Turbomachinery Blade and Stage Vibrations by Patrick McComb An Engineering Project Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements for the degree of MASTER OF ENGINEERING Major Subject: MECHANICAL ENGINEERING Approved: _________________________________________ Professor Ernesto Guitierrez-Miravete, Project Adviser Rensselaer Polytechnic Institute Hartford, CT December, 2013 i © Copyright 2013 by Patrick McComb All Rights Reserved ii CONTENTS LIST OF TABLES ............................................................................................................. v LIST OF FIGURES .......................................................................................................... vi LIST OF SYMBOLS ...................................................................................................... viii ACKNOWLEDGMENT .................................................................................................. ix ABSTRACT ...................................................................................................................... x 1. Introduction and Background ...................................................................................... 1 1.1 Gas turbine Engines ........................................................................................... 1 1.2 Blading ............................................................................................................... 2 1.3 Free Vibration .................................................................................................... 3 1.4 Forced Vibration ................................................................................................ 5 1.5 High Cycle Fatigue (HCF) ................................................................................. 8 1.6 Clipping .............................................................................................................. 9 2. Modeling and Methodology ...................................................................................... 10 2.1 Finite Element Analysis ................................................................................... 10 2.2 Finite Element Model ....................................................................................... 11 2.3 Boundary Conditions ....................................................................................... 13 2.4 2.3.1 Blade Alone.......................................................................................... 13 2.3.2 Fixed Sector Boundaries ...................................................................... 14 2.3.3 Cyclic Symmetry.................................................................................. 15 Solution ............................................................................................................ 17 2.4.1 Free vibration ....................................................................................... 17 2.4.2 Forced Vibration .................................................................................. 17 3. Results and Discussion .............................................................................................. 21 3.1 Free Vibration .................................................................................................. 21 3.1.1 Blade Alone.......................................................................................... 21 3.1.2 Fixed Sector ......................................................................................... 22 iii 3.2 3.1.3 Cyclic Symmetry.................................................................................. 23 3.1.4 Frequency comparison ......................................................................... 27 Forced Vibration .............................................................................................. 29 3.2.1 Frequency and Modeshape comparison ............................................... 29 3.2.2 Post processing ..................................................................................... 31 3.2.3 Mode 1 ................................................................................................. 33 3.2.4 Mode 2 ................................................................................................. 37 3.2.5 Mode 4 ................................................................................................. 41 4. Conclusions................................................................................................................ 46 5. References.................................................................................................................. 48 6. Appendix A: Matlab .................................................................................................. 49 7. Appendix B: ANSYS files ......................................................................................... 50 8. Appendix C: Cyclic Symmetry Modeshapes ............................................................. 53 iv LIST OF TABLES Table 1: Varying Boundary Condition Frequency Comparison ...................................... 27 Table 2. Clipped Blades Frequency Comparison ............................................................ 30 Table 3. Frequency, Deflection and Phase Angle ............................................................ 32 Table 4. Blade Amplitude Summary ............................................................................... 47 v LIST OF FIGURES Figure 1. Simple Brayton Cycle ........................................................................................ 2 Figure 2. Gas turbine engine compressor and turbine ....................................................... 3 Figure 3. Magnitude of Receptance ................................................................................... 7 Figure 4. Phase Shift of Receptance .................................................................................. 7 Figure 5. Blade Geomtry and Break-up ........................................................................... 11 Figure 6. ANSYS Solid45 elements ................................................................................ 12 Figure 7. Rotor and Blade mesh ...................................................................................... 13 Figure 8. Blade Alone Boundary Conditions .................................................................. 14 Figure 9. Fixed Sector Boundary Condititons ................................................................. 15 Figure 10. Cyclic Symmetry Sector Nodes ..................................................................... 16 Figure 11. Cyclic Symmetry Boundary Conditions ........................................................ 16 Figure 12. Forced Response Loading .............................................................................. 18 Figure 13. Baseline Blade ................................................................................................ 19 Figure 14. Clipped Blade 1 .............................................................................................. 19 Figure 15. Clipped Blade 2 .............................................................................................. 20 Figure 16. Blade Alone Modeshapes 1-9 ........................................................................ 21 Figure 17. Fixed Sector Modeshapes 1-9 ........................................................................ 23 Figure 18. 1st Bending and 1st Torsion @ ND 0 .............................................................. 24 Figure 19. 1st Bending and 1st Torsion @ ND 12 ............................................................ 24 Figure 20. Mode 2 @ ND 2 and ND 12........................................................................... 25 Figure 21. Mode 1-9 Nodal Diamter Map ....................................................................... 26 Figure 22. Cyclic Symmetry Modeshapes 1-9 ................................................................ 27 Figure 23. Modes 1-9 Campbell Diagram ....................................................................... 29 Figure 24. Baseline and clipped Modes 1, 2, 4 ................................................................ 31 Figure 25. Resonant Response Amplitudes ..................................................................... 33 Figure 26. Baseline Blade Mode 1 Forced Responses .................................................... 34 Figure 27. Clipped Blade 1 Mode 1 Forced Response .................................................... 35 Figure 28. Clipped Blade 2 Mode 1 Forced Response .................................................... 36 Figure 29. Mode 1 Deflection Comparison ..................................................................... 37 Figure 30. Clipped Blade 1 Mode 2 Forced Responses................................................... 38 vi Figure 31. Clipped Blade 1 Mode 2 Forced Response .................................................... 39 Figure 32. Clipped Blade 1 Mode 2 Forced Response .................................................... 40 Figure 33. Mode 2 Deflection Comparison ..................................................................... 41 Figure 34. Clipped Blade 2 Mode 4 Forced Responses................................................... 42 Figure 35. Clipped Blade 2 Mode 4 Forced Response .................................................... 43 Figure 36. Clipped Blade 2 Mode 4 Forced Response .................................................... 44 Figure 37. Mode 4 Deflection Comparison ..................................................................... 45 vii LIST OF SYMBOLS m = mass c = damping coefficient k = stiffness coefficient αΊ = acceleration αΊ = velocity x = displacement F = force ω = forcing frequency ωn = natural frequency ωd = damped natural frequency ζ = damping ratio α(ω) = receptance r = frequency ratio As = real coefficient Bs = imaginary coefficient Cs = magnitude φs = phase angle xs(t) = steady state response [M] = mass matrix [C] = damping matrix [K] = stiffness matrix {αΊ} = acceleration vector {αΊ} = velocity vector {x} = displacement vector {F} = force vector ND = Nodal Diameter viii ACKNOWLEDGMENT I would like to thank my family and friends, especially my wife Kate for her motivation and support through my degree and this paper. I would also like to thank Professor Gutierrez-Miravete for his knowledge and guidance completing this paper. ix ABSTRACT Blades or integrally bladed rotors in gas turbine engines are subject to dynamic loading and vibrations. Vibration analyses provide a set of information that is valuable in the design of blades, and this data is used to determine overall durability of blade designs. Free vibration analysis of blades is important in identifying frequencies of vibration and critical speed ranges. The free vibration analysis can be performed with different levels of complexity, and simplified loading can simulate more complex situations. Forced vibration analysis builds upon free vibrations by analyzing the displacement or corresponding stress in a blade under a given cyclic load. The simple forced response is important in trying to understand the dynamic response due to complex pressure loading within a gas turbine machine. x 1. Introduction and Background 1.1 Gas turbine Engines The jet age began in the late 1930’s when Hans von Ohain and Frank Whittle developed their first ideas for what has now become the modern jet engine. Since then, the design of jet engines has evolved in size and complexity; however, one aspect that has remained consistent from the onset is the use of rotating blades or turbomachinery. In most modern day jet engines as well as other land based gas turbine engines, understanding the complex mechanical and aerodynamic interaction of turbomachine blading has become paramount to the success of engine design and durability. Modern day axial flow engines typically consist of an inlet, a fan/compressor section, a combustion section, a turbine section and an exit nozzle. Both the compressor section and the turbine section typically consist of several rows, or stages, of blades that vary in size and count from front to back and the two sections are connected by a shaft. Each blade stage is followed by a row of vanes to make up the full stage. The stages of compressor blades suck air into the engine and compress the air, increasing its pressure and temperature for high energy combustion. After the combustion the high energy gas expands through the stages of turbine blades, which turns the shaft and provides the power for operability of the upstream compressor. This process is known as a Brayton cycle engine, seen in Figure 1, and would not be possible for large axial flow machines without the use of many compressor and turbine blades. 1 Figure 1. Simple Brayton Cycle [1] 1.2 Blading Both compressors and turbines typically consist of several stages of blades that are preliminarily designed for optimal aerodynamics, ignoring design and structural constraints. Depending on the application optimal aerodynamics can imply certain levels of efficiency for given thrust and operability requirements. Compressor blades decrease in size from the entry stage to the exit to provide constant mass flow as the air density increases due to compression. Turbine blades increase in size from entry to exit for the same reason as the gas expands. In both cases, each stage consists of a set number of blades that are distributed around a rotor disk which is connected to the rotating shaft. Each blade is initially designed to input or extract a set amount of work at an ideal aerodynamic efficiency for optimal operation. Figure 2 shows a large multi-stage compressor connected by a shaft to a smaller multi-stage turbine section. 2 Figure 2. Gas turbine engine compressor and turbine [2] However, just because a blade is ideal for aerodynamic operation does not mean it is appropriate for the use in a gas turbine engine, especially in a jet engine. The blades in jet engines are subject to extreme loads, many of which can cause vibrations that must be tolerated throughout the operation of the engine. For jet engines, operational safety is extremely important as in many cases many human lives are at stake. It is very important to understand both the cause of the vibrations and the response of the blades to different dynamic loadings. 1.3 Free Vibration One of the most important aspects in the design of turbomachine blades is understanding the dynamics of blade vibration. In reality a blade is a complex system and its dynamics can be represented as a multi-degree of freedom system, where [M] is the mass, [C] is the damping, and [K] is the stiffness. [π]{π₯Μ } + [πΆ]{π₯Μ } + [πΎ]{π₯} = πΉ 3 When used in the Finite Element Method, [M] represents the mass matrix created for each discrete element. Similarly, [C] represents the damping matrix and [K] represents the stiffness matrix. To understand the basics of free vibration a single degree of freedom system is considered first. In this case the differential equation of motion is ππ₯Μ + ππ₯Μ + ππ₯ = πΉ Here m is mass, c is the damping coefficient and k is the stiffness (coefficient of elasticity). The first step in understanding the vibration of a system is to understand the free undamped vibration, or vibration due to initial disturbance of the system. The equation of motion simplifies to [π]{π₯Μ } + [πΎ]{π₯} = 0 And the corresponding single degree of freedom equation simplifies to: ππ₯Μ + ππ₯ = 0 The equation of motion can be solved as an eigenvalue problem where the eigenvectors are the modeshapes or displacements of vibration and the eigenvalues are the natural frequencies of the blade. In many cases it is important to evaluate the free vibrations unloaded, and loaded with certain forces experienced in turbomachinery such as centrifugal loading and the steady pressure loading of the gas passing over the blades. This free vibration analysis is used with other design aspects of a machine to determine potential areas where rotational speed and forcing drivers can create vibrations which the blade material cannot tolerate. 4 1.4 Forced Vibration A free vibration analysis can identify potential regions of vibrational concern; however, it only produces an eigenvalue solution that does not let one predict the amplitude of a vibration due to cyclic loading, like that experienced in a jet engine. In jet engines there are complex disturbances that can be approximately simulated by applying a simple load cyclically as described by the equation below for a single degree of freedom system. ππ₯Μ + ππ₯Μ + ππ₯ = πΉπππ ππ‘ In this equation F is the magnitude of the force and ω is the frequency of the force being applied to the system. This equation can then be rearranged by dividing through by the mass to: π₯Μ + 2πππ π₯Μ + ππ 2 π₯ = πΉ πππ ππ‘ π Where: π = 2πππ π π = ππ 2 π π = √1 − ( ππ 2 ) ππ The transient response of this system can be described with the solution: π₯π (π‘) = πΌ(π)πΉπ π₯ππ‘ = πΉ(π΄π πππ ππ‘ + π΅π π πππ₯ππ‘) The response is harmonic with the frequency of the force and described in real and imaginary terms. In the above equation α(ω) is known as the receptance, which is a function of the frequency response. The receptance can be calculated using the equation: 5 1 1 2 ππ π π πΌ(π) = = −π 2 + 2πππ π₯π + ππ 2 1 − π 2 + 2ππ₯π Where r is the ratio of the forcing frequency to the natural frequency. π= π ππ More practically the receptance can be described in terms of the coefficients As and Bs as: 1 − π2 πππ 2 π΄π = (1 − π 2 + (2ππ)2 2ππ πππ 2 π΅π = (1 − π 2 + (2ππ)2 The real and imaginary solution can be described in the form: π₯π (π‘) = πΉπΆπ cosβ‘(ππ‘ − ππ ) In this description Cs is the magnitude and φs is the phase shift of the solution and are calculated as: πΆπ = √π΄π 2 + π΅π 2 = 1 πππ 2 √(1 − π 2 + (2ππ)2 2ππ 1 − π2 The magnitude and phase shift of the solution can be plotted versus the frequency ratio r. ππ = π‘ππ−1 The shape of the response is dependent on the damping ratio as seen in Figures 3 and 4 plotted in Matlab (Appendix A). 6 Figure 3. Magnitude of Receptance Figure 4. Phase Shift of Receptance The forced response solution has a peak amplitude when: π = ππ √1 − 2π 2 The forced response solution at peak amplitude is known as resonance. Blades in jet engines are subject to resonant responses resulting from cyclic loading due to pressure loading of upstream and downstream disturbances. These disturbances include speed harmonics, vane and rotor pass frequencies, and other aerodynamic and acoustic disturbances. For example, if an upstream stage has 40 vanes, as a blade rotates 1 time around the rotor it is subject to a similar pressure disturbance 40 times, every time it passes by the wake of each vane. The rotating blade will vibrate at a natural frequency 7 harmonic relative to rotor speed and vane count. It is extremely important in the design of turbomachine blading to understand the frequency and amplitude of responses due to different cyclic loadings. 1.5 High Cycle Fatigue (HCF) It is important to understand the cyclic forced vibrations in jet engines in order to avoid failures due to high cycle fatigue. Fatigue is the repeated loading of a structure that over time can cause a failure. As a blade vibrates, the fatigue cycles on a blade add up and eventually cross a threshold dependent on blade material properties that cause a blade to crack and eventually fail. The best way to avoid high cycle fatigue is to avoid resonances; however, in application of real life machines, it is impossible to avoid all resonances in the range that the machine must operate. Typically, resonances are avoided in speed ranges where an engine may operate most often. As a result, forced vibration analysis becomes very important in the design of blades at off design conditions, where the engine may not operate most of the time. This analysis helps to understand the frequencies and amplitudes of vibration when these resonances occur. The blade predicted displacements correspond to a relative stress, which can be compared to material capability to understand the robustness of a design. The robustness of the design can vary, but typically blades are designed to stay below a certain stress value as a result of resonance to assure that a fatigue failure can never occur. 8 1.6 Clipping Although forced vibration analysis is used in the design, the complex nature of the loading can produce a prediction of resonant amplitude that does not match that measured in a test environment. In this situation, forced vibration is valuable as a tool to produce relative magnitude results dependent on geometry changes with the same load. In some cases, blades can be altered in minor ways, sometimes known as clipping, to change the natural frequency or the magnitude of the blade response (or both), to make a design less susceptible to a fatigue failure. 9 2. Modeling and Methodology 2.1 Finite Element Analysis Although simple hand calculations can be very effective in the preliminary understanding of a blade design, with the modern computing power today, the best way to perform vibration analyses is with the Finite Element Method. Finite element models (FEMs) can be made to approximate blade geometries, and can then be loaded to simulate real life operational conditions. Static and dynamic analyses are then used to help understand the failure mechanism such as HCF. Finite element analysis takes a real structure and approximates it by breaking it up into a number of discrete elements with a finite number of degrees of freedom. In a free vibration analysis the number degrees of freedom of a given model corresponds to the number of natural frequency modes that can be extracted, which in most cases is more than is necessary. The equations for the vibrating system with multiple degrees of freedom neglecting damping as discussed earlier is: [π]{π₯Μ } + [πΎ]{π₯} = 0 In this scenario [M] and [K] are the mass and stiffness matrices respectively, which are generated from the model elements and nodes. The {x} is a vector of displacements of each degree of freedom of the system. If the solution is assumed to be harmonic then the resulting solution is an eigenvalue problem in the form: det([πΎ] − π2 [π]) = 0 The solution of the eigenvalue problem includes both the eigenvalue (natural frequencies) and the eigenvectors (mode shapes). The equation for the forced vibration analysis with damping is: 10 [π]{π₯Μ } + [πΆ]{π₯Μ } + [πΎ]{π₯} = {πΉ} Finite element analysis is used to determine the corresponding magnitudes of deflection at a given frequency and phase angle. 2.2 Finite Element Model Commercial modeling software, Unigraphics NX6 and ANSYS version 12.1, were used to produce the geometry and finite element model for this analysis. The blade model represents a cyclic symmetric sector of an integrally bladed rotor (IBR) typically used in modern compressors. To produce the geometry a simple full rotor disk was revolved 360 degrees. Then a simple blade was place onto the rotor disk and merged together. Twenty-four blades were placed symmetrically around the disk. A 15 degree sector was used to produce a symmetric sector. Finally, the rotor sector was broken up strategically using the curvature of the blade to produce clean sweep meshable geometry. The resulting geometry for this analysis can be seen in Figure 5. Figure 5. Blade Geomtry and Break-up 11 The model was then meshed using ANSYS solid45 elements and default key options. These elements are 3-D 8 noded brick/hexagonal elements, with three translational degrees of freedom at each node seen in Figure 6. Figure 6. ANSYS Solid45 elements The best way to mesh cyclic symmetric geometry is to sweep the mesh through the rotation. This will assure that the mesh on each cyclic edge of the rotor will have the same mesh, which is important when running cyclic symmetry. The final meshing step is to sweep mesh the blade. It is important to specify an appropriate mesh size for the model to produce good frequency results, while minimizing run time. The model was meshed using a general element edge length of .2”. The blade itself was meshed with 4 elements through the thickness in order to properly capture the bending stiffness using solid elements. The fidelity of the mesh can be verified by refining the mesh until frequencies remain constant and converged. The final blade mesh has 5256 elements and 6602 nodes and is shown in Figure 7. 12 Figure 7. Rotor and Blade mesh The blade and rotor were assumed to be made from a titanium alloy (Ti 6-4) with a modulus of elasticity of 16.5 MSI, Poisson’s ratio of .35, and a density of .160 lb/in3. 2.3 Boundary Conditions 2.3.1 Blade Alone The simplest way to model a blade and rotor is to model just the blade. This method is used in blade design optimization, when solution accuracy may not be as critical as run time. This model only has 924 elements and 1345 nodes. Physically, the blade alone solution tends to be stiffer and have higher frequencies since it essentially assumes that the rotor is infinitely stiff. In this model, the blade nodes on the bottom of the blade are constrained in all 3 translational degrees of freedom as shown in Figure 8. 13 Figure 8. Blade Alone Boundary Conditions 2.3.2 Fixed Sector Boundaries Another simplified way to model the boundary conditions of the model is to fix the sector boundary nodes in all degrees of freedom. In this scenario more of the elasticity of the rotor structure is considered in the solution, however, the model will still be ignoring some of elastic properties of the rotor. In the case where the rotor is very stiff, the solution can typically be very similar to an N/2 cyclic symmetry solution (where N is the number of blades in the rotor). The model with these BCs can be seen in Figure 9. 14 Figure 9. Fixed Sector Boundary Condititons 2.3.3 Cyclic Symmetry ANSYS has the capability to model a full 360 degree rotor by using cyclic symmetry such that the full rotor does not have to be modeled. This reduces model complexity and run time while, allowing for full interaction between the rotor and blade. To set up the model for cyclic symmetry the two sector edges of the rotor have to have matching meshes by either sweep meshing or using a mesh copy. ANSYS has an automated procedure that detects the coordinate system of symmetry and the proper sector edges to produce the appropriate constraint equations seen in Figure 10. This produces a full 360 degree rotor model in which all sectors are identical. 15 Figure 10. Cyclic Symmetry Sector Nodes Several rows of nodes on the inside diameter of the rotor are fixed in all degrees of freedom to simulate the rotor tied to an arbitrary shaft. It is important not to select nodes on the cyclic symmetry boundary. These boundary conditions are shown in Figure 11. Figure 11. Cyclic Symmetry Boundary Conditions Unlike the first use of the simpler boundary conditions, the cyclic symmetry solution provides analytical results for the interaction between the rotor vibration and the blade vibration. The blade frequencies are solved corresponding to a particular disk mode or nodal diameter (ND). Different rotor nodal diameters cause the blade mode shapes and frequencies to change. Depending on the purpose of the analysis it is 16 important to track the blade frequencies and modes for each nodal diameter from ND 0 to ND N/2, N being the number of blades. For example, a stiff wise bending blade mode may couple more with the vibration of the disk, and may be a resonance to avoid. 2.4 Solution 2.4.1 Free vibration ANSYS has a variety of methods to solve modal analysis. The model in this project was solved using the Block Lanczos mode extraction method, which is efficient at solving large eigenvalue problems using the sparse matrix solver. Linear or non-linear prestressed static analysis can be solved before the modal solution to capture the effects of stress stiffening, spin softening, and large displacements. Also, thermal affects can be included in the model using temperature dependent material properties and thermal mapping. Each step of increased complexity increases run time. As a result, the solutions presented here ignore both thermal and pre-stress effects on the rotor and blade system. 2.4.2 Forced Vibration The forced vibration solution uses the frequency and displacement results from the linear free vibration analysis. The fixed sector boundary conditions were selected above to perform this analysis for simplicity. It will be demonstrated that the rotor modeled is significantly stiff enough to assume similar modal results to the cyclic symmetry N/2 analysis. An arbitrary 10 lb force is applied at the leading edge tip of the blade normal to the blade surface, which is used as a simple harmonic force in the analysis. This is a simple loading case; a more complex multi-point load, or pressure load could be used in 17 a real application to more accurately simulate the loading. Figure 12 shows the forced response model with the applied force. Figure 12. Forced Response Loading ANSYS uses a harmonic analysis to solve the forced vibration problem. The mode superposition solution method was used because it is the fastest of the methods and is recommended by ANSYS. The sparse solver was also used in the solution and the output format of the solution is real and imaginary. The damping ratio for the harmonic solution was assumed to be .01, which is a standard value for metals, but could be varied. The harmonic solution can be solved for a set number of sub-steps over a range of frequencies. The frequency range can sweep and capture a large number of resonances, or the range can be narrowed to focus on a known resonant frequency. A control file for performing this analysis is found in Appendix B. Three models were used to complete the forced response investigation. All three used the same rotor model, but each had a different blade. The second and third blade models are the same as the baseline blade except with small clips of the trailing edge blade tips. Clipped blade 1 has material removed .058” chord wise from the trailing edge 18 and 1” down the span. Clipped blade 2 has material removed .116” chord wise from the trailing edge and 1” down the span. The blade models can be seen in Figured 13-15. The blade response of these three models will be compared using the forced response analysis. Figure 13. Baseline Blade .058” x 1” clip Figure 14. Clipped Blade 1 19 .116” x 1” clip Figure 15. Clipped Blade 2 20 3. Results and Discussion 3.1 Free Vibration The free vibration mode shapes and frequencies will be presented below for each set of boundary conditions. All of the results will be compared and summarized. 3.1.1 Blade Alone The simplest solution results are from the model with blade alone boundary conditions. The first nine modes were extracted within the first 20,000 hertz. The nodal solution combined X, Y, and Z deflections (USUM) were used to plot the modeshapes seen in Figure 16. Corresponding frequencies are summarized in Table 1 on page 24. Figure 16. Blade Alone Modeshapes 1-9 21 The first two modes (M1 and M2) are first bending and first torsion respectively, with mode shape complexity typically increasing with each mode. Identify blade modes can be more difficult as a blade shape gets more complex; however, for this simple model M5 is second torsion and M9 is third torsion, while M3 is second bending and M6 is third bending. M4 and M8 have a stiff wise bending component. 3.1.2 Fixed Sector The first nine modes within the first 20,000 hertz were also extracted for the fixed rotor sector boundary conditions. The nodal solution USUM (X,Y,Z) deflections were used to plot the modeshapes seen in Figure 17. Corresponding frequencies are summarized in Table 1 on page 24. 22 Figure 17. Fixed Sector Modeshapes 1-9 The first 9 mode shapes with the fixed sector boundary conditions are the same as the blade alone, except modes 7 and 8 have flipped in order. Also, as expected, the addition of the rotor sector has decreased the natural frequencies of each mode. For example, blade alone M1 is ~1374 Hz, while the fixed sector M1 is ~1264 Hz. 3.1.3 Cyclic Symmetry Cyclic symmetry provides rotor and blade vibration combined solutions. The /CYCEXPAND command can be issued to expand the number of sectors around the rotor from 1 to a full 360 degree rotor. If the rotor is expanded to a full 360 degrees the full rotor modeshape can be viewed, as well as the interaction with the blades. The 23 analysis was run for every nodal diameter from 0 to N/2, which is ND 12 for this system. Figure 18 shows the blade first bending mode and first torsion mode respectively at ND 0 with all of the blades vibrating in phase. Figure 18. 1st Bending and 1st Torsion @ ND 0 Figure 19 shows the blade first bending (1B) mode and first torsion (1T) mode respectively at ND 12 with all of the blades vibrating out of phase. Figure 19. 1st Bending and 1st Torsion @ ND 12 Even though the blade mode shapes are similar for ND 0 and ND 12 the system vibration and frequency can be different. In some cases the disk is stiff enough and blade frequencies only have small variations from the blade alone simulation, but certain blade 24 and rotor modes can interact to dramatically change the system vibration. Figure 20 shows the mode shape for mode 2 at nodal diameters 2 and 12. Figure 20. Mode 2 @ ND 2 and ND 12 The modeshape at ND 12 is 1T, but at ND 2 the mode shape has changed to a stiffwise bending mode driven by the rotor, indicated by the displacement seen in the rotor. The M2 ND2 image shows the lighter blue shades into the rotor section, implying a larger amplitude of displacement and more vibration interaction. The blade and disk interaction has changed the modeshape. Also, not only is the mode shape affected, but the frequency will vary for each nodal diameter. A good way to track mode sensitivity to nodal diameter is to plot nodal diameter versus frequency for each mode, sometimes referred to as a nodal diameter map. Figure 21 shows a nodal diameter map for ND0 to N/2 for the first 9 modes, each colored symbol representing a different mode. 25 Figure 21. Mode 1-9 Nodal Diamter Map This chart shows how the lower ND can have a dramatic effect on the system fundamental frequencies. This can be extremely important in understanding certain resonant crossing in the engine. One can also observe that as the higher nodal diameters approaching ND 12 are less influenced by the rotor vibration and the ND 12 mode shapes and frequencies can be approximated by the fixed sector boundary conditions. To plot mode shapes for a single sector, a phase sweep is performed to find the phase angle with the peak vibration amplitude. Then each modal deflection is plotted at that phase angle. Figure 22 shows the first 9 modes up to 20,000 Hz for nodal diameter 12. These mode shapes are comparable to those for the fixed sector in Figure 17. The mode shapes for the other nodal diameters can be found in Appendix C. 26 Figure 22. Cyclic Symmetry Modeshapes 1-9 3.1.4 Frequency comparison Each set of model boundary conditions produces different frequency predictions for the first 9 modes, summarized in Table 1. Table 1: Varying Boundary Condition Frequency Comparison Mode 1 2 3 4 5 6 7 8 9 Delta (Fixed Sector to Blade Alone (Hz) Fixed Sector (Hz) Cyclic ND 12 (Hz) Cyclic ND12) 1373.898809 1263.495429 1259.04 0.35% 3853.311373 3652.017458 3636.17 0.44% 5017.390281 4556.490203 4525.04 0.70% 6012.422854 5251.624095 5197.94 1.03% 9011.647358 8637.582689 8617.83 0.23% 12005.7108 11256.90835 11192.8 0.57% 15589.20011 13139.89331 12908.9 1.79% 15985.13839 13715.12062 13497.9 1.61% 16238.25959 15448.72193 15408.3 0.26% 27 Table 1 shows that the blade alone model produces higher frequencies than the other methods as a result of the assumption that the rotor is infinitely stiff. However, even though the fixed sector model makes some simplified assumptions, the frequencies are very similar to the nodal diameter 12 frequencies, with the largest difference being about 1.8% for mode 7. All other modes except mode 8 had less than a 1% delta in frequency. Looking at the mode shapes it can also been observed that the fixed sector and ND 12 model have nearly identical mode shapes. Since the fixed sector result correlate so well to the cyclic symmetry model, for simplicity the following forced response analysis is performed using the fixed sector model. In the jet engine application it is extremely important to plot the natural frequencies versus rotor speed, which is known as a Campbell or interference diagram. The Campbell diagram was introduced by Wilford Campbell who used this tool to understand the interaction between modal frequencies and excitation forces. The diagonal lines beginning at the origin are called driver lines or engine orders and represent the potential disturbances that could affect the rotating blades. Figure 23 is an example of a Campbell diagram where every 10 engine orders are called out as potential drivers of concern. The horizontal lines corresponds to the modes of concern shown in Table 1. 28 Figure 23. Modes 1-9 Campbell Diagram Every mode and driver line crossing represents a potential resonant response; however, known drivers like vane and blade counts of concern can be identified in a design. In this simulation frequencies are the same for each speed range, however, thermal and prestress affects can influence the blade frequencies in a true engine application. The conditions will affect the speed at which particular resonances can occur. 3.2 Forced Vibration 3.2.1 Frequency and Modeshape comparison The first step in running the forced vibration analysis was to run a free vibration analysis with fixed sector boundary conditions for each of the three blades being investigated (Figures 13-15). Table 2 shows the frequencies and deltas for the first 9 modes; Modes 1, 2, and 4 are highlighted because these modes were investigated using the forced vibration analysis. Delta 1 is the frequency delta between the baseline and clipped blade 1. Delta 2 is the frequency delta between the baseline and clipped blade 2. 29 Table 2. Clipped Blades Frequency Comparison Fixed Sector (Hz) Clip1 (Hz) Clip2 (Hz) 1263.495429 1269.8 1278.6 3652.017458 3672.1 3695.4 4556.490203 4606.9 4661.1 5251.624095 5257.6 5273.7 8637.582689 8722.4 8814.8 11256.90835 11259 11286 13139.89331 13155 13200 13715.12062 13767 13835 15448.72193 15513 15592 Delta 1 0.50% 0.55% 1.11% 0.11% 0.98% 0.02% 0.11% 0.38% 0.42% Delta 2 1.20% 1.19% 2.30% 0.42% 2.05% 0.26% 0.46% 0.87% 0.93% Clipped blade 1 has a frequency delta of .5% to 1% when compared to the baseline blade. Clipped blade 2 has a frequency delta of 1% to 2% when compared to the baseline blade. Since material was clipped from the trailing edge tip, these results confirm the natural frequencies should increase. Similarly, clipped blade 2 has a larger frequency delta since it has a larger clip. Although clipping can be done to intentionally change the modeshape, the intent of this study was to understand the change in response to a particular mode due the material removal. Figure 24 shows that Modes 1, 2, and 4 remain the same for the 3 blade geometries. Mode 1 is first bending, Mode 2 is first torsion and Mode 4 is second bending. These modes were selected because they all have considerable deflection at the leading edge tip; mode 3 has a node line through the leading edge tip so it was not investigated in this manner. 30 Figure 24. Baseline and clipped Modes 1, 2, 4 3.2.2 Post processing After each harmonic analysis run is complete it needs to be post processed by both the time history post processor and the general post processor. The time history post processor is used to define the frequency and phase angle to expand the peak amplitude results with the expansion pass solution. The frequency (.rfrq) results file needs to be opened in the time history post processor. The leading edge tip node tangential deflection is added as a variable to track phase angle versus deflection and select a frequency and angle to expand the solution. The tangential deflection was chosen because it had the largest amplitude for these modes. Table 3 shows the tangential deflections and corresponding phase angles for the baseline blade swept from 1250 to 1300 Hz. 31 Table 3. Frequency, Deflection and Phase Angle Mode1 - Baseline Frequency (Hz) 1250 1256 1260.8 1262.5 1263.1 1263.4 1263.5 1263.6 1263.8 1264.5 1266.2 1271.1 1284.6 1300 UY (in.) Phase angle ( Λ ) 0.036168 -42.6445 0.045217 -58.8116 0.051192 -77.5298 0.052096 -85.2486 0.052191 -88.0802 0.052194 -89.1008 0.052188 -89.6737 0.052177 -90.2465 0.052145 -91.2671 0.051969 -94.0987 0.050848 -101.817 0.044377 -120.536 0.026298 -148.51 0.016520 -160.306 An expansion pass solution can then be run to produce the full solution at the frequency and phase angle with the largest corresponding deflection, in this case, 1263.4 Hz and -89.1degrees. Once the expansion pass is complete the overall deflections can be plotted to represent peak magnitude of the forced response. Figure 25 is a plot of tangential displacement (UY) of the LE tip node versus frequency for the baseline blade. Each spike on the chart is a resonance which corresponds to modal frequencies M1, M2, M4 and their amplitude. This analysis was run from 0 to 8000 Hz to capture the first 4 modes. Mode 3 again does not show up here because there is no displacement of the LE tip node. 32 Figure 25. Resonant Response Amplitudes 3.2.3 Mode 1 Figures 26 through 28 show the amplitude versus frequency plots for each blade from 1250 to 1300 Hz and corresponding peak amplitude. The plots show the resonance increasing in frequency and show a small varying peak tip node amplitude. 33 Figure 26. Baseline Blade Mode 1 Forced Responses 34 Figure 27. Clipped Blade 1 Mode 1 Forced Response 35 Figure 28. Clipped Blade 2 Mode 1 Forced Response Figure 29 compares the peak amplitude of deflection for the 3 blades analyzed. For first bending, the peak tip deflection increases as the clip size is increased. 36 Figure 29. Mode 1 Deflection Comparison 3.2.4 Mode 2 Figures 30 through 32 show the amplitude versus frequency plots for each blade from 3600 to 3700 Hz and corresponding peak amplitude. The plots show the resonance increasing in frequency and show a small varying peak tip node amplitude. 37 Figure 30. Clipped Blade 1 Mode 2 Forced Responses 38 Figure 31. Clipped Blade 1 Mode 2 Forced Response 39 Figure 32. Clipped Blade 1 Mode 2 Forced Response Figure 33 compares the peak amplitude of deflection for the 3 blades analyzed. For first torsion, the peak tip deflection decreases as the clip size is increased opposite of the first mode. 40 Figure 33. Mode 2 Deflection Comparison 3.2.5 Mode 4 Figures 34 through 36 show the amplitude versus frequency plots for each blade from 5200 to 5300 Hz and corresponding peak amplitude. The plots show the resonance increasing in frequency and show a small varying peak tip node amplitude. 41 Figure 34. Clipped Blade 2 Mode 4 Forced Responses 42 Figure 35. Clipped Blade 2 Mode 4 Forced Response 43 Figure 36. Clipped Blade 2 Mode 4 Forced Response Figure 37 compares the peak amplitude of deflection for the 3 blades analyzed. For second bending mode, the peak tip deflection for the baseline blade and smaller clip are about the same. For the larger clipped blade the peak deflection than decreases. 44 Figure 37. Mode 4 Deflection Comparison 45 4. Conclusions Understanding the characteristics of turbomachinery blade vibration is essential in the design of gas turbine engines. It is even more important in jet engines when the livelihood of possibly hundreds of passengers is on the line. Free vibration analysis can be run using a variety of methods and boundary conditions ranging from the simple blade alone analysis to the more complex cyclic symmetry model with non-linear prestress. Each type of model can have value in the blade design. While the cyclic symmetry model may capture the most representative physics, it may be too large or complicated for optimization work. It was shown that in some situations it is necessary to understand the complex interactions between the individual blade vibrations coupled with the rotor. However, in the case of the simple rotor and disk presented here, simplified boundary conditions were sufficient to capture modal frequency results within 1-2% of the more costly cyclic symmetry analysis. The forced vibration analysis uses the results of the free vibration analysis and predicts blade responses to different harmonic loadings. The true unsteady aerodynamic loading may be difficult to model, but can be simplified to excite blades in certain modes. One way this analysis can be valuable is by evaluating the response of blades with slight geometric variations. By loading the blades in a consistent manner, deflections (or corresponding stresses) can be compared to understand and improve a blade design Table 4 summarizes the deflection amplitudes for each of the three blades analyzed, with Delta 1 being the change from baseline to clipped blade 1 and Delta 2 from the baseline to clipped blade 2. 46 Table 4. Blade Amplitude Summary Mode 1 2 3 Baseline (in.) 0.052988 0.010053 0.00397 Clip 1 (in.) 0.05327 0.009913 0.003973 Clip 2 (in.) 0.053422 0.009789 0.003947 Delta 1 0.53% -1.39% 0.08% Delta 2 0.82% -2.63% -0.58% First bending increases deflection with clip size from about .5% to .8% with the larger clip. The torsion is more sensitive than either first bending mode and deflection decreases from about 1.4% with the smaller clip to 2.6% with the larger clip. The second bending mode shows an interesting trend, with the smaller clip having very little affect at all, while the larger clip shows a decrease in deflection of about .6%. Overall, these deflections only track one node on the blade, but do show a trend for how a small change in geometry can cause a change in predicted deflection. The results show how much detail must be put into understanding blade vibration. A small clip may help to improve one mode deflection, but could increase the response on a different mode. At this point in the blade design it is usually very important to gather data from true running conditions to correlate the amplitudes of blade response for different modes at different resonant crossing speeds. The more information a blade designer can gather from vibrations analysis and machine testing the more robust a design can be, and the safer each engine can perform. 47 5. References Images [1] http://web.mit.edu/16.unified/www/SPRING/propulsion/notes/node27.html [2] http://www.powergeneration.siemens.com/products-solutions-services/ [3] ANSYS 12.1 Theory Reference, ANSYS Corporation, 2009. [4] Friswell, M.J., Penny, J. E. T., Garvey, A. D., Lees, A. W. (2010). Dynamics of Rotating Machines. Cambridge University Press, New York, New York. [5] Hassan, Mohammed (2008). Vibratory Analysis of Turbomachinery Blades. Rensselaer Polytechnic Institute Master’s Report, Hartford, CT. [6] Hartog, J.P Den (1985). Mechanical Vibrations. Dover Publications, Inc. , New York, New York. [7] Hill, Philip G, Peterson, Carl R. (1992). Mechanics and Thermodynamics of Propulsion. Addison-Wesley Publishing Company, Inc., Reading, Massachusetts. [8] Singh, M., Vargo, J., Schiffer, D., Dello, J. (2002). Safe Diagram – A Design and Reliability Tool for Turbine Blading. Dresser-Rand Company, Wellsville, NY. [9] Synder, Daniel (2011). A Modeling Study of the Sensitivity of Natural Frequency of Vibration to Geometric Variations in a Turbine Blade. Rensselaer Polytechnic Institute Master’s Report, Hartford, CT. [10] Timoshenko, S., Young, D.H., Weaver Jr., W. (1974). Vibration Problems in Engineering, Fourth Edition. John Wiley & Sons, New York, New York. 48 6. Appendix A: Matlab %Matlab Script to plot receptance for various damping ratios% clear; Xo=1; Co=1; m=1; On=1; xsi=.1; xsi2=.2; xsi3=.5; O=On*sqrt(1-(xsi^2)); r = 0:.01:2; Cs=((1/(m*On^2))./sqrt((1-r.*r).*(1-r.*r)+(2*xsi*r).*(2*xsi*r))); Cs2=((1/(m*On^2))./sqrt((1-r.*r).*(1-r.*r)+(2*xsi2*r).*(2*xsi2*r))); Cs3=((1/(m*On^2))./sqrt((1-r.*r).*(1-r.*r)+(2*xsi3*r).*(2*xsi3*r))); plot(r,Cs,r,Cs2,r,Cs3) %Matlab Script to plot phase angle for various damping ratios% clear; Xo=1; Co=1; m=1; On=1; xsi=.1; xsi2=.2; xsi3=.5; eta=.2; O=On*sqrt(1-(xsi^2)); r = 0:.01:2; n=2*xsi*r; d=1-r.*r; t=n./d; phi=atan2(n,d); n2=eta; d2=1-r.*r; t=n2./d2; phi2=atan2(n2,d2); plot(r,phi,r,phi2) 49 7. Appendix B: ANSYS files !This file can run both a modal blade alone analysis or the fixed sector boundary analysis !with meshed geometry and defined boundary conditions. /filname,ibr resume,ibr,db /prep7 !om = 837.758 nmodes = 9 fqlow = 0.0 fqhi = 20000.0 !Enter Preprocessor !Define angular velocity (omega) !Run for 9 modes !Low end of frequency range 0 Hz !High end of frequency range 20,000 Hz MP,DENS,1,.000414 MP,EX,1,1.65e6 MP,PRXY,1,.35 !Define Material Density !Define Material Modulus !Define Material Poisson !Modal Solution! /solu antype,2 eqslv,spar !Enter Solution !Modal Solution !Use Sparse Solver !Define frequencies & modes! modopt,lanb,nmodes,fqlow,fqhi,,off mxpand,nmodes,,,yes !omega,om,,, !Define Frequency Range !Number of Modes and element results !Apply Angular velocity if necessary save solve finish /exit,nosa 50 !This file can run modal analysis cyclic symmetry IBR sector model !with meshed geometry and defined boundary conditions. /clear /filname,ibr resume,ibr,db /title, ibr, !om = 837.758 nmodes = 18 fqlow = 0.0 fqhi = 20000.0 ndlow = 0 ndhi = 12 ndinc = 1 !Define angular velocity (omega) !Run for 18 modes (2X for first 9) !Low end of frequency range 0 Hz !High end of frequency range 20,000 Hz !Low end of Nodal Diameter Range !High end of Nodal Diameter Range !Run each Nodal Diameter increment /prep7 !Enter Preprocessor mp,dens,1,.000414 mp,ex,1,1.65e6 mp,prxy,1,.35 !Define Material Density !Define Material Modulus !Define Material Poisson allsel,all cycopt,toler,3e-3 csys,6 cyclic,24,360/24,6,cyc,1 !Select the whole model !3 thousandths of a inch cyclic tolerance !Cylindrical coordinate system-rotation about X !24 blade sector, 15 degrees each !Modal Solution! /solu antype,2 eqslv,spar !Enter Solution !Modal Solution !Use Sparse Solver !Define frequencies & modes! modopt,lanb,nmodes,fqlow,fqhi,,off pstress,0 mxpand,nmodes,,,0 !omega,om,,, cycopt,hindex,ndlow,ndhi,ndinc,0, !Define Frequency Range !No Pre-stress !Number of Modes and element results !Apply Angular (x) velocity if necessary !Define Nodal Diameter range solve finish save 51 !This file runs a modal analysis with fixed sector boundary conditions !Defines a point load at the tip LE of the blade !Runs a harmonic modal superposition analysis based on frequency inputs, and modal solutions. /clear /filname,ibr resume,ibr,db /output,ibr,output /prep7 forcen=6511 appf=10 !Modal Solution! /solu antype,2 eqslv,spar dmprat,0.01 modopt,lanb,9,,,,OFF mxpand,9,,,1 solve save,,,,,,model !Application of the Harmonic Force! /prep7 nwpave,6511 nwplan,-1,6511,6523,6577 cswpla,15,0,1,1, csys,15 nrotat,6511 F,forcen,FY,appf CSYS,0, finish !Harmonic Solution! /solu antype,harmic hropt,msup,4,1,yes hrout,on,on,on harfrq,1250,1300 nsubst,25 kbc,1 save solve finish /eof !Enter Preprocessor !Node for applied force !Magnitude of point force at above node !Enter Solution !Modal Solution !Use Sparse Solver !Damping ratio for forced response !Define Frequency Range !Number of Modes and element results !Enter Preprocessor !Node of work plane origin !Nodes defining work plane normal to blade surface !Define local csys (coordinate system) !Work in local csys !Rotating nodal coordinate to csys 15 !Apply harmonic force !Returning to global csys !Enter Solution !Run Harmonic solution !Run modal superposition, 4 modes, viscous damping !Output Real/Imaginary, Cluster, Mode Contributions !Frequency range to explore !Number of substeps per freq. range !Step changed loads 52 8. Appendix C: Cyclic Symmetry Modeshapes 53 54 55 56