Estimating Head Trauma Metrics in Pedestrian Collisions Using FrequencyDomainBased Deep Neural Architectures

Due to the growing number of vehicle–pedestrian collisions around the world, pedestrian safety has become a major concern in modern car design. The Head Injury Criterion (HIC) is a widely accepted measure for determining how bad head injuries are when people are hit by cars. In the past, physical headform impact tests done according to standard testing procedures were used to get HIC values. These experimental methods are dependable, but they take a lot of time, cost a lot of money, and aren't very flexible when it comes to testing different vehicle design options. So, there is a growing need for effective predictive methods that can accurately guess HIC values in the early stages of design without the need for a lot of physical testing. This research introduces a Frequency Response Function (FRF)-based Deep Neural Network (DNN) model for forecasting HIC through dynamic stiffness measurements. The study commences by examining the intrinsic relationship between dynamic stiffness properties and HIC values through a simplified steel plate framework. The analysis of the steel plate was done to see if frequency response and stiffness-related features could be used as predictive parameters. The initial results validated a quantifiable correlation between dynamic stiffness behaviour and HIC, suggesting that frequency-domain attributes may function as significant inputs for machine learning-driven injury prediction. After the initial validation, the suggested method was used to look at the hoods of two real cars. Real vehicle hoods are not like the simple steel plate. They have complicated shapes, support structures, and parts underneath them, like engines, brackets, and support frames. Experimental assessment indicated substantial disparities between the steel plate correlation and actual hood outcomes. The main reason for these differences was how the hood panel and the lower structural parts interacted during the impact. The hood's penetration into these underlying parts changes the effective stiffness that the impacting headform feels, which in turn changes the HIC values. To deal with these complicated physical problems, a Deep Neural Network model with many features was made. The model used dynamic stiffness parameters from Frequency Response Function analysis and other structural features to show how impact behaviour and injury outcomes are related in a nonlinear way. Also, a penetration estimation algorithm was made to figure out how much the hood bent into lower parts during impact events. We added this penetration depth feature to the neural network as an extra input parameter. This made the predictions much more accurate and cut down on the number of mistakes. The improved DNN model was much better at learning about complicated structural interactions and predicting HIC values than traditional correlation-based methods. The proposed framework combines frequency-domain mechanical analysis with cutting-edge AI methods to create a hybrid approach that fills the gap between experimental testing and data-driven modelling. The findings of this study underscore the efficacy of FRF-based Deep Neural Networks as valuable supplementary instruments in vehicle design and pedestrian safety evaluation. The model makes it possible to quickly test changes to the hood, materials, and structure without having to do headform impact tests over and over again. An AI-assisted predictive system like this can help automotive engineers make better hood structures to protect pedestrians better in the early stages of design. This will lead to safer vehicle development and lower testing costs. In summary, this study shows that combining dynamic stiffness analysis, penetration estimation algorithms, and deep learning methods is a promising way to accurately and intelligently predict the Head Injury Criterion in pedestrian crash situations