METHODS FOR EVALUATING THE PREDICTIVE ACCURACY OF STRUCTURAL DYNAMIC MODELS

THE DEVELOPMENT OF METHODS FOR EVALUATING THE PREDICTIVE ACCURACY OF STRUCTURAL DYNAMIC MODELS IS PROPOSED. PARTICULAR APPLICATION WILL BE DIRECTED TOWARD LARGE SPACE STRUCTURES, SUCH AS THE NASA SPACE STATION, WHICH CANNOT BE FULLY TESTED IN A GROUND TEST LABORATORY. BOTH GROUND TESTING AND ON-ORBIT IDENTIFICATION OF AS-BUILT STRUCTURES WILL BE REQUIRED. THIS PROPOSAL ADDRESSES GROUND TESTING PROCEDURES WHICH ARE CRITICAL TO THE SUCCESS OF ON-ORBIT IDENTIFICATION. ALTERNATIVE APPROACHES FOR EVALUATING PREDICTIVE ACCURACY WILL BE INVESTIGATED, INCLUDING FIRST ORDER STATISTICAL METHODS, NUMERICAL SIMULATION METHODS AND FUZZY SET METHODS.THE FIRST ORDER STATISTICAL APPROACH IS OUTLINED. IT WILL BE APPLIED TO SPECIFIC NUMERICAL EXAMPLES IN PHASE I. THE NUMERICAL EXAMPLES WILL CONSIDER BOTH SUBSTRUCTURE AND SUBSCALE TESTING (USING SIMULATED DATA) OF A SINGLE-KEEL CONFIGURATION OF THE NASA SPACE STATION. PREDICTIVE ACCURACY WILL BE EVALUATED IN TERMS OF SELECTED AMPLITUDE AND PHASE CHARACTERISTICS OF COMPLEX FREQUENCY RESPONSE. THE PREDICTIVE ACCURACY BASED INDIVIDUALLY ON SUBSTRUCTURE AND SUBSCALE TESTING, AND A COMBINATION OF THE TWO WILL BE COMPARED. THE FEASIBILITY OF DEVELOPING A KNOWLEDGE BASED EXPERT SYSTEM TO ASSIST IN THE EVALUATION OF PREDICTIVE ACCURACY AND GUIDE ANALYTICAL AND EXPERIMENTAL PROCEDURES FOR IMPROVING PREDICTIVE ACCURACY WILL BE EXPLORED.