Optimization of vehicle front end geometry for adult and child pedestrian protection

Abstract

Motor Vehicle Crash statistics globally indicate that pedestrians make up the second largest category of fatalities after vehicle occupants. Pedestrian kinematics during the crash event with the vehicle has been shown to significantly affect the injury mechanism contributing to severe injuries to the head in particular. The paediatric population stands at significantly higher risk of sustaining heavy casualties during pedestrian vehicle impact compared to adults as they face an additional fatal risk of the vehicle running over them following the initial impact. This work aims to achieve an optimized vehicle front end profile which caters for improved protection for both adult and child pedestrian groups and simultaneously avoiding the Run-over scenario. A hybrid vehicle front end model is developed, and subjected to extensive validation. It is found that despite the simplified structure, the model's deformability provides excellent kinematic accuracy to better capture vehicle-impact, pedestrian fall patterns and corresponding injury values. The hybrid case model records similar impact locations with the verification models for the three tested speeds of 40,32 and 25 km/h. The HIC values for these three cases showed an error margin of + 200.The fall pattern of the case model closely conforms to the PHMS verification model. The model offers the distinct advantages of relatively fast processing speed as well as ease of modifications due to its simple profile, which satisfy the criteria necessary for a multi-parameter optimization study. The processing speed is reduced approximately by 705 times achieving 99.85% efficiency in CPU time in comparison to a full FE vehicle model using the similar processing capacity. Design of Experiments (DoE) using the Central Composite Design (CCD) is initially utilized in which, a total of 100 computational runs are generated. The Head Injury Criteria (HIC) results from the simulations are tabulated as the response functions. Polynomial Response Surface Method (RSM) is used to generate mathematical models. Thereafter, the Latin Hypercube Sampling (LHS) design is used with 80computational runs and the mathematical models are generated using the Radial Basis Function (RBF). A comparison is made between the CCD-RSM models and the LHS-RBF models. The CCD-RSM models fitness is at 82.66% with a RMSE of 0.058 and the LHS-RBF has a fitness of 99.91% and a RMSE of 0.044. This clearly indicates that the LHS-RBF pair is best suited for optimization work. Optimization is performed using Genetic Algorithm. Unconstrained optimization is carried out separately for adults and for 6 year old child. A combined Adult-Child optimization is carried out as well. The individual adult optimized design and the child optimized design are shown to be not mutually applicable to each other i.e., HIC for Adult-Opt is 115.09 and using the similar optimized vehicle for the child records a HIC of 1797.4. The combined optimized profile however indicates high probability of Run-over scenario occurring for the child pedestrian, which invalidates the design. Thus, the Run-over occurrences from the DoE data are mapped using Logistic Regression and the resultant mathematical model is introduced as a constraint for the combined optimization. The final optimized model is shown to achieve a safe vehicle front-end profile with Combined-opt showing an observed HIC of 181.92, and Adult and Child-opt each respectively record a HIC of 209.34 and 195.47 successfully addressing both adult and child pedestrians, while simultaneously avoiding Run-over scenarios