Thermal Optimization of a Planar Wide-Range Oxygen Sensor

Thermal Optimization of a Planar Wide-Range Oxygen Sensor

C. Scott Nelson
Delphi Automotive Systems

Introduction

Exhaust sensors, and in particular oxygen sensors, have been used in automotive exhaust systems for over 25 years. Typically, these sensors have been stoichiometric or switching sensors. Switching sensors will only indicate whether the air/fuel (A/F) ratio is rich (excess fuel) or lean (excess oxygen), but they do not tell how rich or how lean. With the advent of increasingly stringent emission regulations, and lean burn (gasoline 2- and 4-cycle direct injection as well as diesel) applications, there is a need for knowing the precise air/fuel ratio or oxygen content. A linear or wide-range air fuel sensor has the ability to measure air/fuel ratios between 9:1 to air.

Since the majority of hydrocarbons and carbon monoxide emissions are given off within the first 30 seconds after a gasoline engine cold start, there has been a drive to decrease the time in which it takes an exhaust sensor to reach operating temperature. This has been accomplished using two basic techniques: increasing the power level to the sensor heater, and decreasing the mass of the sensor element. The latter is accomplished by changing the geometry of the ceramic element from a conical shape to a planar format. In a planar element, the heater is often integrated into the sensor whereas a conical element uses a separate heater to heat the sensor element. This combined with a mass of about 50% less than a conical element, allows the planar sensor to reach operating temperature in less than 10 seconds.

Because the heater is integrated into the ceramic element, the heater voltage can be controlled very precisely through closed loop feedback. It is very important in a wide-range element to maintain the element temperature to within a very narrow band of operating temperature in order to maintain precise A/F ratio accuracy.

Another feature of a planar element is its ability to provide a sealed air reference. The sensor provides its own electrochemical oxygen “pump” in both rich and lean exhaust mixtures, preventing the possibility of air reference contamination and allowing for a simplified package.

An exhaust sensor package must meet three minimum requirements: it must be functional, durable and protect the element. It is also highly desirable to produce the smallest package (typically height of the sensor is most critical) possible. There are several reasons for a short package, especially notable are for ease of packaging and durability. The shorter the package, the less susceptible the sensor is to vibration and impact.

This paper will discuss several thermal optimization strategies for lowering the temperature in critical areas of a sensor in order to reduce the overall height of the package. Examples of these strategies will be put into practice and evaluated using finite element analysis (FEA). This will be followed by a comparison of the FEA results and actual test data.

Thermal Strategies

In determining the packaging vertical dimensions for a sensor, typically the most important parameter to determine is the maximum steady-state temperature to which the sensor will be exposed. In exhaust sensors, there are usually two critical areas to observe when determining the height of the sensor. First, all materials should be below their maximum use temperatures when the sensor is exposed to its steady-state maximum temperatures. The component that is most sensitive is usually the wire seal at the top of the sensor. This material is usually made of some type of elastomer in order to seal around the wires; if the material reaches temperatures in excess of its operating temperature, it will begin to degrade and can cause contaminants to enter the sensor. The second area of concern is the terminal contact interface. Depending on the material, and type of interface, terminal materials can greatly increase their susceptibility to oxidation with increasing temperature. Thus, the two areas this paper will examine when optimizing the sensor packaging to reduce the temperatures in critical areas are the element-terminal connection and the wire seal.

The primary equation that controls heat transfer inside a sensor package is:

(1)

Where Q is the heat energy transferred, A is the cross-sectional area perpendicular to the direction of heat flow, k is the thermal conductivity of the material, L is the length of the heat path, T1 is the starting temperature and T2 is the temperature at the distance L.

Although equation (1) is rather simple, it is seldom fully optimized. Often the area A is chosen based on previous sensor designs, and the thermal conductivity k is based on previous materials used, thus the length L in equation (1) is used to reduce the heat energy that reaches the critical components. Equation (1) can also be used to maximize heat transfer away from a component.

In sensor design, there are three basic strategies used for minimizing temperatures at a specific component (given a certain set of boundary conditions): restrict heat flow up the sensor, promote radial heat flow to the ambient environment and increase heat flow through a component. First, and foremost, is to prevent heat from traveling up the sensor. This is accomplished by using low thermal conductivity materials. For example, in the metal category, 400 series stainless steels have about 1.4 times the thermal conductivity as 300 series stainless steels. Alumina has about two to three times the thermal conductivity of steatite. Of course, there are often reasons for choosing a particular material over another, other than thermal conductivity; however the key is to design around these other reasons.

Cross-sectional area is also critical in restricting heat transfer. As shown in equation (1), area is directly proportional to the amount of heat transfer. Comparing alumina and a 300 series stainless steel can show an example of how this could be applied. Because alumina is quite brittle, it generally cannot be made as thin as a metal component. Although 300 series stainless steel has about twice the thermal conductivity as alumina, if the thickness of the alumina is more than twice the thickness as the metal, the net heat transfer will be greater with the alumina.

Another form of a low thermal conductivity material is an air gap. If air gaps are used correctly, they can significantly reduce the heat transfer up the sensor due to its ultra low thermal conductivity. However, if the air gaps are not used correctly, they can actually raise the temperature at the top of the sensor. A sensor should be designed so that the air gaps are perpendicular to the main heat flow path, and minimize air gaps parallel to the heat flow path above the shell (or the portion used to connect to the exhaust pipe). This allows the heat to escape from the center of the sensor to the outside surface, which allows the ambient air to remove heat through convection and radiation. In the radial direction, the heat should not be blocked from reaching the outside surface, the thermal conductivity should be maximized.

When trying to conduct the heat in the radial direction out to the ambient air, effort should be made to maximize conduction by making as large of contact area as possible. For example, often there are two or three shields (inner and outer and/or upper) that make up the main body of a sensor. A solid connection (through welding, crimping or interference fit) allows heat to transfer through conduction instead of through convection. Further, the solid connection should be made over as great of a horizontal surface as possible. For example, if an alumina insulator is crimped in by a metal shell or shield, there is typically only a point of contact made instead of surface contact. This in effect greatly decreases the area A in equation (1), decreasing the amount of heat transferred between the two materials.

Another method of lowering the temperature in a specific area, as long as it is not at the top of the sensor, is to actually increase the thermal conductivity of the component. Heat transfer is similar to water flowing through a garden hose. If the middle of the hose (sensor) is severely restricted (low thermal conductivity), the pressure (heat) will be high at the location of restriction. By decreasing the amount of restriction (increasing thermal conductivity) the pressure (heat) will not build up as much in that location. Of course, this means more heat will reach the components above the area of concern, but it allows yet another tool for optimization.

Finite Element Analysis

ALGOR's steady-state heat transfer software was used to optimize the sensor packaging. The model takes into consideration conduction, convection and radiation. Since the analysis was evaluated at maximum temperature conditions, the temperature of the sensing element will be greater than the normal operating temperature and thus the heater will be inactive and does not need to be taken into consideration.

Although the sensor is not truly symmetrical, in order to simplify the analysis, the sensor is represented as a 2-D axisymmetric model. In situations where a component was not symmetrical, an equivalent area was created so that when it was swept 360 degrees, both the model and actual component would have the same volumes.

Temperature boundary conditions used in the model were taken from worst-case vehicle conditions. Convection coefficients were taken from Chen, et al. [1] for wide open throttle conditions as shown in Table 1.

Although the exact convection and radiation temperatures actually vary along the surface of the sensor more than shown in Table 1, it still gives an excellent approximation of the true temperatures in the sensor and allows for a quantitative comparison of different thermal optimization techniques.

The boundary conditions shown in Table 1 are extremely harsh conditions. Specifically, the environmental conditions of 150°C air temperature and convection coefficients of 1.517(10^-5) W/(mm² °C) should be noted. This convection coefficient is equivalent to free convection. While it is not advisable to expose an exhaust sensor to these conditions, depending on the vehicle packaging configuration, some sensors have been known to approach these conditions.

As an illustration of the aforementioned thermal optimization techniques, a common sensor construction will be used as a starting point. Instead of constantly shortening the sensor with each iteration, the target height will be used for all models and the terminal and seal temperatures will be lowered to the target temperatures. As much as possible, only the necessary components will change from iteration to iteration. Figure 1 shows a cross-section of the base sensor design. As shown, the sensor consists of an inner and outer lower shield (high temperature stainless steel (ss)), planar element (alumina), shell (ss 400 series), lower ceramic insulator (alumina), compressed talc, upper insulator (alumina), upper shield (ss 400 series), connector (steatite) and environmental seal (viton). As the design progresses, only the upper and lower insulator, talc, and to some extent, the upper shield will vary.

As previously mentioned, the two most critical areas on this type of sensor are the terminal contact temperature and the seal temperature. Therefore, these are the two temperatures that will be compared.

Figure 2 shows the FEA model of the sensor shown in Figure 1. Running the model with the boundary conditions shown in Table 1, the steady-state terminal temperature is 435°C and the seal temperature is 324°C. Figure 3 shows the model results.

As an example of how material changes affect temperatures, if the lower and upper insulator’s material were changed from alumina to steatite, the resulting temperatures at the terminal and seal would be 388°C and 292°C respectively. As an additional example, if the upper shield were changed from ss 400 series to ss 300 series, the temperatures would be reduced to 381°C and 281°C, respectively.

An example of area reduction would be to reduce the cross-sectional area of the upper shield from 0.457 to 0.381 mm. This reduces the terminal temperature to 373°C and the seal temperature to 274°C. Even though the heat flow is being restricted in the upper shield, which is outside the direct heat flow path of the terminals, the terminal temperature is also reduced. As a general guideline, the temperatures of a sensor are roughly the same for a horizontal plane through the sensor. Thus, changing the temperatures in that horizontal plane will affect the temperatures elsewhere in that plane.

An example of increasing radial conductivity to the ambient environment would be to eliminate vertical air gaps by closing the ceramic around the element and making a solid contact with the upper shield. This results in terminal and seal temperatures of 370°C and 273°C respectively. Of course, in this particular design, this procedure is only illustrative, since it would be difficult to eliminate air gaps around hard ceramics, however, this technique will be used in the next sensor configurations.

In order to further decrease the temperatures at the terminals and seal, a different kind of material is needed. If a ceramic fiber mat were to be used instead of the upper insulator, the thermal conductivity can be greatly reduced.

The mat material used in this sensor is made of an alumina fiber pre-form with vermiculite, that is crushed to fill the area of the inner shield (much like a talc, except at a lower pressure) preventing any air gaps. Since there are no binders to burn off, the mat actually exerts more pressure when heated due to the vermiculite. The mat has intimate contact with the element for its entire length; so although the thermal conductivity of the mat is low (not wanting to accept heat from the element), there is a large amount of surface contact with the element thus pulling heat away from the center of the sensor. The mat also forms a cushion around the element, helping to dampen vibrations and impacts to the sensor. Figure 4 shows the thermally optimized sensor. Figure 5 shows the FEA model.

The lower insulator shown in Figure 1, is required to be as tall as shown in order to withstand the large compression forces needed to compress the talc (without cracking due to stress risers). If a metal disk were used in place of the insulator, it could withstand the compaction forces (if properly sized), and have a much lower profile, and allow for more mat material to be used. Thus, by adding a metal in place of a ceramic, even though it has approximately 3 times the thermal conductivity (in the location used) as ceramic, because it is less than 1/3 the height, and allows for a much less conductive material than ceramic to be used, the net heat transfer will be decreased.

In order to compress the talc using a mat support in place of the upper insulator, an inner shield is used to surround the mat and apply pressure around the periphery of the top metal disk. The talc is then compressed between the two metal disks.

The mat support system allows for all of the previously mentioned thermal management techniques to be employed. If the upper and inner shields are crimped together, there is a direct conductive path from the center of the element to the ambient environment.

Performing FEA on the sensor shown in Figure 5, the resulting temperatures at the terminal and seal would be 355°C and 253°C respectively. The FEA model results are shown in Figure 6.

The seal is now well below the maximum continuous temperature threshold for this material, since the analysis was conducted under worst-case conditions, this indicates that the seal can withstand a slightly higher heat flow. This brings us to the last thermal optimizing concept examined in this paper; by changing the connector material from a steatite to an alumina, the thermal conductivity is increased, allowing more heat to flow up the sensor. However, it should also decrease the temperature at the terminal connection slightly. Running FEA shows that by increasing the thermal conductivity at the connection slightly, the terminal connection temperature is reduced to 352°C and the seal temperature is increased to 256°C. This procedure should be done after all other changes have been made. This allows the sensor to be fully optimized at the maximum temperature boundary conditions, allowing for the shortest sensor possible.

Table 2 summarizes the temperature results from the various configurations. As shown, the mat supported sensor has the lowest temperatures at both the terminal connection and the seal. Although it is primarily due to the low thermal conductivity of the material, the other optimization techniques also play a role.

As previously mentioned, the environmental conditions around the sensor are critical. The previous results were modeled using free convection only. If a 0.75 m/s (~1.7 mph) airflow is allowed to flow across the sensor upper shielding, the terminal temperature is reduced to 338°C and the seal temperature is 249°C. If the airflow is increased to 1.8 m/s (4 mph), the temperatures are reduced to 323°C and 210°C for the terminal and seal respectively. Table 3 shows a comparison of the same sensor configuration, with different airflows across the upper shield.

Comparing the airflows in Table 3, the role that airflow plays in sensor temperatures is obvious. Even though each row of results uses ambient air temperatures of 150°C, the critical temperatures decrease considerably with only a slight breeze since the convection coefficient is greatly influenced by the airflow. This comparison shows how critical it is to package an exhaust sensor so that it has proper airflow.

The shell temperature, also called the hex temperature, is often used to make reference to the exhaust temperature. This can often be misleading when comparing different designs. The hex temperature is influenced by the thermal conductivity and heat path of the overall package. This can be seen from the summary results as shown in Table 2. As shown, even though the exhaust temperature remains the same for each iteration, the hex temperature varies according to the design.

Finite Element Analysis Verification

Verifying the thermal model by comparing FEA results to test data is necessary at the conclusion of any analysis. Conditions for testing were made as close as possible to the FEA modeling boundary conditions. However, it is difficult to determine the exact convection coefficients and radiation emissivity, so this is often the reason for temperature differences. Comparison of different sensor configurations would not be affected however, since the same boundary conditions were applied to each model.

Table 4 shows how the sensor shown in Figure 5, with steatite connectors, compares with actual test data. The FEA model assumes 0.75 m/s (1.7 mph) airflow across the outside of the sensor. Since it is difficult to place a thermocouple at the exact center-bottom location of the seal without actually reading an air temperature, the thermocouple was fed through a hole alongside the wire stopping about 0.5 mm from the bottom of the seal. The terminal temperature measured 349°C and the seal temperature was 220°C. The seal temperatures in Tables 2 and 3 were measured in the exact center-bottom location of the seal. Since the seal has a very low thermal conductivity, temperature varies significantly throughout the seal. Therefore, in order to better approximate the location of the thermocouple, the surrounding nodes in the FEA model were measured to show the range of temperature in that location. As shown, the FEA modeled temperatures are fairly close to actual. FEA results show 345°C for the terminal temperature and a range of 214 – 242°C with an average temperature of 228°C for the seal (again, not knowing the exact location of the thermocouple). Maximum differences are in the seal area and are within 4%. This is understandable, due to the difficulty in determining the exact amount of surface contact made between several components, convection coefficients and ambient temperatures.

Summary

It has been shown that FEA analysis is very useful in thermally optimizing a sensor. Three main temperature reducing techniques were used: restricting temperature up the sensor (low thermal conductivity materials, reducing cross-sectional area, utilizing air gaps), promoting radial conduction (eliminate transverse air gaps, solid material contact) and allowing heat to pass through a critical component (increasing thermal conductivity above the critical component). From the starting sensor configuration, the temperatures were reduced by 83°C at the terminal contacts and 68°C at the seal. Comparison of FEA modeling and actual data shows that they are within 4%. Non-traditional sensor components may be necessary to achieve lower-profile sensors that retain the same temperature requirements as their longer counterparts. It has also been shown how critical airflow around the outside of the sensor is in minimizing temperatures. It is important to model the worst-case conditions of a sensor in order to prevent out gassing of the seal and oxidation at the terminals. Although thermal optimization is only one step in making a durable sensor, it plays a major role in contributing to a world class design. The thermally optimized sensor is currently undergoing an extensive battery of tests, and is showing that it is an extremely durable design.

References

David Chen, et al., "Optimization of Oxygen Sensor" (SAE 2000-01-1364).

Biographies

Mr. C. Scott Nelson is a Senior Project Engineer for Delphi Automotive Systems, Delphi-E&C Division, Customer Solution Center. He graduated with a BSME from Lawrence Technological University and received his MSE degree from Purdue University. In his 10 year career, he has specialized in advanced exhaust aftertreatment systems.