A Hybrid Prediction Fault Location Model for Copper Wire Manufacturing Process

Abstract

This paper presents a novel prediction of fault location during copper wire manufacturing using a hybrid Nonlinear Autoregression Neural Network (NARNN) and Markov chain model. A four (4) year daily primary data spanning from 2018 to 2022 consisting of 261502 data points obtained from a cable manufacturing company in Ghana was used for the prediction. A comparison between the suggested model and decision tree algorithm was done. To assess the predictive effectiveness of the two models, performance indicators including Mean Absolute Deviation (MAD), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) were used. As determined by their evaluation criteria, the findings revealed that the suggested hybrid model had superior data fitting and accurate prediction capabilities.