Number of Blockage Prediction for Sanitary Sewer Networks (Case Study: Isfahan Region 2)

Wastewater network is an inseparable part of urban life. Due to importance of this network as one of the urban infrastructure, the failure of this system will lead to stopping service, causing many social, economic and environmental consequences. Hence, assessing the wastewater networks condition and its failure is an important approach for managing it. Generally, failure of system means any condition which is lead to stopping service. In general, artificial intelligence methods are used as a low-cost method to predict failure. In this research, genetic programming (GP) is used to predict the number of blockage (hydraulic failure) in the wastewater network and its results are compared with the results of the artificial neural network (ANN). As a case study, here, a part of Isfahan wastewater network is investigated. The parameters such as age, pipe length, slope and depth as input data and the number of blockage are considered as the output data of the model. In this research, the number of blockage data in the wastewater network at 1394 and 1395 are used, in which the 70% of the data is used for training and 30% for the test. These data are classified in three way leading to three model. In the first model, data are classified based on the slope and in two other models the data are classified according to the cover depth. The results show that all models predicts the number of blockage with good accuracy. In addition the accuracy of the result of GP model is better than the ANN model. For example, for GP model, the values of R2 and RMSE for the second model at the training stage are 0.97 and 0.8 and at the test stage are equal to 0.94 and 0.69, respectively. However these values for ANN model are 0.96 and 0.95 at the training stage and 0.87 and 0.96 at the test stage respectively. These results show the superiority of the GP model in comparison with ANN model in which the results of second proposed model are better. The results of these proposed model can be used for preventive maintenance, prioritization of sewage network repairs and inspections, and finally to prevents the occurrence of suddenly accidents.