01 April 2003
Thermal protection allows for advanced control systems in gas turbines.
By Paul Ivey
Measurement of unsteady pressure is a requirement in many proposed and existing aeroengine active control systems. In compressors, the precursors to stall and surge flow instabilities have been identified. These precursors are characteristic pressure fluctuations that can trigger a quick-acting control function.
Consequently, the requirement for a control loop taking real-time unsteady measurements of fluctuating pressure in the engine to drive preventive action reducing the surge margin required by the engine has been demonstrated. The benefits clearly are increased overall engine performance and operating envelope.
Water-cooled transducers commercially available underwent testing using isothermal and heat transfer experiments to establish the best form for the cooling adapter. Mounting arrangements for the transducer in the engine were the most important.
A condition to this arrangement was the thermal protection required to permit the transducer to survive without seriously limiting the transducer bandwidth. Using one of these water-cooled arrangements, design tools using Finite Element (FE) analysis were constructed and validated.
Once validated, an improved design concept used advanced heat transfer mechanisms taken from turbine blade cooling concepts.
The main problem was to provide an unsteady pressure measurement in an environment whose temperature is far higher than the operating temperature range of the pressure transducer.
However, there are additional important factors to recognize. These factors indicate the improvement required over the commercially available water-cooled adapter.
The coolant media is air (available in most engine applications, whereas water is not, particularly in flight).
The physical size of the cooling adapter must be smaller.
The diaphragm of the transducer must be close to the hot gas flow.
It must be economic to manufacture.
The design equation is smaller + closer to gas + air cooled = enhanced design.
Specifications occurred on two levels. One was a "target" level, an ideal level to achieve unless satisfactory reasons prevent it, and an "acceptable" level, where anything less would be a failure. The final design was intended to fall somewhere between the two. Following are points from the specification.
Physical size envelope/operating temperature
A survey of current and proposed instrumentation locations covered a range of engines, from the smaller military aircraft or helicopter engines to the large civil aircraft engines. From this large amount of data, we arrived at a loose determination of a "generic" size using current openings for boroscope inspections as a guide.
Designing for a transducer to withstand temperatures in the combustion section of the engine (typically 1,700°C) is currently an unrealistic objective.
The current thought is that the use of transducers in the turbine section of the engine at temperatures up to 1,200°C to measure the unsteady pressure forces on highly stressed blades is achievable using film cooling technology.
The problem is that the use of a film of cooling air in front of a transducer has a large potential for measuring error. The study of an air-cooled, commercially available, unsteady pressure transducer capable of operation at temperatures around 900°C is realistic.
Flush mounting the transducer in the adapter provides the greatest range of measured frequencies. However, flush mounting severely restricts the temperature at which survival is possible, and accepting a measurement limitation through recessing the transducer is unacceptable for practical applications in the compressor. Therefore, the design needs to be a compromise between achieving sufficient cooling of the transducer, by recessing it back from the flush position, and providing a measurement over the widest possible frequency range. The frequency range can be predicted as a function of the size of the recess by calculating the point at which a recognized resonant frequency will occur:
Helmholtz resonant frequency Fh:
a = speed of sound in air
L = recess length
D = recess diameter
r = recess radius
The resonant frequency is a function of recess length. The selection of a 4-millimeter recess length came about because of the maximum measuring requirement of 20 kilohertz (kHz) and 5–10 kHz being acceptable to maintain sufficient scope for cooling. The resonant frequency should occur after 13 kHz, giving at least 10 kHz of effective measuring bandwidth, known as "flat [response] up to 10 kHz."
From the design specification, the original validated water-cooled design was modified using the specified physical size envelope and the required recess of the transducer.
We designed a new base for air-cooled operation. This design was very simple and similar in its heat transfer features to the water-cooled arrangement.
A thermal analysis used the validated FE model. Assuming measurement in a flow at 500°C, the base design prediction showed that the front face of the transducer diaphragm would be at a temperature of 207°C ±25°C, with "simple" airflow into and out of the jacket.
Two major concerns were revealed:
- Heat transfers very easily up the recess in front of the transducer, perhaps the most direct heat path from the hot gas to the vulnerable transducer.
- Heat transfers very easily into the adapter as a whole from the hot gas and the wall supporting the transducer, raising the temperature of the whole assembly to a dangerously high level.
These two heat path concerns forced us to concentrate on the redesign. Applying enhanced heat transfer augmentation transferred from heat exchanger and gas turbine engine blade cooling technology, the objective was to construct a new model and conduct experiments that would demonstrate an improved cooling effectiveness, permitting air-cooled transducer survival at flow temperatures up to and beyond 800°C.
Key tasks for this design were to select heat transfer tools to address the critical heat paths between the hot flow and the transducer. The following methods of heat transfer augmentation went in the final design.
Thermal barrier coatings
The benefits of reducing the transfer of heat to the cooled component in the first place are so great that many people use thermal barrier coatings for gas turbine blades. The predicted transducer diaphragm temperature for a flow temperature of 500°C with only a ceramic adapter shell was 148°C.
Multipass of coolant
Double and triple pass systems see use in cooled turbine blades. The inclusion of extra coolant passes also increases the internal surface area to increase the heat transfer. The predicted transducer diaphragm temperature with a ceramic adapter shell and two-pass coolant system was 104°C.
Turbulators disturb the hot boundary layer of the coolant in contact with the component wall, causing it to mix with—and more effectively transfer heat to—the bulk coolant flow and raise the level of turbulence, again enhancing heat transfer. Vertical rather than horizontal ribs appeared on the first pass because of anticipated blockage effects, reducing the effect of the impingement cooling at the end of the down passages.
On the second pass, we used horizontal ribs on both sides of the passage because it is an advantage to retard the flow as well as increase the internal surface area to maximize heat transferred into the coolant. The predicted transducer diaphragm temperature with a ceramic adapter shell, a two-pass coolant system, and rib-roughened internal surfaces was 87°C.
The impingement cooling principle is effective over only a very small relative area and so is typically restricted to specific areas where very high rates of heat transfer are required. To use impingement cooling over a wider area requires a large array of holes. However, the greater the number of holes, the lower the relative velocity of the impinging coolant. The cooling also becomes less effective.
The major heat path up the recess in front of the transducer required impingement cooling. We decided that a discrete rib, attached to the wall with air impinging on it, was a more effective way of drawing heat away from the recess into the coolant.
We felt this area needed the most effective cooling. Uniform impingement was not feasible because the impinging jet would not be at a high enough velocity when it reached this area. The predicted transducer diaphragm temperature with a ceramic adapter shell, a two-pass coolant system, rib-roughened internal surfaces, and discrete impingement was 80°C.
The predicted transducer temperature was then at a satisfactorily low level to proceed with the assembled concept. The front end of the adapter is so small that to have an axisymmetric design was impossible.
We then decided to have a series of impinging "cells" at the front end. This meant there were two planes in the design: an "up plane" and a "down plane." The down plane is where the air enters and flows down to the discrete impinging fins. The up plane is where the air flows away from the site of impingement and up to exit.
There are eight down planes coinciding with the eight impinging holes and eight up planes in between the eight impinging holes. A screen appears at the inlet of the recess to greatly reduce the heat transferred to the transducer from the hot gas flow. The magnitude of radiated heat transfer was relatively small at a flow temperature of 800°C, where it was 8% of total heat transfer.
Air enters axisymmetrically into the inner channel at the top, then accelerates through eight long holes, drawing heat away from the transducer side. The relatively high-velocity coolant impinges onto eight discrete ribs near the bottom of the adapter, drawing heat away from the recess. The air spills over the discrete ribs into the bottom of the adapter, then moves in either direction, where it goes up on the outside of the adapter.
Just before the air enters the rib-roughened passage area, it again becomes axisymmetric. It then passes through the passage, roughened on both ceramic and metal sides, with heat transferring easily into it. The air then exits in the outside channel at the top.
Thermal design validation
The Kulite XTE-190 transducer has an operating temperature range up to 235°C. We attached thermocouples to the cooling adapter so that in addition to measuring the material temperature of the transducer in three different positions, air temperatures occurred in several different positions in an attempt to assess the performance of the cooling design itself. We replaced the transducer with blank attached with a thermocouple. A portable compressor supplied the adapter cooling air; the airflow rate was 0.0142 cubic meter/minute.
Tests at a source mass flow rate (0.124 kilogram/second) maintained steady, stable combustion at increasing flow temperatures of 500°, 650°, 800°, and 900°C. Air flowed across the transducer mounted in a boss in the pipe wall. The heated air then vented to atmosphere. Test results showed that even when the flow temperature rises to 900°C, the maximum temperature of the transducer is 182°C. This experimental value, given the facility uncertainty, is well within the operating range of the transducer and is also well predicted by the model. These were encouraging results and suggested there is further scope for an ability to survive, using these conditions and configuration, at even higher flow temperatures than 900°C.
Pressure measurement validation
The ability to measure over a wide bandwidth becomes restricted once you recess the transducer back from the flow. The use of a protective screen in front of the transducer will add to this restriction.
Directing a high-velocity air jet over the front of the new design would provide excitation over a broad band of frequencies. If a Fourier transform occurs on a sample response from the transducer, this would indicate the presence of any resonant frequencies likely to cause a reduction in the transducer's frequency response.
The voltage output underwent sampling and time averaging for the whole period and then normalization to indicate only the unsteady pressure variation.
By applying a Fourier transform to the purely unsteady signal, you can obtain a plot of unsteady voltage against frequency. If the resolution used in the transform was sufficiently low, the frequency of the resonant peak(s) became obvious.
The next phase was to conduct a similar series of tests to evaluate the unsteady pressure measurement characteristics of the transducer in the air-cooled adapter at elevated flow temperatures.
Using a combustion rig, static pressure tapping and water manometer would indicate the static gauge pressure of the hot gas flow in the measuring plane of the rig. A multimeter provided an additional indication of the DC voltage output from the transducer. Measurements from the manometer, multimeter, and a dynamic and time-averaged static sample were taken at flow temperatures of 20° (ambient), 400°, 500°, and 600°C.
Survival of the transducer has been shown up to 900°C flow temperature, with good agreement with FE model predictions and potential to increase the flow temperature further and stay within the transducer operating temperature range (up to 235°C). When we insulated the assembly from atmospheric cooling, simulating a more aggressive heat transfer environment typical of that in a gas turbine engine, survival is shown at 800°C flow temperature, with good agreement with model predictions and potential to increase the flow temperature further. The frequency response is a borderline pass or fail with respect to the specification. If the resolution of the transform is increased, the response is flat up to 4.55 kHz. TT
Behind the byline
Dr. Paul C. Ivey is at the school of engineering at Cranfield University, Bedfordshire, U.K. His e-mail is P.C.Ivey@Cranfield.ac.uk.