鍋爐設(shè)計(jì)外文翻譯---燃煤鍋爐的個(gè)案事故研究

1、<p>　　附錄A </p><p><b>　　英文原文</b></p><p>　　A. Kusiak and A. Burns, Mining Temporal Data: A Coal-Fired Boiler Case Study, Proceedings of the 9th Inter

2、national Conference, KES 2005, Melbourne, Australia, September 14-16, 2005, in R. Khosla, R.J. Howlett, L.C. Jain (Eds), Knowledge-Based Intelligent Information and Engineering Systems: Vol. III, LNAI 3683, Springer, Hei

3、delberg, Germany, 2005, pp. 953-958. </p><p>　　Mining Temporal Data: A Coal-Fired Boiler Case Study</p><p>　　Andrew Kusiak and Alex Burns</p><p>　　Intelligent Systems Laboratory, In

4、dustrial Engineering</p><p>　　3131 Seamans Center, The University of Iowa</p><p>　　Iowa City, IA 52242 – 1527, USA</p><p>　　andrew-kusiak@uiowa.edu</p><p><b>　　Ab

5、stract</b></p><p>　　This paper presents an approach to control pluggage of a coal-fired boiler. The proposed approach involves statistics, data partitioning, parameter reduction, and data mining. The

6、proposed approach was tested on a 750 MW commercial coal-fired boiler affected with a fouling problem that leads to boiler pluggage that causes unscheduled shutdowns. The rare-event detection approach presented in the pa

7、per identified several critical time-based data segments that are indicative of the ash pluggage. </p><p>　　1 Introduction </p><p>　　The ability to predict and avoid rare events in time series

8、 data is a challenge that could be addressed by data mining approaches. Difficulties arise from the fact that often a significant volume of data describes normal conditions and only a small amount of data may be availabl

9、e for rare events. This problem is further exacerbated by the fact that traditional data mining does not account for the time dependency of the temporal data. The approach presented in this paper overcomes these concerns

10、 </p><p>　　windows. </p><p>　　The approach presented in this paper is based on the two main concepts. The first is that the decision-tree data-mining algorithm captures the subtle parameter rel

11、ationships that cause the rare event to occur [1]. The second concept is that partitioning the data using time windows provides the ability to capture and describe sequences of events that may cause the rare failure.

12、</p><p>　　2 Event Detection Procedure </p><p>　　In the case study discussed in the next section rare events can be detected by applying the five step procedure. These five steps include: <

13、/p><p>　　Step 1: Parameter Categorization </p><p>　　The parameter list is divided into two categories, response parameters and impact parameters. Response parameters are those that change values d

14、ue to a rare event or a failure, e.g., an air leak in a pressurized chamber. Impact parameters are defined as parameters that are either directly or indirectly controllable and may cause the rare event. These are the p

15、arameters that are of greatest interest for the determination of rare events. </p><p>　　Step 2: Time Segmentation </p><p>　　Time segmentation deals with partitioning and labeling the data into

16、time windows (TWs). A time widow is defined as a set of observations in chronological order that describe a specified amount of continuous observations. This step allows the data mining algorithms to account for the tem

17、poral nature of the data. The most effective method to segment the data is by determining/estimating the approximate date of failure and set that as the last observation of the final time window. </p><p>

18、　　Step 3: Statistical and Visual Analysis </p><p>　　This step involves statistical analysis of the data in each time period that was designated in the previous step. Process shifts, changes in variation, an

19、d mean shifts in parameters are helpful in indicating that the appropriate time windows and parameters were selected. </p><p>　　Step 4: Knowledge Extraction </p><p>　　Data mining algorithms dis

20、cover relationships among parameters and an outcome in the form of IF … THEN rules and other constructs (e.g., decision tables) [1], [5]. Data mining is natural extension of more traditional tools such as neural networks

21、, multivariable algorithms, or traditional statistics. In the detection of rare events, the decision-tree and rule-induction algorithms are explored for two significant reasons. First, the algorithms generate explicit k

22、nowledge in the form understandabl</p><p>　　Step 5: Analysis of Knowledge and Validation </p><p>　　This step deals with validation of the knowledge generated by the data mining algorithm. If a

23、validation data set is available it should be used to validate the accuracy of the rules. If no similar data is available then unused data from the analysis or a 10-fold cross-validation can be utilized [6]. </p>

24、<p>　　3 Power Boiler Case Study </p><p>　　The approach proposed in this research was applied to power plant data. Data mining algorithms are well suited for electric power applications that produce hu

25、ndreds of data points at any time instance. </p><p>　　This case study deals with an ash fouling condition that causes boiler shutdowns several times a year on a commercial 750 MW tangentially-fired coal boi

26、ler. The ash fouling causes a build up of material and pluggage in the reheater section of the boiler. Once the build up becomes substantial the boiler performance is negatively affected. This leads to the derating an

27、d the eventual shutdown of the boiler. The cleaning of the boiler during the shutdown requires 1 to 3 days. This problem is ma</p><p>　　The list of 173 parameters, which included both response and impact p

28、arameters, was analyzed. The list was reduced to include twenty-six impact parameters. This parameter categorization and reduction was accomplished with the assistance of domain experts as well as statistical analysis s

29、uch as correlation and multivariate analysis. </p><p>　　The initial step for time segmenting the data was to determine an approximate date for the failure event. In this application the failure event was d

30、efined by the date when the boiler was derated due to the pluggage. The cause of the shutdown was confirmed through visual inspection of the affected region. This date was then set to be the last day of the final time w

31、indow (TW6). </p><p>　　The windows were set to be approximately one week long. A week was chosen for several reasons. First, the boiler was inspected approximately one month prior to its derating. During

32、 the inspection the reheater section of the boiler was completely free of ash. This information provided the knowledge that the pluggage required less the one month to manifest itself to the point of shutdown. It was hyp

33、othesized that the pluggage requires several days to build up. Based on this information one week</p><p>　　Using the derate date and a one-week-long time window, the data was divided into six time windows s

34、hown in Figure 1. Time window 1 (TW1) was included to ensure that there was adequate data to describe normal operating conditions. </p><p>　　There appears be a process shift between time windows 3 and 5 in

35、 Figures 1. The west tilt demonstrates a mean shift during window three and the hot reheat steam temperature displays a mean shift as well as a large increase in variation starting in time window four and culminating in

36、 window five. The results of this analysis lead to the hypothesis that the events that lead to the eventual pluggage occur between time windows three and five. It also confirms the selection of parameters and window<

37、/p><p>　　The data mining approach was then applied to the data set to predict the predefined time windows (decision parameter). The algorithm produced a set of rules that described the parameter relationships

38、in each time window. </p><p>　　The knowledge extracted by the algorithm had an overall 10-fold classification accuracy of 99.7%. The confusion matrix (absolute classification accuracy matrix) is shown in

39、Figure 2. The matrix displays the actual values and the values predicted by the rules during the cross-validation process. </p><p>　　It can be seen from the data in Figure 2 that there are few predicted va

40、lues that are off by more than one time window from the actual window. The results provided in the confusion matrix provide a high confidence in the proposed solution approach. </p><p>　　Another test data

41、set was extracted from the week following time window 1 and was labeled time window 2 (Test TW2). The last portion of the data (Test TW3) was obtained from the week after the generator was derated and the outcome was lab

42、eled time window 6 (TW6). The total test set contained over 30,000 observations. </p><p>　　The rules and knowledge that were extracted from the original data set were then tested using the test data set.

43、For purposes of analysis time windows 1 – 3 were considered normal and time window 4 – 6 were considered faulty. The resulting confusion matrix is shown in Figure 3. </p><p>　　The rules accurately predicted

44、 the normal cases, but they were not as effective in predicting the fault cases. This is most likely explained by the fact that the test data labeled, time window 6, was extracted after the boiler had been derated. The

45、 derating of the boiler significantly changes the combustion process and was not included in the original data set. In spite of this, the overall classification accuracy of the test data set is greater than 89%. The hig

46、h cross-validation accuracy in</p><p>　　4 Future Research</p><p>　　Event detection for control advisory systems has also been successfully demonstrated for applications that are dynamic and inv

47、olve rare and catastrophic events [4]. Finch et al. [2] developed expert diagnostic information system, MIDAS, to alert users to abnormal transient conditions in chemical, refinery, and utility systems [3]. </p>&

48、lt;p>　　The approach presented in this research produced rule sets that can be utilized for the development of a meta-control system. Integrating concepts from expert advisory systems and intelligent power control syst

49、ems will form the meta-control system architecture for the avoidance of the ash pluggage. </p><p>　　5 Conclusion </p><p>　　In this paper a data mining approach to predict failures was proposed

50、and successfully implemented. The research utilized parameter categorization and time segmentation to overcome the limitation of traditional data mining approaches applied to temporal data. The proposed approach produce

51、d a knowledge base (rule set) that accurately described the subtle process shifts and parameter relationships that eventually may lead to the detection and avoidance of failures. </p><p>　　This approach wa

52、s applied to a commercial tangentially-fired coal-boiler to detect and avoid an ash fouling pluggage that eventually leads to boiler shutdown. The approach produced a rule set that was over 99.7% accurate. The knowledg

53、e base was also validated with a separate test data set that has predicted failures with accuracy of over 89.8%. </p><p>　　The discovered knowledge will be used to develop an advance warning system reducin

54、g the number of boiler shutdowns. The intelligent warning system will have a significant economic impact. This translates into reduced cost to the consumer and a more efficient power industry. </p><p>　　Ref

55、erences</p><p>　　1. Quinlan, J.R., “Induction of decision trees,” Machine Learning, vol. 1, no. 1, pp. 81-106, 1986. </p><p>　　2. Branagan, L. A., and Wasserman, P. D., "Introductory use

56、of probabilistic neural networks for spike detection from an on-line vibration diagnostic system", Intelligent Engineering Systems Through Artificial Neural Networks, vol. 2, pp. 719-724, 1992. </p><p>

57、　　3. Finch, F. E., Oyeleye, O. O., and Kramer, M. A., "Robust event-oriented methodology for diagnosis of dynamic process systems", Computers & Chemical Engineering, vol.14, no. 12, pp. 1379-1396, Dec, 199

58、0. </p><p>　　4. Pomeroy, B. D., Spang, H. A., and Dausch, M.E., "Event-based architecture for diagnosis in control advisory systems", Artificial Intelligence in Engineering, vol. 5, no. 4, pp. 174

59、-181, Oct, 1990. </p><p>　　5. Pawlak, Z., Rough Sets: Theoretical Aspects of Reasoning About Data, Boston: Kluwer, 1991. </p><p>　　6. Stone, M. "Cross-validatory choice and assessment of

60、statistical predictions," Journal of the Royal Statistical Society, vol. 36, pp.111-147, 1974. </p><p><b>　　附錄B</b></p><p><b>　　中文翻譯</b></p><p>　　燃煤鍋爐的個(gè)

61、案事故研究</p><p><b>　　摘要</b></p><p>　　這篇論文講述了控制燃煤鍋爐堵塞的方法。在這個(gè)方法中涉及了統(tǒng)計(jì)學(xué)，數(shù)據(jù)分析，參數(shù)改變和數(shù)據(jù)采集等。這個(gè)所提到的研究方法是以750MW工業(yè)燃煤鍋爐為實(shí)驗(yàn)測(cè)試對(duì)象。研究由于污染問題導(dǎo)致的不定期堵塞問題。在這篇論文中描述的小概率事件的測(cè)試方法，識(shí)別了一些緊要時(shí)刻以時(shí)間數(shù)據(jù)為依據(jù)的片斷，這些片斷對(duì)灰堵塞有

62、指示作用。</p><p><b>　　1介紹</b></p><p>　　在時(shí)間系列數(shù)據(jù)中預(yù)測(cè)和避免小概率能力是極富有挑戰(zhàn)性的，通過數(shù)據(jù)采集的方法使得具有這種能力。在實(shí)際研究過程中出現(xiàn)的問題通常是：有大量空間的正常數(shù)據(jù)描述而只有很少量的可以使用的有效數(shù)據(jù)是關(guān)于小概率事件的。由于實(shí)際上傳統(tǒng)的數(shù)據(jù)采集方法不能解釋時(shí)間同步數(shù)據(jù)使得這個(gè)問題變得更加難以解決。在這篇論文中講述

63、了通過限定時(shí)間窗口來克服這一問題的方法。</p><p>　　論文中所呈現(xiàn)的方法是以兩個(gè)主要概念為基礎(chǔ)的，一個(gè)是引起小概率事件發(fā)生的三維數(shù)據(jù)采集運(yùn)算法則中獲取敏感參數(shù)的關(guān)系。另一個(gè)概念是分布使用的數(shù)據(jù)用時(shí)間窗口提供了獲取和描述可能引起少數(shù)失敗的一系列事件的能力。</p><p><b>　　2事件發(fā)展程序</b></p><p>　　在這個(gè)個(gè)案

64、研究中，通過使用五個(gè)步驟程序來討論在下個(gè)區(qū)段中小概率事件是否有被發(fā)現(xiàn)的可能性。</p><p>　　這五個(gè)步驟程序包括：</p><p>　　第一步驟：參數(shù)編輯方法</p><p>　　參數(shù)目錄可分為兩種：一種是反應(yīng)參數(shù)，另一種是影響參數(shù)。反應(yīng)參數(shù)是那些變化數(shù)值后能引起小概率事件發(fā)生或失誤的參數(shù)。舉例來說一個(gè)壓力空間內(nèi)的空氣泄漏事件。影響參數(shù)被定義為是一種能直接或間

65、接控制或引起小概率事件的參數(shù)，這些參數(shù)是決定小概率事件的關(guān)鍵參數(shù)。</p><p><b>　　第二步驟：計(jì)時(shí)分割</b></p><p>　　時(shí)間窗口的分布和標(biāo)志的數(shù)據(jù)。時(shí)間窗口被定義為在按時(shí)間順序組成的一系列觀察中，由他描述一個(gè)特別數(shù)量的連續(xù)觀察。這個(gè)步驟允許用數(shù)據(jù)采集運(yùn)算法則來解釋數(shù)據(jù)的當(dāng)前性質(zhì)。最有效的分割數(shù)據(jù)的方法是估計(jì)失敗的大約時(shí)間，而且把這個(gè)時(shí)間設(shè)置為時(shí)

66、間窗口的最后觀察時(shí)間。</p><p>　　第三步驟：統(tǒng)計(jì)和視覺分析</p><p>　　這個(gè)步驟涉及對(duì)在早期制定的每個(gè)時(shí)間階段的數(shù)據(jù)的統(tǒng)計(jì)學(xué)分析和過程轉(zhuǎn)換，變化時(shí)的改變以及在參數(shù)中的意義轉(zhuǎn)變。在指定大概的時(shí)間窗口和參數(shù)選擇中是十分有用的。</p><p><b>　　第四步驟：知識(shí)抽取</b></p><p>　　數(shù)據(jù)

67、采集的運(yùn)算法則以：“IF…….then”法則的形式發(fā)現(xiàn)了參數(shù)和結(jié)論的關(guān)系。數(shù)據(jù)采集是很多傳統(tǒng)工具的天然延伸，像是內(nèi)神經(jīng)網(wǎng)絡(luò)，多變數(shù)運(yùn)算法則或傳統(tǒng)的統(tǒng)計(jì)學(xué)。在小概率事件的發(fā)現(xiàn)中決定—樹和規(guī)則—?dú)w納法的運(yùn)算法則因?yàn)橐韵聝蓚€(gè)重要的理由被探究。首先運(yùn)算法則通過使用者以一種可以理解的形式產(chǎn)生的明確的結(jié)果。使用者能夠了解收集到的信息評(píng)定信息的用處而且學(xué)習(xí)新的有趣的概念，其次數(shù)據(jù)采集運(yùn)算法則在很多領(lǐng)域中都顯示出很高的準(zhǔn)確性。</p>&

68、lt;p>　　第五步驟：知識(shí)和確認(rèn)的分析</p><p>　　這個(gè)步驟通過數(shù)據(jù)的采集運(yùn)算法則來處理知識(shí)產(chǎn)生的確認(rèn)性。如果一個(gè)確定性的數(shù)據(jù)集合是可以獲得的，那么這個(gè)數(shù)據(jù)應(yīng)該被用來使規(guī)則的確定性有效。如果沒有可以獲得的相似的數(shù)據(jù)那么分析得來的為使用的或者迭代的超確定性數(shù)據(jù)可以被利用。</p><p>　　3.電站鍋爐的事故和研究</p><p>　　這個(gè)研究中所使

69、用的方法也適用于電站廠的數(shù)據(jù)研究。數(shù)據(jù)采集運(yùn)算法則通過產(chǎn)生數(shù)以百計(jì)的數(shù)據(jù)，在任意時(shí)間例證點(diǎn)使電站廠的使用得到很有效的研究。</p><p>　　事故研究主要處理了灰污染的問題。這種灰污染能使一個(gè)750MW的切線點(diǎn)燃燒煤工業(yè)鍋爐在一年中停產(chǎn)很多次?；椅廴臼沟缅仩t在再熱器部分產(chǎn)生了灰堵塞。一但這種漸變累積成實(shí)質(zhì)的質(zhì)變，鍋爐表面將受到負(fù)面影響。這些都導(dǎo)致了鍋爐的效率減弱和逐漸堵塞以至停產(chǎn)。在停產(chǎn)期間，鍋爐需要停止運(yùn)行一

70、到三天。由于實(shí)際上，沒有辦法在實(shí)際運(yùn)行過程中徹底的使鍋爐檢查區(qū)域封閉，從而來決定產(chǎn)生的灰量和水量，使得這個(gè)問題的研究變得更困難了。此外，在分析中的所使用的所有參數(shù)是在標(biāo)準(zhǔn)規(guī)格中的。因此，沒有顯而易見的單個(gè)數(shù)據(jù)是引起堵塞的唯一原因。這篇論文中所提到的研究對(duì)這個(gè)問題十分重視。</p><p>　　數(shù)據(jù)通過173個(gè)不同鍋爐參數(shù)被搜集。參數(shù)包括流程，壓力，濕度，控制，要求等等。在三個(gè)月中，數(shù)據(jù)每隔1分鐘的間隔收集一次。數(shù)

71、據(jù)搜集隨著再熱器部分堵塞而整個(gè)鍋爐還沒有塞滿的時(shí)刻開始。大約三個(gè)月的時(shí)間完成了數(shù)據(jù)搜集工作。當(dāng)鍋爐由于堵塞而不得不停爐時(shí)這個(gè)數(shù)據(jù)集合包括了168.000個(gè)觀察結(jié)果。173個(gè)參數(shù)表格被分析，它包括反應(yīng)參數(shù)和影響參數(shù)兩部分。這個(gè)表格縮減為26個(gè)參數(shù)。在這個(gè)領(lǐng)域?qū)＜业膮f(xié)助下，這個(gè)參數(shù)編目的方法研究和減少量研究也完成了。同時(shí)，也完成了相關(guān)關(guān)系和多變量分析的統(tǒng)計(jì)分析。</p><p>　　時(shí)間分布數(shù)據(jù)的初始步驟是為了確定一

72、個(gè)更失敗的日期。在這個(gè)使用中，當(dāng)鍋爐由于堵塞而降低效率時(shí)，失敗事故通過日期被限定通過觀測(cè)檢查影響區(qū)域，停產(chǎn)的原因被確定，而后，這個(gè)日期被定為最終時(shí)間的最后一天。</p><p>　　窗口被設(shè)定成大約一個(gè)星期長(zhǎng)。以一個(gè)星期為期限是由很多原因的。首先，鍋爐在它效率下降前的一個(gè)月就被檢驗(yàn)了，在檢驗(yàn)期間，鍋爐的再熱器區(qū)段沒有一點(diǎn)積灰。這個(gè)信息提供了一個(gè)知識(shí)，那就是不用一個(gè)月就能顯示出堵塞，在停爐發(fā)生中的重要性一般假設(shè)，堵

73、塞需要幾天才能產(chǎn)生?；谶@些信息，一個(gè)星期被認(rèn)為是一個(gè)適當(dāng)時(shí)間窗口，一周也為數(shù)據(jù)采集運(yùn)算法則提供了一個(gè)充分的觀察數(shù)據(jù)。</p><p>　　使用減額日期和一個(gè)單周長(zhǎng)時(shí)間窗口，數(shù)據(jù)被分為6個(gè)時(shí)間窗口在圖1中被顯示了。第一時(shí)間窗口包括確定有適當(dāng)?shù)臄?shù)據(jù)描述正常操作情況。</p><p>　　這里顯示了一個(gè)第3時(shí)間窗口和第5時(shí)間窗口的過程轉(zhuǎn)換，傾斜表示在第3窗口點(diǎn)平均轉(zhuǎn)換和再熱系統(tǒng)的蒸汽溫度的這個(gè)

74、分析的結(jié)果產(chǎn)生了一個(gè)假想；導(dǎo)致最后堵塞的事件發(fā)生在第3時(shí)間窗口和第5時(shí)間窗口之間，它也確定了參數(shù)和窗口大小的選擇，然后數(shù)據(jù)采集方法被用于數(shù)據(jù)集合中來預(yù)測(cè)預(yù)先定義的時(shí)間窗口，運(yùn)算法則產(chǎn)生了一組規(guī)則，這組規(guī)則描述了在每一個(gè)時(shí)間窗口的參數(shù)關(guān)系。被運(yùn)算法則吸取的知識(shí)有了全部10層分類準(zhǔn)確性的10%。在跨確認(rèn)期間進(jìn)行中點(diǎn)陣式顯示陣式的價(jià)值和被規(guī)則預(yù)測(cè)的價(jià)值。</p><p>　　在表格2中可以看到，從來自實(shí)際窗口的1個(gè)時(shí)間

75、窗口有很少的預(yù)測(cè)值。結(jié)果提供在混亂點(diǎn)陣式提供高信心的解決方式。另外的測(cè)試數(shù)據(jù)來自于接下來一周的時(shí)間窗口1中和被標(biāo)記過的時(shí)間窗口2中，最后的部分?jǐn)?shù)據(jù)來自于發(fā)生器效率降低域的一周和結(jié)果被標(biāo)記的時(shí)間窗口6中，總測(cè)試結(jié)論包括了超過30.000的觀察點(diǎn)。在原始數(shù)據(jù)中取得的規(guī)則和知識(shí)被測(cè)試使用以及測(cè)試數(shù)據(jù)設(shè)定。</p><p>　　由于分析的原因，時(shí)間窗口的1~3 個(gè)窗口被認(rèn)為是常態(tài)，而時(shí)間窗口4~6窗口被認(rèn)為是有誤的產(chǎn)生的

76、混亂點(diǎn)陣式在表3中顯示。規(guī)則能準(zhǔn)確的預(yù)測(cè)常識(shí)事件，但是，在預(yù)測(cè)過失事件時(shí)就不那么有效了。這很可能通過實(shí)驗(yàn)測(cè)得數(shù)據(jù)被解釋，在鍋爐減產(chǎn)時(shí)，時(shí)間窗口6被吸取。</p><p>　　鍋爐的減額法著重的改變了燃燒過程，在原始數(shù)據(jù)中不包括，盡管這樣，全部實(shí)驗(yàn)數(shù)據(jù)的分類準(zhǔn)確性大于89%。較多的跨確認(rèn)準(zhǔn)確性指明了在導(dǎo)致鍋爐飛灰污染，堵塞，減弱和逐漸停爐的過程中規(guī)則準(zhǔn)確地吸取了變化。</p><p><

77、;b>　　4.結(jié)論</b></p><p>　　在這篇論文中，數(shù)據(jù)采集接近預(yù)測(cè)失敗是達(dá)到目的的。成功的實(shí)現(xiàn)了研究采用的參數(shù)和時(shí)間分割法克服了傳統(tǒng)數(shù)據(jù)采集方法的限制。文中提到的方法，產(chǎn)生了一個(gè)知識(shí)庫(kù)（規(guī)則組）。它確切的描述了敏感的程序改變和能導(dǎo)致發(fā)現(xiàn)和避免失敗的參數(shù)關(guān)系。這個(gè)方法適用于工業(yè)切線點(diǎn)燃的燃煤鍋爐的檢驗(yàn)和避免由于飛灰污染引起的灰堵塞。這方法生產(chǎn)了一個(gè)規(guī)則集合，這個(gè)規(guī)則有大于99.7%的準(zhǔn)