Crash tests and dummies can be used to investigate the behavior of vehicles and passengers in car accidents. The dummies are becoming more realistic and the challenges posed by automated driving are becoming greater.

In recent decades, the safety of car occupants and other road users has improved enormously. Crash tests make a significant contribution to reducing the number of victims in road traffic. The key measurement tools for assessing the risk to vehicle occupants and road users in these accident tests are located in the vehicle itself: crash test dummies. These are “human-like dolls (or sometimes just body parts)”, as Thomas Kinsky and Christian Kleeßen from Humanetics explain in the article Female dummies and occupant dummies for autonomous driving (page 29) from ATZ 4-2022. While the dummies were initially rigid and rough, they are becoming increasingly human and changing with the demands placed on them by automated driving.

Editor’s recommendation

01.12.2023 | Cover story

“We want to conduct up to 500 crash tests per year”

The opening of a new test center for vehicle safety does not happen every day. After seven years of soil remediation and construction of the incampus technology park, the engineering teams were able to move into the largest building on the site on September 15, 2023. The highlight in Ingolstadt is the crash arena with a column-free area of ​​50 × 50 m and opposing runways with a total length of 250 m. In the ATZ interview, Dr. Martin Friedrichsen from Audi explains how safe batteries in electric vehicles are, why virtual simulations are essential for test planning and what he thinks of female crash test dummies.

50th percentile man most commonly used

Test specimens or crash test dummies are expected to meet certain requirements, as the Springer authors around Kai-Uwe Schmitt explain in the chapter Methods of Trauma Biomechanics (page 46f) of the book Trauma Biomechanics. This ranges from anthropometry and biofidelity – i.e. the requirement to be as close as possible to real people in terms of body size, mass, mass distribution, moments of inertia, (sitting) posture and biomechanical behavior – to instrumentation (sensitivity and measurement options) to repeatability and durability.

The most commonly used crash test dummies today come from the Hybrid III family. They were developed more than 40 years ago and have been in use for decades. “The 50th percentile adult male, whose anthropomorphic data were obtained from the US population in the 1960s (standing height: 1.75 m, weight: 78.2 kg), is the dummy most commonly used in the automotive industry,” explain Schmitt et al.. Other dummy types available are the 5th percentile female (1.51 m, 49.1 kg) and the 95th percentile male (1.87 m, 101.2 kg). There are also dummies that represent children of different ages.

Overall, the number of dummy types has increased significantly in recent years – also due to the many different types of collision (frontal, rear or side impact) that are tested. A torso block for the development of steering devices, a leg model for pedestrian protection and a head cap for the interior are also available as a “partial model” for component testing, according to Springer author Xiangfan Fang in the book chapter Requirements for the body due to passive safety.

Two new models from the Thor family

Since 2016, there has been a newer generation, the crash test dummy called Thor (Test Device for Human Occupant Restraint). It resembles humans even more, has more authentic movements and measures more precisely. This new model, as a hardware and software version, has begun to replace the Hybrid III in frontal impact crash tests, explain the engineers at Humanetics. In their article mentioned above, they present two new models from the Thor family: the Thor-5F dummy, which represents a 5th percentile woman, and the Thor-AV dummy, a dummy model optimized for highly automated and autonomous (SAE Level 5) vehicles.

Accident researchers have long criticized the fact that ergonomics and safety technology in vehicles are primarily geared towards men. Women and people of small stature in particular are considered to be at a disadvantage and often suffer more serious injuries in traffic accidents. And the data collected by Humanetics also showed that the safety requirements for small women differ from those of medium-sized men. The new Thor-5F dummy therefore offers improved biofidelity compared to the corresponding Hybrid III woman (HIII-5F). With a height of around 1.51 m and a weight of around 47.5 kg, the Thor-5F corresponds to around the 5th percentile of the female population – that is, only 5% of the female population is smaller and lighter than this dummy.

Thor-5F represents female body type better

However, Thor-5F also continues to follow “the approach commonly used in vehicle safety today of using the 5th percentile of women and the 50th and 95th percentile of men as a basis for developing restraint systems,” according to Humanetics. The male median dummy serves as the basis for tests, the other two dummies are intended to cover the lower and upper extremes so that cars can be developed for all occupants.

In contrast to the HIII-5F, the Thor-5F was not scaled from the 50th percentile male, but simply made smaller. “For the first time, this female dummy is based exclusively on anthropometric data from women. The body stature and shape therefore represent a typical woman much better,” says Humanetics. The American traffic safety authority National Highway Traffic Safety Administration (NHTSA) sees a 45% improvement in the Thor-5F biofidelity ranking compared to the HIII-5F.

The risks of injury are human

According to Dr. Martin Friedrichsen from Audi, there is a misunderstanding in society when it comes to gender-specific dummies. A dummy does not measure an injury directly, but rather physical quantities such as force, deformation and acceleration, which in turn have to be “translated” to humans using injury risk functions. “In our opinion, women do not need a new dummy with adapted dimensions, but rather more in-depth basic research into the injury symptoms in women compared to men,” explains Friedrichsen in an interview with ATZ. “We want to conduct up to 500 crash tests a year.”

Mercedes-Benz sees things similarly. The five percent dummy may have the anatomy of a woman, but in terms of injury risk it represents the group of small people – regardless of whether they are women or men, says Hanna Paul, Head of Mercedes-Benz Dummy Testing. “In other words: the injury risks are human – not male or female,” says Paul. The same applies to the other dummies. The NHTSA is currently researching the extent to which injury risks depend on gender. The first published results show that gender is not the most important factor influencing the frequency of injuries. The dummies used today are therefore effective measuring tools for the development of safety systems.

Dummies for autonomous driving

In addition to designing cars for different types of passengers (tall men, small women, children), vehicle developers today also have to consider alternative seating concepts that will be used in future vehicles with automated driving functions and in autonomous vehicles. This is because the seating position in these vehicles is no longer necessarily forward-facing and upright. As Humanetics reports, dummies used today are only of limited use for such purposes, as they are seated dolls that are not well suited to any position other than the upright seating position.

In order to evaluate passenger safety in other seating positions, Humanetics has therefore developed the Thor-AV dummy. The current Thor-50M and the new Thor-5F served as the basis. Both dummies were modified so that angles with a seat back inclination of at least 60° could be examined. Initial tests with new dummies were positive. It was shown “that the laboratories do not have to relearn how to handle these new dummies and that both dummies can be seen as a further evolutionary stage of the existing dummy generation,” says Humanetics.

Human Body Models as a supplement to dummies

In the future, however, physical dummies alone will no longer be sufficient. Virtual human models with computer simulations are intended to support conventional dummies in the evaluation and approval of future vehicles, as Virtual Vehicle Research explains in the article Future occupant safety in accidents involving passenger cars from ATZ 1-2020. Virtual Vehicle Research has coordinated the research project “OSCCAR: Future Occupant Safety for Crashes in Cars”, which has developed a virtual, simulation-based approach to assessing the safety of vehicle occupants in future accident scenarios.

In order to take greater account of the individual proportions of women and older people in restraint systems, IAV relies on AI methods. In the ATZ article Optimized crash safety through AI and synthetic data, the development service provider explains how, in addition to recognizing seat occupancy and body posture, the gender and age of all vehicle occupants can also be included in the design of personalized restraint systems.

Physical crash tests remain relevant

A combination of physical crash tests with virtual simulations also helps to ensure vehicle safety in automated vehicles even with shorter development cycles and complex crash scenarios and configurations, Siemens describes in the article Safety in electric and autonomous cars – Simulation of complex crash scenarios from ATZ 7-8-2020. With its MBS software Simcenter Madymo, the company provides an active human model for fast safety simulations for occupants. Real dummy crash tests are expensive and time-consuming and can only be carried out late in the vehicle development process. This is where simulation helps to save time and money.

However, Mercedes-Benz does not believe that simulations will soon make physical crash tests superfluous. Calculating the kinematics and deformation behavior can significantly reduce the number of full-vehicle tests and speed up development. But according to Hanna Paul, there is no way around vehicle crashes. There are several reasons for this: crash tests are necessary to confirm the simulations, which are carried out with many assumptions. They are also required by law or ratings. Paul sums up: “The interaction between sensors, crash and dummy behavior can best be verified in full-vehicle tests.”