### Project proposal:
__Development of a rapid evaluation and optimization method for automotive airbag system design__

##### [中文版本](https://hackmd.io/@SamuelChang/SylfZzUnJx)

### Abstract

The development of customized vehicle airbag systems presents significant challenges, requiring coordinated efforts across inflator ballistics adjustment, airbag-module engineering, and system integration. This process spans multiple R&D departments, specialized firms, and market sectors. While inter-institutional collaboration is crucial, current specifications inadequately define critical gas dynamics and performance thresholds for inflators. Consequently, the conventional hierarchical and sequential development approach often results in protracted project timelines due to iterative optimization loops, incurring substantial costs in prototyping, testing, and numerical simulations.

This project proposes a solution to this challenge through a bottom-up model chain that simulates test environments across all development levels. This approach establishes comprehensive component-to-system performance relationships. Coupled with a top-down optimization process, the model chain rapidly identifies optimal design parameters, dramatically narrowing the search space during fine-tuning. Besides, since all necessary parameters are obtained as a result of the optimization, the predicted performance in each test environment can be evaluated and evolved as a target for each development level. Related teams can work in parallel. This can not only reduce project execution time, but also break the current sequential development structure and improve efficiency while controlling costs.

---

### Introduction
Achieving a 5-star [NCAP](https://www.euroncap.com/en) rating is a key milestone for new vehicle models, signifying that the highest standards in occupant safety systems have been met. This evaluation encompasses active safety systems, pedestrian protection, and—most critically—passive safety equipment, such as vehicle airbag systems.

However, developing customized airbag systems remains highly challenging due to divergent regional NCAP regulations and unique interior designs for each vehicle model. The process often requires multiple iterative loops involving costly prototyping, testing, and numerical simulations to identify a rough optimal solution. These challenges are further compounded by the rise of autonomous driving systems, where passengers may adopt non-standard seating positions during a crash, potentially rendering the traditional restraint system design principles obsolete.

Among all airbag components, the inflator is the core element responsible for system activation. The mechanism is highly complex, engineered to generate a precise gas volume within milliseconds. Inflators are broadly classified into three types: pyrotechnic inflators, which rely on chemical reactions to produce gas and heat; cold gas inflators, where high pressurized gas is stored in the inflator cartridge; and the hybrid inflators, which combine the both mechanism. All inflators are triggered by pyrotechnic initiators, either to ignite chemical reactions or rupture a sealing membrane. Due to their differing underlying mechanisms, each type exhibits distinct gas dynamics (mass flow rate and gas temperature evolution).

The inflator performance directly impacts critical safety metrics, e.g., cushioning of body movement, thoracic and lumbar load mitigation (for both adults and children) and effectiveness in Out-of-Position (OoP) scenarios. Thus, defining inflator characteristics is the most crucial step in the development chain, as it has the highest influence on overall safety performance.

The standard test procedure for inflators, as defined in both the AK-LV-03 and USCAR-24[[2]](https://www.dinmedia.de/en/standard/sae-uscar-24-3/376304261) specifications, is the tank test. In this test, the pressure evolution inside a sealed container is measured during inflator deployment. By analyzing the pressure curve characteristics, e.g., the maximum pressure level, the time from triggering to peak pressure, and the pressure slope within the first 5 milliseconds, airbag engineers assess whether the inflator’s performance is too aggressive or too sluggish for the intended airbag system.

Moving from inflator testing to airbag-module integration, the inflation test is a standard procedure for evaluating system performance. Here, the focus lies on the airbag’s deployment dynamics. During the test, internal pressure is monitored, and high-speed cameras capture the inflation process. Using this data, airbag developers adjust the bag volume and vent size, which can sometimes compensate for suboptimal inflator characteristics.

To differentiate between frontal and side-impact protection, pendulum impact tests are often employed for system calibration. In [frontal impact tests](https://www.mdpi.com/energies/energies-14-05288/article_deploy/html/images/energies-14-05288-g008-550.jpg), a pendulum is released from an elevated position and cushioned by the inflating airbag, while pressure and pendulum trajectory are recorded. By varying impact velocities, developers can evaluate the airbag’s energy-absorption capabilities and fine-tune triggering delays to optimize performance.

For side-impact simulations, the pendulum starts in a lowered position and is propelled by the deploying airbag. Again, bag pressure and pendulum movement are tracked, replicating side-collision scenarios and validating the effectiveness of side airbags.

The sled test represents the final stage of system-level validation, simulating a crash scenario using a rapidly decelerating test sled. This standardized setup incorporates anthropomorphic dummies of varying sizes, with measurements taken for head displacement, chest compression, and thoracic loading. Unlike full-vehicle crash tests, the sled test isolates the restraint system’s performance, providing the closest approximation to NCAP assessment conditions without interior design variables.

The entire development process—from customized inflator tuning to system integration—is hierarchical and sequential, as shown in the sketch below. If a system performs poorly, reverting to inflator re-engineering may be necessary, requiring months of retesting and reevaluation before returning to system-level validation. Consequently, development projects can span several years, often yielding lessons learned rather than immediate progress.

<img src="https://live.staticflickr.com/65535/54426141147_ac1f5463ff_b.jpg" width="90%">

Moreover, the technical complexity of airbag systems demands cross-disciplinary collaboration between R&D teams or specialized firms. While top-down specifications are crucial, many existing standards inadequately define the necessary physical gas dynamics and performance thresholds. Significant gaps remain in these specifications, warranting further research and refinement.

### States of the Art

Scientific contributions in this field address multiple perspectives of airbag inflator development.

Regarding thermodynamic analysis in the inflator, mathematical models are developed to quantify the thermal-chemical interactions during the deployment. [Butler et al. [3]](https://www.sciencedirect.com/science/article/abs/pii/0360128593900094) proposed a 0-dimensional interior ballistic model to describe the thermal mechanism in inflators based on generalized design principles. [Materna [4]](https://www.sae.org/publications/technical-papers/content/920120/) further synthesized ballistic functions of pyrotechnic inflators and corresponding modeling strategies for each functional component. Building on these foundations, commercial software now integrates capabilities to simulate pyrotechnic and hybrid inflators, including deflagration processes[5]. Additionally, Kang and Wang[6] developed numerical models for hybrid inflators, while [Shieh et al.[7]](https://journals.sagepub.com/doi/10.1243/0954407011525386) and Wu[8] advanced filter mechanism modeling by incorporating 1D flow simulations, later applying this to dual-stage inflators.

For the inflator testing methods, thrust measurement[[9]](https://www.dynamore.it/en/downloads/papers/09-conference/papers/E-III-01.pdf) [[10]](https://publica-rest.fraunhofer.de/server/api/core/bitstreams/bc85f019-d128-40d2-81cd-0b97197d4da8/content) has emerged as a supplementary method for characterizing inflator performance. This approach deploys the inflator in a bell-shaped open vessel, measuring the thrust force generated by gas expulsion through the nozzle. The resulting force profile serves as a direct indicator of inflator behavior. Compared to tank tests, thrust measurements offer greater agility, as the test vessel is significantly smaller than standardized tank volumes.

For system-level suitability assessment, the Force INdicating Assessment Tool (FINAL ton) test[11] was developed as a simplification to full inflation tests. Positioned between inflator and airbag development, this method deploys the inflator in a standardized bag with fixed venting, yielding objective data with reduced measurement uncertainty. The resulting pressure curves exhibit higher correlation with system-level performance metrics than tank test data, while providing cleaner signal outputs than conventional inflation tests.

[Schäfer et al.[12]](https://www.sae.org/publications/technical-papers/content/2023-01-5029/) contributed extensively to inflator characterization methodologies, emphasizing mass flow rate and gas temperature profiling across inflator types. Their work proposed type-specific handling protocols and cross-validation across test devices to ensure computational accuracy.

When focusing on system-level simulation techniques, finite element (FE) and multi-body (MDS) simulations are industry standards for predicting airbag inflation and crash dummy kinematics. Advanced computational methods implemented in commercial software[13]. For flow patterns in the airbag inflation process, uniform pressure method (UPM), finite point method (FPM) and Corpuscular Particle Method (CPM) are the general methodologies modeling the gas flow condition within a defined volume. These are the meshless concepts implemented in the structural FEM regime.

Arbitrary Lagrange-Eulerian (ALE)[14] is the alternative concept to capture the airbag inflation, which couples the finite element method for solid deformation and movement with the meshed Eulerian fluid domain. By solving the transport equation of gas continuum, the movement of airbag is explicitly resolved via fluid-structure interaction technique. Thus, details of gas movement, as well as gas dynamic phenomena, e.g., under-expansion wave and gas impingement, are captured instead of simply only the pressure representation of UPM and FPM/CPM. However, ALE requires very high computational cost, for which it is not commonly utilized as industrial standard yet.

Such computational methods are expensive and account for all possible details, including how the airbag is folded into its module, the material properties of the module casing, and even the size of the car brand logo on the steering wheel center - all of which can influence results. Moreover, running a single simulation case may require several hours of parallel computing. As a result, these computational models can capture detailed flow patterns along with dummy part trajectories and squeezing forces throughout the entire crash scenario. This capability provides comprehensive insight into all potential design features, but simultaneously makes the optimization process extremely time- and resource-intensive.

### Objectives of the Project
Within this work, a __model chain__ and __optimization process__ are proposed to to systematically describe and automate the development cycle through analytical methods

<br/>

<img src="https://live.staticflickr.com/65535/54431265565_ed9b57ddb1_b.jpg">

<br/>
<br/>

The model chain simulates the key characteristics of an airbag deployment process with generic boundary conditions. This obtains the deployment of an inflator (general inflator model for pyro, cold gas and hybrid inflators), inflation process of an airbag and the collision with the passive objective with possible steering parameters. By focusing only on performance-critical variables, the modeling approach establishes a direct relationship between inflator characteristics and standardized test outcomes, enabling rapid evaluation.

#### Model chain development
The model chain is developed following a bottom-up approach, progressing from inflator-level to system-level analysis. This framework is structured into four key phases:

1. __inflator characteristic evaluation__ in three inflator types:
- pyrotechnic inflator with generic propellant burn rate scenario
- stored gas inflators with common inert gases of He/Ar mixture
- hybrid inflators covering both direct ignition (pyrotechnics in inert gas) and separated chamber designs

Main output of these inflator models are the patterns for mass flow rate and temperature profile, with correponding pressure curves in the tant test environment.

2. __airbag inflation procedure__ for adjustable bag volume and vent size incorporated with inflator mass flow and temperature profile.

Underlying physics behind airbag inflation process is the energy conservation related with euqation of state and gas dynamic properties on the vent. Neglecting the local gas flow patterns, the dynamics of inflation can be approached using numerical methods in a 0D manner. Results of this model is the pressure evolution in the bag. Relevant characteristics such as maximum pressure level, time to maximum pressure will be clearly revealed. Generic model for airbag-module case rupturing will be created to describe the mechanical interaction, which has to be calibrated according to realistic data.

3. __pendulum absorption__ for both frontal and side protection setups

The pendulum test mechanism can be viewed as an inflation process obstructed by a passive object. The modeling work will be conducted in a 0D-1D regime, focusing on pendulum dynamics and its interaction with the inflating bag. Simplified assumptions will be made to describe bag deformation. The model outputs will include bag pressure, deformed bag volume, and pendulum trajectory.

4. __sled scenario__ for dummies of different sizes, incorporating seat belt functionality.

#### Optimization process
Following validation in generic systems, optimization algorithms such as Particle Swarm Optimization (PSO) or Genetic Algorithms (GA) can be implemented to identify optimal and robust design parameters. The cost function for the optimization process may incorporate both the impact on dummy body parts in sled tests across varying initial velocities and pendulum trajectories for frontal or side protection configurations.

The optimization results will determine key parameters such as __system trigger delay timing__, __optimal airbag shape, volume and vent hole sizing__, and __inflator gas mass flow rate and temperature profiles__ for performance adjustment. By leveraging the model chain outputs, these optimized characteristics automatically correlate with performance metrics from corresponding physical tests.

### Project Benefits
The proposed modeling approach, operating in 0D (temporal evolution) and 1D (temporal-main axis) regimes, offers significantly lower computational costs compared to detailed crash test simulations. This efficiency eliminates the need for high-performance computing resources when executing the model chain and optimization process in real-world projects.

At the meta level, this integrated model chain and optimization framework can break the current hierarchical and sequencial structure of the product development paradigm. Since the best fit solution obtains design parameters of all the development levels (inflator ballistics, airbag-module assembly and system integration), the related assignments in each level can be carried out parallelly. With the predicted performance on the related test environment as target, hardware test results can be assessed in an objective manner. This can reduce enomrous project execution time and unnecessary prototype build-ups and testings.

<img src="https://live.staticflickr.com/65535/54427252013_204f9f9b42_b.jpg">

Of course, fine-tuning in detailed simulations is still eligible. With restricted search fields as outputs of the optimization process, 3D multibody and finite element simulations can further focus on detailed design. High-quality optimization systems can be achieved more easily.

### Time and budget plan
The proposed works are expected to be realized within 3.5 years.
...

### Potential participants:
- OEM: [VW- Florain Linke]
- Restraint system developer: [JSS](https://www.joysonsafety.com/de/), [Yanfeng](https://www.yanfeng.com/en)
- Inflator manufacturer: [JSS](https://www.joysonsafety.com/de/), [iSi](https://www.isi.com/de-AT/automotive/inflator), [Mosa Industral Corp.](https://www.twmosa.com/?lang=en)
- Research institute: [FH Aalen(Prof. Aschenbrenner)](https://www.hs-aalen.de/de/users/24092)
- Measurement system developer: [HuDe](https://root.hude.com/de/home/) or equivalent firms?

### Reference
2. _USCAR INFLATOR TECHNICAL REQUIREMENTS AND VALIDATION_, SAE USCAR-24-3:2023-12-12 .
3. P.B. Butler, J. Kang, and H. Krier. _Modeling and numerical simulation of the internal thermochemistry of automotive airbag inflators_ Progress in Energy and Combustion Science, 19(5):365–382, 1993.
4. P. Materna. _Advances in analytical modeling of airbag inflators_. Technical report, SAE Technical Paper, 1992.
5. K.-S. Im, Z.-C. Zhang, and Grant O Cook Jr. _Airbag inflator models in ls-dyna_. In 14th International LS-DYNA Users Conference, Livermore Software Technology Corp., Livermore, CA, volume 94551, 2016.
6. J. Kang and J.T. Wang. _H-isp-a hybrid inflator simulation program_. SAE transactions, pages 2035–2044, 1999.
7. W.H. Hsieh, L.Y. Sun, J.K. Chen, and S.W. Wang. _Theoretical simulation of combustion processes of airbag inflators_. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 215(1):1–9, 2001.
8. W.T. Wu, W.H. Hsieh, C.H. Huang, and C.H. Wang. _Theoretical simulation of combustion and inflation processes of two-stage airbag inflators_. Combustion science and technology, 177(2):383–412, 2005.
9. H. Kratz, J. Fernández, A. Heym, and A. Kato. _Characterization of the performance of airbag inflators by thrust measurement method_. Takata Petri AG, IPASS, pages 295–307, 2009.
10. P. Yurrita, N. Edelmann, T. Klahn, and J. Neutz. _Massflow determination of airbag inflators_. In 14th International Symposium on Sophisticated Car Occupant Safety Systems, 2018. Mannheim, Germany.
11. F. Wollny. _New evaluation methods for inflator performance_. In 14th International Symposium on Sophisticated Car Occupant Safety Systems, 2018. Mannheim, Germany.
12. Schäfer, T., Chang, C., and Neutz, J., _Assessment of Airbag Inflator Characterization Methods for Numerical Prediction in the Automotive Restraint System Applications_, SAE Technical Paper 2023-01-5029, 2023, https://doi.org/10.4271/2023-01-5029.
13. J.T. Wang and D.J. Nefske. _A new cal3d airbag inflation model_. SAE Transactions, pages 697–706, 1988.
14. D. Fokin, E. Dessarud, C. Ljungqvist, and Saab Automobile AB. _Simulation of curtain airbag with arbitrary eulerian-lagrangian method_. In 6th LS-DYNA Forum, 2007.

---

### Principal Investigator
## Chi-Yao Chang, Dr.-Ing.
<img style="float: right" width="20%" height="20%" src="https://i.imgur.com/ZuJhZ8h.jpg">

- 2022 Mar. - Now
Senior Development Engineer, Gas Metering, Honeywell-Elster
- 2016 Oct. - 2022 Feb.
Lead Engineer, CAE, inflator core and applications
Aschaffenburg Inflator Center, Joyson Safety Systems
- 2014 Feb. – 2016.Sep.
Post -Doctoral research, CFD windfarm site assessment,
Institute for Wind energy and Energy System technology,
Fraunhofer Society
- 2009 Apr. – 2014 Jan.
PhD, chair of fluid mechanics and aerodynamics ,
Technical University Darmstadt, Germany