چكيده به لاتين
The water network is one of the most important water organizations' assets. This importance includes various economic and social dimensions. Therefore, having correct and optimal planning for repairs and maintenance of the water network can effectively reduce costs and increase the water distribution quality. Optimal planning requires complete knowledge of the water distribution network and efficient management tools that, in addition to presenting the current situation, can predict the water distribution network's future behaviour. In recent years, the use of artificial intelligence methods to solve complex problems has expanded, and the validity of these methods has been proven. Due to the complexity of the water distribution network and its behaviour, machine learning methods will have very efficient results. In this research, the fuzzy clustering method and neural fuzzy inference system model have been used to predict water distribution pipes' failure rate. To increase the accuracy of fault rate prediction, the data is divided into six clusters using the fuzzy mean clustering method based on the status index, age, diameter and length. Cluster information is used to predict the failure rate, and the information is divided into three categories: training, experimental and verification. The results of statistical indices RMSE = 0.025, SD = 0.025 and R2 = 0.95 in the experimental category confirm the validity of the model. Also, considering that more than 90% of the network pipes are not damaged, this information causes excessive training of the model. Deleting this information solves the problem, but the developed model will only predict the failure rate of pipes with a history of failure. Therefore, it is suggested that some information about the pipes be preserved in the training group without any damage.
The water distribution network of Montreal's city, Canada, has been used to develop the model and validation. The information used includes diameter, year of installation, pipe length, material, failure statistics, environmental conditions and operating conditions.
