1

Experimental Investigation of Hub Leakage Flow

through Stator Knife Seals

Jordan K. Lewis Purdue University, West Lafayette, IN, 47907, USA

When looking at the efficiency of a high speed compressor, secondary leakage flow through the

stator hubs can influence the compressor’s performance. Compressors work by passing a fluid between

alternating rotational and stationary reference frames further diffusing the flow in each stage. The static

pressure in the system increases with each stage as the flow diffuses. This state of high pressure is not

easy to achieve or maintain because compressors create an adverse pressure gradient which makes the

fluid want to resist the energized state. The fluid will always want to flow from high pressure to low

pressure, resulting in flow leaking backwards through the compressor.

Figure 1: Schematic showing leakage flow through stator knife seals

These leakage flows are considered secondary flows because they are not the primary flow

through the outlet of the compressor. However, these secondary flows still impose a significant effect on

the overall efficiency of the compressor. Computational fluid dynamics (CFD) programs are useful tools

for predicting how a compressor design under various operating conditions will affect the performance.

CFD works with set boundary conditions for the flow through the system; however, boundary conditions

modeling secondary leakage flows are difficult to model and implement. For this reason, CFD is less

accurate when modeling flow through a compressor because the CFD mesh will be unable to account for

the secondary flows leaking through the stator hub [1].

The compressor used in this study is the Purdue 3 Stage Axial Research Compressor. It consists

of an Inlet Guide Vane followed by 3 stages of rotors and stators with a 24 in. tip diameter. The design of

the Purdue 3 Stage consists of shrouded stators in which the inner stator shroud sits below the inner

diameter of the primary flow path. The leakage flow through the shrouded stator cavity is minimized with

the use of knife seals to reduce the area between the spinning hub and stator vane.

2

Figure 2: Diagram of Purdue 3 Stage shrouded stator cavity and knife seals.

The purpose of this experiment was to quantify how much flow is leaking through the shrouded

stator cavities and to provide data that can be used as boundary conditions to calibrate and validate a CFD

mesh of flow through the compressor.

The primary measurement device in this experiment was a piezoresistive Kulite pressure

transducer. The transducer is activated when mechanical strain is implemented on piezoresistors placed in

a Wheatstone bridge configuration because the mechanical strain caused by pressure alters the resistivity

of the transducer. The change in resistance modifies the voltage output from the transducer which can be

calibrated to determine the pressure imposed on the transducer [3]. An LQ-062 Kulite transducer was

placed on each side of the knife seal for stators 1 and 2. Each pressure transducer measured the resulting

voltage at each operating condition and through a voltage pressure calibration, the difference between the

upstream and downstream pressure measurements was calculated. A former study by Wellborn in a low-

speed axial compressor suggested that a difference in static pressure could be used to determine the

leakage mass flow rate through the knife seal clearance. Wellborn suggested using an equation derived

from the conservation of linear momentum and scaled with the primary mass flow rate through the

compressor [4].

The experimental setup consisted of machining a rectangular cavity into each stator vane. A

rectangular block with two holes for each transducer was also machined to fit precisely in each stator

cavity. It was important to use well-machined parts within a tolerance of 0.001 in. so that the effect of the

transducers in the cavity would be minimal on the overall performance of the compressor. The two

transducers were positioned inside of the rectangular block and a black epoxy resin was poured over the

top to provide a secure backing. The epoxy filled rectangular blocks were then baked at 160 degrees

Fahrenheit to further set the epoxy.

Once the epoxy hardened, the rectangular plates were placed in a sealed calibration chamber to

determine the conversion between the voltage output of the transducers and the actual pressure acting on

them. The chamber was slowly filled to 5 psi with data points being taken intermittently in a LabView

code relating the sensor output voltage to the pressure inside the chamber. The pressure in the chamber

was then decreased from 5 psi back to zero to account for any hysteresis in the system. With a complete

calibration data set, the points were plotted and a line of best fit was used to determine the calibration

factor for all four sensors. The coefficient of determination for the best fit line was 1.0000 verifying a

linear relationship for the calibration.

3

Figure 3: Voltage to pressure calibration plot for LQ-062 Kulite transducers.

With the sensors fully calibrated, the blocks were installed in the stator assembly and the 36

gauge wire was carefully glued along the stator vane and passed through an exit hole in the casing. The

operating conditions for the compressor were controlled through a LabView code and the data collected

from the transducers was relayed through the LabView code to be saved for processing. The data were

sampled at a rate of 300 kHz with an anti-aliasing analog low-pass filter at a cut-off frequency of 100

kHz. The testing consisted of running the compressor at speeds of 68%, 80%, 90%, and 100%. Data

points from the transducers were taken intermittently as the throttle was adjusted to provide a range of

corrected mass flow rates. The experiment was repeated multiple times on different days and on opposite

sides of the compressor to account for leakage deviations from eccentricity of the rotor, manufacturing

tolerances, and other geometric defects. The data were processed and the change in pressure across the

knife seals in stators 1 and 2 were plotted. The solid red and dashed black lines represend the mean values

for experiments conducted on opposite sides of the compressor. Range bars were included on the plots to

represent the variation between the maximum and minimum changes in pressure with day to day

repeatability.

Figure 4: Plots of the change in pressure across the knife seals for stators 1 and 2 with respect to the normalized corrected flow

rate.

4

Figure 5: Total pressure ratio map of the Purdue 3 Stage [2].

The resulting lines denoting the knife seal pressure difference at constant speeds can be combined

with a total pressure ratio map to quantify the flow leaking through the knife seals at each operating point

of the compressor [2]. The measured secondary leakage flow can then be used to calibrate and validate a

CFD model of secondary flows because the CFD has comparable experimental data allowing known

boundary conditions to be created for the mesh. These data are significant for the future development of

CFD due to the ability to better predict losses in a compressor system. In addition to the opportunities for

future CFD technology, some trends in the leakage flow were also discovered through this experiment.

Particularly, in stator 2, there is a slight decrease in the knife seal pressure difference as the stability limit,

represented by the dashed line in Figure 5, is approached. Further experimentation and analysis are

recommended to determine the reasoning for why this phenomenon occurs.

5

Sources

1. Ball, P.R., 2013, "An Experimental and Computational Investigation on the Effects of Stator Leakage

Flow on Compressor Performance," MS Thesis, School of Aeronautics and Astronautics, Purdue

University.

2. Berdanier, R.A., 2015, "An Experimental Characterization of Tip Leakage Flows and Corresponding

Effects on Multistage Compressor Performance," PhD Dissertation, School of Mechanical

Engineering, Purdue University

3. Carter, S., Ned, A., Chivers, J., & Bemis, A. (n.d.). Selecting Piezoresistive vs. Piezoelectric Pressure

Transducers. Retrieved April 23, 2016, from

http://www.kulite.com/docs/technical_papers/Piezoresistive_vs_Piezoelectric.pdf

4. Wellborn, Steven Robert, "Effects of shrouded stator cavity flows on multistage axial compressor

aerodynamic performance" (1996). Retrospective Theses and Dissertations. Paper 11344