Long-Term Pipeline Failure Prediction Using Nonparametric Survival Analysis

Australian water infrastructure is more than a hundred years old, thus has begun to show its age through main failures. Our work concerns approximately half million pipelines across major cities that deliver houses and businesses, serving over five customers. Failures on these buried assets cause damage properties supply disruptions. We applied Machine Learning techniques find cost-effective solution the pipe failure problem in cities, where average 1500 of failures occur each year. To achieve this objective, we construct detailed picture understanding behaviour network by developing model assess predict likelihood breaking using historical records, descriptors pipes other environmental factors. results indicate our system incorporating nonparametric survival analysis technique called ‘Random Survival Forest’ outperforms several popular algorithms expert heuristics long-term prediction. In addition, statistical inference quantify uncertainty associated with predictions.