Lithium-ion battery systems are an energy source for a variety of electric-vehicle applications due to their high energy density and low discharge rates. Battery packs, whether made of prismatic, cylindrical, or pouch cells, are cooled by common automotive thermal management systems.
The rapid detection of battery pack coolant-system leaks during production operations is essential for meeting necessary safety and service-life requirements. Industry standards for measuring leak rates for both glycol-based and refrigerant-based cooling systems, however, currently do not exist.
This article discusses how leaks in water-glycol cooling circuits can be detected reliably and quantitatively through detection of escaping test gas as an indicator of ethylene glycol leaks and how the test gas leak rates correlate to the liquid leakage of the cooling liquid. Influencing variables such as leakage channel diameter, pressure difference, and viscosity are considered, and go/no-go leak rates are described.
In this paper, the tightness requirements that are necessary in a coolant circuit that is not operated with pure water are considered, but with a mixture of water and glycol. Two typical automotive applications here are the cooling of the engine block in an internal combustion engine and the cooling circuit in a traction battery box for cooling the battery cells.
The requirement for leak tightness of the coolant circuit in the engine block is defined as less critical than the requirement in the cooling circuit of traction a battery enclosure. In the application case of the engine block to be cooled, coolant loss must not exceed specific limits prior to any required replenishment. In the case of the application for cooling a battery enclosure, the requirements are defined much more critically. Here, damage or short circuits to the battery cells must be prevented. Leakage coolant from the cooling circuit can cause a battery fire.
Independent of the leak tightness required in each respective application regarding the loss of liquid, a requirement must be defined for the leak tightness during leak testing with test gas. In this paper, the smallest acceptable cross-section or diameter of a leakage channel for the coolant glycol is derived and the leakage rate value to be assigned for the test gas leakage test is given. Compared to the requirements regarding IP67, three essential differences must be considered in the application case of a coolant circuit.
In the coolant circuit, overpressure of up to 5 bar prevails during operating conditions, whereas the requirements for IP67 generally consider an effective force at the leakage channel corresponding to a pressure difference of 1100 mbar against 1000 mbar. With increasing pressure difference at the leakage channel, the leakage rate increases correspondingly with the same leakage channel geometry and significantly more medium leaks out than under the IP67 test conditions and thus requirements for test criteria for testing a coolant circuit must be defined more strictly.
Furthermore, the temperature in the coolant circuit is significantly increased during operation, which in turn affects the viscosity of the medium. As the temperature increases the viscosity decreases, which in turn increases the leakage rate.
The difference in temperature from room temperature to the operating temperature in the coolant circuit changes the viscosity by up to an order of magnitude, which correspondingly increases the leakage rate.
Third, the property of surface tension or wetting angle of the liquid in a leakage channel and its wall affects the channel geometry of both the leakage flow, which may be prevented due to blockage of the leakage channel. Thus, when setting rejection limits for leak testing, the property of the liquid medium used must be considered.
Theory of Leak-Channel Behavior
The blocking of a leak channel with a liquid, e.g., a water-glycol mixture, depends mainly on the surface tension (σ), the contact angle (θ) between the solid-state material and the fluid, and the maximum overpressure (p).
The leakage channel radius at which a liquid can no longer escape from a leakage channel is described in the appendix. Using equation (1), the leakage channel radius at which the leak channel is blocked by the liquid due to capillary forces and prevents the liquid escaping the tube may be calculated.
p = pressure inside the drop of liquid
σ = surface tension of the liquid
θ = contact angle Theta
r = radius of the leak channel
In the case of a cooling system, the maximum overpressure (p) varies typically between 2.5 to 5 bar overpressure, depending on the cooling system. The surface tension (σ) of pure water and ethylene glycol is given with 72.7·10-3 N/m and 48.0·10-3 N/m.
In our experiments, glass capillaries are used because only glass capillaries are available in such a range of small inner diameters. In our previous SAE-paper a contact angle for glass capillaries of 25° for pure distilled water was used.
However, this paper describes the wetting properties of different liquids: pure distilled water, a water-ethylene-glycol mixture and pure ethylene glycol. Thus, we use a slightly different contact angle for glass, substituting the contact angle of quartz. Instead of a contact angle for glass of 25° we used the contact angle for quartz of 29° since glass has a quartz content of 80 percent.
Using this equation, the leakage channel radius at which the glass (quartz) leak channel theoretically is blocked by using pure distilled water and pure ethylene glycol can be calculated.
Figure 1 shows the results of the blocked leak channel diameter at different overpressures of a cooling system using derived surface tension and contact angles for pure distilled water and pure ethylene glycol.
The contact angle (θ) between the solid material, the liquid overpressure (p) and the surface tension (σ) of the liquid affects the blocking behavior of a leakage channel. Thus, the hole’s diameter from which a leakage channel blocks with a liquid essentially depends on the material used, the surface tension of the cooling liquid, and the overpressure of the cooling system.
If the radius of the leak channel is bigger than the calculated radius, fluid will escape and a leak channel has been formed. If the radius of the leak channel is smaller or equal to the calculated radius, the leak channel is blocked by the fluid due to the capillary force and no fluid will escape. Thus, the cooling system is leak tight if the radius of leak channel is smaller or equal to this calculated radius.
To simulate a leaky cooling system, a special test setup was developed in which glass capillaries with a length of 30 mm and different inner diameters were adapted. With this test setup, it is possible to count the number of liquid drops from different leak channel diameters.
The experimental setup shown in Figure 2 consists of a liquid reservoir, a water pump, several stainless-steel pipes and adapters, an inspection glass with a paddle wheel inside, a pressure gauge, five glass capillary slots, a pressure regulator, and several plastic tubes for connecting the individual components.
By reducing the flow rate at the pressure regulator, a pressure difference to the atmosphere of 1 bar, 2 bar, 3.5 bar, and 5 bar were set, which were read on the pressure gauge in the middle of the adapter construction. The different coolant compositions are distilled water, a 1:1 solution of distilled water and ethylene glycol (EG), and 99 percent ethylene glycol.
Over time, different numbers of drops develop at the glass capillary ends during the experiment. These drops drip off once reaching a certain size. The number of dripping drops provides information about the leakage at the various capillary diameters at the various pressures.
We conducted experiments to determine if theory fit practice when pressurized fluid is expressed through a variety of orifices. Both distilled water and a mixture of water and water/ethylene glycol were used as these are typical automotive coolants. In theory, as water/ethylene glycol has a different viscosity than distilled water, we expected that pressure and the inner diameter of test capillaries would produce significant differences.
The first experiment looked at the number of droplets dripping off a glass capillary during one hour in the system with distilled water as the coolant at various operating pressures at 20 °C. It can be seen that the system pressure has a significant influence on the number of drops, just like the inner diameter of the capillaries.
It is also noticeable that with the 5μm capillary, no drops fall off during the measurement time and no visible droplet emerges from the capillary tip. With the 10μm capillary, at an overpressure of 5 bar against the atmosphere, only one drop drips off during the measuring time. At 2 bar and 3.5 bar overpressure one droplet develops at the capillary tip but does not drip off.
The number of drops as a function of the overpressure shows a good linear correlation as expected from the Hagen Poiseuille law for liquids. In a second experiment, the dripping behavior of the various capillaries with a solution of distilled water and ethylene glycol in a mixing ratio of one-to-one at various operating pressures at 20° C was investigated. It can be seen that the number of drops has clearly decreased in contrast to the previous measurement with pure distilled water.
Here, within the measuring time of 60 minutes, not a drop fell from the 10μm capillary for every pressure difference. In the case of the 5μm capillary, no drop formation could be determined visually. Transferring the measurement results into a double logarithm scaled diagram (Figures 3a and b), the dependence of the drop rate to the fourth power of the radius can also be seen in a good approximation, just like the influence of the pressure in the system which is described in the theory.
Water has a particularly low dynamic viscosity of 1.002 mPa/s and pure ethylene glycol has a high dynamic viscosity of 19.83 mPa/s at a temperature of 20° C. A water-ethylene glycol solution of 1:1 (50 percent volume distilled water and 50 percent ethylene glycol) has a dynamic viscosity of 4.1 mPa/s.
Summary and Conclusions
The measurements show very clearly that the enumerated liquid droplets or liquid leakage rate is linearly dependent on the pressure difference, with the pressure always being at atmospheric pressure. This linear relationship is very clear in Figure 4. In addition, it is easy to see in the graph that the magnitude of the leak rate decreases as the viscosity of the liquid increases. The leakage rate for glycol shows a factor of 10 times lower and for the 50/50 water glycol mixture a factor of 2.3 lower than for pure water, all other parameters being equal.
As the operating pressure of the coolant in the cooling circuit increases, the amount of liquid leakage increases linearly. Temperature influences leakage rate because the viscosity of the coolant decreases with increasing temperature. For example, the leakage rate for glycol as a coolant increases by a factor of five when the operating temperature rises from 20 °C to 80 °C. This is true over long periods of time. In terms of longer periods of time, even a few drops of loss over the time scale of minutes or hours mean significant loss of coolant during a year and should be avoided accordingly.
However, in the case of the cooling circuit of a combustion engine under operating conditions, especially at high operating temperatures, losing a few drops of coolant does not have a harmful effect on the immediate environment because of the evaporation rate.
Unfortunately, the same cannot be said of battery cells, modules, and packs. The harmful effect of even a small amount of liquid, droplets or as water-bearing vapor, is much more dangerous to any traction battery module or pack. There, the liquid coolant or water vapor can destroy battery cells or generate a short circuit. The leak test of cooling circuits of traction battery packs and for fuel cells must be done with higher reliability compared to tests of cooling systems of combustion engines.
The corresponding limiting leakage rates for such leakage channels are presented in the paper; the leakage rates specified here for gas pre-testing are an order of magnitude significantly below the detection limit of classical test methods such as pressure decay or mass flow leak testing. However, when using modern test gas methods with forming gas or helium as the test gas, reliable detection of critical leakage rates is possible.
This article was written by Marc Blaufuß, Application Engineer, Leak Detection Tools, and Daniel Wetzig, Research Manager, both at Inficon GmbH (Köln, Germany). This article is a condensed and edited version of SAE Technical Paper 2022-01-0716. For more information, visit here .