A New Approach in Improving Traffic Accident Injury Prediction Accuracy

Abstract

This paper focused on the effect of intrusion magnitude and maximum deformation location in improving the accuracy of Injury Severity Prediction (ISP) for Advanced Automatic Crash Notification (AACN) system. This study used 545-passenger vehicles involved in Car-to-Car side impact data from NASS CDS (CY: 2004-2014). Variables mentioned in Kononen’s 2011 ISP algorithm are considered as base model. In addition to Kononen’s variables, magnitude of intrusion and maximum deformation location are added in the proposed model. As the location of maximum deformation moves away from the B pillar to end regions (front or back), the percentage of serious injury reduces drastically. Similar trend is verified in both accident analysis and FE numerical simulation results. Addition of intrusion magnitude and location of maximum deformation as additional injury predictors helped to improve the proposed model sensitivity, overall accuracy by 16%, 3.12% respectively without any change in specificity value.