Performance prediction of a production line with variability based on grey model artificial neural network

Performance prediction of a production line is an important part of production process control. In this study, a series-parallel and multi-product hybrid production line with variability is investigated. The performance is determined according to product type, rework ratio, bottleneck rate, batch number and equipment random failure have properties of multi-variable, poor information, nonlinear, strong coupling, long time-delay, high order, etc. The analysis showed that variability causes its performance prediction results to be not accurate, even may lead to erroneous results. Treating the line as a “black box”, the prediction model is established by combining grey model (GM) with artificial neural network (ANN), to make most use of the advantages of the ability of a small amount available data mining and self-learning. Numerical simulation experiments are performed to demonstrate our theoretical analysis, and simulation analysis shows that results of performance prediction are satisfactory. Furthermore, the strength of the proposed method is illustrated by comparison with the common ANN in the same scenarios. Finally, conclusions and future research directions are discussed.