Bayesian Regularizers of Artificial Neural Networks applied to the reliability forecast of internal combustion machines in the short-term

Predictive as well preventive maintenance are tools of programs that aim to increase or maintain the life expectancy an equipment through computational techniques and tools. Bearing in mind power generation industry has a high rate with machines / electric generators stopped, this research aims develop model for predicting Reliability Key Performance Indicator (KPI) identify how available will be time span 22 days, methodology used based on analyzes tests artificial neural network (ANN) architectures using Bayesian Regularizers training algorithm, alternating transfer functions layers hidden find best state convergence minimum Root Mean Square Error (RMSE) value calculated between real simulated outputs. According results obtained by training, validation test steps, algorithm presented RMSE 0.0000104202 99.9% correlation values, thus is able which machine have greatest efficiency less within defined span.