Predicting Capacity of Defected Pipe Under Bending Moment with Data-Driven Model

Water mains which suffered from corrosive environment and various loads/effects. This leads to the simultaneous occurrence of the decrease of the pipe capacity and the appearance of significant bending moments within the pipe. Various investigation in literature focused on the stress of defected pipe due to burst pressure for oil and gas pipe under the high internal pressure, rather than the bended water mains. Primitive studies on this problem are at the observation step without providing an applicable model for practice. Since finding analytical solution for corrosion pipe is a challenging task because of the localized of the defects, the Finite Element Analysis (FEA) is an effective alternative. The critical drawback of FEA is that the problem needs to analysis separately with computational expense and required highly skilled experts. To ease these difficulties, a data-driven model is developed based on the database generated from FEAs and labeled by their results. The application of machine learning techniques improves the accuracy of the conventionally statistical regression models such as linear regression due to the flexible of the model and thus a trained data-driven model is a practical approach to solve the problem. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.