A risk-based soft sensor for failure rate monitoring in water distribution network via adaptive neuro-fuzzy interference systems

Our plain-English summary of A risk-based soft sensor for failure rate monitoring in water distribution network via adaptive neuro-fuzzy interference systems, by M. Gheibi, R. Moezzi, H. Taghavian, et al (Scientific Reports, 2023).

The study

• Authors: M. Gheibi, R. Moezzi, H. Taghavian, S. Wacławek, N. Emrani, M. Mohtasham, M. Khaleghiabbasabadi, J. Koci, C.S.Y. Yeap, J. Cyrus
• Journal: Scientific Reports
• Published: 2023;13(1):12200
• Cited by: 3 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Water Distribution Networks (WDNs) are considered one of the most important water infrastructures, and their study is of great importance. In the meantime, it seems necessary to investigate the factors involved in the failure of the urban water distribution network to optimally manage water resources and the environment. This study investigated the impact of influential factors on the failure rate of the water distribution network in Birjand, Iran. The outcomes can be considered a case study, with the possibility of extending to any similar city worldwide. The soft sensor based on the Adaptive Neuro-Fuzzy Inference System (ANFIS) was implemented to predict the failure rate based on effective features. Finally, the WDN was assessed using the Failure Modes and Effects Analysis (FMEA) technique. The results showed that pipe diameter, pipe material, and water pressure are the most influential factors. Besides, polyethylene pipes have failure rates four times higher than asbestos-cement pipes. Moreover, the failure rate is directly proportional to water pressure but inversely related to the pipe diameter. Finally, the FMEA analysis based on the knowledge management technique demonstrated that pressure management in WDNs is the main policy for risk reduction of leakage and failure.

© 2023. The Author(s).

In plain English

This engineering study examined pipe bursts and leaks in zone B of Birjand's water network in Iran, where pipes are mainly polyethylene (PE) or asbestos cement (AC). It did not study asbestos exposure or human health. The authors analysed 2009 failure records, measured pressure at 14 points, fitted curves, built a prediction model (ANFIS) from 30 events, and had ten experts score risks. PE pipes failed distinctly more often than AC pipes, though the AC pipes were older. After the authors set an extreme rate in one rehabilitated 160 mm PE section to zero, the highest rate was in 32 mm PE pipes: 168 failures over 4,513 m. Failures rose sharply above 50 m of water pressure. Pipe diameter (P less than 0.0001) and pressure (P = 0.0117) were statistically significant factors; pipe type alone was not (P = 0.2015). The model's weighted overall R-squared was 0.94; a table gives testing correlation 0.83 and R-squared 0.68, though the text calls 0.83 the R-squared. The authors concluded that pressure management is the main strategy for reducing failure risk, and noted they lacked data on pipe age and slope.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Engineering case study of pipe failures in one zone of a city water network. It combined curve-fitting regression, ANOVA, an adaptive neuro-fuzzy inference system (ANFIS) prediction model, and an expert-scored Failure Mode and Effects Analysis (FMEA).
• Who was studied: Zone B of the Birjand water distribution network, Iran (33 nodes, 44 pipes). The authors analysed all failure events in 2009, used 30 events from a later dataset for the ANFIS model, and surveyed ten experts at the Water Management Company of Birjand City.
• Exposure: No human exposure was studied. The factors examined were pipe material (polyethylene versus asbestos cement), pipe diameter (32 to 350 mm in zone B) and water pressure.
• Main results:
  • In zone B, PE pipes had distinctly higher failure rates than AC pipes, even though the AC pipes were relatively older.
  • One area of 160 mm PE pipe made of low-quality manufactured pipeline had an extremely high failure rate. Because that pipeline had been rehabilitated, the authors set its rate to zero. After that adjustment, the highest failure rate was in 32 mm PE pipes: 168 failures over 4,513 m of pipe, a rate of 0.037. The text gives this rate per km per year, while the table gives it per metre per year.
  • Failure probability rose gently at lower pressures and sharply once pressure exceeded 50 m of water. The authors selected Fourier (R-squared 0.97) and Gaussian (R-squared 0.96) curves to describe failure versus pressure.
  • In the ANOVA, pipe diameter had a P value of less than 0.0001 and pressure had a P value of 0.0117. Pipe type on its own had a P value of 0.2015, so it was not statistically significant by itself. The authors ranked the factors as pipe diameter, then pressure, then pipe type.
  • The ANFIS model had a correlation coefficient of 0.99 in training. The authors computed a weighted overall R-squared of 0.94. For the testing data, Table 8 gives a multiple R (correlation) of 0.83 and an R-squared of 0.68, while the text describes 0.83 as the testing R-squared.
  • In the expert FMEA, exceeding the recommended pressure threshold of 55 m of water column had a risk priority number of 384. Improper operation of pressure relief valves scored 240, and bursting of PE pipes during network modifications scored 168.
  • Pipe age was not analysed quantitatively, but the authors report that new pipes had higher failure rates.
• The authors' conclusion: The authors concluded that pipe material, pipe diameter and pressure have the highest impact on failure rate, and that smaller PE pipes fail more than larger AC pipes. They also concluded that pressure management is the main strategy for reducing failure risk.
• Limitations the authors state:
  • The study lacked data on other aspects of the network, such as pipe age and slope.
  • The study had no dynamic system for real-time risk assessment of failure.
  • Pipe age was not quantitatively analysed.

Topics

Medical terms in this study, explained in our Lexicon: Asbestos-Cement Pipe.

Related studies

• Improving pipe failure predictions: Factors affecting pipe failure in drinking water networks — N.A. Barton, T.S. Farewell, et al. (2019)
• Combining a gravitational search algorithm, particle swarm optimization, and fuzzy rules to improve the classification performance of a feed-forward neural network — M.L. Huang, Y.C. Chou (2019)
• Evidence of arsenic release promoted by disinfection by-products within drinking-water distribution systems — S.S. Andra, K.C. Makris, et al. (2014)
• Stimulation of 2-methylisoborneol (MIB) production by actinomycetes after cyclic chlorination in drinking water distribution systems — M. Abbaszadegan, M. Yi, et al. (2015)
• Monitoring of airborne asbestos fibers in an urban ambient air of Shahryar City, Iran: levels, spatial distribution, seasonal variations, and health risk assessment — F. Taghizadeh, A.J. Jafari, et al. (2019)
• Monitoring of airborne asbestos fibers in an urban ambient air of Mashhad City, Iran: levels, spatial distribution and seasonal variations — A. Moteallemi, M. Minaei, et al. (2020)

Browse more research in Scientific Reports, or return to medical research.