Metodologias para determinação de periodicidade ótima de manutenção preventiva sob a suposição de reparo imperfeito em manutenção corretiva

Abstract

Reliability has played a key role in system development and consequently in increasing competition company. Thus, the appropriate choice of a model for a repairable system is critical to reducing expenses and related risks to failures. In general, an optimal maintenance policy is sought to reduce total cost maintenance. This work presents alternative methodologies for determination of optimal periodicity of Preventive Maintenance – PM under the assumption of Repair Imperfect – IR in Corrective Maintenance – CM. When a case study has more than a system under study, estimate the number of failures up to a certain downtime 𝑇 directly depends on the maintenance policy chosen. In addition, to estimate the expected time where a failure occurs can be done in different ways and generate different results too. In this work we use the model 𝐴𝑅𝐴1 – Arithmetic Reduction of Age memory 1, in addition to statistical methods such as maximum likelihood estimation and Monte Carlo simulationHere we propose three alternatives methodologies for obtaining the optimal periodicity: the first objective is to determine the point estimation by the cost function and the interval estimation by the source of variability of the Monte Carlo Simulation; the second aims to estimate the average function by the method proposed by Jack (1997), adapting it to 𝐴𝑅𝐴1; and the third methodology objective an estimation of failure times using the recursively estimated Mean Cumulative Function – MCF. Λ̂(𝑡). The practical situation studied in Toledo et al. (2016) is revisited, and the results obtained compared and analyzed. The proposed methodologies demonstrated alternative analysis situations regarding the behavior of the systems under study, directly impacting the decision making for choosing the optimal maintenance policy. The estimation of the optimal periodicity time as well as the determination of the expected failure times proved to be more consistent with the 𝐴𝑅𝐴1 arithmetic reduction model.