Figure "Designing brush seals" (Ref. 7.3.3-1): The depicted methodical procedure during design and construction of a brush seal is based on understanding of the physical processes that influence brush function, experience with brush seal behavior in gas turbines, and testing rig results. This systematic approach is used for brush seals that are designed to seal air. The depicted method makes designs for the use of brush seals in hot and cool engine areas possible.
If the necessary data for brush design are not available and/or were not provided by the brush seal manufacturer, then they must be determined by separate tests, which are often expensive and time-consuming. This is the case with special, application-specific tribo-systems with rotor coatings. The direction of the leakage flow must prevent loose bristles or bristle particles from damaging the roller bearings.
The depicted designed system makes it possible to take advantage of the excellent seal effectiveness and mechanical security (relative to labyrinths) for the engine. However, the seal effectiveness of brush seals does not compare with that of floating ring seals, which are used primarily for sealing oil/air in bearing chambers (Fig. "Brush seal limits").
Different analytic disciplines are used during design. Analysis of the stresses in the brush and analysis of the security against failure due to dynamic fatigue in the HCF region are both conducted within the framework of the geometric data, which take into account the available space for installation, the friction heat, and the seal requirements. The actual leakage air flow during operation must be determined with the aid of data that were empirically derived from testing rig experiments.
The same is true for wear data (Fig. "Brush seal determines operation"). If necessary, these must be determined for the specific tribo-system (e.g. special coatings on the rub surface of the rotor; also see Ref. 7.3.3-3).
Coefficients of friction and the calculated evaluation of the friction heat created during rubbing are covered in Refs. 7.3.3-2, 7.3.3-4, and 7.3.3-6).
Several iteration steps are necessary to fulfill the requirements for tolerable pressure differences, sufficiently small whirl in front of the brush, and good performance under extreme conditions and unsteady operating conditions.

Figure "Brush seal determines operation": During rubbing, the heat created and its dissipation control the heating-up of the brush and, therefore, its reliability (Fig. "Brush seals rules preventing damages"). The top diagram (Refs. 7.3.3-2 , 7.3.3-4 and 7.3.3-6) shows the increase of the friction heat along with the excess brush length. The shorter the “effective” bristle length (depending on the brush stiffness, which is also influenced by the friction inside the brush and against the backing plate), the stiffer the brush (higher friction forces). The higher the friction forces, the more friction heat is created. Large leakage air flows can dissipate more heat and prevent dangerous overheating of the brush. In order to minimize heat creation, the rotor coating must be as smooth as possible and suitably matched with the brush material. If the rotor coating dissipates heat well, it is also good for the brush. Uncoated rotors must also be viewed with regard to this positive aspect.
The middle diagram (Ref. 7.3.3-3) shows that the coefficient of friction, which is an important value with regard to heat creation during rubbing, is clearly, material-specifically, dependent on the rubbing speed. The given data were determined for “miniature brush seals” with 0.25 mm radial excess and a testing rig temperature of 426 °C. The coating-specific, very different progression of the coefficient of friction is surprising.
The low coefficient of friction of NiCrAlY brushes is due to the tough oxide film on the contact surface, as well as a lubricating film of Ba- and Ca-fluoride. This film forms on the rub coating of the rotor, which is made of a thermal-sprayed Cr-carbide coating with Ba and Ca additives.
Al₂O₃ and Cr₂O₃ rub-tolerant coatings exhibit micro-spalling (grain outbreaks), which create a high coefficient of friction and heavy brush wear (Fig. "Brush seal track damages"). Co-alloys such as Haynes 25 are known for their low coefficient of friction at temperatures under 400°C. The steep increase in the coefficient of friction above 400°C is due to a structural change.
The bottom diagram shows brush wear opposite the wear of various thermal-sprayed rotor coatings (Ref. 7.3.3-5). The data were taken from a testing rig with complete brush seals running for 100 hours at room temperature.

Figure "Brush seal bristle oxidation" (Ref. 7.3.3-3): The left diagram shows the typical oxidation behavior of a nickel alloy:
A relatively short incubation time “I”, during which the weight of the probe stays almost unchanged.
At first, accelerated weight increase in region “II”, where an oxide film forms. In region “III” the weight increase is continual, and the weight-time curve has a corresponding linear progression. Here, the oxidation is balanced with oxygen diffusion through the oxide layer.
The right diagram shows the oxidation behavior of three potential bristle alloys at an annealing temperature of 1038°C. The oxidation resistance depends on the type of oxide film that forms
The NiCrAlY alloy with the highest oxidation resistance forms strongly-adhering dense(Al or Cr)₂O₃ oxides, which are stabilized by rare earth metal supplements.
On the other hand, Inconel 718 and Haynes 25 (Co-alloy) form non-stabilized Cr₂O₃ films, which are considerably less oxidation resistant. These oxides do not form a diffusion barrier for oxygen ions. Therefore, the oxidation process continues, which is the deciding difference between these materials and those that form oxygen-sealing oxide layers.

Figure "Brush seal leakage by backing plate": The gap between the backing plate and rotor (fence height) requires special attention during brush design. A small fence height is vital to ensure a low leakage air rate (top left diagram). However, it must not touch the rotor, which could result in self-increasing rubbing (bottom left diagram). If this cannot be ensured, then the rotor must be protected with a sufficiently thick and wide, insulating, wear- and heat-resistant coating. The coating must not be worn through to the base material under any circumstances (top right diagram).
The thickness/stiffness of the backing plate must be selected so that no plastic deformation occurs due to the pressure differences throughout the designated life span at operating temperatures. The bottom right diagram depicts a variation (Ref. 7.3.3-2) where the backing plate was extended with an abradable coating. In this way, contact should not result in damage to either the rotor or the seal.

Figure "Brush seals for high pressure ratio" (Ref. 7.3.3-7): For a short distance, the leakage flow through the brush is deflected radially inward by the bristles and the backing plate. It then blows the bristles inwards towards the rotor and decreases the leakage gap. This is referred to as “blow down” or “pressure closure” (top left diagram). Friction inside the brush is caused by the bristles laying against one another and against the backing plate. This stiffens the brush and prevents the bristles from springing back after the rotor “moves back” (clearance increase). In order to minimize the friction against the backing plate and, therefore, the brush hysteresis (Fig. "Brush seal leakage behavior"), the backing plate is machined out at the lay-on surface (top right diagram) so that the bristles are free and can spring back due to blow down (LHS brush=Low Hysteresis Seal). In this way, the dependency of the brush stiffness on the differential pressure is also minimized considerably (bottom diagram). The drawback of this brush type is that it is especially sensitive to bristle flutter. Vibrations and tangling of the bristles are cause by whirl in the flow in front of the brush (Fig. "Brush seal problems by turbulence") and around the backing plate. A possible remedy for this is mounting a flow deflector at the front plate (top right diagram).

Figure "Preventing bending of brush seal bristles": Flow deflectors can be made from sufficiently strong perforated metal sheets or porous metal felts. Another possible remedy may be affixing a cover, separate from the brush (Fig. "Designing surrounding of brush seals").

Figure "Design for removal of brush seals": When mounting and removing brush seals, special precautions are necessary in order to avoid the following damaging situations from occurring:

• Overstressing of the brush (deformation) by locally acting powerful mounting forces.

• Rotating the brush against the angle of the bristles while they are in contact with the rotor during installation.

• Wrong direction of rotation after installation, if the rotor is in contact with the brush.

• Bending and/or splaying of the bristles due to improper handling of the brush before or during installation.

• Damaging the brush and/or the housing during removal due to the brush splaying in the housing.In order to avoid these situations, suitable tools for installation and removal are necessary, as are easily accessible, sufficiently large working surfaces for these tools.
  

Figure "Designing surrounding of brush seals": In order to prevent brush seal damage due to bristle flutter (see Fig. "Design for removal of brush seals"), unallowable whirl must be suppressed in the space in front of the brush. In the case shown in the top diagram, the brush was rapidly destroyed by uneven wear and entanglement, because the rotating screw heads caused dangerous whirl.
The middle diagram depicts a screw cover fastened onto the rotor to minimize whirl. The bottom diagram shows another configuration which does not prevent whirl, but a ring-shaped cover protects the front side of the brush.

Figure "Brush seals testing design parameters": The complex loads on a brush seal combined with various simultaneously-acting damage mechanisms prohibit satisfactory, purely theoretical procedures in order to ensure that a configuration is sufficiently safe (Fig. "Designing brush seals"). For this reason, developing brush seals requires testing in order to understand the influence of the operating parameters and estimate the operating life span.
Technical literature describes various testing rigs for brush seals. The assemblies are different, depending on the properties that are to be tested (e.g. seal effectiveness, rubbing behavior, maximum pressure differences, sensitivity to flow whirl). If one wants to inspect the realistic seal effectiveness of a brush seal in operation, it must be done in a testing rig in which the entire seal is being tested, in its original size with realistic pressure ratios, leakage gaps, and temperatures. Due to the necessary high pressure (long clearance gap), these tests are very elaborate. This is especially true if the pressurized air must be heated in order to simulate hot part conditions. Testing rigs with only one brush (top left diagram) have a problem of high thrust loads on the rotor. If two brushes are installed opposite one another to compensate for the thrust loads and the air inlet is between them (top right diagram), then the testing rig becomes even more elaborate and requires twice the amount of leakage air.
With long-duration tests for determining the wear behavior of the brush/rotor tribo-system or wear (fretting) and oxidation of the brush, the expense and effort required for testing full-sized brushes is unbearable. Therefore, special testing rigs were designed for these tests, which only inspect a brush segment in a pressure chamber, which can be moved against the rotor along with the brush (bottom diagrams). This configuration has the advantage of needing only small amounts of air, because the gap is closed during rubbing. If one wall of the chamber is made of armored glass, then the brush can be observed during operation, and its behavior can be documented. The rub track on the rotor can be directly observed throughout the entire test, which allows changes to be recognized immediately during operation or after altering testing parameters. If the pressure in the chamber puts too much radial force on one side of the rotor, a balancing force must be applied to the other side (bottom right diagram).