Turbine Sensors

By Joshua McConkey & Reamonn Soto

As market forces push towards de-carbonization and lower cost of operation, efficiency and reliability of gas turbines have become key differentiators among competitors. For Brayton Cycle machines, those efficiency gains mostly come via higher combustion temperatures, lower volumes of cooling flows, and larger blades. As those parameters are pushed to new limits, this also puts pressure on component reliability.

Intimate knowledge of component temperatures and strains is key to designing around this complex series of trade-offs. But as these components are made of materials that can survive ultra-harsh environments within high performance gas turbines, sensors must be able to mechanically survive the temperatures and g-loads inside turbines while also performing electrically or chemically.

TEMPERATURE MEASUREMENTS

Platinum thermocouples (Type S) are the only way to get accurate real-time measurements on the most extreme stationary hot gas path temperature readings. Technically, they can function up to about 1,800°C. But effective lifetimes above 1,650°C are limited as foreign material diffusion, annealing, and other material failures lead to cracking and intra-grain conductivity failures in a short amount of time. In gas turbine operational conditions, a lifetime of less than 20 hours is typical. Adding 5% and 20% or 0% and 13% rhodium to the platinum leads to create a Type R thermocouple can increase functionality to 1,730°C for hundreds of hours. However, rhodium costs >$2,000 per ounce in recent years. By comparison, platinum is less than $1,000/oz. Type R also has a much lower sensitivity than most other thermocouple types, at only 10µV/°C.

Most thermocouple manufacturers recommend that the long-term operating temperatures of Type S and Type R thermocouples only intermittently approach 1,600°C, with continuous use limited to below 1,400°C. Regardless, thermocouples are the most widely used instrumentation within the hottest parts of gas turbines. They are often also installed on rotating parts.

The thermocouple wires are routed along the rotor to get to a cooler, exposed section of the rotor where active electronics on the rotor take in the thermocouple signals, converting them into short-range wireless signals. These electronics can only work up to 230°C, so are typically installed on the rotor between the turbine compressor and the generator. Thermocouple wires are thus routed from the hot turbine section through the entire compressor. Each disk must be machined to allow the passage of the wires as they route along the rotor along a large proportion of the rotor. While this can be done on a fielded engine, the intrusive nature of the installation and the impacts on disk and rotor lifetime usually limit the use of rotating thermocouples to lab or test bed use.

For temperatures on the much cooler compressor, thermocouples are a useful, long-duration option. Wiring down the rotor and through multiple blade disks is still a tedious and expensive engineering task, but much more tractable the closer the instrumented blades are to the compressor inlet. Compressor blades are more sensitive than turbine blades to mass loading, so the thermocouple installation must be as low-mass as possible to prevent unwanted or unrepresentative vibration modes. On smaller and warmer compressor blades near the outlet, the size of the bonded thermocouple and its wiring are of aerodynamic concern. In these cases, thin film resistive devices are sometimes used. These wired sensors have low profiles and are becoming available at higher temperatures.

THERMAL PLANTS

Temperature Indicating Paints (thermal paints) are used in hot gas path surface temperature measurements. They provide large surface temperature indications rather than point measurements. They change the chemical and thus optical properties of coatings. These changes are a function of temperature and time. Changes are exponential, with temperature and usually quite fast in time.

For a typical lead-, cobalt- or chromium-based thermal paint, activation times at temperatures of interest (e.g., 1,200°C) are usually five to seven minutes. Longer exposures muddy the reading because the color changes of a paint at 1,200°C for 7 minutes are the same as they are for 1,150°C for an hour. Thus, the paints cannot be left in the engine for a run longer than 60 minutes or readings lose accuracy. Further, these paints are applied in thin coats and are not sintered onto the blades. Therefore, adhesion can be poor after more than a few minutes. If paint flakes off, information is lost. Thus, an engine campaign of only a few minutes must be performed with painted hot gas path parts, then the engine must be shut down and partially disassembled. Parts must be removed so painted parts can be analyzed and temperature maps created. Analysis takes from one to 10 months.

The maximum temperature of commercially available thermal paints is usually below 1,530°C. Accuracy is a concern for temperature indicating paints. Some studies show accuracy is on the order of +/- 0.5% of reading, while others show errors on the order of +/- 5%. Some vendors have patented formulations that change color above these temperatures, but they generally keep these paints in-house. Heavy metals such as lead and chromium used in many paints present a health hazard. The fragility, cost of running dedicated, short engine tests requiring tear-down, health concerns, questionable accuracy, and long lead times for measured data are among the disadvantages of these paints.

CRYSTALS AND CAMERAS

Thermal crystals are sub-millimeter grains of silicon carbide that have been neutron-activated to have certain distributions of different sizes of crystallographic dislocations within each grain. These grains are embedded in the surfaces of turbine components, and the engine is run for 100 to 200 hours. Components are then removed from the engine, and thermal crystals are de-embedded from the components, carefully labelled, and sent back to a laboratory for analysis.

In this analysis, x-ray diffraction is used to read the size distribution of the remaining crystal dislocations. As a particular grain sees higher temperatures, different sizes of dislocations within it become healed, meaning the dislocations are annealed or re-crystalized back into near-perfect atomic order locally. By analyzing the before and after size distribution, an inference of maximum temperature is established. While this seems to be a complex method of measuring maximum exposed temperature, vendors claim accuracy at temperatures around 1,400°C +/- 3°C

Thermal crystals have cost and testing inconvenience disadvantages. The longer the test, the more ambiguous and inaccurate results become as dislocation relaxation is a function of temperature and time. A long engine test at 1,100°C may give the same result as a short engine test at 1,200°C. Embedding and carefully removing crystals is laborious. It can take months to obtain the results.

Alternatively, infrared cameras can be used for hot gas path measurement. This technique offers 2D temperature measurements of turbine surfaces in real time, unlike thermal paint or thermal crystals. The components glow with blackbody radiation based on surface temperatures. The camera looks at a range of infrared colors, and attempts to rebuild the Wein displacement graph of that item. If this can be done, the uncorrected temperature of the item can be known. But a correction must be applied based on wavelength, temperature, and the material’s emissivity.

Unfortunately, the emissivity of the materials within a gas turbine varies greatly, and there is a lot of reflection of infrared light. Coatings, too, inhibit accuracy. Finally, infrared camera lenses have sensitivity to contamination, vibration, temperature, and thermal expansion. These reasons combine to make the use of infrared cameras in turbine hot gas paths somewhat rare.

Further methods for measuring air temperatures in hot gas paths are Filtered Raleigh measurements and Tunable Diode Laser Absorption Spectroscopy (TDLAS). They provide optical measurements of different molecular absorptions and reflections of the hot gas itself. While they show promise, they are expensive at this time.

PASSIVE RF WIRELESS

Passive wireless RF technology based on high temperature electrical resonators is a new approach. These sensors can be braze-bonded or Rockide-bonded to turbine blades. Typical sensor sizes are 10mm x 8mm x 0.3mm, resulting in a mass below 0.3 grams and a bonded profile height of less than 0.4mm (16 mils), which help maintain expected blade mass loading and laminar flow across the blade surface. The sensors require no attached antenna, as they are their own antenna. They are bonded in a way that has high heat transfer with the blade. A fixed antenna inside or outside the hot gas path pings the sensor, which resonates at a frequency representing its temperature. They work above 800°C in the field on blades and up to 1,700°C in laboratory settings. The antenna is available in forms that intrude into the hot gas path, if desired. The antenna and flexible cable can withstand temperatures up to 800°C. Non-flexible versions work up to 1300°C. Each antenna can read up to 5 sensors simultaneously. No external cooling is required on the antenna or cable probes, which are 2-7mm in diameter. The wireless range is 4-5cm in fielded systems, though 15-30cm range versions are possible at lower temperature applications.

The system can provide real-time blade metal temperatures across a range of 0-800°C. These sensors can also be used in air temperatures much higher than 800°C, and on early turbine blade rows. They can be read through an overcoat of flame-sprayed thermal barrier coating. Wireless range limits blade metal temperature readings to blade tips with a non-intrusive antenna or the use of an intrusive hot gas path antenna if the sensor is more than 5cm from the hot gas path inner casing. Up to five sensors on five different blades at the same radii can be read by the same antenna.

STRAIN MEASUREMENT

Strain gages provide blade design feedback. These small resistive devices change their electrical parameters linearly as a function of how much they have been stretched in a certain direction. As they are fundamentally electrical, response times can be fast, allowing for blade vibration monitoring. For larger blades, vibration modes need to be characterized across frequencies from 50 to 1000 Hz in most cases. For smaller blades, dynamic strain frequencies need to be read up to 20,000 Hz or even 100,000 Hz for smaller aviation turbines.

Static strain is also important for turbomachinery. Slower measurements of strain (1-10Hz) are used to infer linear strain that changes with rotation speed or shaft power. This enables the characterization of blade disks, which much carry the centripetal load of up to 60 blades rotating at high speeds, generating tremendous radial loads.

As efficiency goals become more aggressive, reduction of disk mass is targeted. Strain gages can confirm new designs function safely over different speed and temperature regimes. Rotor shaft tension in the axial direction can be measured with strain gages, and shaft power can be inferred from looking at the strain in the circumferential direction of the surface of the shaft. These applications require the wiring through multiple blade disks and along the rotating shaft. But wire breaks and changes in line impedance can cause loss or drift of measurement values.

Dynamic strain signal conditioners power the sensors and read their resistance changes. New vendors have strain sensors and signal conditioners available now for less than $200, but many in the turbomachinery industry report that these less expensive options simply don’t work, or are unreliable at this time. Staid suppliers of dynamic and static strain gages and signal conditioners are currently the suppliers that most Instrumentation Engineers trust. Those gages can range in price from $40 to $200 each, and the signal conditioners are typically $1,500- $3,000 each. While these prices are not low, they pale in comparison to the cost of installing the strain systems on rotating equipment. To mitigate this, some attempts have been made to perform optical wireless measurements of strain on rotating blades and disk. Differential Speckle Photography and other systems have been attempted with varying success.

A recent addition is Digital Image Correlation, which is coming down in price. It uses stereoscopic cameras to view a sample as it changes over time, and directly measures strain in a non-contact method by comparing the positions of the minute surface variations over time. Two cameras are usually necessary. However, no robust versions are commercially available for use in the harsh environments of turbomachinery.

Wireless static strain gages offer an alternative. They are conformable and can be shaped to fit snugly onto rotating parts, with an effective conformability radius of less than 3 cm. They report strain by resonating at different frequencies and can be read from 5 cm at high rotation speeds. They are limited to temperatures of less than 150°C and can only be read up to 10 times per second. This is a limitation of the electronics, not the system; electronics systems are under development to enable true wireless dynamic strain.

Joshua McConkey is Chief Technology Officer at Sensatek Propulsion Technology, a maker of wireless strain gauges and passive wireless RF sensors.

Reamonn Soto is CEO of Sensatek Propulsion Technology.

For more information, visit www.sensatek.com or call 850.321.5993