El uso sostenible del suelo y la ZAP

A moderately simple mathematical model for estimating the linear and angular head accelerations that occur in a vehicle-pedestrian collision was described by Vilenius et al in their 1993 paper. The model used the stiffness of the vehicle structure impacted, the offset of the centre of mass from the force vector and the mass and moment of inertia of the human head. The accuracy of the model was compared with cadaver testing. Reasonable agreement was found with the linear acceleration values whilst a lack of data precluded an evaluation of the angular acceleration values.

Whilst determining head impact acceleration is useful, the above method requires a reasonable understanding of the kinematics of the collision being studied. Where this

to iteratively derive the pedestrian kinematics. Examples of full-body mathematical pedestrian models include (Linder, 2004):

• Chalmers (Yang, 1997)

• MADYMO (TNO, 2001)

• JARI (Japan Automobile Research Institute)/JAMA (Japan Automobile Manufacturers Association) (Konosu, 2002; Neale et al, 2003; Sugimoto and Yamazaki 2005)

• Adelaide University (Anderson and McLean, 2001; Anderson et al 2005)

• Honda (Okamoto et al, 2000; Shin et al, 2006)

Partial-body mathematical human models of note (and relevant to this thesis) include:

• WSUHIM/WSUBIM (Wayne State University Head/Brain Injury Model) (Zhang et al, 2001 & 2003).

• ULP (Louis Pasteur University) Model (Willinger et al, 1999; Willinger and Baumgartner, 2003).

The benefits of mathematical human models for injury appraisal include:

• Flexibility in experimental and loading conditions

• Insight into internal dynamic mechanisms

• Flexibility in human model sizing, shape, gender and other characteristics typical of the actual human population

• Incredible range of possibility for virtual instrumentation

Table 4.6 shows the different instrumentation in the MADYMO Human Model and the POLAR-II test dummy. Both the MADYMO Pedestrian model and the POLAR-II have been validated against full scale cadaver tests (Yang, 1997; 2002; Kerrigan et al, 2005). Some differences are apparent in the instrumentation of the two models with the MADYMO model generally having a somewhat wider range of measurement capability. Exceptions to this include the ribcage and abdomen deflection measurement capability of the POLAR-II dummy. POLAR-II’s T12 vertebrae and pelvis accelerometers appear to be matched by similarly located accelerometers in the

Measurement Type Location MADYMO Human Model Ver 6.3 POLAR II Pedestrian Dummy

(Source: MADYMO Human Model Ver 6.3 Manual) (Source: Rangarajan, 2000)

Velocity Head C.G.

Cardan Output Hip R, P, Y

(Roll, Pitch, Yaw) Knee R, P, Y

Ankle R, P, Y

Force and Torque Lower Torso F, T

Upper and Lower Neck F, T

Upper and Lower Leg F, T

Load Cell Lower Torso R, L, F/R, A

(Resultant, Lateral, Forward/ Upper and Lower Neck R, L, F/R, A R, L, F/R, A

Rearward, Axial) Upper and Lower Leg R, L, F/R, A R, L, F/R, A

Comparison of MADYMO Human Model and POLAR II Pedestrian Dummy Instrumentation

Table 4.6 Comparison of MADYMO Human Model and Polar II Instrumentation

Despite the wide range of measurement capability within the MADYMO model the validated injury parameter set for the pedestrian model is fairly limited and includes HIC, 3 millisecond and Viscous Criterion.

Other full-body mathematical models:

• The JARI model appears to be capable of displacement, velocity and acceleration measurements of the head, hand, pelvis, knees and feet (Konosu, 2002). Whilst this is sufficient to derive HIC and several other head injury parameters there appears to be limited ability to determine the potential for thorax and abdomen injury and extremity joint injuries. Neale et al (2003) compared the kinematics and head impact velocities of the TNO MADYMO pedestrian model, the JARI pedestrian model and full-scale cadaver tests. It was found that the greater biofidelity of the joints of the TNO MADYMO pedestrian model compared to the JARI model resulted in considerable differences in the predictions offered by the two models. The lack of joint biofidelity in the JARI model may have resulted in greater model accuracy in regard to head impact velocity due to the inadvertent simulation of muscle tension. Conversely, the lack of biofidelity of the lower extremities of the JARI model resulted in considerable kinematic differences from that predicted

by the TNO MADYMO model. Inspection of the impact sequence indicated right-femur fracture in the TNO MADYMO model when impacted by a simulated SUV vehicle at 40 km/h. Research by the author suggests that such a fracture is indeed highly likely at that speed in a vehicle-pedestrian collision involving an SUV. The JARI model did not appear to suffer a serious leg fracture and this changes the predicted pedestrian kinematics considerably after 270 ms. This suggests that the JARI model may not be suitable for lower extremity injury simulation. New developments have seen the development of the JAMA (Japan Association of Automobile Association) pedestrian model (Long and Anderson, 2005), a finite element model designed for LS-DYNA and PAM-CRASH solvers (See Chapter 3, Section 3.3.1 and Section 3.3.3).

• The Adelaide University model was designed with a focus on head impacts.

Its neck model was based on the findings of human volunteer tests (Anderson and McLean, 2001) in order to provide better biofidelity. Both linear and angular head acceleration can be recorded (Anderson et al, 2005). It has been validated using the results from cadaver tests, particularly in regard to femur and pelvic fracture.

• A finite element model developed by Toyota R&D Labs and Toyota Motor Corporation called THUMS (Total Human Model for Safety) (Sugimoto and Yamzaki, 2005; Snedeker et al, 2005). THUMS has been designed for use in the PAM-CRASH environment and has been validated using instrumented cadavers.

• A finite element model of the POLAR-II dummy is under development (Shin et al, 2006). Due to POLAR-II’s high cost the ability to accurately simulate the crash-test dummy has obvious cost-benefits. Because it is not a human model current finite element methods should be able to replicate the dummy with a high degree of accuracy. However, once finite element models of human bodies are sufficiently accurate the need for POLAR-II and its corresponding mathematical model will be negated.

• The WSUHIM (Wayne State University Head Injury Model) model, an FE human head model, was developed using brain injury data from people injured whilst playing American football (Zhang et al, 2001 & 2003). It can model the

throughout the brain, as well as linear and rotational acceleration. The researchers at Wayne State University believe that too much attention is focused on brain acceleration as the predominant parameter in brain injury and that more attention needs to be paid to intracranial pressure and brain stress and strain.

• The ULP (Louis Pasteur University) Model (Willinger et al, 1999; Willinger and Baumgartner, 2003) was created to address the perceived deficiencies of the Head Injury Criterion (HIC). The ULP Model is a finite element model of the skull and brain which models both the interaction of the skull and brain using fluid-structure interaction and skull damage from bone fracture. Injury potential is determined from intracranial pressure, Von Mises stress and cerebral-spinal fluid internal energy (CSFIF). When evaluating motorcyclist head injuries, Von Mises stress was found to be a good indicator of concussion, CSFIF was noted to predict sub-dural haematoma and the FE model accurately forecast skull fracture.

Anderson and McLean (2001) compared the head-impact speed using the JARI, MADYMO and Adelaide University full-body mathematical pedestrian models in simulated collisions with three different vehicles and two different pedestrian postures. The vehicle shapes represented were a flat-fronted vehicle, a passenger car and an SUV. Anderson obtained the most consistent results across the different pedestrian models and postures for simulations using the passenger car model. In these cases the head impact speed varied between 22% and 82% for a given impact speed. For simulations using the SUV model, the head impact speed predicted by the different pedestrian models varied by between 130% and 210% for a given impact speed whilst for the flat-fronted vehicle the head impact speed for a given impact speed varied by between 76% and 200%. Based on these results there are several possible conclusions:

• The models are inaccurate

• Head impact during a vehicle-pedestrian collision is highly variable, particularly for vehicles with a relatively high leading bonnet edge.

• A combination of the above

Anderson concludes by stating the need for further investigation.

4.7.6 Comparison of Mathematical Modelling and Real-World Pedestrian