The consequences of average curve generation: implications for biomechanics data.

One method of understanding the general mechanical response of a complex system such as a vehicle, a human surrogate, a bridge, a boat, a plane, etc., is to subject it to an input, such as an impact, and obtain the response time-histories. The responses can be accelerations, velocities, strains etc. In general, when experiments of this type are run the responses are contaminated by sample-to-sample variation, test-to-test variability, random noise, instrumentation noise, and noise from unknown sources. One common method of addressing the noise in the system to obtain the underlying response is to run multiple tests on different samples that represent the same system and add them together obtaining an average. This functionally reduces the random noise. However, if the fundamental response of each sample is not the same, then it is not altogether clear what the average represents. It may not capture the underlying physics. This paper evaluates the use of transducer time-histories for developing an underlying response when there is variation in the time-histories that is not due to random noise, but to a fundamental aspect of the response. Although the examples used are from NCAP tests, the analysis has direct application to the development of Anthropomorphic Test Devices (ATDs) when the underlying response to which the ATD is designed is obtained from impact tests on Post Mortem Human Surrogates.