Original Date: 01/27/1997
Revision Date: 01/18/2007

Lawrence Livermore National Laboratory (LLNL) uses its expertise in Decision Sciences and Systems Research to solve or prevent a customer’s problem. By defining and structuring the customer’s requirements, LLNL can determine the correct blend of personnel and skills for various situations. The Laboratory’s expertise includes reliability, availability, and maintainability; decision analysis; probability; statistics; simulation; optimization; economic modeling; and database design.

The LLNL Decision Sciences and Systems Research expertise has been applied to the National Ignition Facility by developing a reliability and availability allocation model which used Bayesian techniques to allocate goals and quantify uncertainties in the early project phase. The model identifies sources of relative availability and reliability concerns; assigns preliminary subsystem goals to achieve overall goals; examines implications and insights from allocation; and iterates and updates the process. Availability concerns are delays which cause missed shot opportunities. Shot reliability concerns are ruined or failed shots.

The allocation process begins with a formal query of experts with relevant knowledge in frequency of problem occurrence, length of delay, and frequency of ruined shots. LLNL spot-checks the information using Nova and Beamlet data. The allocation calculation reflects the relative performance of subsystems (Table 3-1). The allocations are evaluated relative to their achievability. Uncertain achievability requires additional data to fill in the gaps. Cost and schedule concerns are given formal consideration. The experts are rechecked on interpretation of the data and implications and insights are examined, such as significant sources of missed and failed shots; initial performance specifications for designers and vendors; and opportunities for improved design or repair policy. By using the reliability and availability allocation model, LLNL can budget the overall availability and reliability goals of subsystems and components to identify the initial design specifications, achieve the required performance, and identify performance concerns and areas for improvement.

The LLNL System Sciences have been applied to the alternative strategies for long-term management or use of depleted uranium hexafluoride, stored in the cylinder yards in Paducah, Kentucky; Portsmouth, Ohio; and Oak Ridge, Tennessee. By developing a computer cost model, the system was able to integrate the cost estimates for the individual options of each alternative strategy and then provide the cost of the strategies.

The LLNL System Sciences has also developed statistical sampling methods for compliance auditing. Originally, LLNL’s property management used a wall-to-wall inventory (50,000 items) method which was quite costly. The Laboratory began statistical sampling in 1995 and developed a sophisticated low-cost sampling design in 1996. For 71 property center representatives with varying numbers of assigned property items, a statistical sampling was set up to sample a portion of the representatives and then a portion of the items under the stewardship of the representative. This design minimized the total cost while achieving a specified accuracy. Investment in this sampling design reduced LLNL’s cost from 4,945 hours to 110 hours.

LLNL’s Decision Sciences and Systems Research expertise serves a broad range of programs and technologies. Applications include safety analysis; vulnerability; surety; and technical risk analysis.

For more information see the Point of Contact for this survey.