Ydenius 1 ACCURACY OF FOLKSAM ELECTRONIC CRASH RECORDER (ECR) IN FRONTAL AND SIDE IMPACT CRASHES Anders Ydenius (1) Helena Stigson (1, 2) Anders Kullgren (1, 3) Cecilia Sunnevång (4 , 5) 1) Folksam Research, Stockholm, Sweden 2) Division of Insurance Medicine, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden 3) Department of Applied Mechanics, Vehicle Safety Division, Chalmers University of Technology, Göteborg, Sweden. 4) Autoliv Research, Vårgårda, Sweden 5) Division of Surgery, Department of Surgical and Perioperative Sciences, Umeå University, Umeå, Sweden Paper no.13-0368 ABSTRACT Estimation of crash severity from crash recorders is important in the evaluations of vehicle crashworthiness. The number of cars fitted with on- board crash recorders is increasing. The majority of these recorders are integrated with airbag sensors that usually have limitations regarding e.g. recording time and sampling rate. The aim with this study was to evaluate the accuracy of an Electronic Crash Recorder (ECR) compared to laboratory accelerometers. The ECR records car body acceleration during a crash event. The ECR is part of a large accident data collection system where 10,000 units per year are installed in various car models in Sweden. The ECR has the possibility to record acceleration in longitudinal and lateral impacts and also in multiple events. The ECR also meet requirements like recording data 30 ms prior to pulse start (t 0 ) and recording time up to 500 ms with 1 kHz sampling rate. The focus was to evaluate the accuracy in a wide range of impact speeds and with different pulse shapes. A series of 12 sled tests were conducted with delta-V between 12.3 and 73.5 km/h. In each test the sled was fitted with 10 ECRs as well as a laboratory accelerometer. Five ECRs were fitted in the longitudinal direction and five in the lateral. In total 120 ECR recordings were evaluated. Acceleration data were filtered according to CFC60 as defined in SAE J211. Change of velocity, mean and peak accelerations were derived from the filtered acceleration. No systematic error was found regarding delta-V. The systematic error of mean acceleration in the longitudinal direction was 0.4 g (3.5%). For all tests the standard deviation for delta-V in the longitudinal direction was 0.8 km/h (1.9%). The corresponding value for the lateral direction was 1.4 km/h (3.9%). The standard deviation for mean acceleration was 0.2 g (1.7%) in the longitudinal direction and 0.4 g (3.0%) in the lateral direction. In general no major differences in standard deviation between low and high speed crash tests were found. Overall the evaluation of the ECR showed that a low cost accelerometer device gives accuracy close to a laboratory accelerometer. INTRODUCTION Knowledge of crash severity is important in crash injury analyses. The link between injury outcome and crash severity is essential for both car manufacturers and road designers. Traditionally crash severity, often change of velocity, is calculated with energy based reconstruction software based on measurements of structural deformation of the car. Historically the most frequently used ones are CRASH3, SMASH and WinSmash, where WinSmash is the most recent one. The error of such reconstruction software has been shown to be large, with underestimations of delta-V up to 33% (Smith and Noga 1982; O'Neill et al. 1996; Lenard et al. 1998; Gabler et al. 2004; Niehoff and Gabler 2006) . Using crash recorders may have a profound impact on vehicle crashworthiness by providing delta-V, mean and peak acceleration to be used in crash reconstruction and analyses. This helps car manufacturers to improve automotive safety more effectively, but also to evaluate benefits of new safety technology. The most used severity parameter in crashworthiness analysis is delta-V. Mean acceleration is also used and shown to have a good correlation to injury risk (Ydenius 2009). Peak acceleration is not commonly used in crash