí)框架下條件證據(jù)更新的故障檢測(cè)方法[J].計(jì)算機(jī)應(yīng)用研究,2015,32(8):2370-2373.

[10] 柳新軍,張東來,李安壽,等.基于BIT技術(shù)的PCU故障診斷和性能監(jiān)測(cè)系統(tǒng)[J].電子技術(shù)應(yīng)用,2019,45(5):34-37.

[11] 宋玉琴,朱紫娟,姬引飛.基于粗糙集優(yōu)化的信息融合故障診斷系統(tǒng)[J].電子技術(shù)應(yīng)用,2015,41(8):143-145,148.

[12] OUKHELLOU L,DEBIOLLES A,DENCEUX T,et al.Fault diagnosis in railway track circuits using dempster-shafer classifier fusion[J].Engineering Applications of Artificial Intelligence,2010,23(1):116-123.

[13] ZHANG Q H,HU Q,SUN G X,et al.Concurrent fault diagnosis for rotating machinery based on vibration sensors[J].International Journal of Distributed Sensor Networks,2013,75(8):93-102.

[14] UX B,ZHOU Z,WEN C L.Data fusion algorithm of fault diagnose considering sensor measurement uncertainty[J].International Journal on Smart Sensing and Intelligent Systems,2013,6(1):170-190.

[15] WEN C L,XU X B,LI Z L.Research on unified description and extension of combination rules of evidence based on random set theory[J].Chinese Journal of Electronics,2008,17(2):279-284.

[16] DUBOIS D,PRADE H.Updating with belief funcations, ordinal conditional functions and possibility measure[C].Proceedings of the 6th Annual Conference on Uncertainty in Artificial Intelligence,Cambridge,Amsterdam:Elsevier Science,1990:311-329.



作者信息:

朱玉華,曲萍萍

(沈陽(yáng)工業(yè)大學(xué) 化工過程自動(dòng)化學(xué)院,遼寧 遼陽(yáng)111003)

此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載。