Portal de Periódicos da CAPES

Reliability assessment of passive systems using artificial neural network based response surface methodology

2020; Elsevier BV; Volume: 144; Linguagem: Inglês

10.1016/j.anucene.2020.107487

1873-2100

R.B. Solanki, Harshavardhan D. Kulkarni, Suneet Singh, P. V. Varde, Ajit Kumar Verma,

Fault Detection and Control Systems

Many advanced Nuclear Reactor designs deploys passive systems for enhancing safety Passive systems rely on natural driving forces, such as natural circulation, gravity, internal stored energy etc., which do not require external power sources. The driving forces are weak, hence, the phenomenological failures becomes equally important as compared to mechanical failures for passive systems. Substantial efforts are underway towards improving reliability assessment methods for passive systems, however, consensus is not yet reached. Simplified and subjective assumptions made for generating the Response Surface may not necessarily be true for the complex systems. This paper describes ANN Based Response Surface (ANNBRS) methodology developed for reliability assessment of passive systems. The methodology utilizes Principle Component Analysis (PCA) to reduce dimensionality of system and optimizing efforts in designing training data for ANN model. The ANNBRS methodology is applied to Isolation Condenser System (ICS) to demonstrate improvements over the existing Response Surface based approaches.