Abstract
Mission Success Probability (MSP) is a critical assessment index for the mission-oriented maintenance and assurance capability of complex equipment. The traditional MSP calculation is based on numerical simulation, which is computationally intensive and inefficient, and assumes a fixed probability distribution for the failure parameters of each module. In actual engineering, the system component may experience life consumption or failure, its failure distribution parameters will change, and the replacement cycle varies. At the same time, the designer needs to adjust the component design parameters at any time during the equipment design to optimize the success rate of the mission. Consequently, it is a meaningful research topic to achieve rapid assessment of MSP in variable scenarios to support utilization and maintenance decisions. In this paper, an end-to-end neural network proxy model oriented to rapid assessment of mission success probability is proposed for series-parallel systems, the design and training methods of the model are described. By comparing it with the traditional simulation-based approach. The results show that the proposed proxy model can significantly improve the computational efficiency while having high accuracy, and can support the rapid evaluation of mission success probability under variable scenarios.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call