Abstract

Due to the poor conditions of wastewater networks, there is an increasing need in the capital investments allocated for enhancing their condition. As per the Canadian Infrastructures Report Card, one third of the total lengths of sewer pipes in Canada is in fair to very poor condition (Canadian Infrastructures Report Card, 2016). As such, there is an urgent need for inspection planning tools, with which decision makers could assess the condition of pipelines and identify pipes with higher risk of failure. These tools are potentially of service in prioritizing and optimizing inspection activities that lead to decisions regarding appropriate courses of action, especially in cases of limited resources and funding.
The goal of this research is to develop an optimization model for scheduling the inspection of sewer pipelines by performing defect-based risk assessment. The risk of failure is determined to identify critical pipe sections; by combining likelihood and consequence of failure values using
the Sugeno Fuzzy Inference System. The developed optimization model determines the inspection sequence of pipeline sections in addition to optimizing the utilization of inspection crews by minimizing both time and cost of inspections. The risk assessment model is divided into two sub
models: likelihood and consequences of failure. Structural and operational defects and pipeline characteristics in an existing sewage network are used to develop the likelihood model that
determines the structural, operational and overall condition ratings of pipelines.
Method-wise, Bayesian Belief Network (BBN) is used to develop a static condition assessment model using probabilities of occurrences and conditional probabilities. Moreover, time dimension is introduced to the developed BBN model using logistic regression as temporal links
which are required to convert BBN into Dynamic Bayesian Network (DBN). The accuracy of the model’s prediction is examined through referencing of actual data, where the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for the BBN model are 0.67, 1.06, 0.56 and 1.05, 1.60, 0.95 for structural, operational and overall conditions, respectively. The second sub-model representing the consequences of failure is developed to determine the impact of sewer pipelines’ failure using Cost Benefit Analysis (CBA). Developing this sub model involves identifying and analyzing costs of failure and benefits resulting from avoiding such failures. In order to validate the CBA model, actual costs from a real failure incident are compared
with the proposed model's outputs. During the implementation of the CBA model, it is found that the indirect costs resulting from sewer pipelines’ failure represent a significant portion of the total failure costs.
The proposed risk assessment model is validated using actual data derived from inspected sewer pipelines. Cost savings of around 67% could be achieved if the risk assessment model is applied and deployed over ongoing inspection practices followed by municipalities. A Mixed Integer Linear Programming (MILP) model is developed to optimize scheduling of inspection activities by including sewer sections, time and cost of inspections. This model is developed using GAMS and solved using CPLEX to maximize the number of sections and minimize time and cost.
The output from the MILP model is compared to the results of another model solved using the Genetic Algorithm (GA) approach. It is found that the MILP model could perform better than the GA model in terms of optimal solutions. Additionally, a resulting inspection cost reduction of approximately 38% could be achieved when utilizing the MILP model. It is expected that the proposed inspection scheduling model could help decision makers
better assess the condition of sewer pipelines and improve their decision-making on proactive or reactive measures. The proposed model could help allocate budgets more efficiently in addition, to being an enabler for better inspection programs, particularly in cases of limited funds and task forces.