Reliability prediction model of a repairable system with random failure: Case study of a brick manufacturing company

• Main Authors: Chikuruwo, Mary N. H., Urombo, Jack, Manyangadze, Milton, Chikowore, Takawira R., Musiyarira, Harmony K.
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial neural networks; Breakdown; Bricks; Conveying; Downtime; Manufacturing; Neural networks; Prediction models; Predictive maintenance; Reliability; Strategy; Training
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0126395

This research results from the repeated occurrence of unanticipated breakdowns on a brick manufacturing company's screw conveying system components utilizing a breakdown maintenance strategy. This strategy usually results in either labour or materials being unavailable when required and further exacerbates the problem by increasing equipment downtime and leading to the ultimate unnecessary production delay. This aspect of unpredictability can also affect budgets and cost planning. It may eventually lead to a loss in revenue and, with time, a loss in the market share for the company. In the current system, the overall equipment effectiveness (OEE) is 66.9%. If the system component failures are to be predicted, a proper maintenance strategy can be put in place and system downtime reduced. Therefore, this research aimed to model the reliability prediction of a conveying system for a brick manufacturing company using artificial neural networks (ANN). In this research, the company database was used to extract data for training the ANN. The reliabilities of each component were obtained by using a Weibull calculator specifically designed for components failure history and validation of a neural network model. The tool used for training artificial neural networks (ANN) is MATLAB with an interfacing platform in a GUI. A simulation procedure verified that ANN does effectively predict how reliable the components in a system are.