Evaluating and Predicting Overall Equipment Effectiveness for Deep Water Disposal Pump using ANN-GA Analysis Approach

Abstract

This study proposes the Artificial Neural Network with a Genetic Algorithm analysis approach to investigate the Overall Equipment Effectiveness of the deep-water disposal pump system. The ANN-GA model was developed based on six big losses over eighteen successive months of the operating period to evaluate the current and future performance of the DWD system. 70% of the data was used for training and 15% for each data validation and testing. The DWD system faces frequent failure issues, significantly impacting its performance, so it is important to reveal the main causes of these failures to manage them properly. ANN-GA is applied to make a linear trend prediction and assesses the confidence and accuracy of the results obtained. Analysis of ANOVA (variance) was adopted as an additional decision tool for detecting the variation of process parameters. ANN-GA results showed that the current OEE value ranges between 29% to 54%, whereas the predicted future system performance average is approximately 49%, which reflects the poor performance of the DWD pump system in the future compared to the world-class target (85%). ANN-GA analysis results indicated were very close and matched with the actual values. The model framework and analysis presented are used to develop a decision support tool for managers for early intervention to minimize system deterioration, reduce maintenance costs and increase productivity. Furthermore, it allows early identifying the potential area of improvement to support continuous improvement (CI) objectives by identifying and eliminating unnecessary maintenance activities. The proposed model framework uses the ANN approach to identify the current state and predict the future of the system performance to ensure confidence in the results. The contribution of the paper will be helpful for experts like managers, reliability engineers, and maintenance engineers to identify the state of the system's performance in advance.