In the rapidly evolving landscape of electric vehicles (EVs), the safety of the EV battery pack has become a paramount concern. As EV adoption surges globally, the risk of thermal runaway and fire due to mechanical intrusion during collisions, particularly in side pole impacts, poses significant challenges. Traditional computer-aided engineering (CAE) methods for crash analysis often involve full-vehicle models, which are computationally expensive and time-consuming, hindering rapid design iterations. Conversely, evaluating the EV battery pack in isolation using standardized tests may not accurately replicate real-world crash scenarios, leading to potential safety oversights. To address these issues, I propose a decoupling analysis method based on submodeling, which balances iterative design speed and assessment accuracy for the EV battery pack. This approach leverages submodeling techniques within explicit dynamics simulations to focus on the EV battery pack’s response during side pole collisions, enabling efficient and precise evaluation of structural performance.
The core idea of submodeling is to extract a critical region of interest—here, the EV battery pack and its immediate surroundings—from a full-vehicle model and analyze it with boundary conditions derived from the full simulation. This method relies on Saint-Venant’s principle, which states that the effects of localized loads diminish with distance, allowing for accurate local analysis if the submodel boundaries are sufficiently far from the area of concern. In this study, I implement the submodeling method using LS-DYNA, a finite element software for explicit dynamics. The process involves two main steps: first, conducting a full-vehicle side pole crash simulation to obtain boundary conditions; second, performing a submodel simulation with these boundary conditions applied to the truncated model. This decoupling allows for rapid exploration of different EV battery pack designs without re-running the entire vehicle model, significantly reducing computational time and resources. The workflow for design iteration using this method is straightforward: after validating the submodel against the full model, multiple design variants of the EV battery pack can be tested within the submodel framework, accelerating development cycles while maintaining fidelity to real crash conditions.
To demonstrate the effectiveness of this method, I applied it to a side pole crash scenario based on the C-NCAP (2021 revised version) protocol, excluding dummies to focus solely on the EV battery pack’s mechanical response. The full vehicle model included all relevant components, with the EV battery pack mounted beneath the cabin floor. In a side pole impact, a rigid pole strikes the vehicle laterally, creating concentrated forces that can intrude into the EV battery pack, potentially damaging cells and triggering short circuits. The submodel was defined to encompass the entire EV battery pack, the driver’s seat (due to load transfer through mounting points), surrounding body sections larger than the pole diameter, and the rigid pole itself. Boundaries were selected at locations distant from the intrusion zone to minimize their influence on the EV battery pack’s behavior, primarily relying on battery pack mounting points to drive motion. This setup ensures that the submodel captures the essential dynamics while isolating the EV battery pack for detailed study.
Validating the consistency between the submodel and full model is crucial for reliability. I assessed three key parameters: the internal energy of the EV battery pack, which reflects overall energy absorption and computational accuracy; the intrusion depth of the battery pack side rail inner wall, a critical measure of crashworthiness; and the intrusion of the battery pack top cover, indicating vertical deformation. The internal energy, denoted as \(E\), can be expressed as the work done by forces during deformation, often approximated in crash simulations as the sum of elemental strain energies. For intrusion, the displacement \(d\) at specific points relative to a reference plane is monitored. To quantify consistency, I compared these parameters at 55 ms, a time point capturing peak responses. The results showed excellent agreement, with errors below 2.5% for all metrics, confirming that the submodel accurately replicates the full model’s behavior for the EV battery pack.
An important aspect of submodeling is the time interval for boundary condition extraction, which affects both accuracy and computational efficiency. In LS-DYNA, boundary data can be output at user-defined intervals. While smaller intervals closer to the simulation time step yield higher precision, they also generate larger files. I investigated three intervals: 0.005 ms, 0.050 ms, and 0.500 ms, labeled as Submodel 1, Submodel 2, and Submodel 3, respectively. The impact on the EV battery pack’s internal energy and intrusion was minimal across these intervals, as shown in the following tables. This indicates that for engineering purposes, a moderate interval like 0.500 ms is sufficient, balancing accuracy with storage requirements. The submodel simulations required only about 20% of the computational time of the full model, highlighting the efficiency gains for EV battery pack analysis.
| Model | Internal Energy (J) | Error (%) |
|---|---|---|
| Full Vehicle Model | 21,165.289 | Baseline |
| Submodel 1 (0.005 ms) | 21,212.839 | 0.22 |
| Submodel 2 (0.050 ms) | 21,337.680 | 0.81 |
| Submodel 3 (0.500 ms) | 21,236.259 | 0.34 |
| Model | Intrusion (mm) | Error (%) |
|---|---|---|
| Full Vehicle Model | 29.399 | Baseline |
| Submodel 1 (0.005 ms) | 29.438 | 0.13 |
| Submodel 2 (0.050 ms) | 29.589 | 0.65 |
| Submodel 3 (0.500 ms) | 29.452 | 0.18 |
| Model | Intrusion (mm) | Error (%) |
|---|---|---|
| Full Vehicle Model | 22.845 | Baseline |
| Submodel 1 (0.005 ms) | 22.727 | -0.52 |
| Submodel 2 (0.050 ms) | 23.036 | 0.84 |
| Submodel 3 (0.500 ms) | 23.411 | 2.48 |
The robustness of boundary conditions is another critical factor, as design changes to the EV battery pack should not significantly alter the boundary responses if the submodel is to be used for rapid iteration. To test this, I modified the thickness of the side rails, which absorb approximately 84.5% of the EV battery pack’s internal energy, by ±0.1 mm and ±1.0 mm, representing minor and major design variations. Using the original submodel boundary conditions, I evaluated the impact on internal energy and intrusion. The results, summarized in the tables below, show that for minor changes (±0.1 mm), the submodel predicts trends correctly with small errors, demonstrating robustness for engineering design selection. However, for major changes (±1.0 mm), errors increase substantially, particularly in internal energy, indicating that boundary conditions may become invalid for large modifications. This suggests that the submodel method is best suited for incremental design changes, and monitoring internal energy shifts can help decide when a full-model reanalysis is needed for the EV battery pack.
| Thickness Change | Full Model Energy (J) | Submodel Energy (J) | Error (%) |
|---|---|---|---|
| +0.1 mm | 21,460.861 | 21,980.355 | 2.42 |
| +1.0 mm | 23,697.651 | 30,059.426 | 26.85 |
| -0.1 mm | 20,833.434 | 20,511.911 | -1.54 |
| -1.0 mm | 17,267.794 | 14,616.566 | -15.35 |
| Thickness Change | Full Model Intrusion (mm) | Submodel Intrusion (mm) | Error (%) |
|---|---|---|---|
| +0.1 mm | 29.046 | 29.080 | 0.12 |
| +1.0 mm | 22.735 | 26.847 | 18.09 |
| -0.1 mm | 29.846 | 29.900 | 0.18 |
| -1.0 mm | 31.279 | 33.530 | 7.20 |
| Thickness Change | Full Model Intrusion (mm) | Submodel Intrusion (mm) | Error (%) |
|---|---|---|---|
| +0.1 mm | 22.969 | 23.572 | 2.63 |
| +1.0 mm | 20.547 | 23.178 | 12.80 |
| -0.1 mm | 23.748 | 22.947 | -3.37 |
| -1.0 mm | 22.060 | 30.080 | 36.36 |
The consistency and robustness analyses underscore the value of submodeling for the EV battery pack. From an energy perspective, the internal energy \(E\) of the EV battery pack in the submodel closely matches that of the full model, with errors governed by factors like boundary selection and time interval. Mathematically, this can be expressed as \(E_{\text{sub}} \approx E_{\text{full}} + \epsilon\), where \(\epsilon\) is a small error term. For intrusion, the displacement \(d\) at critical points follows a similar relationship: \(d_{\text{sub}} = d_{\text{full}} + \delta\), with \(\delta\) being minimal for well-defined boundaries. The submodel’s computational efficiency is a key advantage, as the simulation time scales with model complexity. If \(T_{\text{full}}\) is the time for the full vehicle model and \(T_{\text{sub}}\) for the submodel, the ratio \(R = T_{\text{sub}} / T_{\text{full}}\) is approximately 0.2 in this case, demonstrating significant savings. This efficiency enables rapid iteration on the EV battery pack design, such as optimizing material thickness or geometry, without the overhead of full-vehicle simulations.
In discussing the implications, it is important to note that the EV battery pack’s safety hinges on preventing intrusion that could compromise cell integrity. The submodel method provides a high-fidelity tool for assessing this risk under realistic crash conditions, unlike standardized battery-level tests that may not replicate vehicle interactions. For instance, in side pole impacts, the localized force \(F\) from the rigid pole induces deformation in the EV battery pack structure, which can be modeled using stress-strain relationships. The stress \(\sigma\) in components like side rails relates to strain \(\epsilon\) via material laws, such as \(\sigma = E \epsilon\) for linear elasticity, though crash simulations often use plastic models. By capturing these details in the submodel, designers can evaluate metrics like intrusion velocity and energy absorption more accurately, enhancing the EV battery pack’s crashworthiness.
However, there are limitations to consider. The submodel assumes that boundary conditions remain valid for design changes, which holds only for moderate modifications, as shown in the robustness test. For major redesigns of the EV battery pack, a full-model update may be necessary. Additionally, the current study uses homogenized models for battery cells, which are isotropic and simplified. In reality, EV battery pack cells are anisotropic and multi-component, affecting their mechanical response. Future work could integrate more detailed cell models to improve accuracy. Despite this, the submodel method offers a practical balance, providing faster turnarounds than full-vehicle analysis while being more accurate than isolated battery pack tests for evaluating the EV battery pack.
In conclusion, the submodel-based decoupling analysis method effectively addresses the trade-off between speed and accuracy in EV battery pack crash performance evaluation. By validating consistency with full-vehicle models and demonstrating boundary condition robustness for minor design changes, this approach enables efficient iteration on the EV battery pack structure. The use of explicit dynamics simulations ensures realistic representation of side pole collisions, a critical scenario for EV safety. With computational time reduced to about 20% of full-model analyses, the method supports rapid development cycles without compromising on fidelity. For engineers focusing on the EV battery pack, this submodeling technique provides a valuable tool for optimizing crashworthiness, ultimately contributing to safer electric vehicles. As EV technology advances, further refinements in modeling, such as incorporating anisotropic cell behavior, will enhance the method’s predictive capabilities for the EV battery pack.
To summarize the key equations and relationships, the energy absorption of the EV battery pack can be described as \(E = \int_{0}^{t} P \, dt\), where \(P\) is the power dissipated through deformation. The intrusion depth \(d\) is often correlated with the impulse \(I = \int F \, dt\) applied to the EV battery pack. In submodeling, the error in these quantities due to boundary approximation can be modeled as a function of distance from the intrusion zone, per Saint-Venant’s principle. For practical application, the submodel method for the EV battery pack is recommended for design variations where the change in internal energy \(\Delta E\) is less than a threshold, say 5%, to ensure boundary condition validity. This guideline helps maintain the decoupling advantage while safeguarding accuracy in EV battery pack development.