An Artificial Neural Network meta model for availability of systems in series with buffers under imperfect repair

The series system reliability and availability are considerably deteriorated as the number of components increases. Redundancy are use to over come such problem. However in liquid supply systems (oil supply) and line production systems that afford to provide buffers in between components of the system could be considered as a competing more economical solution to the problem of drastic decrease of system reliability and availability of multi-components connected in series. Considering the provision of buffers, the rated capacity of components should be increased by a sufficient margin to be able to provide the buffers with quantities necessary and sufficient to keep the flow in the system uninterrupted in case of components failure. Modeling availability of such system is a challenging task especially if repair is considered imperfect. Usually simulation models are used for such model but it is slow when it comes to be used in optimal allocation for components reliability and maintainability. An Artificial Neural Network (ANN) is a promising alternative for such optimal problem due to fast output and capability of modeling non-linear model. In the proposed work the availability model of series system with in-proces buffer under imperfect repair is investigated using an Artificial Neural Network (ANN). The ANN model is verified and shows that it outperform the multiple regression analysis.