Application of monte carlo method for failure prediction: a tool to support maintenance management

• Autor(a) principal: de Oliveira, Lucas Guedes
• Data de Publicação: 2019
• Outros Autores: Paiva, Emerson José de, Paiva, Anderson Paulo de
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Produção Online
• Texto Completo: https://www.producaoonline.org.br/rpo/article/view/3091
• topic: Maintenance Engineering. Monte Carlo Simulation. Failures Prediction. Applied Statistics.; Engenharia de Manutenção. Simulação de Monte Carlo. Previsão de Falhas. Estatística Aplicada.

The maintenance field has undergone important changes in the last decades, oriented, mainly, by the evolution of the managerial concepts in the companies. Although it has been treated for long periods as an onerous sector for organizations, maintenance is pointed out in the latest literature as a huge source of competitiveness. To explore its potential, however, Maintenance Management must incorporate Engineering as the driving force for routine processes and improvements. In a practical way, the company must work to avoid failures or, at least, to foresee them. In line with this strategic vision, the present work engages in the development and validation of a failure prediction system. Using the Monte Carlo Method, this article integrates a quantitative study of modeling and simulation. From mathematical and statistical concepts, different failure prediction series were formulated and comparative analyses were performed on their precisions. As results, it was verified the effectiveness of the method in determining the moment of occurrence of failures from numerical simulations and evidenced the optimal regions of prediction of each proposed series. Among the main contributions of the study, we highlight the higher precision of the series simulated by the Monte Carlo Method in relation to the series estimated from the historical average of the data, despite the good adjustment of these series in selected areas of the real curve. Future works will investigate the behavior of other models of a series of failures, generated from new combinations of the proposed parameters.