An Approach Based on Bayesian Networks for GO-FLOW Methodology

The GO-FLOW methodology is a reliability analysis methodology [1]. It is success-oriented system analysis technique and is capable of evaluating system reliability and availability of system with complex time-sequence problems and phased mission problems. The modeling technique produces a chart which consists of signal lines and operators, and represents engineering function of the components/subsystems/system. Moreover, it is able to model and analyze complex systems which contain multi-state under different functional scenario [2]. Now, some software and approaches are able to support the calculation of GO-FLOW methodology. However, the traditional modeling methods are no longer appropriate to calculate the complicated and time-sequence models. For example, (i) traditional approach initially assumes that the signals are independent to each other. However, the necessary consideration of multiple shared signals demands recalculation the model; (ii) if each states of operators are restricted to binary values, i.e. success and fail, the system that contains 100 operators may expand to 2100 status combinations. It would bring much pressure on the computing system; (iii) it needs to be calculated for each time interval based on the traditional probabilistic approach. Because of the restrictions above, the GO-FLOW methodology is kept from being widely used. Aiming at the restrictions, an approach based on Bayesian networks is proposed in this paper. The proposed method calculates the reliability of system by simultaneously considering: (i) time-sequence and multi-states problems; (ii) multiple shared signals; (iii) the efficiency and applicability of the approach. The rest of this paper is organized as follows: Section 1 presents the brief description of the Bayesian networks and GO-FLOW methodology. Section 2 gives a detail mapping rules from operators into Bayesian network (BN). Lastly, application case study is descripted in Section 3 demonstrates the effectiveness of the proposed method and Section 4 concludes the paper and discusses some future work.