Development of smart searching algorithms for vulnerability and uncertainty analyses in probabilistic risk assessments
In order to evaluate the risk inherent in a complex system, including a reasonable accounting of the many uncertainties that characterize both the environments to which the system is exposed and the responses of the system to these environments, risk analysts are forced to use statistical approaches. If the system is very well designed for safety, then simple convolution or sampling approaches may have to be augmented by intelligent searching schemes. An excellent example is the prediction of the probability of an accidental nuclear detonation for a nuclear weapon system. Under practically all situations, accidental nuclear detonation is virtually impossible because modern nuclear weapon systems are designed to preclude it. They have a series of strong links and weak links that are guaranteed to fail in a prescribed order when exposed to virtually all credible abnormal environments, such as fires, impacts, punctures, crushes, external pressures, lightning or chemical attack. However, no engineered system is perfect, and under certain peculiar conditions involving combined environments that are spatially directs in the worst possible way, the strong links may fail before the weak links and an accidental nuclear detonation may conditions through an intelligent searching process and to determine whether their probability of occurrence is high enough to be of concern. In the weapon system application, LHS generates millions of sample members, each consisting of a different set of parameter values, and we must employ intelligent searching techniques to reduce this set to a more workable number. We do this through a discriminator subprocess, which we will describe.
- Research Organization:
- Sandia National Labs., Albuquerque, NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC04-94AL85000; AC04-76DP00789
- OSTI ID:
- 10103549
- Report Number(s):
- SAND-93-1236C; CONF-940312-33; ON: DE94001825; BR: GB0103012
- Resource Relation:
- Conference: 2. Probabilistic safety assessment and management conference (PSAM),San Diego, CA (United States),20-24 Mar 1994; Other Information: PBD: [1993]
- Country of Publication:
- United States
- Language:
- English
Similar Records
Development of fast-running thermal and structural response models for probabilistic analysis of complex systems
Vulnerability assessment of a space based weapon platform electronics system exposed to a thermonuclear weapon detonation