3.1 Field crash tests characteristics
3.1.1 Scope of available data
A report was written for each field crash test providing the following data: type of test (in the field, simulation), detailed characteristics of the test, barrier data (e.g. type, barrier width, post spacing), vehicle data (type, mass, impact speed and impact angle and speed and angle of departure), barrier parameters after impact (normalised working width, dynamic deflection, maximal sustained deformation, length of contact between vehicle and barrier, length of barrier damage, number of barrier segments that need replacing, number of damaged posts, incident severity indices (ASI, THIV), video and photographic documentation (field tests), comments and drawings for specific results such as barrier intrusion (numerical tests).
3.1.2 Set of field crash tests
A set of 25 tests was analysed combining LifeRoSE (3 tests), RoSE (9 tests) and barrier manufacturers’ tests (13 tests). Tests carried out under RoSE and LifeRoSE include: TB32 (vehicle with a mass of 1500 kg, impact speed 110 km/h, impact angle 20o) for a road wire rope barrier for a section of a barrier installed on a curve with a radius of 400 m (two impacts), TB32 for a road steel barrier for a section of a barrier installed on a horizontal curve with a radius of 400 m (two impacts), TB11 (vehicle with a mass of 900 kg, impact speed 100 km/h, impact angle 20o) for a bridge barrier with a kerb 14 cm high (two impacts), TB51 (vehicle with a mass of 13,000 kg, impact speed 70 km/h, impact angle 20o) for a bridge barrier with a kerb 14 cm high, TB32 for the connection between a road wire rope barrier with a steel barrier, TB51 for a road steel barrier and a lighting pole placed within the barrier’s working width, TB41 (vehicle with a mass of 10,000 kg, impact speed 70 km/h, impact angle 8o) for a concrete barrier, TB32 for a road wire rope barrier (impact angle 7o), TB32 for a road steel barrier (impact angle 7o), see Fig. 1.
For the purposes of the analysis, a set of numerical tests was used simulating vehicle impact with barriers. The set comprised 570 tests conducted under this LifeRoSE project (173 numerical tests) and RoSE project (397 numerical tests) totalling 139 numerical tests with a concrete barrier, 149 numerical tests with wire rope barrier and 282 numerical tests with steel barrier.
3.2 Numerical tests characteristics
The main objective of safety barrier modelling and simulation tests is to develop numerical models of crash tests. The work conducted in 2016 included an extensive review of the literature, adjusting vehicle numerical models to the needs of the project and preliminary numerical tests. Building on these results in 2017 further research included simulations of virtual crash tests using a commercial system of the FEM, the LS-Dyna. LS-Dyna is a well-known and widely used commercial FEM code, designed to deal with highly nonlinear, short lasting phenomena. In the presented research, explicit dynamics solver is used, due to its robustness in solving problems with a very large number of degrees of freedom, many material laws included and possible complex contact behaviour. Another advantage of LS-Dyna is parallelization, which makes calculations scalable up to hundreds of cores. Nevertheless, an explicit dynamics algorithm is conditionally stable. It means that the time step in calculations should be very small and it is dependent, among others, on the characteristic size of the finite element. Furthermore, the time step size must be a compromise between the total calculation time and the level of details in the model. In a typical LS-Dyna model, beside geometrical data, dozens of materials definitions are included, with hundreds of parameters defined. Some of them ale collected from laboratory tests, some of them are prescribed basing on research papers, some of them are defined according to common engineering knowledge. This causes the fact that numerical analysis must be conducted by an experienced research team, with great knowledge of FEM and material laws theory and their implementation. Huge amount of post-processing data must be carefully revised to prove lack of numerical issues, like non-physical penetrations, so called “shooting nodes” or hourglass deformations.
A reliable simulation may be a basis for extended parametric analyses with changing conditions such as the initial conditions (vehicle speed and impact angle). This, however, should be treated with some caution with the results tested each time for reliability and any possible numeric errors. The experience of project participants which they gained by consulting constructors who conduct simulations for barrier manufacturers, shows that the process, although reverse to validation (i.e. anticipating actual behaviour based on numerical simulation, see Fig. 2), is correct and reliable.
Even the most detailed of models is of no value unless its results are compared to a real test. Numerical solutions must be confronted with site test results in a qualitative and quantitative check. This process validates a solution. To ensure that barrier validation was performed properly, the numerical simulation should be consistent with the site test in the following points: maintaining the vehicle on the lane, vehicle rollover, maintaining the vehicle within the so called exit box, wheel trajectories, damage to vehicle suspension, damage to barrier longitudinal elements (guardrails), barrier dynamic deflection, vehicle intrusion, intrusion of parts of barrier into the vehicle, deformation of the entire object of study.
As barrier fragments are being damaged, the sequence of this should be the same in the site test and simulation. The reliability of numerical tests is defined by criteria of a quantitative comparison of the parameters:
dynamic deflection: |DMcrash test – DMnumerical test| ≤ 0,05 m + 0,1DMcrash tests,
working width: |WWcrash test – WWnumerical test| ≤ 0,05 m + 0,1WWcrash tests,
intrusions: |Wcrash test – Wnumerical test| ≤ 0,05 m + 0,1Wcrash tests,
ASI: |ASIcrash test – ASInumerical test| ≤ 0,1,
THIV: |THIVcrash test – THIVnumerical test| ≤ 3 km/h.
In each case, the difference between the simulation and site test moment reaching their maximum values cannot be greater than 0.05 s.
Figure 3 shows the standard test TB32. A passenger car (1500 kg, 110 km/h, 20° angle) hits a W-beam steel barrier. The collision (length of contact, vehicle redirection) and functional parameters (ASI, THIV, working width) were represented reliably. Figure 4 shows the passenger car crash (1500 kg) in a non-standard test (110 km/h, impact angle 50°). In this case, with no field test results, it is advisable to thoroughly check the simulation for non-physical phenomena (such as element overlay, “glued” contact surface).
Figure 5 shows the result of a bus hitting (13,000 kg) a concrete barrier. Major destruction to parts of the concrete are clearly visible. In reality, while concrete may no longer be able to redirect loads, it continues to retain a specific volume. In the simulation, a “vanishing’ effect was observed with extensive parts of the structure gone which allowed the vehicle’s intrusion into the barrier. In this case, further and more in-depth studies into structural effort are required.
About 10% of the numerical tests involved two specific effects of vehicles hitting a barrier: the vehicle was wedged or immobilised under the barrier, the barrier was broken through or the vehicle went under or over the barrier). Both cases occur primarily when the vehicle hits a segment of the barrier at a high angle or with very high kinetic energy (Fig. 6).
As defined in the objective, the tests selected were designed to help to:
- 1.
Identify the most relevant factors of vehicle movement that have an effect on barrier deformation or destruction and change barrier functional parameters;
- 2.
Develop simple mathematical models to estimate the effects of barrier design parameters and vehicle movement parameters on the extent of damage and change in barrier functionalities.