失效模式在台灣地區配水管線之應用

Application of analytic models for pipe failure in Water Distribution Network in Taiwan Area

黃科勤(Ko-Chin Huang)

指導教授 ： 黃志彬

國立交通大學/工學院/環境工程系所/博士(2020年)

摘要

國內自來水管線於108年底總長度為63,484公里，累積汰換管線經費約433億元，完成汰換管線約4,898公里，惟漏水率仍達14.49%；國際水協(IWA)將「監測漏水、早期修復」作為主動控制漏水之重要行動準則，而「設施更新汰換、選用適當管材」更是重要方案；如能將汰換管線投資在管線弱點區域，有效預防漏水，每年只要降低漏水率1%，即可節省約4千萬公噸自來水的滲漏流失，有效降低水資源的浪費。
本研究歸納國內配水管線失效主要原因，分為「管線外部」、「管線內部」及「管線失效反應」等三類，依據台灣自來水公司「修漏管理系統」及「地理資訊圖資管理系統」資料庫進行數據分析，探討環境、管材、管徑、管齡及水壓等各項因子，篩除資料庫中缺漏及錯誤數據，運用「風險矩陣法」、「類神經網路法」、「環境因子高風險潛勢法」及「失誤樹法」等四種預測技術，建立國內預測配水管線失效模式。
經由上述四種失效預測模式評析論述，「風險矩陣法」運用於80 mm(含)以上管徑，為惟一考量失效後造成之影響程度之模式，而「類神經網路法」初步以「竹南區域」分析，需進一步再深入研究，「環境因子高風險潛勢法」為最簡單模式，惟當失效率相同時，須以其他模式深入分析，而「失誤樹法」則可有效預測高風險潛勢之配水管線弱點區域；另依各模式計算結果，分別評估出風險等級，以規劃標的管線短、中、長期之汰換及檢測計畫，區分為高風險、次高風險、中風險分三種等級，率定檢測頻率，而低風險則不需採取措施及監測。據此可提升投資效益，降低配水管線漏水率，使水資源有效運用。

關鍵字

配水管線 ； 管線失效 ； 風險矩陣法 ； 類神經網路法 ； 失誤樹法

並列摘要

At the end of 2019, the total length of water distribution pipeline in Taiwan was 63,484 kilometers, and the cumulative cost of pipeline replacement came to NT$43.3 billion. Although 4,898 kilometers of pipelines have been renewed, the water leakage rate remains as high as 14.49%. The International Water Association (IWA) has issued "monitoring leakage and early repair" as an important action criterion for proactively water leakage control, as well as “facilities renewal/ replacement and selection of appropriate materials" to be a critical approach. If the pipeline renewal projects are invested in the weak pipeline areas, water leakage can thus be effectively prevented, and even 1% of leakage rate decrease can save about 40 million metric tons of water annually, thereby effectively reducing the waste of water resources.
In this study, we summarized the causes for the failure of water distribution pipelines and classified them into three main categories: "outside the pipeline", "inside the pipeline" and "pipeline failure response". We analyzed data based on the database of “Repairing Management System" and "地理資訊 Map Management System" provided by Taiwan Water Supply Company, and investigated various factors, such as environment, pipe material, pipe diameter, pipe age and water pressure, etc. After filtering out missing and incorrect data in the data base, and by applying four prediction techniques, including "risk matrix method", "neural network method", “high-risk potential environmental factor method" and "fault tree method" , we have established the failure predicting modes for domestic water distribution pipelines.
We comment on the four prediction models mentioned above. The "risk matrix method" is applied to pipeline above 80 mm (inclusive) in diameters, and is the only model that takes the degree of influence caused by pipeline failure into account. The "neural network method" was initially utilized for analysis in Zhunan district, and must do further research. The " high-risk potential environmental factor method " is the simplest model, however , if the failure rates are the same , other model must be introduced for further analysis. The "fault tree method" is adopted to effectively predict the weak areas of pipelines with high risk potential. We evaluated the risk levels according to the calculated results of each model respectively, and used them to scheme the short, medium, and long term detection plan for object pipelines. The detection frequencies were classified as high risk, sub-high risk, and medium risk, while low risk class has no need to take measurement or monitoring. Accordingly, the investment efficiency can be improved and the leakage rate of pipelines will be reduced, improving the efficiency of water resources.

並列關鍵字

water distribution network ； pipe failures ； risk matrix method ； artificial neural networks ； fault tree analysis

參考文獻

Amuji, H.O.; Ogbonna, C.J.; Ugwuanyim, G.U.; Iwu, H.C.; Nwanyibuife, O.B. Optimal water pipe replacement policy. Open Journal of Optimization 2018, 7, 41-49.

Andreou, S.A.; Marks, D.H.; Clark, R.M. A new methodology for modelling break failure patterns in deteriorating water distribution systems: Applications. Adv. Water Resour. 1987, 10, 11-20.

Baybutt, P. Guidelines for designing risk matrices. Process Saf. Prog. 2018, 37, 49-55.

Berardi, L.; Giustolisi, O.; Kapelan, Z.; Savic, D.A. Development of pipe deterioration models for water distribution systems using EPR. J. Hydroinform. 2008, 10, 113-126.

Bogárdi, I.; Fülöp, R. A spatial probabilistic model of pipeline failures. Periodica Polytechnica Civil Engineering 2011, 55, 161-168.