Using Differential Evolution with Neural Networks Forecasting Model Creating for Pipeline Corrosion

Abstract

Pipeline corrosion is among the most critical and precarious causes of pipeline incidents which is observed year after year. As these pipeline incidents give rise devastating harms to people as well as to the economy and ecosystem of a country. Monitoring this component, pipeline operators have installed a more systematic and comprehensive program for pipeline inspection by different sensors for the attainment of data that may be helpful to gauge the existing pipelines state. However, in this corrosive process different factors are involved which cause erosion, therefore, current inspection methods are not sufficiently particular in the measuring process. Hence, a prediction model, capable to measure precise corrosion damage mechanisms is required to develop. The most apposite method to be adopted for such model is Artificial Neural Networks (ANN). Among the existing works on ANN, a critical research has proved the requirement to develop time effectiveness of the technique. A hybrid prediction model is developed in this research which can measure particular corrosive mechanisms. An elementary ANN Model is enhanced by incorporating the Differential Evolution (DE) algorithm in order to acquire an improved and ideal performance. The obtained hybrid model will be tested with industrial dataset of world to approve its time effectiveness as compared to the elementary