Finite Elements in Analysis and Design 29 (1998) 173186

Three-dimensional nonlinear nite element analysis of dovetail joints in aeroengine discs

Abstract Three-dimensional nonlinear nite element analysis is made of the dovetail region in aeroengine compressor disc assemblies using contact elements. The study is devoted to examining the eect of the critical geometrical features, such as ank length, ank angle, llet radii and skew angle upon the resulting stress eld. Frictional conditions at the interface between the disc and the blade are also examined. The nite element predictions were validated using three-dimensional photoelastic stress freezing results. Comparisons with the two-dimensional nite element analysis made earlier by Papanikos and Meguid (Fatigue Fract. Eng. Mater. Struct. 17 (5) (1994) 539550) of the same geometry reveal certain inadequacies. Specically, the earlier analysis underestimates the maximum equivalent stress along the interface by as much as 40%. This could have serious implications concerning the safety margins of the disc assembly. 1998 Elsevier Science B.V. All rights reserved. Keywords: Three-dimensional; Finite elements; Aeroengine disc; Contact; Photoelasticity

1. Introduction Aeroengine compressor disc assemblies are subjected to high thermo-mechanical loads and contain highly stressed components. Whilst blade loss can be contained within the engine casing on failure, the catastrophic failure of a compressor disc, on the other hand, could cause the larger fragments of the disc to puncture the engine casing. The consequences of such a failure are particularly costly resulting in the destruction of the engine and ultimately in the loss of life. Aeroengine compressor discs (Fig. 1) have basically three critical regions for which lifetime certication is necessary: the dovetail-rim region, the assembly holes or weld areas and the hub

* Corresponding author.  On leave from Technische Universitat Munchen. 0168-874X/98/$19.00 1998 Elsevier Science B.V. All rights reserved PII S 0 1 6 8 - 8 7 4 X ( 9 8 ) 0 0 0 0 8 - 0

174

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 1. Typical aeroengine compressor disc assembly.

region. The loads associated with these regions are the centrifugal forces of the blades, the self-generated loads applied by spacers and assembly bolts, and thermal stresses. These thermomechanical stresses are not constant but vary during a ight causing fatigue of the disc. In the majority of cases, cracks are initiated in the dovetail region due to the fretting action at the blade/disc interface. The stress analysis of the dovetail-rim region of aeroengine discs has received the attention of several investigators. Of particular interest to this study is the work of Kenny et al. [2] and Nurse and Patterson [3]. Their work was concerned with stage two fatigue crack growth paths in r-tree xtures. Since no allowances were made for contact elements in their model, approximate contact pressures were assumed at the interface between the blade and the disc. The numerical analysis at dovetail joints were also treated by Boddington et al. [4]. In their work, a technique is developed to model the relative motion at the interface of the assembly, and thus accounts for contact at the interface. Parks and Sanford [5,6] conducted two- and three-dimensional photoelastic analyses of the blade/disc r-tree region of a turbine disc. Centrifugal, circumferential and antiplane bending loads were applied at the centroid of the blade for the three-dimensional model, while only centrifugal loads were applied to the two-dimensional model. Their results revealed that the stresses at the central region of the three-dimensional disc were approximately twice those found in the twodimensional study and that the stress concentration in the central region is balanced by a large reduction in the stress at both ends of the llet. This reduction was associated with the exibility of the dovetail joint at both ends. Durelli et al. [7] conducted a comprehensive study on turbine blade attachments. They concluded that the most important forms of loading are the radial centrifugal force due to blade loading and the bending of the blade due to the gas pressure. The use of a small-scale test specimen which reproduces the stress state and consequently the failure behavior of the dovetail blade root xing was introduced by Ruiz and his collaborators in [8,9]. Utilising both photoelastic and biaxial fatigue testing, Ruiz et al. [9] found that the hoop loads had a relatively small eect on the peak stresses and the stress distribution in the immediate vicinity of the dovetail. He and Ruiz [10] utilized the small-scale test specimen introduced in [8,9] to predict the location of the severe damage, and of crack initiation at the dovetail blade root xing.

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

175

They proposed the use of a fretting damage parameter (FDP), as a means of predicting the location of crack initiation. Their results indicate that the maximum value of FDP is reached at the tangential point between the straight ank of the dovetail and the root radius. In [11], Ruiz et al. used high sensitivity Moire interferometry to determine relative normal and tangential displace ments at the dovetail boundary. They found that in response to the change-over from a stick to a slip, the dovetail joint behaved in an asymmetrical manner under increasing load. The above review shows that existing research in aeroengine disc stress analysis has been primarily concerned with the two-dimensional treatment of the problem. Due to skew angle eects and thickness variations in a real disc, a triaxial stress state exists, which can be conveniently evaluated using three-dimensional stress analysis. In this study, attention is devoted to examining the eect of the critical geometric features and interface conditions in the dovetail region upon the stress distribution at the blade/disc interface. These features include: inner and outer llet radii R and R, ank length l, ank angle , skew angle and the coecient of friction at the blade/disc interface (see [1] for the denition of these features). Three aspects of the work were accordingly examined. The rst was concerned with the two-dimensional nite element analysis of the stress eld in an aeroengine disc assembly. The second was concerned with the three-dimensional nite element analysis of the assembly, thus enabling the examination of the eect of the skew angle upon the triaxial state of stress present in the disc. The third was devoted to the validation of the results using photoelastic stress freezing results. This article is divided into four sections. Following this brief introduction, Section 2 describes the details of the nite element modelling. Section 3 provides a detailed analysis of results and Section 4 concludes the paper.

2. Finite element modelling The main emphasis of this study is the three-dimensional nite element analysis of the dovetail region in aeroengine compressor disc assemblies. However, in order to: (i) provide a reference for comparison with the three dimensional analysis, (ii) provide additional information that the authors did not account for in their earlier work [1], and (iii) ensure the completeness of the present article, additional results concerning the two-dimensional model were included. Accordingly, two- and three-dimensional nonlinear nite element analyses were conducted in order to assess the eect of the critical geometric features on the stress eld of the dovetail region of an aeroengine compressor disc assembly. The geometries of the disc selected in this study are shown in Fig. 2. Throughout this work, the authors utilised the ANSYS nite element code. The two-dimensional analysis was conducted for the mid-section of the disc, thus pertaining to plane stress conditions. In view of disc symmetry, only one sector of the disc was modelled. In order to satisfy the compatibility conditions, the blade and the disc should be discretised such that the nodes and elements at the common boundary match in number and in position. Because of the curved boundary of the model, eight-noded quadrilateral and six-noded triangular plane stress elements were selected. The two-dimensional model is incapable of predicting the stress variations across the thickness of a realistic disc. Furthermore, it is incapable of assessing the eect of the skew

176

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 2. Disc and dovetail geometry examined.

angle on the stress eld. Accordingly, the accurate prediction of the stress state in the disc requires three-dimensional modelling of the compressor disc assembly. For the case of a disc with a straight dovetail slot, the same sector of the disc as in the two-dimensional analysis was modelled. Due to lack of symmetry, the nite element analysis of a disc with a skew dovetail slot was conducted using a submodelling technique; detailed in section three. Both 20-noded hexahedral and ten-noded tetrahedral elements were used in the three-dimensional analysis. In order to estimate the optimum number of elements in each model, convergence tests were carried out. Both the accuracy of the solution and computing time were taken into consideration. Fig. 3 shows typical three-dimensional meshes used to test convergence.

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

177

Fig. 3. Convergence tests for the three dimensional geometry.

To model the blade/disc interface conditions, contact elements were used. The two- and three-dimensional contact elements used in this study adopt a node-to-segment interface model. The amount of the open gap or the gap penetration of the contact node on the target plane is calculated along with the point of projection of the contact node. Contact is indicated when the contact node penetrates the target surface dened by the target nodes. The penetration represented by the magnitude of the gap is a violation of compatibility. In order to satisfy contact compatibility, forces are developed in a direction normal to the target that will tend to reduce the penetration to an acceptable numerical level. In addition to normal forces, friction forces are developed in a direction tangent to the target plane. The Coulomb friction representations require the specication of the coecient of sliding friction . To satisfy contact compatibility two methods can be employed: the rst uses a penalty parameter, while the second uses a combined penalty-Lagrange multiplier method. The penalty method enforces compatibility approximately by means of a contact stiness through the imposition of a penalty parameter. The combined approach which is used in this study satises compatibility to a user-dened precision by the generation of additional contact forces that are referred to as Lagrange forces. A major problem in the implementation of contact elements is the assignment of values to the normal (K) and tangential (K) stinesses, which govern the convergence and accuracy of the

178

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 4. Variation of contact stiness with relative nodal displacement: (a) normal stiness K , and (b) tangential  stiness K . 

solution. Fig. 4a and b show the variation of K and K with displacement. The values of K and K are required to be very large. However, the use of excessively high values of K and K results in ill-conditioned global stiness matrices, leading to numerical errors and divergence. On the other hand, the use of smaller values of K and K results in convergence to the wrong solution allowing for interpenetration and incorrect estimates of the stick and slip regions [12]. In the current study, the appropriate values of the normal and tangential stinesses were selected from convergence tests where no interpenetration was allowed. It was found, following numerous test runs, that the appropriate values of K and K for the considered geometries are 10 and 10, respectively. 3. Results and discussion In this study, the dovetail geometries outlined in section two were analysed using the nite element method. The material properties used for the modelling of the blade and the disc were that of titanium alloy Ti-6Al-4V. The mechanical properties of Ti6Al4V are: Youngs modulus E"114 GPa, Poissons ratio "0.33, and density "4429 kg/m. Another important parameter which plays a major role at the dovetail region is the coecient of friction between the blade and the disc. It has been shown in [13] that the coecient of friction for Ti6Al4V varies during the fretting fatigue life of the component from 0.1 to 0.4 at room temperature. The majority of the work examined in this study was conducted using "0.25. However, to investigate the eect of dierent interface friction at the blade/disc contact region upon the resulting stress eld, the coecient of friction was varied from 0.0 to 1.5 [9]. All the models examined were subjected to centrifugal loading only with an angular velocity of 1000 rpm. Due to the presence of nonconservative frictional forces, the loading was applied incrementally and the load step in each increment was automatically calculated within the nite

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

179

element code. The contacting surfaces were modelled using contact elements, as described in section two. No attempt has been made to accurately model the blade except insofar as providing the necessary centrifugal loading and the associated eect at the interface. In addition, to simplify the analysis, the eects of thickness and temperature variations of the disc upon the stress distribution at the dovetail region were not considered. 3.1. Experimental validation with three-dimensional photoelasticity In order to validate the results from the three-dimensional nonlinear nite element analysis, three-dimensional stress-freezing photoelastic spin tests were conducted on an existing disc. The photoelastic disc details were as follows: ank length"6 mm, ank angle"45, thickness" 30 mm. Both straight and skew dovetail slots were examined and compared with the nite element predictions. The photoelastic model was made of Araldite CT200 epoxy and it was built to full size. Prior to testing, it was necessary to determine the appropriate spin speed so that adequate number of fringes are introduced in the disc to enable the accurate measurement of the resulting fringe pattern. This was found to be 1000 rpm. Block-type blades were used to exert radial pull on the disc periphery and to simulate the blade centrifugal force. The three-dimensional photoelastic model was tested to the softening temperature of the Araldite material (135C), soaked at this temperature for 2 h, and then cooled to room temperature at a uniform rate of 2.5C per hour. Fifteen slices, each 2 mm thick, were taken at right angles to the axis of the dovetail slots by means of a high-speed diamond cutting wheel. A polariscope with circularly polarised light beam and the normal incident method were used for the measurement of the isochromatic fringes. Fig. 5a shows typical isochromatic fringe patterns for the dovetail region of three slices. The maximum shear stress distribution at the lower contact line for the case of a straight dovetail slot is shown in Fig. 5b. In view of disc symmetry, the results for only one half of the disc are plotted. Comparison between the nite element predictions and the photoelastic results reveals a maximum discrepancy of about 10%. 3.2. Two-dimensional nite element analysis In order to evaluate the eect of the ank length of the blade and ank outer and inner radii of the disc upon the resulting stress eld, several geometries were modelled. The eect of friction at the blade/disc interface for sliding and sticking contact upon the resulting stress eld was investigated by varying the values of the coecient of friction. Values of 0.0, 0.25, 0.5, 1.0 and 1.5 were chosen. Sticking contact between the blade and the disc occurred at the lower contact point for "1.0 and 1.5, while the two bodies were found to be entirely in sliding contact for the smaller values. The results obtained from the two-dimensional analysis reveal the following: 1. an increase in the outer radius of the disc increases the stresses at the upper contact point, whereas the maximum value changes only by a small amount, 2. a change in the inner radius of the disc does not aect the stress distribution, 3. a change in the ank length of the blade relocates the point of maximum stress concentration on the disc boundary, but leaves the values of the peak stresses virtually unchanged,

180

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 5. Photoelastic analysis of a compressor disc: (a) typical fringe patterns, and (b) distribution of maximum shear stress along thickness direction at the lower contact line.

4. in the case of sliding contact, an increase in the coecient of friction decreases the disc boundary peak stresses, and 5. in the case of sticking contact, the peak stresses increase due to separation at the upper part of the interface. 3.3. Three-dimensional nite element analysis 3.3.1. Straight dovetail slot To create the three-dimensional model for a straight dovetail slot, use was made of the two-dimensional model. The three-dimensional geometry was then created by extruding the two-dimensional sector in the direction normal to its plane. In view of symmetry of geometry and loading, only one half of the disc (t"10 mm) was modelled. The deformation of the aeroengine disc assembly due to the rotational loading of the blade and the disc is depicted in Fig. 6. The von Mises stress distribution along the blade/disc interface at two dierent thickness locations is shown in Fig. 7, together with the stress distribution obtained from the two-dimensional analysis. The stress distribution for the three-dimensional model reveals that the stress level in the middle of the disc is much higher than that at the disc surface. As can be seen more clearly from Fig. 8, the magnitude of the peak stress at the lower contact line increases by as much as 40% from the disc surface to the disc central plane. This trend of the stress eld is in agreement with the photoelastic results described in Section 3.1. The two-dimensional analysis underestimates the

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

181

Fig. 6. Deformed geometry of an aeroengine compressor disc assembly (undeformed geometry indicated by dashed line).

stress level along the contact region. The maximum value of the von Mises stress obtained from the two-dimensional investigation at the lower contact point is 7% smaller in magnitude than the maximum value predicted by the three-dimensional analysis (Fig. 8). The von Mises stress distribution at the lower contact line for dierent ank angles is shown in Fig. 9. The results for only one half of the disc thickness are plotted. The three-dimensional results show that the stresses at the lower contact region increase for smaller values of the ank angle. This observation is in agreement with the results obtained from the two dimensional analysis. The three-dimensional model with the straight dovetail slot was also analysed for dierent values of the coecient of friction. The three-dimensional analysis reveals that the blade and the disc experience sliding contact for all values of the coecient of friction investigated (01.5). As can be seen in Fig. 10, an increase in the coecient of friction decreases the disc boundary peak stresses. This eect of the coecient of friction on the stress eld at the blade/disc interface was identied earlier by the two-dimensional model.

182

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 7. Von Mises stress along the interface at dierent thickness locations.

Fig. 8. Von Mises stress across thickness at the lower contact line.

3.3.2. Skew dovetail slot In the case when a skew angle of 20 was assumed, use was made of submodelling. In this case, the entire disc assembly was discretised using the coarse mesh shown in Fig. 11a. In the initial runs, the blades and the disc were joined at the interface and no allowances were made for contact eects. The submodel shown in Fig. 11b was then developed with the appropriate boundary conditions.

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

183

Fig. 9. Eect of ank angle on the von Mises stress across thickness at the lower contact line.

Fig. 10. Eect of the coecient of friction on the von Mises stress across thickness at the lower contact line.

Due to anti-symmetry, only the stresses at one side of the blade/disc contact region were examined. The distribution of von Mises stresses along the blade/disc interface for the front and back surfaces is shown in Fig. 12. The gure indicates a large dierence between the corresponding surfaces in the stress values at the lower contact region. Fig. 13 demonstrates the increase of the von Mises stress at the lower contact line from the front surface to the back surface of the disc, where the peak value occurs. It is clear from Fig. 13 that the two-dimensional analysis underestimates the maximum equivalent stress along the interface by as much as 40%. This could have serious implications concerning the safety margins of the disc assembly.

184

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

Fig. 11. Discretised geometry of (a) the full model, and (b) the submodel.

Fig. 12. Variation of von Mises stress along the interface of a skew dovetail slot.

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

185

Fig. 13. Variation of von Mises stress across thickness at the lower contact line of a skew dovetail slot.

4. Conclusions Three-dimensional nonlinear nite element analysis is made of the dovetail region in aeroengine compressor disc assemblies to evaluate the eect of the critical geometric features and frictional interface conditions in the dovetail region. Whilst the two-dimensional analysis is a useful starting design tool, it cannot be used in the prediction of the stress state in a complex aeroengine disc assembly. The results of this work demonstrate the inadequacy of the two-dimensional analysis to model the disc. The study also reveals the following: 1. the maximum stress concentration occurs at and just below the lower contact point between the blade and the disc, 2. the ank angle, ank length and coecient of friction can signicantly change the blade/disc interface stress distribution, 3. the three-dimensional results reveal large stress variations through the disc thickness, which cannot be predicted by a two-dimensional analysis. The two-dimensional results underestimate the peak stresses, and 4. the skew angle has a signicant inuence on the stress distribution and the magnitude of the peak stress at the blade/disc interface.

References
[1] P. Papanikos, S.A. Meguid, Theoretical and experimental studies of fretting-initiated fatigue failure of aeroengine compressor discs, Fatigue Fract. Eng. Mater. Struct. 17 (5) (1994) 539550. [2] B. Kenny, E. Patterson, M. Said, K. Aradhya, Contact stress distributions in a turbine disc dovetail type joint A comparison of photoelastic and nite element results, Strain 27 (1991) 2124.

186

P. Papanikos et al. / Finite Elements in Analysis and Design 29 (1998) 173186

[3] A.D. Nurse, E. Patterson, Experimental determination of stress intensity factors for cracks in turbine discs, Fatigue Fract. Eng. Mater. Struct. 16 (1993) 315325. [4] P.H.B. Boddington, K. Chen, C. Ruiz, The numerical analysis of dovetail joints, Comput. Struct. 20 (1985) 731735. [5] V.J. Parks, R.J. Sanford, Experimental stress analysis of the TF-30 turbine engine third-stage fan-blade/disc dovetail region, NRL Report, NRL 8149, August 1977. [6] V.J. Parks, R.J. Sanford, Three-dimensional photoelastic stress analysis of the dovetail region of the tf-30 turbine engine third-stage fan, NRL Report, NRL 8276, December 1978. [7] A.J. Durelli, J.W. Dally, W.F. Riley, Stress and strength studies on turbine blade attachment, SESA Proc. XVI (1) (1957) 171186. [8] P.H.B. Boddington, C. Ruiz, A biaxial fatigue test for dovetail joints, Proc. ASME, Int. Conf. on Advances in Life Prediction Methods, Albany, April 1983, pp. 277283. [9] C. Ruiz, P.H.B. Boddington, K.C. Chen, An investigation of fatigue and fretting in a dovetail joint, Exp. Mech. 24 (3) (1984) 208217. [10] M.J. He, C. Ruiz, Fatigue life of dovetail joints: verication of a simple biaxial model, Exp. Mech. 29 (2) (1989) 126131. [11] C. Ruiz, D. Post, R. Czarnek, Moire interferometric study of dovetail joints, ASME J. Appl. Mech. 52 (1985) 109114. [12] R.D. Cook, Finite Element Modeling for Stress Analysis, Wiley, New York, 1995. [13] M.M. Hamdy, R.B. Waterhouse, The fretting wear of Ti-6Al-4V and aged Inconel 718 at elevated temperatures, Wear 71 (1981) 237248.