Prediction of the number of failures of repairable elements aimed at homogenizers in the dairy sector with the Monte Carlo method

Abstract

An attempt is made to determine a reliability indicator for repairable assets, performing a probabilistic analysis on repairable elements, homogenizers in the dairy sector. Various failure times were analyzed in recent years. A goodness-of-fit test was performed to determine the distribution that best fits the behavior of the data, an important step to obtain greater certainty in predictive reliability. In the case study, the Kolmogorov-Smirnov test was used. Then, the operating time histories and the number of failures that the asset had in that period were collected, the distribution that best aligned with the data was the Weibull distribution.

For the prediction of the calculation of the number of failures, the Monte Carlo method was used, random values were assigned to the function of the accumulated number of failures F(t). The values of the times obtained were compared with the time tm (mission time), the values that exceeded the mission time occur when the equipment does not fail. Therefore, as many predictions as history running times must be made, the times that did not exceed the mission time were added, thus determining the number of failures; Performing several data runs, the average was calculated, to obtain a better statistical result.