Abstraction and simulation of EV battery systems—resilience engineering by biological transformation

While the demand for electric vehicles (EVs) is continuously growing, safety issues still remain, specifically related to fire hazards. This research aims to improve the resilience of battery systems in EVs by transferring concepts found in biology to a bioinspired battery system. Due to the complexity of modern battery systems, the biological concepts cannot be applied directly. A simplified simulation battery system for EVs is modelled, which contains the essential battery components necessary to understand both, software and battery dynamics. This is used as a baseline model to study the effects of typical heat-related disturbances. Subsequently, this simulation model is modified to demonstrate the transfer of biological concepts underlying specifically the hypersensitization and vasospasm mechanisms related to wound healing, and to test the effects of disturbances and alterations comparable to damages caused by vehicle accidents. As a battery system’s mass and volume should not be increased by additional hardware, the biological concepts target the interaction within, and the composition of, the system, while leaving single components relatively unchanged. It is found that small bioinspired alterations to the battery system can have significant impacts on their vulnerability to common hazards.


Introduction
The global market of electric vehicles (EVs) increased significantly during the last years with about 12 000 EVs sold globally per year in 2012 to about 12 000 EVs sold per week in 2021 (IEA 2022). Projections show that this market will grow even further by a factor of 20 until 2030 (IEA 2022). Enhancing battery safety is thus a major challenge in terms of the further development and widespread use of EVs (Feng et al 2018). As battery systems are characterized by a high level of complexity, including different chemical and technical aspects (He et al 2012, Kaufman et al 2013, Khalik et al 2021, solutions to solve safety challenges can be manifold and on various abstraction levels.
Nature offers a wide range of inspirations that can be incorporated into different areas of modern technical systems. In the course of biological transformation, i.e., the use of biological materials, structures or processes (Neugebauer 2020), many bio-inspired interfaces and structures have been developed that optimized mechanical properties and enabled new architectural and robotic designs (Huang et al 2019, Shin et al 2019, Fiorello et al 2020, Zimmerman and Abdelkefi 2020, Fratzl et al 2022, Xu et al 2022, Rhee et al 2022. Numerous bio-inspired algorithms have been designed that facilitate artificial systems to selforganize and find the best solutions, eventually leading to artificial or swarm intelligence (Meng et al 2014, Tandiya et al 2019, Filipič et al 2020, Crespo-Mariño and Segura-Castillo 2022. In the manufacturing industry, the transfer of biological principles like recycling and circular flow has made production more efficient and sustainable (Wu et al 2016, Miehe et al 2018.
Despite these successful examples of biological transformation, the application and transfer of biological concepts is not always immediately realizable. The challenges companies and individuals face in this context are manifold. On the one hand, prerequisites and conditions can differ between biological and technical systems. For battery systems commonly used in EVs, the system`s mass and volume play a critical role. An increase in weight is proportional to energy consumption and this to the operating distance (van Hoek et al 2010). Such limitations can often be neglected in biological systems or have been compensated over time by the evolution of other mechanisms. In addition, biological mechanisms are perfectly adapted to the environment and function efficiently under the specifically given conditions. A deviation from these conditions may have an impact on efficiency, which must be taken into account for the technical transfer (Stepankova et al 2013, Kumar et al 2016. A previously published study presented a novel attribute-based method that can be used to transfer biological protection mechanisms to a typical battery system (Bessler et al 2021). Rather than copying the mechanisms directly or attaching biological organisms to the battery system, the mechanisms underlying wound healing and pain reflex response were abstracted and the properties of the human body that enable these mechanisms were identified. Subsequently, these properties or attributes were fulfilled in the battery system by technical modifications. Based on these modifications, certain reactions of the battery system could be adjusted that increased the resilience of the system.
To understand possible safety hazards arising in a battery system, first, knowledge of the components and the system's composition is required. Therefore, some studies investigated how to simplify models and what is the respective impact on computational complexity (Hongwen et al 2012, Khalik et al 2021. In the context of biological transformation, the technical system has to be abstracted in a similar way to the biological one in order to find suitable bioinspired modifications and to implement them. After that, effects and functions of the modification can be analysed. After presenting a simple battery system (section 2) and common safety hazards (section 3), we introduce a methodology (section 4) to simulate an entire battery system that, despite of its simplicity, can reproduce the reaction to relevant safety hazards. Specifically, during this study and based on this simplified system, the heat development in functional and in disturbed battery systems are studied in simulations of the load-cycle process. Finally, the effects of bio-inspired modifications to the battery system are analysed and discussed, showing that small bioinspired modifications can result in a battery system less prone to hazards resulting from temperature effects in typical battery systems.

Battery system
A simple battery system, which serves as the baseline to demonstrate the effects of bio-inspired modifications, and its essential elements are illustrated in the Figure 1. Block diagram of an exemplary battery system as implemented in the simulation. The loading unit converts alternating current coming from the charging station into direct current, which can be stored in the battery unit. The direct current is converted back to alternating current by an inverter, before it can be used by an engine. Red represents the electrical circuit; green lines are representing sensor signals; blue lines are communication wires. block diagram in figure 1 (see also Kaufman et al 2013, Qian et al 2018. The core of the battery system consists of two major parts, the battery unit and the battery management system (BMS).
The battery unit contains the individual battery cells grouped into modules and provides the current supply. Each battery cell is equipped with sensors to measure their voltage U, the current I, and the temperature T of the battery enclosure. During charging and under load, the individual batteries heat up. Hence, the battery unit necessarily needs to contain a cooling system (see figure 2).
To cool the elements, most manufacturers install a pipe system below the cells. A cooling liquid flows through those pipes and thus, during its cycle, the cooling liquid heats up. Hence, electric cars must include thermal reservoirs to reduce the liquid's temperature. Further, a significant part of the energy stored in the battery cells must be utilized to operate the reservoirs.
The second main component of the battery system (figure 1) is the BMS. The function of the BMS as a central control point is quite similar to the central nervous system of a living being. This is the first analogy with respect to an approach aimed at increasing the resilience of the battery system through biological transformation. The BMS maintains normal operation, prevents malfunctions, or acts to restore normal conditions when malfunctions occur. To achieve this, the BMS processes all sensor signals recorded Figure 2. Unmodified cooling system beneath the battery unit. The cooling fluid provides a slowdown of the heating process to all cells. Due to the arrangement of the cooling tubes and the one-side filling, the battery cells in modules 1 and 8 will heat up less than, e.g., the cells in modules 4 and 5 (left side). Modified cooling system with additional valves and drains against fluid leakage, e.g., after a breakdown (right side). originally at the battery unit and then passes them on to the other elements.
The BMS not only controls that the battery unit is charged and discharged in a balanced manner, i.e., that all cells are loaded equally according to their wear, but also that neither overcharging nor excessive discharging occurs. If the battery cells were to operate outside a certain charge interval permanently, they would age much faster and therefore deliver less accurate readings (Perner andVetter 2015, Jongerden et al 2017). If any measured values exceed critical thresholds, the BMS also has the task of activating appropriate emergency switches. This disconnects the battery unit from the power circuit, i.e., the load and the charging unit, in order to prevent additional hazardous heat production. Finally, the battery control unit connects the battery system with other systems in the vehicle, amongst others the crash sensors and temperature sensors. In this way, the vehicle and environmental status can be included in the control system. For simplicity, in the present study, a minimal system as shown in figure 1 is sufficient to test the effects of a disturbance to a battery system leading to a potential hazard and showcase the potential improvements through biological transformation. Additional components or more complex BMS could be added to the simulation but would not change the overall conclusion about the potential of transferring biological mechanisms to technical systems.
In the following paragraph, we describe some of the most common issues to occur after such a disturbance takes place.

Common safety hazards
A battery system is susceptible to faults. Disturbances can not only affect the functionality and efficiency of the system, but also pose safety risks.
Complications may arise in the case of accidents because mechanically damaged battery casings can leak acids and toxic gases. In addition, short circuits may occur at breakage and compression points, which can generate excessive heat and discharge sparks (Kim and , Chen et al 2021, Lai et al 2021. In the worst case, this can cause parts of the system to catch fire, which in turn can destroy battery casings and allow further acids and gases to escape (Jin and , Lai et al 2021.
A significant safety risk is excessive heating. When current flows inside the individual battery cells, i.e., when their state of charge changes, the chemical processes generate heat. It is in the interest of manufacturers to make long-distance travel as attractive as possible and not to delay the charging process with additional cooling breaks. However, rapid recharging leads to heating just as well as continuous discharging, which might lead to overstressing the cooling system.
High loads are particularly critical in the case of high ambient temperatures, easily occurring on a hot summer day. Also very low ambient temperatures might cause malfunctions, because if the internal temperature of individual battery cells falls below −30 • C, chemical bridges can form in the cells (Abada et al 2016). When the cells are loaded again, short circuits can occur, which can lead to overheating. Of course, modern vehicles have very reliable cooling systems, so such critically high temperatures are only to be expected due to severe malfunctions caused by, e.g., car crashes or other physical accidents. However, in general, the heating is proportional to the load. Therefore, it is controllable both in terms of the speed and in terms of the amount of temperature rise.
In the worst case, heat development can get critical and lead to a thermal runaway which cannot be stopped or slowed down. If the cooling system is not sufficient to counteract the heat development anymore, the chemical processes that regularly take place inside the cells increase the reaction rate, triggering further exothermic reactions that in turn generate heat (Jongerden et al 2017). Since this heat increases the reaction rate, a cycle of heat generation is created. The heat production then no longer occurs due to stress, but due to exothermic chemical processes.
A failure of the BMS is particularly problematic since the entire control system is centralized here. If the BMS fails, there is no longer any recording of the measured values and thus no more monitoring of parameters indicating potential thermal runaway. There is no more a system to trigger the emergency switch and to interrupt charging and discharging processes, which can irreversibly damage the battery cells due to overcharge and excessive discharge.
Since the methodology described in the following section aims at mitigating safety hazards that impair the entire system, we focus on system-wide issues and neglect challenges in the construction of single battery cells.

Methodology
The following section describes how the charging and discharging process of a common battery system can be simulated and how individual components such as a cooling system can be implemented. This is the baseline model for the subsequent modifications inspired by biological principles. The simulation is performed using the MATLAB software tool Simulink and its extension Simscape. Combining both tools enables the mapping of the battery control unit, the electrical components and the cooling system in just one simulation (The MathWorks 2019).
When simulating the battery system's power source, the pros and cons of including an alternating current source could be debated. An alternating current power source depicts the most common power outlet, but also increases the simulation time immensely, since e.g., a typical sinoid curve with a period of T = 2 · 10 −2 s requires very small simulation steps. If the simulation instead includes a direct current source, which EV owners have access to at most public charging stations, the simulation time can be reduced approximately twenty-fold.
The complexity of the charging unit also increases in models that use alternating current. In the DC model, the charging unit only needs a switch to disconnect the charging current, whereas in the AC model at least a transformer, a capacitor and a bridge circuit of diodes is required (see figure S1). While there are no significant differences with respect to the aims of the study, the computational effort is less for the DC model compared to the AC simulation.
The loading unit provides current to the battery unit, of which the simulation is described subsequently. Since even the exact simulation of a single battery cell, including its material characterization and the chemical processes inside, can be very resource-intensive, it is advisable to use equivalent circuit models in the simulation. These models can very well depict the dynamics of charging and discharging as well as the ambient temperature rise but neglect chemical processes and temperature gradients within the individual cells. To keep the effort low, the model presented in (Kaufman et al 2013) was chosen to simulate the individual cells. The values to be set for the resistors and the capacitor, contributing to the equivalent circuit model, were also determined empirically in (Kaufman et al 2013) and adopted here. In the simulation presented here, the battery cells are divided into eight modules connected in parallel, which in turn contain 12 individual cells connected in series (modelled after a BMW i3 battery system, see Schoewel and Hockgeiger 2014). Table 1 shows the data for the battery unit. When executing a simulation, it might be noticeable that the maximum achievable performance in the simulation is significantly below the performance of a comparable vehicle. This is often not due to incorrectness, but is explained by the Peukert effect (Peukert 1897). According to this effect, the power P taken from the batteries is proportional to the current I and the amount of time t during which it is discharged. As a consequence of the Peukert effect, it is possible to achieve a higher power output if the battery unit was discharged over a longer period of time. To reduce the computational effort in the simulation, a high discharge rate was chosen, reducing the available power of the unit in the simulation.
As most common, the individual cells in our simulation are wired in series, while the modules are connected in parallel. The fact that the battery cells within the individual modules are wired in series ensures that all cells are charged and discharged with the same current intensity and thus no irregularities occur even if the internal resistance of the cells deviates (e.g., due to aging or production irregularities). In addition, the series connection makes it possible to keep the current intensity within the modules low but has the disadvantage that higher voltages are required for charging. In addition, the parallel connection of the modules ensures that if one module fails, the remaining modules can continue to operate, and the total voltage required for charging is not increased any further. While the parallel connection of the modules achieves the desired high capacities C, the total current of the battery unit also increases, which is a safety risk especially in case of accidents. More complex ways of interconnecting the battery cells or reconfiguration models to increase efficiency and safety are, however, the focus of various research projects and can therefore be expected to be featured in future vehicles (Eckstein et al 2011, Huria et al 2011, Jin and Shin 2012, Abada et al 2016, Savard et al 2018.
One of the system's most important elements is the emergency switch, as there are potentially fatal scenarios, which can be terminated by its activation. The position of the emergency switch is chosen so that it can disconnect both the charging unit and the load from the battery unit. Since neighbouring cells usually have similar temperatures due to the heat exchange between them, in our simulation, the emergency switch is triggered as soon as one single cell reaches the temperature T = 90 • C. The emergency switch can also be triggered if threshold values of the voltage and current are exceeded. In general, the measured values are strongly interdependent and influence each other's behaviour. For example, high temperatures favour high currents and a rapid voltage change heats up the battery cells. Therefore, if one of the measured values is incorrectly processed, it is to be expected that the emergency switch will nevertheless be triggered promptly.
Because of the already high complexity and resource intensity of the simulation presented here, the batteries are simulated as thermal mass points. In the simulation, the individual cells are thus cooled regularly instead of only from below as in reality. Thus, the cooling system in the simulation is slightly more effective and heat-caused fatigue effects to the battery cells affect the entire cell regularly. It is however possible to consider that the cooling liquid heats up after passing through each cell. The cooling system cools the cells in the back of the battery unit, meaning the cells placed the furthest away from the pipes' entrance (e.g., cells of module 4 and 5, see figure 2), less effectively. In the simulation, the cooling system does not require any energy, which is admittedly not compatible with reality. Therefore, an additional power source is inserted to extract electrical charge from the battery unit. The amount of charge withdrawn is proportional to the deviation of the temperature of the cells from the temperature of the cooling or heating fluid.
For a cooling effect, the corresponding current was calculated by and for heating effects with Here, α is a scaling factor (unit Ampere per degree Celsius) with which the current can be manually adjusted to the desired magnitude. In the simulation, a simple combination of Simulink and Simscape blocks is sufficient to implement the current demand of the cooling system (see figure S2).
At the end of the battery system, there is an inverter connecting the system to the load. To generate alternating current, a simple resonant circuit consisting of a coil, a capacitor and a resistor was chosen. However, in order not to increase the simulation time any further, the generated frequency was chosen significantly lower than the AC frequency fed into the charging unit. Similar to the elements presented before, the simplest possibility was chosen, which still fulfils its function-the generation of alternating current-but does not fully correspond to a realization in reality (see figure S3).
The BMS is divided in this study into five subsections. The first section is for the voltage management.
Here the voltages applied to the individual cells are recorded, the total voltage of the modules (identical to the total voltage of the battery unit) is calculated and the voltages of the weakened and unweakened cells are compared.
The next section of the BMS records the current I of the battery cells. This is necessary even though the BMS records the voltage U, because both values lead to the internal resistance R. R is not constant but increases depending on the degree of wear of the cell. In addition, a part of the load management, namely the determination of the discharge current intensity and the pulse duration, is carried out in this section.
The third section is responsible for the calculation of the performance of the system. For this purpose, the voltage U of the battery unit is multiplied by the current I flowing out of the battery unit during the discharge phases (P = U × I). The total voltage during a simulation period is then obtained by adding the power of the individual discharge intervals.
Next, the fourth section is dedicated to the determination of the state of charge. Here the individual values for each cell are recorded as a basis for the maximum-, minimum-and average state of charge calculations. With a homogeneous composition of the battery unit and adequate cooling, these three values should hardly differ from each other. These values are used in the simulation to define the charge and discharge cycles. The charging unit is automatically disconnected when the minimum state of charge reaches 100% and reconnected when the maximum state of charge drops to 10%. For models where the maximum and minimum states of charge are very different, it is recommended to use the average state of charge instead. In this way overcharging and excessive discharging can be limited and at the same time the loss of power can be kept low (stopping discharging at an average state of charge of 10% occurs earlier than stopping at a maximum state of charge of 10%). In our simulation, Simulink-blocks automatically initiate the charge and discharge cycles (see figure S4).
Besides, the user can adjust the desired length of the breaks between the processes and the stateof-charge (SoC) levels at which the charging or discharging is terminated. The fifth and final section of the BMS is dedicated to the temperature management. Here, Simulink-elements record all temperatures, define the cooling and ambient temperature, and determine the maximum, minimum, and average temperature. Since a value only needs to be recorded during a simulation step, instead of being processed further immediately, algebraic loops can be avoided, and the overall simulation time can be significantly reduced be adding delays.

Simulation results of a typical discharge cycle
The voltage applied to the battery unit is one of the central measured values. Since the modules are wired in parallel, the voltage at each of the modules is the same. The voltage curve shown in figure 3 shows a typical discharge process.
At the beginning, the voltage decreases continuously and slightly, which is explained by the correctly accounted energy consumption of the cooling system. After this phase, the cell is discharged in regular short pulses, which is recognizable by the steep drops of the voltage curve. Between these short discharge pulses there are pauses (no acceleration while driving at constant speed). It is easily observable that the height of these plateaus decreases after each discharge. The BMS uses the measured decrease of the height of the plateaus, the internal resistance and the temperature of each cell to determine the state of charge as described in (Kaufman et al 2013). Another important measured value is the SoC, shown in figure 4.
A total of four exemplary charging and discharging processes with intermediate rest phases are shown here. In a typical simulation cycle, the loading unit first charges the battery unit from its ideal storage state-all cells at approx. 40% state of charge and approx. 20 • C case temperature-to 100% state of charge.
After charging, there is a short pause, during which the cooling system remains activated, and a significant cooling of the entire battery unit is observable (figure 4). The load then gradually discharges the battery unit to an average state of charge of 10%. Discharging is not carried out constantly here, since it is assumed that the charge is mainly used during acceleration processes, but not during constant speed or braking manoeuvres. As observed before, when looking at the voltage curve, during the rest phases, a decrease of the SoC due to the energy demand of the cooling system is visible. After the charge and discharge processes are repeated four times, the battery unit is finally completely discharged, and the cooling system is disconnected. Theoretically, with this configuration, it would be possible to reach a state of charge in the negative percentage range, which is from a physical or mathematical perspective incorrect. As mentioned before, this is due to the fact that not the entire capacity of the battery cells is used and overdischarge, however detrimental to the life time of the cells (Gao et al 2018), is possible. The state designated as SoC = 0% is only defined so by its definitionthere is still charge in the battery cells, but the SoC reaches a level deemed too low for further discharge. In addition, it is worth mentioning that, when looking at the SoC values of all 96 cells, there are slight differences. Especially during the last cycle, not all cells are equally loaded. This is because the arrangement of the cells and the arrangement of the cooling system cause temperature fluctuations-an effect that is intensified with continuous loading. This effect is realistic and thus it is appreciated that the simulation is able to depict it.
Further, it is observable that modules and cells in the centre of the battery unit heat up significantly more than cells on the outer edge. As expected, the cells in module 1 and module 8 heat up the least, since this is also where the inlet of the cooling system is located (see figure 2). When looking at the average cell temperature (figure 5), one observes that the average temperature remains within an acceptable interval, even though the cycles can be considered extreme due to the short breaks and consecutive rapid recharging and even though the temperature is by no means constant.
In the following, the temperature measurement will be used as a representative parameter to assess the resilience of the battery system. Ideally, the individual battery cells should always be operated within a temperature interval of approx. 15 • C to approx. 55 • C (Doughty and Roth 2012). The following function describing a temperature factor F is defined for each of the cells: T: Temperature battery cell α : Constant t: Time (t 0 starting time, t f final time) T ideal : 25 • C, ideal cell temperature T min : Lower limit temperature interval, here 15 • C T max : Upper limit temperature interval, here 55 • C F: Temperature factor θ (x): Heaviside Theta-function If the temperature of the battery cell is within the comfort interval, both Heaviside functions result in zero. The function therefore remains constant. This can be seen in figure 6.
However, if the temperature is lower than T min , the first Heaviside function becomes one; if the temperature is higher than T max , the second becomes one. Therefore, in these periods, the integral grows. Further, the magnitude function guarantees that the value increases the more the temperature deviates from T ideal . The calculation of this factor can be done directly in the simulation (see figure S5) or by exporting the temperature measurement to a regular MATLAB file.
The presented function describing the temperature factor evaluates the ability of the system to operate within the desired temperature interval. If the function remains low during the whole simulation time, the system fulfils the ability to not heat up in a dangerous way. However, if the temperature factor F increases, it is indicated that the safe temperature interval has been left and, thus, that there is a possible safety risk. In the following, the battery system is considered to be resilient if it exhibits a low temperature factor F, even during disturbances and stress situations, i.e. if the system has the ability to maintain safe operation in extreme situations.
Later in this text, we will compare the unmodified battery system's factor F to two systems' factors F, which contain alterations inspired by biological principles, which excel in their flexible nature (Bessler et al 2021). The alterations may be defined as passive as they do not impair the system's regular operation and do not increase the system's physical mass in a significant way.
Prior to introducing the changes aiming to increase the resilience of the battery system, we will present a simulation circle affected by disturbances, its effects on the system and a possible method to simply portray this in a simulation.

Simulation of loading cycle subjected to disturbances
In order to evaluate the resilience of the system, two situations are introduced whose effects are comparable to those of an accident.
The first failure scenario does not originally assume the complete failure of elements, but only irregular heat generation in parts of the battery unit. The direct damage caused by the scenario is therefore initially limited and the functionality of the system is only slightly affected. This creates a situation from which potentially greater hazards can arise, but which also offers a starting point for mitigating measures.
Here, 25% of the battery cells, divided into a whole module and half of each of the two directly adjacent modules, are affected. The simulation was created in such a way that the completely affected module can be randomly selected at the beginning of a simulation run. However, in order to be able to compare the effects of the scenario in different models, the irregularity was fixed to module 6 and thus also to the lower half of module 5 and the upper half of module 7 (see figure 2). Heating starts about 30 000 s after the start of the simulation, during the second discharge at an average state of charge of the battery unit of approximately SoC = 80% and follows the function After the scenario has started, the charging and discharging cycles are continued until the additional heat development reaches a maximum temperature of 90 • C. This temperature triggers the emergency switch and disconnects any load. However, since there is an irregularity, actuating the emergency switch only eliminates the heat generated by operation, not the additional heat due to the misfunction. Another area to which the scenario extends, is the BMS. Even so, to be able to observe the damage caused by the disturbance, all measurements should be recorded correctly, it is valuable to look at a partly failing BMS, as this illustrates the risk posed by the presence of a single central control unit. If the entire control system is concentrated in one place, as is common in most battery systems, physical damage to this location means that the entire control system will fail (Savard et al 2018). Further risks arising from centralization are discussed (He et al 2016, Erhard 2017, Han et al 2020. The damage to the BMS in the simulation is independent of and in addition to the heating. It affects the processing of the measured values. In detail the parts of the BMS listed in table 2 are affected by the damage. An 'affected' signal still exists and is recorded for external observation, but, in the simulation, is superimposed by a sinusoidal interference function. Thus, the correct measurement is no longer available to the BMS, just as during a real misfunction of the sensors. Besides, if a battery cell reaches a threshold temperature of 220 • C, it enters the phase of thermal runaway (see figure S6). In the simulation, this can be seen from the fact that the temperature rise from this value is exponential and reaches several 100 • C within a few seconds. To simulate this, a simple addition, i.e. the thermal blocks shown in figure S7, was made to the battery cells as presented in (Kaufman et al 2013).
Compared to the first scenario, the intensity of the second scenario was chosen significantly higher; the time of the accident scenario was kept at approx. SoC = 80% during the second discharge. Again, 25% of the battery cells should be affected, but now the sensors also fail. This is realistic if the cells are physically destroyed, because in reality the sensors are located directly at the cells.
The 'destroyed' battery cells should no longer supply electricity, but increase their heat production. In the simulation, therefore, they are replaced at the time of the 'accident' by a heat source and a simple power source. This simple power source emits short pulses at regular intervals to imitate short circuits or flying sparks. The BMS is involved in this scenario to the same extent and in the same way as in the first scenario (table 2). Compared to the previous scenario, the cooling system is now more gravely affected by this scenario, since it is located directly next to the battery unit. In order to transfer the scenario to the cooling system, the cooling for the entire module is immediately suspended for the three affected modules. Fifty minutes later, the remaining cooling is deactivated to resemble leaking coolant and destroyed cooling pipes in the simulation.
Here, after the damage has occurred, the charging and discharging cycles are continued until the emergency switch is triggered. In this way it is possible to observe the system's ability to operate with damaged components. However, considering the scale of this scenario, a continuation of the cycles is unrealistic. After an accident resulting in the deformation of large parts of the vehicle, it can be expected that most drivers stop and not resume driving.
Finally, it should be mentioned that the simulation of this second accident scenario also influences the normal mode. The power sources that replace the destroyed cells must be present in the simulation from the very beginning. When the scenario is started, the system only switches between the battery cells and the power sources, but this does not work without increasing the overall resistance (The MathWorks 2019). Even if the scenario is to occur at a fixed point, it is therefore important to insert the option in each module so that all modules have the same capacity up to the scenario initiation time. This concludes the description of the two disturbances. Now, we will introduce two biologyinspired alterations to the battery system that attempt to increase the system's resilience. Then, we will compare the actual simulation results including the disturbances.

Biology inspired alterations
In this section, the modifications of our battery system derived from biology are presented and their influence on the resilience of the system is examined. In order to make the changes more attractive, the modifications were selected so they are reactive, or even passive protective measures, i.e. they affect the normal operation of the system as little as possible and do not involve any loss of performance or the development of new elements (Bessler et al 2021).
As an approach, our study looks at healing processes in the human body, which are always present and activatable. Still, as long as there is no injury, their maintenance does not strain the organism or requires significant amounts of energy. Besides, most healing processes are highly flexible and thus display a high level of resilience (Scheithauer andRiechelmann 2003, Reinke andSorg 2012). If the underlying principles are studied and completely understood, they may be transferred to technical systems, which thus benefit from the same flexibility. The conceptual derivation of the bio-inspired modifications were previously described by (Bessler et al 2021).

Hypersensitization of the temperature measurement
As the first modified model, the effects of a modification of the battery system corresponding to the biological protection principle of hypersensitization (Arcourt et al 2017, Bessler et al 2021 are tested. These modifications will influence the first accident scenario, of which the biological counterpart is an inflammatory reaction (Basbaum et al 2009).
Positive with regard to the implementation potential of this modification is that it only affects signal processing and does not require any additional hardware (no additional costs), so that the performance of the battery system remains unaffected in normal mode, i.e., when no unusual scenario occurs.
In the BMS of the modified model, a total of two extensions are installed. The first one serves the detection of a disturbance, the second one the reaction to the disturbance (since the time of the accident can be defined in the simulation, detection is only required for the actual BMS). The detection was based on the biological process of chemotaxis (Scheithauer andRiechelmann 2003, Bessler et al 2021). In normal mode, with an activated cooling system, it is expected that the temperature of each cell will only increase when charge is either added or removed. If, however, a positive temperature gradient is measured during a resting phase, this is sufficient as a signal to trigger hypersensitization, i.e. the reaction. In its abstracted, Figure 7. Diagram showing the abstraction of hypersensitization as applied in the battery system. If the hypersensitization is active, e.g., after the crash system registered an accident, even a light stimulus is sufficient to activate the emergency switch. transmitted form, this hypersensitization proceeds as follows (see figure 7): If the BMS measures a weak stimulus in the normal state of the non-modified battery system, i.e. it records a temperature of more than 60 • C, there is no reaction, because the information is immediately inhibited again-the temperature is sufficiently high to leave the 'comfort interval' , but it does not yet justify an activation of the emergency switch (upper diagram figure 7). For this, a strong stimulus would have to be measured in normal mode, i.e. a temperature of at least 90 • C (lower diagram figure 7).
However, if the hypersensitization is activated, all stimuli are additionally amplified. It is therefore possible that the emergency switch is already triggered at a maximum temperature of less than 90 • C. However, this does not mean that a slight increase in temperature is sufficient. The system must already be in an alarmed state and even then, the maximum temperature must still exceed a threshold that can be set by the user (the simulation) depending on the desired degree of sensitization (see figure S8).
In order to achieve a higher stability against false signals, an inhibiting signal was built into the simulation. Since the simulation only considers the battery system and not the entire vehicle, the hypersensitization currently only has the application of triggering the emergency switch. An integration into the overall vehicle system in the sense of an early warning system would be, however, easy to implement.
If it is possible to ensure that hypersensitization is only active in an actual hazardous situation, its simplicity can make it a valuable supplement.

Intelligent bioinspired cooling system
To demonstrate the effect of the transfer approach for the exemplary case of a major accident (second scenario), a vasospasm-oriented cooling system (Scheithauer and Riechelmann 2003, Reinke and Sorg 2012, Bessler et al 2021 is now described. Earlier published aspects of the study looked into healing processes in the human body. That work not only evaluated those processes' contribution to bodily resilience, but also investigated which healing processes were most suitable for abstraction and application to a technical system (Bessler et al 2021). Here, the cooling system is extended by the mechanism of vasoconstriction through additional drains and valves. If a part of the pipes is damaged in a way that coolant can leak, the valves located before the point of rupture are closed and the corresponding additional drains are opened. This allows the liquid circuit to be maintained and at the same time prevents coolant leakage.
As observable in figure 2, this modification has the disadvantage that intact modules and cells behind the valves can no longer be reached by the coolant. This effect is minimal when mainly module 4 and 5 are affected and greatest when either module 1 or module 8 are affected. In the simulation, it would be easy to design a cooling system that can restart after the interruption, but this is difficult to implement in the real system. Due to the limited space available inside a vehicle, only one cooling reservoir can be used and the number of pipes must be kept low.
The cooling system extension also includes a cooling control unit separate from the battery control unit. In the modified model, all tasks previously associated with cooling are decoupled from the BMS and the new tasks created by the valves and additional drains are transferred to a cooling management system (CMS). With a real conversion it is meaningful to place the CMS also spatially separated from the BMS so that a higher chance exists that, in the case of an accident, one of the systems remains. In addition, the resilience can be increased by merely deactivating the tasks in the BMS that were transferred to the CMS and then switching them over again if the CMS fails. For this to be functional, the CMS and the BMS must be interoperable.
Before the improvements achieved are presented, the control of the valves is briefly discussed. If j is the index number of the completely failing module, then the position of the accident is defined with the simple vector where 'cl' means that a valve must be closed and 'op' means that a valve should be opened. Multiplying the matrix M with the 'accident position vector' ⃗ a, results in a new vector⃗ r, which describes the corresponding reaction:

Results and discussion
In this section, we look at the simulation results during disturbances of the unmodified battery system and the two systems modified according to bioinspired principles. Firstly, the average cell temperature is consulted (see figure 8).
All systems display an increase in temperature once the disturbance occurs. It is, however, easily observable that both alterations significantly reduce the increase in temperature.
In the non-modified models, it can be observed that a temperature increase occurs shortly after the time of the accident, which is due to a failure of the cooling system (completely leaked coolant). In the modified models, this increase takes place much later and is due to a complete discharge of the battery unit. The cooling system is still sufficiently intact, but there is no energy left to run it. It is noticeable that the Figure 8. Increase of the temperature T due to the perturbations to the battery system. The continuous line stands for the unmodified reference system to be compared with the hypersensitization. The dashed line includes the modification hypersensitization. The dotted line is the second reference system corresponding to the second modification-the modified cooling system (dashed and dotted line). performance of the modified models is significantly lower than that of the standard models-the charging process is slowed down and the overall power output is lower. Especially when looking at the hypersensitization modification and its reference system, we observe that the performance of the modified system is decreased. In the modified system, the emergency switch is triggered much earlier, which is intended, since the present situation can lead to a thermal runaway (table 3). Further, the unmodified system is able to conduct an additional charge and discharge cycle. With regard to the temperature factor, a significant improvement can be observed, when comparing the modified systems to their reference systems. Figures 8 and 9 show that modifying the cooling system permanently reduces the maximum temperature, after the accident has occurred. It is also noticeable, however, that the temperature difference in both cases is negative before the accident, which indicates that the maximum temperature is increased by the modifications. Although the modified cooling system has the ability to continue working even if some cooling pipes are damaged, it is less efficient due to the additional valves and the energy requirements of the CMS.
This effect could be reduced by optimizing the valves and the CMS, which is subject to further research. While the temperature factors improve significantly in all accident scenarios, performance decreases in the first accident scenario and remains  table 3). This is an indication that the selected modifications do not help the system to return to normal operation.

Conclusions
The research reported herein showed a method how a complex battery system, as commonly used in EVs, can be abstracted and simulated in a suitable way. It was aimed to investigate the potential impact of a failure within the battery system on its performance and safety. Simulations of the charging and discharging processes of the battery showed a significantly higher temperature profile in failure scenarios compared to normal operation. In addition to other misuse failures like mechanical damage, overheating of battery cells can eventually lead to a so-called thermal runaway effect, which poses a major threat not only for e-mobility but also in other domains with increasing number and capacity of batteries. This highlights the requirement of improved safety measures to counteract the hazards raised by higher energy densities. This publication has shown that biology can offer a remedy for such safety-related challenges and that biological safety concepts can be transferred to technical systems. For the transmission method, however, it is necessary to abstract both the biological and the technical system to a similar level. In case of the battery system, modelling the essential elements of the system enabled the simulation of charging and discharging processes, the analysis of failure scenarios, and finally the implementation and investigation of bio-inspired modifications. In the exemplary battery system, we demonstrated that bio-inspired modifications, abstracted from hypersensitization and vasospasm mechanisms in wound healing processes, can improve the battery system safety without significant impact on the normal-state operation.
Technical as well as ecological challenges can be addressed by biological transformation. Both the sustainability and the performance of biological models provoke hope for optimized resource consumption and climate-friendliness of future technologies. Even if the desired effect of the biological transformation is not significant, advantages can often rise in other areas. For example, despite the decrease in performance of the battery system in the simulation, it can be argued that an increase in safety, due to the increased reliability associated with it, can also result in an increase in performance, and vice versa, an increased performance, due to the additional resources available, also promotes safety.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.