Forecasting reliability using non-linear autoregressive external input (NARX) neural network

Abstract

This article is concerned with the study of reliability assessment of the urea plant. A non-linear autoregressive external input (NARX) neural network was developed to predict the reliability using the failure frequency and all possible combinations of the equipment in the plant. The model was fed using field data; ten hidden layers and three subsets for the training were used: 50%, 70% and 85%. The NARX was capable of simulating the system and demonstrates a good performance comparable with the traditional reliability method and real data.