The Application of Neural Network for Predicting Corrotion Rate in Metal Pipe Installation

Abstract

Corrotion is one of the problems that must be considered in the metal pipe installation because it can disturb the operation of the plant. The possibility of the corrotion occurrence can be predicted using neural network system. The black box system in the neural network can be used to calculate several potential causes the corrotion and to predict the corrotion rate. This study had constructed the prediction system of corrotion rate using neural network. The input of the system are material compositions, pH, flow rate and temperature. The material compositions which are used are Carbon (C), Manganese (Mn), Silicon (Si), Phosphorus (P), Sulphur (S), Chromium (Cr), Molybdenum (Mo), Aluminium (Al), Nickel (Ni) and Iron (Fe). The corrotion rate prediction network is using one hidden layer and lavenberg marquardt for the learning algorithm. The Mean Square Error (MSE) which is used to analyze the network performance indicates that both of training and validation show excellence results. The MSE of training is 0,000338971 and the validation is 0,000493117.