ASME Turbo Expo 2004 Control, Diagnostics and Instrumentation

DRAFT

1 Copyright © #### by ASME

ASME Turbo Expo 2004: Control, Diagnostics and Instrumentation

14-17 June,Vienna, Austria

GT2004-53992

REAL TIME FLUTTER MONITORING SYSTEM FOR TURBOMACHINERY

Harbans S. Dhadwal1, Marc Radzikowski1, Dmitri Strukov2 and Anatole Kurkov3

1 – Integrated Fiber Optic Systems, Inc. Stony Brook, NY 11790 2 – Stony Brook University, Electrical and Computer Engineering, Stony Brook, NY 11794

3 – NASA Glen Research Center, Cleveland, OH 44135 ABSTRACT

A fiber optic laser probe based system is described for real time monitoring of flutter in rotating turbomachinery. The system is designed for continuous processing of blade tip timing data at a rate of 10 MB/s. A USB2.0 interface provides un-interrupted real time processing of the data. The blade tip arrival times are measured with a 50 MHz oscillator and a 24-bit pipelined counter architecture. A Graphical User Interface provides on-line interrogation of any blade tip from any light probe sensor. Alternatively, data from all blades can be superimposed into a single composite scatter plot displaying the vibration amplitude of each blade. A hardware platform was developed to simulate a seventy two bladed turbine operating at 15,000 rpm. Blade tip responses from three light probes were generated in an infinite loop, providing reproducible and controlled conditions for testing the vibration monitoring system. Time interval measurements were consistently made with a single count error in a 24-bit count vector. Real time testing was done using a two blade rotor mounted in an evacuated chamber at the Spin Rig Facility at the NASA Glen Research Center. The shaft in this facility was suspended by two radial magnetic bearings and the non-synchronous vibration was communicated to the blades through the magnetic bearing. The shaft motion was much smaller than the blade vibratory amplitude, realistically simulating flutter vibrations. Non-synchronous vibratory amplitudes for the first mode were of the order of twenty mils and for the second mode of the order of a few mils. INTRODUCTION

Blade tip timing techniques are gaining wide acceptance as viable means for characterizing vibrations

of bladed systems in rotating tubomachniery1-4 . Non-intrusive evaluation of the engine blades provides vital information for the long term durability of engine performance. Time event history of each blade can be obtained through the use of either capacitive or optical probes, however, the latter, being more robust and accurate, are generally preferred in the electrically noisy environment of engine testing facilities. Optical blade tip techniques have gone through evolutionary development since their inception in the early 70’s5-12 . In recent years, considerable effort has been devoted to characterizing both synchronous and non-synchronous behavior of blades. Much of this effort has been targeted toward improving the appropriate theoretical models13-14, which are necessary for retrieving the mode structure of the blades.

Quantitative measurements of synchronous and non-synchronous blade vibration have been in use at NASA for several years. However, the current system is based on a commercial timing board that necessitates data analysis after the completion of the run (i.e., not in a real time). Thus, with the current system one could not effectively monitor vibrations. This article describes a real time blade tip timing system that removes this restriction and adds vibration monitoring capability to the current system. One can, therefore, supplement or completely replace the strain gage monitoring with a non-intrusive optical system. Similar systems, which are, however, based on proprietary technology, have been in use by aircraft engine manufacturers. The real time vibration monitoring system described in this paper is based on a 50 MHz oscillator. The system comprises a hardware unit for measuring the time history of all blade events from three independently located light probes, and a graphical user interface (GUI)

DRAFT


Ligh

t Pro

be

inpu

ts

OPR

EdgeDetection Circuit

Register Bank

Register Bank

TCXO Counter

FIFOBank

ReadSequencer

USB TX/RX

SBCR

Com

pute

r

Figure 1: Schematic of the harware unit.

that provides for control of the hardware and processing of the time stamp data for both on-line and off-line processing. The system has been used to monitor controlled blade flutter in real time.

VIBRATION MONITORING SYSTEM Optical front end

Typically, an optical system comprises a number of light probes, which are located in the casing of an engine. The light probes, as described by Dhadwal et al.10, provide coherent illumination of the blade tip through the use of a singlemode optical fiber. The diffuse light reflected from the blade tip is collected by a ring of high numerical aperture optical fibers arranged in an annular ring surrounding the central transmitting optical fiber. The fibers are mounted in a stainless steel tube, and the entire assembly can be used at temperatures of 500 oF. By using non-epoxy based binding techniques, the light probes can be assembled for use at operating temperatures of 1500 oF.

The optical source of choice is a semiconductor laser operating at 780 nm with a power output of 30 mW. Measurements have been performed over distances of 120 m. The passage of the blade produces an optical pulse which is detected by an avalanche photodiode receiver with a 35 ns rise time and a sensitivity of -35 dBm. Typically, blade tip timing data is acquired through the use of either analog-to-digital converters (ADC) or timer boards. The raw data is subsequently processed to determine the vibration amplitude. Timing clock resolution

Blade-tip-timing systems are based on a time interval measurement of a blade edge and a reference edge derived from a notch (or tooth) on the shaft. The latter is often called the once-per-revolution (OPR) pulse. The accuracy of the time interval measurement is determined by the time jitter of both the OPR pulse and blade tip pulse in the absence of vibrations. If the two time jitters are defined by standard deviations σS and σB, for the shaft and blade, respectively, then the corresponding time jitter for the time interval measurement will be

given by ( )2S

2B σ+σ .

This defines the upper limit for the timing clock resolution. Consider a 24 inch span bladed rotor shaft rotating at a nominal speed of 15,000 rpm. Assuming a speed regulation of 0.01% and a root-mean-square (rms) vibration amplitude of 0.5 mil, we find that σS and σB have values of 0.4 µs and 26 ns, respectively. Under

these conditions vibration amplitudes resulting from speed regulation dominate and these can be measured with a clock frequency of 2.5 MHz. Amplitude sensitivity can be further increased by normalizing the instantaneous blade tip amplitude (time) with the instantaneous measurement of the OPR period, resulting in a σS of zero. The practical lower bound of the vibration amplitude is determined by vibrations produces in the absence of forced excitation and away from the blade resonances. For the system discussed here a practical lower limit of vibration is a rms amplitude of 0.5 mil. This can be measured with a clock frequency of 50 MHz. It should be noted that the above discussion has assumed that all other sources of timing jitter, such as, the uncertainty in defining the trigger points are negligible. Hardware design

The primary goal of the hardware unit is to accurately measure the time of all events occurring after the system has been triggered with an OPR pulse. With knowledge of the blade radius and the blade tip speed, the instantaneous position of the blade can be computed. By referencing this measurement to the average blade position, the instantaneous amplitude of the vibrating blade can be computed.

Figure 1 shows a schematic of the hardware unit. It

contains all of the circuitry needed to detect, synchronize, store and transmit the time of interval information to a computer. The system is designed to

DRAFT


accept analog pulses from the output of an optical receiver. The input block uses Schmitt-trigger buffers to condition the analog signals for digital use as well as to filter out low-level noise transients. The hardware was designed to accurately measure the time-of-flight between blade edges and the OPR synchronizing pulse, which originates from the shaft of the rotating assembly. The timing measurements are based on a 50 MHz temperature compensated oscillator (TCXO). The OPR pulse provides synchronization for a 24-bit counter, which measures the time location of all blade events during one revolution. The combination of a 20 ns resolution clock and a 24-bit counter allow the system to successfully measure the performance of most multi-bladed rotating assemblies.

The rate at which time stamp data is generated is

determined by the product of the number of light probes, the number of blades, the count vector size and the rotation rate in revolution per second of the bladed assembly. For example, the data rate is 432 kB/s for a six probe system with a seventy two blade rotor operating at 15,000 rpm. A FIFO (first-in-first-out) memory is used to store the 24-bit count vector along with an 8-bit probe identifier. Upon detection of the OPR pulse the data is moved to a computer via a USB2.0 interface (Universal Serial Data Bus15). The USB2.0 interface has a maximum sustainable data transfer rate of over 50 MB/s, and is plug and play compatible with virtually any modern computer. The USB, being an external connection, allows the entire system to be run from a laptop computer.

The detection and synchronization block

conditions the incoming analog pulses to remove high-frequency noise and to make them compatible with digital switching thresholds; the pulses are then latched and synchronized to the sampling clock (20ns). Additionally, the input stage includes a glitch catcher, which reduces the probability of detecting a ghost blade to negligible levels.

The counter block contains a 24-bit counter,

which is synchronized to the previous block by the sampling clock. When an OPR pulse is detected, it is applied to the reset of the counter. Subsequent edge events latch the output of the counter to the FIFO. Each 24-bit vector, representing a particular time interval, is appended with the corresponding probe identification word. Using this method any change in pulse displacement (greater than 1 LSB) will result in a change of the count value, which can then be detected and displayed by the control software.

The register bank consists of two 24-bit registers for each of the light probes. The two banks are used to latch the count value for each probe edge. Each register is connected directly to the output of the counter in a pipelined configuration.

The reference clock block is based on a 50 MHz

temperature compensated oscillator (TCXO) coupled to a high-current, low-skew clock buffer. This block is used to synchronize the operation of the input stage to the operation of the counters and the register pipeline.

The FIFO memory block is used to temporarily

store the data generated before it is transferred to a computer. The use of 64k x 18 synchronous FIFOs for each light probe allow the hardware to store data for large intervals of time if necessary. This approach is used to prevent the loss of data in the event that the computer is unable to receive its data at the specified time.

A double-redundant FIFO structure enables the

systems to operate during any interrupt flags that may be generated by the USB host controller. This feature works in conjunction with a high-priority programming thread to significantly reduce the likelihood of any data loss. In the unlikely event that data is lost, the discontinuity in the revolution count is an immediate indicator of the severity of the data loss.

The read sequencer block coordinates the

movement of data between the FIFO memories and the USB transceiver. In the event of an interrupt condition, the circuitry automatically locks out the unread FIFO until it has been emptied, while concurrently executing another process, which enables a secondary FIFO to accept the newly arriving data.

The transceiver block contains a third-party

USB transceiver chip as well as an EEPROM (for firmware), and a Type-B physical layer connector.

The status buffer (SB) and control register (CR)

block allows the hardware to be configured and monitored via the USB interface. Hardware simulator test-bed A hardware test bed was designed to simulate the output from three light probes. The simulator, Fig 2a, was programmed to generate an ideal time sequence corresponding to the output of three light probes (LP) measuring a seventy two blade assembly rotating at 15,000 rpm. Fig 2b shows a typical output sequence for the three LPs corresponding to blades widths of 100 ns,

DRAFT


Figure 2b: Output of the hardware simulator.

LP1

LP2

LP3

OPRBlade #1 Last Blade

one revolution

Figure 2b: Output of the hardware simulator.

Clock

ClearCounter

Address BusA

ddre

ssab

le

L

ocat

ions

20 B

it Co

unte

r

EPROM 8-bit

D0 - OPRD1 - LP1D2 - LP1D3 - LP2

D7

D5 - NCD6 - NC

D4 - NCSi

mul

ated

Lig

ht P

robe

Out

puts

Can be slaved or free running

Figure 3: Software/hardware architecture for the vibration monitoring system.

Windows application

Win32 sub-system

Device (function) driver

Bus & host driver

USB host controller

Data acquisition hardware

Personal computer

8051 CPU

USB bus interface

Flutter measurement system

USB peripheral chipUSB wire

Win

dow

s XP

API calls

Software target

Hardware target

200 ns and 300 ns. LP1 and LP2 are separated by a phase angle of 157 µrad. The simulator allows for thorough testing of the blade tip timing hardware and to accurately quantify its performance. The simulator could be run in either a synchronous mode (simulator clock is slaved to the hardware clock) or asynchronously (simulator clock is free running). As expected in the former arrangement the timing measurements were in perfect agreement with the programmed time intervals (see appendix). In the asynchronous case the reproducibility was within one least significant bit of a

24-bit count vector.

Software design

Today, most of the blade tip timing systems acquire data through either a ADC board or a counter/timer

board. Such systems, typically, use a PCI or a VME interface for real time control of the hardware. Our design is based on a custom counter/timer board, which is controlled through the USB interface, making the system portable and machine independent. A friendly graphical user interface (GUI) provides additional control, and on-line and off-line processing capability.

Figure 3 shows a schematic of the three essential components of the software/hardware architecture: 1) programming of the USB peripheral integrated circuit (IC), which resides in the hardware unit; 2) development of a device driver for Windows XP operating system; and 3) development of specific client application software, including a GUI.

A Cypress IC (Part# CY7C68013)16 was used to

provide the USB interface. The Cypress IC contains a serial interface engine (SIE) which controls all USB transfers and essentially hides the lower implementation layers, such as, error correction, and protocol handling. The SIE is controlled through registers which can be accessed by the integrated enhanced 8051 microprocessor. Apart from configuring and controlling of SIE the microprocessor (CPU) can be programmed to interface with other circuitry through I/O ports; this allows control of hardware functions, such as, hardware buffers, control registers and reset. The integrated FIFOs inside the Cypress IC are used for high bandwidth transfers by connecting them directly to the SIE interface, bypassing the CPU.

Large part of the software is written using the C

language and some on 8051 Assembler. It includes a

DRAFT


Figure 4: Window client software architecture.

Real Time Subsystem Acquisition

USB data thread USB control thread

HDD thread User Interface

Offline Subsystem

Data Processing

VCR Functions: Playback,

Pause, Fast/Slow

motion Memory buffer

GUI thread

configuration routine, which is run on start up by the CPU to configure the USB IC in the specified mode. After the configuration step the CPU enters the infinite loop where it constantly checks and services any standard device requests. All other user requests are served by interrupt routines. The USB descriptor tables which provide unique identity and communication interface are coded in assembly language.

The other side of programming a USB device is the device driver and application software on the host. The USB requests should pass through several layers of Windows XP architecture17. While some of them are part of an operating system (system supplied), others must be developed. Each USB device, unless it is compliant with the standard classes (like human interface device or printer) should communicate with application software through a custom device driver. In our present system, we adapted the freeware device driver which comes with Cypress development kit. However, this driver works in a synchronous operation mode. We plan to extend the driver by adding asynchronous (overlapped) mode. Operation in asynchronous mode reduces the delay between two back to back requests from client software. In this mode, the depth of FIFOs used in the hardware unit can be considerable reduced, resulting in a significant cost saving.

The windows client application was developed

using the C++ language. The software, as depicted in Fig. 4, comprises a real time subsystem and an off-line subsystem. The former acquires data from the hardware, saves the raw data to the hard drive and monitors data traffic through a GUI interface. It is responsible for the real time operation of the vibration monitoring system. In the off-line mode, the archived data can be processed and displayed in a VCR-like fashion, i.e. showed in fast or slow motion or frame by frame.

In order to have a responsive interface, while

servicing high priority processes, such as, real-time data logging, the client application is running several Win32 threads. These threads are implemented with Microsoft Foundation Class (MFC) Library18. A USB data thread is given the highest priority and is responsible for transferring the data from the USB to a memory buffer by sending API calls to the device driver. The HDD (hard disk driver) and GUI threads can use the data from the memory buffer to save the raw data to a mass storage device and/or display it on the screen. Real time control of a hardware implemented by the USB control thread processes hardware interrupt requests, such as, FIFO full.

The data, either in real time or in a playback mode,

can be viewed in two formats: the instantaneous and time evolution of a relative displacement to a reference position for any combination of blades/probes. The refresh rate of the display windows is determined by the operating system and the number of open windows. The user may open multiple windows with graphs, zoom scales of individual windows to examine the portions of data, save and print instantaneous values. In the off-line mode user has all the on-line controls but with data coming from the stored raw files, rather than from test engine. The appendix shows some of GUI screen shots.

EXPERIMENTAL RESULTS

The real time vibration monitoring system was tested in NASA’s vacuum spin rig facility at The Glen Research Center. This facility was set up to characterize damping properties of composite blades using a blade tip timing optical system. The real time analysis was not essential for the determination of damping; however, the controlled environment of the facility was ideal for testing the real time operation of the new system.

The shaft in the spin rig facility is suspended with

two radial magnetic bearings. The resonance excitation is accomplished by imparting vibration to the shaft through magnetic bearings. The vibration from the shaft is then transmitted through the hub to a pair of blades that are attached to the hub using a cylindrical dovetail locking arrangement. During rotation the blades are locked into the hub by centrifugal force. The blade span is 22 inches.

At resonance, the induced vibration frequency coincides with the bladed rotor natural mode frequency. The speed of rotation can be arbitrary except for the

DRAFT


320

322

324

326

328

330

332

334

336

338

340

- - - - - - - - - - -

10 12 14 16 18 20-20

-10

0

10

20

End of excitation

Start of excitation

Sweep frequency, Hz

Time, sec

Tang

entia

ldi

spla

cem

ent,

mils

Figure 6: Resonance sweep for blade #1, with probe #1, under bending excitation at 4800 rpm.

restriction that it must not be such that the vibratory frequency is an integral multiple of rotational frequency.

This facility then realistically simulates non-synchronous flutter vibrations that occur in the air breathing engine compression system. Because the vibration is non-synchronous with shaft speed, one optical probe suffices for detection of vibration. However, two optical probes were installed so that frequency of vibration could be independently determined.

Figure 5 shows the blades and the two optical probes

installed in the spin rig. As described by Morrison et al19 the direction of excitation could be locked in a rotating frame of reference. Thus, for the first mode the direction of excitation was perpendicular to the blade stacking axes. The shaft and the hub system were much stiffer than the blade so that the motion imparted to the shaft and the hub was much smaller than to the blades in the neighborhood of the blade resonant frequency. (Also shown in Fig 5 is the optical probe that measured shaft vibrations.) The details of the magnetic levitation and its control are given by Morrison and coworkers19.

An important consideration in the experiment is the frequency aliasing. In this optical system the probe is directed in the radial direction so that the light is reflected from the blade tip only when a blade tip is directly opposite to the probe. Thus sampling rate in this system is not arbitrary but is fixed at once per

revolution. Since the first blade natural frequency is considerably higher than the rotational frequency in the operating range, the data are under sampled and it is important to take this fact into consideration. The data in Fig 6, which were taken to locate the blade resonance, illustrate this problem.

Figure 5: Rectangular cross-section of a blade installed in the spin rig with two blade tip probes, a shaft displacement probe, and the laser that was used to generate the OPR pulse.

DRAFT


Figure 7: Blade vibrations at 4800 rpm in the absence of forced vibrations.

0.00 0.20 0.40 0.60 0.80 1.00

Time (seconds)

-0.50

-0.30

-0.10

0.10

0.30

0.50

Tang

entia

l dis

plac

emen

t (m

ils)

Figure 8: Amplitude response under bending excitation at a frequency of 326 Hz and 4800

0.50 0.88 1.25 1.63 2.00 2.38 2.75 3.13 3.50

Time (seconds)

-40

-20

0

20

40

Tang

entia

l dis

plac

emen

t (m

ils)

Start of excitation

The true frequency sweep during which data were obtained in this figure is from 320 to 340 Hz, but the apparent frequency varies from about 1 Hz at the beginning of the sweep, to about 6.4 Hz at the maximum amplitude, and to about 20 Hz at the end of the sweep. Since the sweep was conducted at 4800 rpm or 80 Hz the 320 Hz is an exact multiple of 80, which explains why the lower end of the sweep is at such a low apparent frequency. The 340 Hz frequency corresponds to 4.25 engine orders (or a multiple of rotational frequency). Thus the apparent frequency is 0.25 engine orders or 20Hz as noted previously. When hunting for resonance we must avoid, therefore, the frequencies that are close to integral multiples of rotational frequency. The other limitation is when apparent frequency reaches the neighborhood of one half of an engine order (which is the Nyquist frequency). In the above example that would occur at 360 Hz, when frequency in the engine orders is 4.5. This limitation occurs when number of points per cycle approaches 2. Practically, these two limitations imply that at certain speeds we would not be able to obtain good vibratory data. In such a case, however, we can choose a speed that is somewhat above or below the particular speed.

The vibration monitoring system was used to obtain non-synchronous measurements for both bending and torsional modes of vibration. A baseline measurement of the system response, i.e. the natural vibration of the two blade structure, was recorded at 4800 rpm. Figure 7 shows a plot of the random jitter of blade #1, as measured by probe #2, in the absence of forced excitation. The rms amplitude of the jitter is 0.019 mils.

Forced excitation was applied to the blade tip through sinusoidal excitation of the magnetic bearings as discussed above. Figure 6 shows a response for a sweep from 320 to 340 Hz at an operating speed of 4800 RPM. Real time monitoring of vibration amplitude can be achieved by either displaying the raw data or by displaying data that is normalized to the mean position of the particular blade. The latter approach allows a scatter type plot of the instantaneous position of all blades in the same window. Such plots provide real time visualization of the vibratory nature of the bladed assembly, adding the capability to predict the onset of uncontrolled instabilities. Separate windows can be opened for each light probe. The reference data can be obtained by averaging over any number of cycles, for example, 200. The instantaneous values of the blade tip amplitudes can be updated every revolution, however, the actual refresh rate is determined by the software load, i.e. number of open windows. A set of screen shots illustrating the features of the vibration monitoring system can be found in the appendix.

Determination of the amplitude, frequency and phase of the vibration mode can be determined by using any number of algorithms being developed for inverting blade tip timing data13-14. Frequency information is easily recovered by determining the phase angle between two light probes. Figure 8 shows a typical amplitude response at 4800 rpm with a forcing frequency of 326Hz.

DRAFT


Figure 10: Amplitude response under torsional excitation at 1106 Hz and 5200 rpm.

5.00 5.50 6.00 6.50 7.00 7.50 8.00

Time, sec

-5

-3

-1

1

3

5

Vib

ratio

n am

plitu

de (m

ils) Start of

excitation

Figure 9: Resonance sweep for blade #2, with probe #1, under torsional excitation at 5200 rpm.

1070

1075

1080

1085

1090

1095

1100

1105

1110

1115

1120

- - - - - - - - - - -

8 12 16 20 24 28-4

-2

0

2

4 End of excitationStart of

excitation

Sweep frequency, Hz

Time, sec

Tors

iona

ldi

spla

cem

ent,

mils

In contrast to bending, vibration amplitudes for torsional forces are an order of magnitude smaller but vibration frequencies are higher. Figure 9 shows the resonance sweep, under torsional excitation, from 1070 Hz to 1120 Hz. Figure 10 shows the amplitude response at an excitation frequency of 1106 Hz. The vibration amplitude at resonance is ~ 4 mils. CONCLUSION In this manuscript we have a described a novel system for real-time monitoring of blade tip vibrations in rotating turbomachinery (see appendix for some screen shots of the system in operating conditions). The usefulness of the system has been demonstrated through the measurement of controlled non-synchronous vibration. Although, not reported here, the system can also provide real time amplitude data for synchronous vibrations. Additionally, blade tip timing data can be processed to characterize vibratory motion of the blade tip, that is, amplitude, frequency and phase shift. ACKNOWLEDGEMENTS

We thank Andrew Provenza for his help in operating

the Spin Rig Facility at the NASA Glen Research Center.

REFERENCES [1] Zielinski, M. and Ziller, G., “Noncontact vibration measurmentds on compressor rotor blades,” Meas. Sci. 11 (2000) 847-856

DRAFT


[2] Gallego-Garrido. J., Dimitriadis, G. and Wright, J.R., “Development of a multiple modes simulator of rotating bladed assemblies for blade tip-timing data analysis,” ISMA 2002, 16-18 September, KLU, Leuven, Belgium [3] Ivey, P.C., Grant K.R., and Lawson, C., “Tip timing techniques for turbomachinery HCF condition monitoring,” The 16th Symposium on Measuring Techniques in Transonic and Supersonic Flow in Cascades and Turbomachines, Cambridge, 23-24 September (2002) [4] Nowinski, M. and Panovsky, J., “Flutter mechanisms in low pressure turbine blades,” Trans. ASME, vol. 122, 82-88 (2000) [5] Nieberding, W.C. and Pollack, J.L., “Optical detection of blade flutter,” Gas Turbine Conf., paper # ASME 77-GT-66 (1977) [6] Kurkov, A. P., and Dicus, J., “Synthesis of blade flutter vibratory patterns using stationary transducers,” Gas Turbine Conf., paper # 78-GT-160, ASME 77-GT-66 (1978) [7] Andrenelli, L., Paone, N. and Rossi, G.L., “Large-bandwidth reflection fiber-optic sensors for turbomachinery rotor blade diagnostics,” Sensors and Actuators A, 32, 539-542 (1992) [8] Nava, P., Paone, N., Rossi, G.L. and Tomasini, E.P., “Design and experimental characterization of a non-intrusive measurement system of rotating blade vibrations,” J. Engr. For Gas Turbines and Power, 116, 657-662 (1994) [9] Reinhardt, A.K., Kadambi, J.R. and Quinn, R.D., “Laser vibrometry measurements of rotating blade vibrations,” Trans. ASME, 117, 484-488 (1995) [10] Dhadwal, H.S., Mehmud, A., Khan, R., and Kurkov, A.P., “Integrated Fiber Optic Light Probe: Measurement of Static Deflections in Rotating

Turbomachinery,” Rev.Sci.Instrum. Vol 67 (2), 546-552 (1996) [11] Dhadwal, H.S, and Kurkov, A.P., “Dual laser probe tip clearance system for turbomachinary," J. Turbomachinery, vol. 121, No. 3, pp.481-485 (1999) [12] Kurkov A.P., and Dhadwal, H.S. ASimultaneous optical measurements of axial and tangential steady state blade deflections,@ ASME TurboExpo 1999, Indianapolis June 7-10, Emerging Sensor Technology Session, Paper #99-GT-310 [13] Carrington, I.B., Wright, J.R., Cooper, J.E. and Dimitriadis, G., “A comparison of blade tip timing data analysis methods,” Proc. Instrn. Mech. Engrs. Vol 125, PartG 301-312 (2001) [14] Heath, S., and Imregun, M., “An improved single-parameter tip-timing method for turbomachinery blade vibration measurements using optical laser probes,” Int. J. Mech. Sci., vol. 38, No 10, pp 1047-1058 (1996) [15] “Universal Serial Bus Specification Revision 2.0,” April (2000) [16] Cypress, Inc., Datasheet, “CY7C68013 EZ-USB FX2™ USB Microcontroller High-speed USB Peripheral Controller final datasheet,” January (2003) [17] Walter Oney, “Programming the Microsoft® Windows® Driver Model,” Second Edition, ISBN: 0-7356-1803-8, December (2002) [18] Jim Beveridge, Robert Wiener, “Multithreading Applications in Win32: The Complete Guide to Threads,” ISBN: 0201442345 , December (1996)

[19] Morrison, C.R., Provenza, A., Kurkov, A., Duffy, K., Montague, G., Mehmed, O., Johnson, D., and Jansen, R.,“Fully Suspended Five Axes Three Magnetic Bearing Dynamic Spin Rig with Forced Excitation,” NASA TP 212694 (2003)

DRAFT


Appendix

Some screen shots of the real time vibration monitoring system

A screen shot of the vibration monitoring GUI. Open windows: system configuration, replay archived data, main menu bar options, data file management and system status.

DRAFT


Scatter plots showing the real time flutter monitoring capability of the system. A separate window is opened for each light probe. Vibration amplitude for all seventy two blades are displayed in real time. The simulator is running in the asynchronous mode.

DRAFT


A snap shot of the real time display showing the amplitude of vibration for each blade tip. This plot was obtained for the hardware simulator operating in the synchronous mode. Since there is no time jitter, the normalized amplitude (instantaneous-average) is zero for all blades.

Scatter Plot

A snap shot of the real time display showing the amplitude of vibration for each blade tip. This plot was obtained for the hardware simulator operating in the asynchronous mode. The normalized amplitude (instantaneous-average) shows a one clock jitter. The window displays a scaled value (decimal). For the actual experiment the amplitude can either displayed in mils or in radians.

Scatter Plot

ASME Turbo Expo 2004 Control, Diagnostics and Instrumentation

Documents

Transcript of ASME Turbo Expo 2004 Control, Diagnostics and Instrumentation