THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS Three Park Avenue, New York, N.Y. 10016-5990 99-61-447 The Society shall not be responsible for statements or opinions advanced In papers or discussion at meetings of the Society or of it Divisions or Sections, or printed in its publications. Discussion is printed only If the paper is published in an ASME Journal. Authorization to photocopy for internal or personal use is granted to libraries and other users registered with the Copyright Clearance Center (CCC) provided SS/article is paid to CCC, 222 Rosewood Dr., Danvers, MA 01923. Requests for special permission or bulk reproduction should be addressed to the ASME Technical Publishing Department. Copyright 0 1999 by ASME M Rights Reserved Printed in U.S.A. 111111111111011111111111 PROBABILISTIC ANALYSES OF HSCT COMBUSTOR LINER COMPONENTS Shantaram S. Pal, and P. L. N. Murthy National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135 Presented at the International Gas Turbine & Aeroengine Congress & Exhtftion Indianan*, Indiana June 7 June 10, 1999
ABSTRACT Deterministic as well as probabilistic analysis results for a typical cylindrical combustor liner segment is presented in this paper. The type of liner segment analyzed in this report is a part of a HSCT combustor. In the analyses, thermal profiles, material stiffnesses, and mechanical loadings are considered random. Probabilistic static, stability and free vibration analysis were conducted and results are presented in the form of cumulative distribution functions for critical location stresses, and buckling load. The results indicated that scatter in thermal loads and thermal expansion coefficients have significant impact on the longitudinal and hoop stresses. Also, the thickness variations appeared to be most significant, in influencing the probabilistic buckling behavior. Collectively, the results provide an insight on dominant variables which could be used for reliable/robust combustor liner design. INTRODUCTION Aerospace propulsion systems are a complex assemblage of structural components that are subjected to a variety of thermal and mechanical loading conditions with uncertainties. Furthermore, inherent variability in material properties and fabrication processes, as well as geometrical imperfections introduce additional uncertainties. Deterministic structural analysis methods fail to address these issues. Consequently, these methods could lead to overly conservative designs or sometimes non-conservative designs. Probabilistic Structural Analysis Methods (PSAM) were developed at NASA Lewis Research Center (ref 1) which enable a formal assessment of uncertainties in loads, material property variations, and geometric imperfections on the structural behavior (stability, frequencies, stresses, etc.). PSAM also provides a systematic way to quantify sensitivities associated with the corresponding uncertainties in the design variables to the system model. The PSAM efforts have lead to the development of a computer code NESSUS (Numerical Evaluation of Stochastic Structures Under Stress) (refs. 2 and 3). In the recent past, NESSUS has been used (ref. 4) to computationally simulate and probabilistically evaluate a typical hot structural component within an engine such as composite combustor liner. The probabilistic analysis results showed that the scatter in the combined stress is not uniform along the length of the combustor. Furthermore, the coefficient of thermal expansion, the hoop modulus of the liner material, and the thermal load profile dominate stresses near the support and the mid-span of the combustor liner. This work was further extended (ref. 5) to evaluate the effects of uncertainties of through thickness thermal load gradients for a typical ceramic matrix combustor liner. The combustor was separately evaluated for local stresses using low, high, and mixed low/high moduli ceramic matrix composite (CMC) combustor. The subsequent probabilistic analysis results indicated that the through thickness gradients, the liner thickness, the coefficient of thermal expansion, and the axial and shear moduli had considerable impact on combined stress with the through thickness thermal gradients having most significant impact. Currently, new concepts are being explored for combustor liners of High Speed Civil Transport (HSCT) system which are subjected to complex environments and have to endure these for thousands of hours with assured reliability. In the past, several deterministic analysis which included detailed heat transfer analysis to identify thermal profiles and deterministic stress analysis to identify critical locations of high stresses have been performed and also actual rig tests for segments have been performed by simulating these loading situations as closely as possible. However, it is a well-known fact that many uncertainties do exist in loading (primary thermal loads due to heat transfer), boundary conditions (end fixity unknowns), and material properties (moduli, thermal expansion coefficients and conductivities). The main objective of this report is to explain the computational simulations and probabilistic evaluation methodologies used for a Single Cup Liner Segment (typical combustor liner configuration) so as to account for the above mentioned uncertainties in a formal way in order to assess the performance of the liner component under complex and uncertain loading conditions as well as other geometry and material related uncertainties. FUNDAMENTAL APPROACH AND CONSIDERATIONS One of the major problems encountered in the development of new concepts for hot section components in an engine, such as a combustor liner is that it has to be economically viable inclusive of life-cycle costs and sufficiently assured reliable life as well as environmental acceptability. Furthermore, the liner material should possess 1 ut-
material characteristics to provide long term durability as well as resistance to possible through thickness and axial (along the length) high thermal gradients. In addition, the recommended ceramic matrix composite materials have properly connected economical fabrication processes to minimize overall manufacturing costs. The presently available methods/programs do not allow us easily to statistically quantify the uncertainties in the above described design parameters. Therefore, using the NESSUS code, dominant parameters can be identified to achieve acceptable vibration frequency values, buckling loads, and combined stress values. FINITE ELEMENT MODEL The finite element model of Single Cup Liner Segment consists of a cylindrical liner segment with 660 nodes 640 bilinear isoparametric variable-thickness shell elements based on Reissner-Mindlin plate and shell theories (see Fig. I). The liner is further subdivided into 33 ring sections each with 22 nodes around the circumference. The shell element is a fournoded quadrilateral in three-dimensional space. The liner is made of a typical ceramic matrix composite material (SiC/SiC). As shown in Fig. I, the outer surface of the liner is exposed to a uniform pressure loading. Heat transfer analyses are available in the form of thermal profiles (see Fig. 2) and are used as a starting point for thermal loads. Furthermore, equivalent homogeneous, orthotropic properties of the liner material are obtained using in-house composite mechanics code CEMCAN (ref. 6). In addition, the temperature, moduli, thermal expansion coefficients, and liner thickness are assumed to have known distributions and standard deviations. In the initial finite element analysis, it was assumed that all the nodes at the support locations were allowed to freely expand both in radial and circumferential directions, but were not allowed to move in the axial direction. However, all other nodes along its length were assumed to rotate freely and displace in all directions. (2) Coefficient of thermal expansion (3) Thickness of the liner (4) Pressure load (5) Through thickness thermal loading along the length of the combustor liner (6) Material properties In the case of the material properties, the stiffnesses were considered random for the analyses. In the present probabilistic analysis these material property variables were assumed to be independent of each other and were assumed to be normally distributed. Heat transfer analyses were available in the form of thermal profiles and were used as a starting point for thermal loads. Furthermore, the through thickness thermal load(s) were used according to the thermal load profiles, both inside and outside of the liner, (see Fig. 2). When the influence of distribution-type is considered important, sensitivity studies can be performed readily to assess their respective effects. Initially, the NESSUS/FEM module was used to analyze deterministically the combustor liner for mean values of the primitive variables. The longitudinal and hoop stress values along the length of the cylindrical liner segment were respectively, shown in Figs. 3 and 4. In the subsequent probabilistic analyses, each primitive variable is perturbed independently and by a different amount. Usually, the perturbed value of the primitive variable is obtained by adding a certain factor of the standard deviation on either side of the mean values. Other details regarding probabilistic finite element analyses and sensitivity calculations are described in ref. 4. In general, the finite element equation for motion is written as: [mi{a }+[C]{0}+[K flu ={F(t)} ) where [ M ], [ C ], and [ K ] denote the mass, damping, and stiffness matrices respectively. These matrices are calculated probabilistically in the NESSUS code. Furthermore, (0), {u}, and { u are the acceleration, velocity and displacement vectors at each node, respectively. The forcing function vector (F ( t )), is time independent at each node. In this paper, the thermo-mechanical static case is considered by setting the mass and damping matrices to zero and considering the forcing function being independent of time in equation (1) such that [Kiln) = {F} (2) Figure 1 Finite Element Model of Cylindrical Liner Segment PROBABILISTIC MODEL The following primitive variables were considered in the probabilistic analysis: (1) Density of the liner material Using the NESSUS code, a linear buckling analysis was performed by making use of the subspace iteration technique to evaluate the probabilistic buckling load. The corresponding matrix equation for the buckling (eigenvalue) analysis for a linear elastic structure is as follows: {[K] - 1. [K 1 ]) {.} = (0) ( 3 ) 2
In the above equation, [ K ] is the standard stiffness matrix, [ Kg ] is the geometric stiffness, X is the eigenvalue, and are the eigenvectors. Furthermore, the vibration frequency analysis is also carried out using the following equation: {[K] = {0} ( 4 ) Finally, the NESSUS/FPI (Fast Probability Integration) module extracts the response variables (buckling loads, vibration frequencies, and stresses) to calculate respective probabilistic distributions and respective sensitivities associated with the corresponding uncertainties in the primitive variables. The mean, distribution type, and percentage variation for each primitive variable is given in Table I. Mode Mean of Eigenvalue No. (Critical Load/Applied Load) 1 1.46 2 106.88 3 106.88 4 116.56 5 118.10 Table II- Probabilistic Buckling Analysis [Single Cup Liner Segment (pressure load lpsi)] Primitive variables Density Coefficient bf Thermal i:e*passitoli1 Thickness Pressui'e load Thermal load (inside) Distribution Mean value Scatter, type ± percent Normal 0.1002 lbs/in.- 5.0 sec2 72:43X1CeiiijiiiIPPC 0.8 in. 0 psi 1455-2400 F Mode No. Mean Value of Vibration Frequency (cps) 1 0.97 2 11.08 3 18.97 4 36.43 5 61.17 Tier-Fern' la (outside) Axial modulus 119 13 modulus Poisson's contribution Shear modulus Shear modulus Shear modulus tli7:14362886 F,. 35.81 msi 35.8 171.5i 5.37 msi 12.8 10.2 is 10.2 msi Table 1 - Primitive Vanables and Uncertainties for Probabilistic Structural Analysis of Combustor Liner (Random input data) Table III- Probabilistic Vibration Frequency Analysis [Single Cup Liner Segment (pressure load lpsi)] DISCUSSION OF RESULTS The combustor liner has to satisfy all the structural reliability and safety requirements against the thermomechanical loadings. Initially, the liner was analyzed probabilistically to obtain the cumulative distribution functions (CDF) of the probabilistic buckling loads for the first five modes (see Fig. 5). The eigenvalue (critical load/applied load) for each of the five modes is given in Table H. The sensitivity factors from Fig. 6 show that the uncertainties in the liner thickness has the highest impact on the probabilistic distribution of the buckling load for the first mode followed by the pressure load and hoop modulus of the liner material. Figures 7 through It respectively exhibit the buckled mode shapes of the liner from mode 1 through 5. Figure 7 indicates rigid body motion whereas Figures 8 and 9 depict the edge buckling phenomena which corresponds to a pair of equal eigenvalues (Table ID. On the other hand, the modes 4 and 5 as shown in Figures 10 and 11 exhibit a combination of circumferential/radial deformations that are typical for cylindrical shell structures. It is important to note that, these mode shapes are due to types of boundary 3 ut-
conditions used at the support location(s) as well as the combination of pressure and thermal loads. Subsequently, the liner was probabilistically analyzed to obtain the CDF's of the probabilistic vibration frequencies for the first five modes and are shown in Fig. 12. The mean values of the first five natural frequencies are shown in Table III. According to Figure 13, the first mode vibration frequency was very sensitive to the scatter of the density and the thickness of the liner material as well as hoop and axial moduli of the material. However at higher probability, the variation in the density of the liner material has the highest impact on the probabilistic vibration frequency for the first mode. The respective first five vibration frequency mode shapes are shown in Figures 14 through 18. The first mode appears to be rigid body type and the modes 2 through 5 involve circumferential/radial deformations. In addition to satisfying the design requirements with allowable vibration frequencies as well as buckling loads, the stress level in the combustor liner should be below some critical stress level. Therefore, the probabilistic longitudinal stress at a critical location on the inside of the liner was determined for thermomechanical loads (see Fig 19). Furthermore, this stress distribution showed a wide scatter between the lowest and highest longitudinal stress. According to Fig. 20, the probabilistic longitudinal stresses were very sensitive to the scatter in the thermal load profile, the variations in the liner material thermal expansion coefficient and axial modulus. However, the CDF's of the probabilistic longitudinal stresses for the outer surface of the liner (see Fig. 21) clearly indicated that the stresses were compressive at the lower probability, whereas, they were tensile at the remaining probability. Lastly, the probabilistic hoop stress at a critical location indicated the reversal of the stresses (from compressive to tensile) from lower to higher probability (see Fig. 22). Once again, the variation in the thermal load profile has the highest impact on the CDFs of the probabilistic hoop stress (see Fig. 23). The CDF's of the probabilistic hoop stress at the critical location of the outside of the liner showed a wide variation and the stress reversible from the lower probability to higher probability (see Fig. 24). material: (2) at higher probability, the scatter in the density of the liner material has the most significant impact on the probabilistic vibration frequencies for the first mode: (3) the non-uniform thermal load profile and the liner material thermal expansion coefficient variations had significant impact on the probabilistic longitudinal and hoop stresses at critical locations along the length of the liner. Collectively, the results provide quantifiable guidance on dominant parameters for reliable/robust combustor liner design and optimum support condition. REFERENCES I. Chamis, C. C.: "Probabilistic Structural Analysis Methods for Space Propulsion System Components," Space System Technology Conference, AIAA, New York, 1986, pp. 133-144. 2. Dias, J. B., Nagtegaal, J. C., and Nakazawa, S.: "Iterative Computational Mechanics of Probabilistic Finite Analysis," Computational Mechanics of Probabilistic and Reliability Analysis, edited by W. K. Liu and Belytschlco, ELME PRESS International, Lautanne, Switzerland, 1989, pp. 211-230. 3. Wu, Y. T.: "Demonstration of New, Fast Probability Integration Method for Reliability Analysis," Advances in Aerospace Structural Analysis; Proceedings of the Winter Annual Meeting, edited by 0. M. Burnside, ASME, New York, 1985, pp. 63-73. 4. Pai, S. S. and Chamis, C. C., "Probabilistic Assessment of Combustor Liner Design, "ASME TURBO EXPO"95, Land, Sea & Air, The 40 th Gas Turbine and Aeroengine Congress/Users Symposium and Exposition, Houston, Texas, 1995. 5. Pai, S. S. and Chamis, C. C., "Quantification of Thermal- Structural Uncertainties in Engine Combustor Liners", ASME TURBO EXPO'97 Symposium &Exposition, Orlando, Florida, 1997. 6. Mital, S. K. and Murthy, P.L.N., "CEMCAN Ceramic Matrix Composites Analyzer User's Guide Version 2.0", NASA TM 107187, April, 1996. CONCLUSIONS The computational simulation and probabilistic evaluation of a typical hot structural component within an engine such as a ceramic composite combustor namely, Single Cup Cylinder, is demonstrated using the NESSUS code. The combustor liner is analyzed for compressive pressure loading and non-uniform thermal loading along its length as well as through the thickness of the liner. The cumulative distribution functions (CDF's) for buckling (eigen-values), vibration frequencies, and longitudinal and hoop stresses are evaluated. The results indicate that: (1) the variations in the liner thickness has the highest impact on the probabilistic distribution of the buckling load for the first mode followed by the pressure load and the hoop modulus of the liner 4
0.00 1.80 3.60 5.40 7.20 900 Length along cylinder (in.) Figure 2 mental Profile Along the Length of Cylindrical Liner Segment 120 240 360 480 600 Egenvalue (critical load/applied load) Figure 5 Cumulative Distribution Functions of Cylindrical Liner Segment Buckling Loads 3-0.50-9 -15 0.00 1.80 3.60 5.40 7.20 9.00 Length along cylinder (lit) Figure 3 Longitudinal Stress Along the Length of Cylindrical Liner Segment - 0.001 0.999 Probability Lent Figure 6 Probabilistic Mode I Buckling Load Sensitivities of Cylindrical Liner Seginent 10 6 si -2 nt. (fl -6 I. -10 I I I I 000 7.20 9.00 1.80 3.60 5.40 Length along cylinder fin.) Figure 4 Hoop Stress Along the Length of Cylindrical Liner Segment Figure 7 First Buckling Mode Indicating Rigid Body Motion 5
Figure 8 Mode Two Indicating Edge Buckling Figure 11 Mode Five Indicating Circumferential/Radial Deformations Figure 9 Mode Three Indicating Edge Buckling 12 24 36 48 Vibration Frequency (cps) Figure 12 Cumulative Distribution Functions of Cylindrical Liner Segment Natural Frequencies 60 - --v!--46zat -...: ---....."" 81"r.1.;, o.... a '.4...taa 4:. a,,0:a.s?,... r,:r; ' /.-;.d.... -...... Figure 10 Mode Four Indicating Circumferential/Radial Deformations, - 0.50 f 0.00-0.50 - I Modulus Thickness 1.4 MI Modulus 04008) 0.102 0.999 PTobabihty Levd Figure 13 Probabilistic Mode 1 Vibration Frequency Sensitivities of Cylindrical Liner Segment mom u
Figure Figure 14 First Vibration Frequency Indicating Rigid Body Motion 17 Fourth Vibration Frequency Indicating CircumferentiaVRadial Deformations Figure 15 Second Vibration Frequency Indicating Circumferential/Radial Deformations Figure 18 Fifth Vibration Frequency Indicating Circumferential/Radial Deformations b. 0.75 1 0.50 e 'O.25 Figure 16 Third Vibration Frequency Indicating Circumferential/Radial Deformations 000-60.00-42_50-25.00-7.50 Stress Ks Figure 19 Cumulative Distribution Function of Node 441 Longitudinal Stress 10.00 7
I.00 0.5C 1 Thermal MutU 1111111 I CTE MENI Picdulu blackins orsint ligabral ii N (14c.") 1.00 0.50 too ciirmt Lkis (Hoop) imi= ad u -0-50 -0.54-1.00 Figure 20 Probabilistic Longitudinal Stress Sensitivities for Node 441 at 0.001 Probability Figure 23 Probabilistic Hoop Stress Sensitivities for Node 601 at 0.001 Probability 0.80 an 1 ;3 0.25 0.00 20. 4.80 2.40 13.60 24.80 36.00 Stress (kul Figure 21 Cumulative Distribution Function of Node 421 (outer surface of the liner) Longitudinal Stress 0.00 20 10 25 Stress ocu) Figure 24 Cumulative Distribution Function of Node 461 (outer surface of the liner) Hoop Stress 40 100 0.00-60 -40-20 0 20 Stress (lai) Figure 22 Cumulative Distribution Function of Node 601 Hoop Stress 8 Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/81871/ on 12/03/2017 Terms of Use: http://www.asme.org/ab