Correlation Grading Methodology for Occupant Protection System Models 2004-01-1631

Computer modeling and simulation have become one of the primary methods for development and design of automobile occupant protection systems (OPS). To ensure the accuracy and reliability of a math-based OPS design, the correlation quality assessment of mathematical models is essential for program success. In a typical industrial approach, correlation quality is assessed by comparing chart characteristics and scored based on an engineer's modeling experience and judgment. However, due to the complexity of the OPS models and their responses, a systematic approach is needed for accuracy and consistency. In this paper, a correlation grading methodology for the OPS models is presented. The grading system evaluates a wide spectrum of a computer model's performances, including kinematics, dynamic responses, and dummy injury measurements. Statistical analysis is utilized to compare the time histories of the tested and simulated dynamic responses. The statistical quantity measurements include the average residual, standard deviation, correlation coefficient, and 0th to 2nd moment relative differences. The correlation quality of overall kinematics and dynamic responses is scored and color-coded from weak, marginal, adequate, good to excellent. The grading system can clearly distinguish the correlation quality for different models. The evaluation of a side curtain airbag component model and a frontal impact system model is presented as examples to demonstrate the applications of the correlation grading system.