THE world today is using increasingly complex systemsand taking them for granted. For the scientists and engineers who design and maintain them, two classes of systems present particular challenges: those that operate under extremely stressful conditions and those that are required to operate for extremely long periods of time. Jet aircraft, spacecraft, satellites, and nuclear reactors encounter enormous stresses. Aircraft and spacecraft also often continue to operate for longer than their designed lifetime. We at Lawrence Livermore have been mandated under the Department of Energy's Stockpile Stewardship Program to extend the life of weapons in our nuclear arsenal well beyond the lifetime for which they were designed. We are also collaborating on the design of the Yucca Mountain nuclear waste repository, which may be required to operate for 10,000 years.
For all of these systems, the penalty for failure is high. Injury, death, even major devastation of our planet could result from errors made during design, construction, operation, or maintenance. Yet to a certain extent, these extraordinarily complex and long-lived systems are beyond testing because we cannot be certain whether we have tested them well enough or long enough. Nevertheless, at Livermore, we subject weapon parts and components to rigorous conditions for extended periods before the weapon is assembled and later during its life in the stockpile. We understand that the stakes are high and that high standards are essential during all phases of the system's life.
So how do we, as responsible stewards of our nuclear arsenal and nuclear waste, respond to these challenges? For complex systems with long lifetimes, our approach comprises four integrated elements: prediction, testing, surveillance, and modification.
As described in the article, "A Better Picture of Aging Materials," we are striving to develop improved behavioral and computational models of system performance and effective life expectancy using the best scientific and engineering knowledge and the world's fastest supercomputers. Only with the powerful computers available to us through the DOE's Accelerated Strategic Computing Initiative can we be confident in the predictions for complex systems.
Experimental validation is typically performed initially to determine material properties, again to operationally test components and subassemblies, and finally to test the performance of an entire subsystem or system. But for nuclear weapons, we cannot perform the third phase of testing in its entirety. Computational modeling must of necessity replace full performance testing, previously accomplished through nuclear experiments. This is where the evolution of validated age-aware material models will become key to assessing the life performance of complex systems.
Surveillance incorporates an array of diagnostic techniques for monitoring the nuclear stockpile or the nuclear repository Because we no longer monitor performance with nuclear testing, our surveillance of the many materials that make up the system has taken on new importance.
Data from surveillance and all material and operational tests are essential for validating our computational models and identifying where models need to be changed and limited resources should be focused. Finally, the predictive models notify us when modifications are needed in a weapon system before its performance is reduced.
This integrated approach to complex systems is in itself complex. It involves many disciplines and can be handled effectively only by an organization with the depth and breadth of experience of a DOE laboratory such as Lawrence Livermore.

Back to September 1999