Michelle M. Bright

An Investigation of Surge in a High-Speed Centrifugal Compressor Using Digital PIV

Abstract Compressor stall is a catastrophic breakdown of the flow in a compressor, which can lead to a loss of engine power, large pressure transients in the inlet/nacelle and engine flameout. The implementation of active or passive strategies for controlling rotating stall and surge can significantly extend the stable operating range of a compressor without substantially sacrificing performance. It is crucial to identify the dynamic changes occurring in the flow field prior to rotating stall and surge in order to successfully control these events. Generally, pressure transducer measurements are made to capture the transient response of a compressor prior to rotating stall. In this investigation, Digital Particle Imaging Velocimetry (DPIV) is used in conjunction with dynamic pressure transducers to simultaneously capture transient velocity and pressure measurements in the non-stationary flow field during compressor surge. DPIV is an instantaneous, planar measurement technique which is ideally suited for studying transient flow phenomena in high speed turbomachinery and has been used previously to successfully map the stable operating point flow field in the diffuser of a high speed centrifugal compressor. Through the acquisition of both DPIV images and transient pressure data, the time evolution of the unsteady flow during surge is revealed.

Stall Precursor Identification in High-Speed Compressor Stages Using Chaotic Time Series Analysis Methods

This paper presents a new technique for precursor identification in high-speed compressors. The technique is a pseudo-correlation integral method referred to as the correlation method. To provide a basis for comparison, the traveling wave energy technique, which has been used extensively to study prestall data, is also briefly presented and applied. The correlation method has a potential advantage over the traveling wave energy method because it uses a single sensor for detection. It also requires no predisposition about the expected behavior of the data to detect changes in the behavior of the compressor. Both methods are used in this study to identify stall precursive events in the pressure fluctuations measured from circumferential pressure transducers located at the front face of the compressor rig. The correlation method successfully identified stall formation or changes in the compressor dynamics from data captured from four different configurations of a NASA Lewis single-stage high-speed compressor while it was transitioned from stable operation into stall. This paper includes an exposition on the use of nonlinear methods to identify stall precursors, a description of the methodologies used for the study, information on the NASA high-speed compressor rig and experimental data acquisition, and results from the four compressor configurations. The experimental results indicate that the correlation method provides ample warning of the onset of rotating stall at high speed, in some tests on the order of 2000 rotor revolutions. Complementary features of the correlation method and the traveling wave energy method are discussed, and suggestions for future developments are made.

Investigation of Pre-stall Mode and Pip Inception in High-Speed Compressors Through the Use of Correlation Integral

Five high-speed compressor configurations are used to identify pre-stall pressure signal activity under clean and distorted inlet conditions, and under steady injection and controlled injection conditions. Through the use of a nonlinear statistic called correlation integral, variations in the compressor dynamics are identified from the pre-stall pressure activity far before variations (modal or pip) are observed visually in the wall static pressure measurements. The correlation integral not only discerns changing dynamics of these compressors prior to stall, but is now used to measure the strength of the tip flow field for these five high-speed compressors. Results show that correlation integral value changes dramatically when the stall inception is modal; and it changes less severely when the stall inception is through pip disturbances. This algorithm can therefore distinguish from the pre-stall pressure traces when a machine is more likely to stall due to pips versus modes. When used in this manner, the correlation integral thus provides a measure of tip flow strength. The algorithm requires no predisposition about the expected behavior of the data in order to detect changing dynamics in the compressor; thus, no pre-filtering is necessary. However, by band-pass filtering the data, one can monitor changing dynamics in the tip flow field for various frequency regimes. An outcome of this is to associate changes in correlation integral value directly with frequency specific events occurring in the compressor, i.e., blade length scale events versus long length scale acoustic events. The correlation integral provides a potential advantage over linear spectral techniques The correlation integral provides a potential advantage over linear spectral techniques because a single sensor is used for detection and analysis of the instabilities.

Eddy current sensor signal processing for stall detection

This paper presents algorithms that use data from eddy current sensors mounted in the engine casing for the purpose of gas turbine engine stability monitoring. The focus of this paper is stall detection. Development of a system to detect and compensate for potentially catastrophic engine failures and instability is the primary objective of this work. A motivating objective of our engine work in general has been the development of mathematically well-founded and efficient signal processing algorithms to perform gas turbine engine blade diagnostics and high cycle fatigue prognosis using a minimal number of blade tip monitoring sensors. Our work to date has focused on the general dynamics eddy current sensor (ECS). Our ultimate goal is to extend the functionality of the ECS system beyond diagnostics to active and automatic real-time control of gas turbine engines. Blade tip sensors such as eddy current, capacitive and optical have been being used for some time now in test stand applications to detect engine faults. It has been demonstrated that they are capable of measuring tip clearance, foreign object damage (FOD), blade vibration, and stall/surge. Much of the data analysis for these methods has been performed off line; however, rapid progress is being made toward the goal of real-time detection for use in onboard flight systems. To date, most signal processing techniques using blade tip sensors have been limited to simple parametric measurements associated with the sensor waveform, e.g., measurement of zero crossing locations for time of arrival information or maxima for tip clearance information. Using this type of parametric information, many computations require more than one sensor per stage. The use of a minimal number of sensors is an extremely important practical consideration since each pound that is added to an aircraft engine adds considerable costs over the life cycle of the engine. Because of this we have focused on developing algorithms that allow the reduction in the number of sensors needed for fault prognosis. We have used new parametric algorithms as well as those that make use of the entire ECS waveform. Using our algorithms we have been able to demonstrate the detection of stall cell precursors using a single ECS. These algorithms have been demonstrated in real-time in tests at the NASA Glenn W8 single stage axial-flow compressor facility. The rotor tested, designated NASA Rotor 67, is a fan with 22 blades

Stall Precursor Identification in High Speed Compressor Stages Using Chaotic Time Series Analysis Methods

This paper presents a new technique for precursor identification in high speed compressors. The technique is a pseudo-correlation integral method referred to as the correlation method. To provide a basis for comparison, the traveling wave energy technique, which has been used extensively to study pre-stall data, is also briefly presented and applied. The correlation method has a potential advantage over the traveling wave energy method because it uses a single sensor for detection. It also requires no predisposition about the expected behavior of the data to detect “changes” in the behavior of the compressor. Both methods are used in this study to identify stall procursive events in the pressure fluctuations measured from circumferential pressure transducers located at the front face of the compressor rig. The correlation method successfully identified stall formation or changes in the compressor dynamics from data captured from four different configurations of a NASA Lewis single stage high speed compressor while it was transitioned from stable operation into stall. This paper includes an exposition on the use of nonlinear methods to identify stall precursors, a description of the methodologies used for the study, information on the NASA high speed compressor rig and experimental data acquisition, and results from the four compressor configurations. The experimental results indicate that the correlation method provides ample warning of the onset of rotating stall at high speed, in some tests on the order of 2000 rotor revolutions. Complementary features of the correlation method and the traveling wave energy method are discussed, and suggestions for future developments are made.Copyright

Investigation of Pre-Stall Mode and Pip Inception in High Speed Compressors Through the Use of Correlation Integral

Five high speed compressor configurations are used to identify pre-stall pressure signal activity under clean and distorted inlet conditions, and under steady injection and controlled injection conditions. Through the use of a nonlinear statistic called correlation integral, variations in the compressor dynamics are identified from the pre-stall pressure activity far before variations (modal or pip) are observed visually in the wall static pressure measurements. The correlation integral not only discerns changing dynamics of these compressors prior to stall, but is now used to measure the strength of the tip flow field for these five high speed compressors. Results show that correlation integral value changes dramatically when the stall inception is modal; and it changes less severely when the stall inception is through pip disturbances. This algorithm can therefore distinguish from the pre-stall pressure traces when a machine is more likely to stall due to pips versus modes. When used in this manner, the correlation integral thus provides a measure of tip flow strength. The algorithm requires no predisposition about the expected behavior of the data in order to detect changing dynamics in the compressor, thus no pre-filtering is necessary. However, by band-pass filtering the data, one can monitor changing dynamics in the tip flow field for various frequency regimes. An outcome of this is to associate changes in correlation integral value directly to frequency specific events occurring in the compressor, i.e. blade length scale events versus long length scale acoustic events. The correlation integral provides a potential advantage over linear spectral techniques because a single sensor is used for detection and analysis of the instabilities.Copyright

Dimension determination of precursive stall events in a single stage high speed compressor

This paper presents a study of the dynamics for a single‐stage, axial‐flow, high speed compressor core, specifically, the NASA Lewis rotor stage 37. Due to the overall blading design for this advanced core compressor, each stage has considerable tip loading and higher speed than most compressor designs, thus, the compressor operates closer to the stall margin. The onset of rotating stall is explained as bifurcations in the dynamics of axial compressors. Data taken from the compressor during a rotating stall event is analyzed. Through the use of a box‐assisted correlation dimension methodology, the attractor dimension is determined during the bifurcations leading to rotating stall. The intent of this study is to examine the behavior of precursive stall events so as to predict the entrance into rotating stall. This information may provide a better means to identify, avoid or control the undersireable event of rotating stall formation in high speed compressor cores.

Attractor dimension analysis of a single-stage-high-speed compressor

This paper presents a study of the dynamics for a single-stage, high-speed compressor core, specifically, the NASA Lewis rotor stage 35. Due to the overall blading design for this advanced core compressor, each stage has considerable tip loading and higher speed than most compressor designs, thus the compressor operates closer to stall. Due to the operation of this compressor close to the stall line, it is important to quickly predict the onset of stall. The onset of rotating stall is explained as bifurcations in the dynamics of axial compressors. Data taken from the compressor during a rotating stall event is analyzed. Through the use of a dimension analysis technique, the attractor dimension is determined during the bifurcations leading to rotating stall. The intent of this study is to examine the behavior of precursive stall events so as to predict the entrance into rotating stall. This information may provide a better means to identify, avoid, or control the catastrophic event of rotating stall formation in high-speed compressor cores.

Desktop chaotic systems: Intuition and visualization

This paper presents a dynamic study of the Wildwood Pendulum, a commercially available desktop system which exhibits a strange attractor. The purpose of studying this chaotic pendulum is two-fold: to gain insight in the paradigmatic approach of modeling, simulating, and determining chaos in nonlinear systems, and to provide a desktop model of chaos as a visual tool. For this study the nonlinear behavior of this chaotic pendulum is modeled, a computer simulation is performed, and an experimental performance is measured. An assessment of the pendulum in the phase plane shows the strange attractor. Through the use of a box-assisted correlation dimension methodology, the attractor dimension is determined for both the model and the experimental pendulum systems. Correlation dimension results indicate that the pendulum and the model are chaotic and their fractal dimensions are similar.