FPGA-Based Stochastic Activity Networks for Online Reliability Monitoring

Stochastic activity network (SAN) is a flexible formalism that permits performing reliability and prognostics analysis of complex degrading systems. However, their use for the analysis of high-reliability equipment including fast evolving and degrading mechanisms requires long simulation times and large hardware resources. This paper presents, a field programmable gate array (FPGA)-based architecture that permits performing online Monte Carlo simulations of SAN models of high reliability systems. The synthesis is automated and the result preserves the structure of the models, permitting model inspection, and validation. The architecture runs without the aid of a central processing unit (CPU) and the simulation results are directly recorded into CPU-accessible random access memory (RAM). The design is validated on reliability monitors for an overheating detector and for degrading underground cables. The resource usage and simulation time are compared to the performance of software solutions that are run on parallel processors and large computer resources. Obtained results confirm that the proposed FPGA architecture can be employed for the accelerated runtime reliability analysis of equipment.