PROBABILISTIC STRUCTURAL MECHANICS RESEARCH FOR PARALLEL PROCESSING COMPUTERS

THIS RESEARCH EFFORT WILL INVESTIGATE THE USE OF PARALLEL PROCESSING COMPUTERS TO SOLVE PROBLEMS IN PROBABILISTIC STRUCTURAL MECHANICS (PSM) THAT ARISE IN THE ASSESSMENT OF RELIABILITY OF COMPLEX AEROSPACE STRUCTURES. PSM PROBLEMS, SOLVED BY MONTE CARLO SIMULATION, ARE INHERENTLY PARALLEL AND ARE IDEALLY SUITED FOR SOLUTION ON PARALLEL COMPUTERS. RELATIVELY LITTLE RESEARCH, HOWEVER, HAS BEEN CONDUCTED IN PROBABILISTIC STRUCTURAL MECHANICS TO EXPLOIT THE POWER OF THESE MACHINES. WE WILL INVESTIGATE NEW AND INNOVATIVE APPROACHES FOR IMPLEMENTING MONTE CARLO METHODS ON A PARALLEL COMPUTER FOR THE SOLUTION OF PSM PROBLEMS. THE OBJECTIVES OF THE PHASE I EFFORT WILL FOCUS ON THE EVALUATION, IN A PARALLEL COMPUTER ARCHITECTURE, OF PROBABILISTIC MONTE CARLO IMPLEMENTATION STRATEGIES, INCLUDING PIPELINING (VECTORIZATION), CONCURRENCY (MULTITASKING), SYNCHRONIZATION, AND VARIANCE REDUCTION TECHNIQUES. WHILE MUCH OF THE RESEARCH IN PSM OVER THE LASTTWO DECADES HAS FOCUSED ON BASIC THEORY DEVELOPMENT, THE NEXT DECADE OF RESEARCH IN COMPUTATIONAL PSM MAY OPEN UP A WHOLE NEW CLASS OF FINITE ELEMENT AND DYNAMICS PROBLEMS TO PROBABILISTIC STRUCTURAL ANALYSIS.