ANN and ANFIS Modeling of Failure Trend Analysis in Urban Water Distribution Network

Abstract

Pipeline leakage is one of the crucial problems affecting urban water distribution system from both environmental and economical point of view. Unfortunately, necessary large databases are not maintained in India for proper replacement of pipes. In this situation, this research purports at using two artificial intelligence techniques such as artificial neural network (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) to access the present condition and to predict the future trend of pipeline network of Peroorkada zone in Trivandrum city, Kerala, where a huge amount is spent every year for leakage rectification. Using different influential input variables, four models (all diameter pipes) have been developed. Also, the effect of each parameter (length, age, and diameter) along with previous year failures and previous year failures alone to current year failures are analyzed. Another two models of selective pipe diameter for considering the influence of prefailures up to last year alone and prefailures up to last year along with length to current year failures are constructed. Prioritizing the pipeline replacement is done for mains having 400 mm diameter and above since network details pertaining to those diameters are available. The performance of the models is evaluated using coefficient of correlation and mean absolute error and is compared to multiple linear regression (MLR) models. Three of them perform well and almost in kind, even though ANN is slightly having an upper hand. The applicability and usefulness of ANN and ANFIS will surely become beneficial for the authorities to take decisions regarding the replacement of pipes and this can in turn increase the efficiency of pipes.