Predicting reliability via neural networks

The objective of this work is to predict the reliability of automotive components and systems from experimental failure data using artificial neural networks. To construct the necessary neural models, the Neural Simulation Tool (NEST), developed by Polytechnic of Milan, has been employed. An operative procedure based on the developed ANN models has been been implemented to predict the trend of the unreliability index R/sup 100/(t), the number of faults in 100 vehicles at time t (number of months from production time), starting from information on the number of vehicles produced and sold and the predicted number of faults up to the previous time t-1. The procedure has been applied on data from the Fiat Car Group, leading to satisfactory results.