An Analytical Study to Measure the Performance 0f Reliability Centered Maintenance Management Using Artificial Neural Network Models: Case Study in Food Industries Company

Effective and successful maintenance management is a key partner in the success of the industrial facilities. This work aims to conduct a thorough assessment of different neural network models in order to determine the most efficient model for classifying machines status and determining their needs as a key part in evaluating the performance of maintenance management. Reliability-Centered Maintenance (RCM) is a system for collecting and analyzing data and identifying critical machines by measuring performance indicators such as mean time between failures (MTBF), mean time to repair (MTTR), failure rate (λ), reliability R (t), maintainability M (t), and availability (Ai). By using collected data on machine performance indicators, ANN models are developed using a feed-forward structure and supervised learning using the Levenberg-Marquardt backpropagation algorithm and sigmoid activation function. Experiments were done by changing the number of neurons inside the hidden layer. The output of the ANN is machine status. The experimental and expected values were very close and had a high correlation coefficient. Through the experiments conducted on the selected ANN models, it was found that the Elman backprop model, which includes 25 neurons within the hidden layer, is the best, as 185 epochs were implemented and the best performance was reached at the mean square error (7.9872×10-30) at epoch 162. By using the suggested model, good results were achieved with improved availability by 76%, reliability by 57%, and maintainability by 71%, and a failure rate reduction of 14% for the selected machines.