Simulation einer Mehrlagenschweißung im Großmotorenbau 2.6.2
Transcrição
Simulation einer Mehrlagenschweißung im Großmotorenbau 2.6.2
2.6.2 Simulation einer Mehrlagenschweißung im Großmotorenbau Samuel Sönnichsen, Andreas Junk (CADFEM) Wärtsilä Switzerland Ltd. Winterthur, Switzerland Summary The crankshaft housing of large two stroke diesel engines (with power up to 80’000 kW) is a welded structure subjected to a cyclic, constant amplitude loading and designed for infinite life on full load. The double wall box design, which is standard design since more than 30 years, requires single bevel butt welds with access only from one side. Geometrical stress concentration on the root side in combination with additional stress raisers due to weld imperfections, such as lack of penetration, lack of fusion or cracks, result in high requirements to the fabrication process, since corrective measures from the inside are not possible. In spite of well known recommendations of welding standards, where favourable effects of post weld heat treated components are assumed, the main components of the structure of Wärtsilä engines are not stress relieved. Residual stresses play an important role in the fatigue behaviour. However, long-term experiences indicate that their effect is rather complex and can be positive as well as negative. The present investigation is aimed at a better understanding of the mechanical behaviour of these joints, particularly regarding residual stress and its effect on the fatigue performance. Fig. 2: column Fig.1: Two stroke diesel engine Fig.3: bedplate The investigation includes the following steps: - Stress evaluation by strain gauge measurement and FEM - Evaluation of residual stress distribution by measurement and weld process simulation (SST) - Fatigue strength analyses Keywords Stress calculation (FEM), Fatigue analysis of welded structure, Residual stress, Weld process simulation (SST) th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 1 0. Introduction The crankshaft housings of large two-stroke diesel engines are welded structures with plate thickness up to 50 mm made of hot rolled carbon steel. The engines operate mainly on full design load (cyclic forces from combustion and inertia of moving 9 parts) which leads up to 10 cycles after 30 years in service. As fatigue failure in service requires very expensive repairs and off-hire, a high survival probability is mandatory. Unfortunately different crack series on the engine structure accentuate the need for extensive investigations to improve the situation for future engines. Some cracks occur on few dynamically loaded areas whereas other highly loaded locations are running without problems. For some reasons Wärtsilä never applied post weld heat treatment (PWHT). On one hand, PWHT avoids potential high tensile residual stress, but on the other beneficial compression stresses are reduced as well. So the over-all effect of PWHT is not clear a-priori, and it is essential to know the residual stress distribution and understand their effect on the fatigue behaviour of the present butt welds. The basic steps of the procedure are shown in Fig. 5. This investigation is aimed at clear welding instructions for the production (controlled process). Fig.4: crack in gear column Strategy: Repeatedly increasing the engine power and performance requires the designer to review the rules continuously Knowledge present Design quality s = safety future s k ac cr k ac cr s Production 1/ Q action 1/ Q k ac cr quality 1/ Q Reliable control of the welding process has a very strong influence on the real fatigue strength and the safety margin of components quality Single side butt weld (without back weld) is not well described in any fatigue design codes diameter: inaccuracies about allowable fatigue limit design (stress) Knowledge Design Production design (stress) design (stress) action Development of technical knowledge for specific welding application (single bevel butt weld) Design improvements in respect of weld ability, stiffness and prevention of stress concentration. result More precise information about fatigue limits and their influencing factors Detailed instruction for production welding (qualified process) Less fluctuation in quality Increased safety margin Fig.5: Schematic representation of strategy th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 2 Decreased stress level Increased safety margin 1. Problem definition 1.1 Design aspects The double wall box design shown in Fig. 6 leads to single bevel butt welds. box box crosshead Fig.6: Crank shaft housing (FEM-Model, right) and section through the double wall box (left) Fatigue endurance and reliability are preconditions for an engine structure. Omitting hot spots and local stiffeners are the most effective measures to avoid damages during life time. However, using thicker plates at potential critical locations is not always suitable to increase the safety margin. Due to the fact that the fabrication process itself has a large influence on fatigue, detailed instructions for the manufacturer help to minimise the variance of welding quality. Feasibility and production friendliness are further aspects that needs to be considered. The numerous years of service experiences of the existing double wall structure (without PWHT) can be regarded and used as long-term fatigue tests, provided they are accompanied by a detailed theoretic investigations, including stress analysis due to service loads and determination of residual stresses. 1.2 Problem of root welding of a single bevel butt weld Concerning the fatigue behaviour, the vicinity of the root of a single bevel butt weld is in general the most critical area. Typical weld defects in this region are shown in Figs. 7-9. Obviously they act like pre-existing cracks. Since the inside of the box is not accessible there is no possibility for corrections from the inside. The stress concentration due to the global geometry in combination with the local geometrical irregularities and uncertainties regarding residual stresses makes a fatigue analysis of such joints difficult. Possible root imperfections: Fig.7: Lack of fusion Fig.8: Hot crack Fig.9: Lack of penetration Due to the difficulties to detect root imperfections such as the ones shown above by Non Destructive Testing (NDT) the structure needs to exhibit a sufficient degree of so-called damage tolerance. This requires a certain representative crack-like defect to be taken into account in the fatigue analysis and assessment. According to Linear Elastic Fracture Mechanics (LEFM), crack propagation strongly depends on residual stress in the affected zone. Compressive residual stresses in the root area can actually prevent a crack from growing, whereas tensile residual stresses would accelerate it. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 3 1.3 Theoretical assumption of residual stress distribution In a multi-layer single bevel butt weld the residual stress distribution as well as the deformations depends mainly on the boundary conditions. If there is no restriction to relative rotation, then each additional weld layer adds a certain amount of distortion, so a considerable rotational angle χ can accumulate, causing tensile stresses T’ in the root area. If the relative movement of the plates is prevented, then the corresponding clamping force gives rise to a bending movement M, which causes compressive stresses in the root area. In the real case, a relative rotation of the welded plates is prevented by the spatial arrangement of stiffeners, so it corresponds to Var.2 rather than 1. M/a a T T M χ T’ C Variant 2: clamped to base plate Variant 1: No kinematical restriction during welding Fig.10: Schematic representation of the effect of the kinematical boundary condition /4/ Mean stress sensitivity and residual stress factor according FKM /3/ ƒÐ AK 2,19 71 R 50 = 1,48 ‡ Transverse butt welds welded from one side without backing bar, full penetration Root controlled by NDT FAT 0 Description = Structural Detail According to FKM /3/ the influence of residual stresses can be classified as: 1,0 Low: stress relieved structures Moderate: thin-walled simple structures High: thick-walled complex structures SULZER 85 (fatigue test) R 1.4 R 29 N/mm2 ,5 =0 25 Fig.11: Haigh diagramm (FAT 71) acc. to FKM This ranking of residual stresses 0 represents only their negative impact -75 -50 -25 0 25 50 75 ƒÐ on fatigue strength in case of tensile m residual stresses. However, if compressive residual stresses prevail in the root region, their effect may be positive, as shown below. 1000 fatigue test : crack fatigue test : no crack S-N-Curve for experimental datas (95% survival probability) 2 stress range (log N/mm ) S-N-Curve for experimental datas (50% survival propability) stress ratio: R = 0 / no PWHT 100 s (R=0) = ± 46 N/mm2 In 1985, SULZER have performed fatigue tests. The results of 2 σAK = ± 46N/mm (specimen not stress relieved) shows higher fatigue limit than excpected according FKM /3/. This specific application utilise the beneficial influence from residual stress. Fig.12: Fatigue test results 10 1.0E+05 1.0E+06 1.0E+07 number of load cycles (log N) th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 4 1.0E+08 2. Stress evaluation 2.1 Strain gauge measurement Strain gauge mesurements are used for verification of calculated engine loads and to assess the accuracy of FEM results. 10 ES 0 0° 30° 60° 90° 120° 150° 180° 210° 240° 270° 300° 330° 360° crank angle (0° = TDC Cyl.1) stress (N/mm2) -10 cyl.4 gear -20 Sx measured FPS Sx measured ES -30 Sx calculated Sx calculated -40 FPS -50 Fig.13: 10RTA96C Comparison of mesured and calculated stresses on gear column 100 50 DMS 89 DMS 91 DMS 93 DMS 95 DMS 97 DMS 101 DMS 103 force cyl.2 force cyl.3 4030 30° 0° 91 330° 93 0 0° strain (mst) 89 30° 60° 90° 120° 150° 180° 210° 240° 270° 300° 330° 60° 360° 95 crank angle (0° = DTC cyl.2) -50 300° cyl.3 97 90° -100 270° 101 cyl.2 120° -150 side plate -200 measuring position Location on top of column less stressed due to better support of inclined side plate. 240° 103 150° 210° 180° Inclined side plate -250 0 600 400 200 0 -200 Fig.14: Exposition of guide shoe force and their impact on stress in side plate On the engine structure all three normal stress components (transverse, longitudinal and shear, see Fig. 15) are present. To measure them a triaxial strain gauge (“rosette”) is required. The evaluation based on nominal stresses allows a specific assessment of all critical load components. Which one of them is crucial depends on the weld joint and the load direction. For a single bevel butt weld with an open root the transverse stress is the most critical load component. Fig.15: Stress components th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 5 -400 2.2 FEM – Calculation All types of loads acting on the component – stationarly and cyclic ones - must be included in the FEM – calculation. Often it is suitable to use a shell-element model for such thin-walled components. Nevertheless a detailed evaluation of the non-linear stress distribution over the plate thickness needs 20-node hexahedron elements. For stress evaluation two different concepts are common: A) Geometric stress: this concept includes all stress raising effects of a structural detail excluding all stress concentration due to the weld profile itself. B) Effective notch stress: this concept calculates the total stress at the root of a reference notch, for which a radius of 1mm and linear-elastic material behaviour is assumed. hot spot geometric stress Stress on surface Radius 1mm F F Fig.15: geometric stress Fig.16: effective notch stress Both these concepts are illustrated in the following by a stress simulation of an actual case of a welded structure where cracks occurred (Fig. 17). 1 side plates Section A - A A A stiffening ribs bending tension compression longitudinal Load 1 fuel pump support 2 2 Fig.17: Overview of crack situation on fuel pump support Stiffening ribs inside of the crankshaft housing prevent a rotation of the side plates (caused by the one-sided guide shoe force). This generates local stress concentrations on the root side of the vertical weld. Crack propagation as observed (see no.1 inside crankshaft housing / and no.2 outside crankshaft housing) is explained by the residual stresses (depending on the welding sequence) and the ones due to the service load (depending on the crank angle offset). Fig.18: View inside box th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 6 SY SX 50 hot spot stress dynamic stress (N/mm2) 40 geometric stress 30 20 coarse mesh fine mesh 10 nominal stress 0 0 50 side sloping plate The geometric stress is mainly influenced by the design of the edge relief in the siffening rib. This edge relief allows for local bending of the side plate. The bending of the side sloping plate, which also has an effect on the stresses in the root area, is not covered by the geometric stress concept. For this purpose, the effective noch concept is preferable. The use of hot spot stress for evaluation is only convenient for comparison of different countermeasures.The only restriction is to use the same FE-mesh size. 100 150 distance (mm) 200 R8 0 Fig.19: Notch stress (1 mm radius) 250 side plate 300 stiffening rib Fig.20: geometric stress (hot spot) A detailed stress analysis including the defect exhibits a stress concentration in the root area (Fig. 19). The safety against fatigue is just sufficient if the weld quality is according ISO 5817 (full penetration) Taking the compressive residual stresses in the root region (see Sect. 3 below) a certain root imperfection is accepable without any rsik of cracking. In fact, experience - many engines are running without problems, although quality deficiences in the root are likely to be present - confirm this assumption of a beneficial influence of the residual stresses on the type of weld considered here. Stress evaluation with FEM program ANSYS strain ANSYS ANSYS command: prnsol,epto ε x ε y ε z ε xy ε xz ε yz y Evaluation εa = ε x ε +ε +ε ε b = x y xy stress ANSYS command: prvect,pdir Principle stress S1 S2 S3 y vector S2 dx1 dx2 dx3 dy1 dy2 dy3 2 εc = ε y x dz1 dz2 dz3 dx1 α dy2 dx2 Def: S1 > S2 > S3 th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 7 εc εa α = a tan (dy1 dx1 ) S1 dy1 εb Selection of both principle stresses which are perpendicular to x each other S3 = 0 3. Residual stress distribution 3.1 Residual stress measurement Residual stresses, enclosed in the welding seam are not of primary interest. There will always be a balance between compression and tension. Much more important with regard to the manufacturing process is the influence to the structure itself. If there are no constraints imposed on the displacements in any directions, deformations due to the welding distortions occur rather than residual stresses. If deformations are restricted by geometrical constraints, then much higher residual stresses are produced. These two conflicting effects have to be optimized for a certain welded structure. An optimised welding procedure has to allow for free shrinkage to avoid tensile residual stresses on one hand, and to restrict those deformations that lead to compressive residual stresses (with positive or negative effect on fatigue) and shape stabillity on the other hand. low residual stresses high Fig.21: Schematic representation of interaction high deformation low between residual stresses and deformation due to degree of restraint degree of restraint free rigid An efficient method to determine residual stress distributions experimentally is the cut-compliance method /5/. Measurements performed on a testweld are desribed in /4/. The measurement setup is shown in Fig. 22, and the corresponding results in Fig. 23. Strain gage F CC-cut Fig.22: Experimental set-up to measure residual stresses at the weld Strain gage R 150 cut from root cut from upper surface Residual Stress [MPa] 100 50 0 0 5 10 15 20 -50 -100 -150 Root Distance from root [mm] Upper Surface Fig.23: Measured residual stresses at the weld root section To cover the root region as well as the upper surface the cutting required by the CC-method was performed from both sides. As shown in Fig. 23, both measurements are in good agreement with each other in the overlapping central region, indicating the reliability of the measurement. Apparently, both the root and the upper surface region are predominantly subjected to compressive stresses. Only in the central region tensile stresses prevail. Note that this measurement was made after releasing the clamps that were applied to prevent distortion (see Fig. 10 and Fig. 32). Thus, in the real structure an additional compressive stress is to be expected in the root region. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 8 3.2 Residual stress calculation 3.2.1 Simulation model The simulation of the residual stress distribution was performed by means of a 3-dimensional model of a representative portion of the weld (Fig. 24). The model included the cross sections of the material in pre-weld situation and the 5 layers. The cross sections were adjusted to the boundaries of the layers as given in micrograph results. The elements representing the filler material were activated with EALIVE according to the process specifications. Fig. 24: Simulation model The temperature simulation was done by a transient thermal analysis. The heat source was simulated with Gauss distribution across the elements of each layer. The position and direction of the heat source was adjusted according to the contour of the weld bead. The structural mechanics boundary conditions were set to represent planar cross sections at both cut sections of the model. These constraints are representing the center portion of a very long weld. The other displacement constraints were chosen according to the inservice condition of the welded part. They do not relate directly to the experimental setup described hereafter. The steel in the weld region and the heat affected zone undergoes changes in the microstructure. This behavior was simulated by two different methods, the STAAZ approach and the Leblond approach as explained below. 3.2.2 Fig. 25: Mesh of weld region Simulation using the STAAZ method The simulation was done using the weld simulation tool SST “Schweißsimulationstool” licensed by CADFEM. It is based on the ANSYS program and includes weld specific simulation algorithms and other features. The STAAZ approach to simulate the microstructural changes was proposed by Ossenbrink, Michailov /1/. This approach simulates the thermal strain due to metallurgical phase transformations. To cover the influence of the thermal history of the Fig. 26: STAAZ approach material a triad of significant parameters is used including the maximum temperature (ST), the time in the austenitic phase (A) and the cool-down gradient (AZ). This methods directly uses dilatogram data. A set of material test data is required including dilatogram tests for different triad combinations. During the analysis an interpolation is made at each location inside the FEM model according to the local triad. This STAAZ approach avoids the uncertainties of CCT derivation and the succeeding steps to transfer the CCT data into numerical functions and parameters. The microstructure contribution to thermal strain is directly included, there is no calculation of microstructure fractions. The calculation of the temperature distribution oder time and the structural mechanics is done in sequence. The transient thermal analysis uses estimated phase change behavior. The temperature distribution and function over time were compared to experimental data. There was a good agreement. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 9 Fig. 27: Temperature as a function of time After the thermal analysis is finished the temperatures and the thermal result triads are transferred to the structural mechanics simulation. This simulation includes the thermal strain influenced by microstructure changes. The mechanical behavior simulates nonlinear stress-strain relationship, depending on temperature. Fig. 28 shows the thermal strain contribution during heatup and cool-down cycle of a position of layer 1. It can be nd seen that the α-γ change during heat-up and the reverse Fig. 28: Stress after 2 layer γ-α change during cool-down occurs at different temperature levels depending on the local triad values of maximum temperature (ST), the time in the austenitic phase (A) and the cool-down gradient (AZ). Fig. 29 shows the distribution of equivalent stresses after the 2nd layer is finished. The results were reasonably comparable to test data. The application of the STAAZ approach has uncertainties when applied to multilayer welds, since the remelting of material and its influence on the thermal strain results in unreliable material data. nd Fig. 29: Stress after 2 3.2.3 layer Simulation using Leblond kinetics For comparison purposes the simulation was also done using the Leblond microstructure kinetics approach /2/. This feature is also available in the SST Schweißsimulationstool (weld simulation tool). The Leblond approach approximates the phase change between the fractions austenite, ferrite, bainite, martensite, pearlite with analytical functions. Usually the parameters are derived from CCT diagram data. Considering the source of CCT diagrams being dilatograms there are several processing steps between the data source and the application. This results in considerable Fig. 29: Leblond kinetics approach uncertainties and approximations using this method. The analysis simulates the change of fractions over time. Thermal and mechanical material values are combined according to the local fraction relation. Using the Leblond microstructure kinetics approach in SST the transient thermal field and the structural behavior are simulated simultaneously in a coupled analysis. The model is meshed with coincident elements for the thermal distribution (solid70) and the structural mechanics (solid185). The coupling is done internally in a weak procedure (load vector coupling). Fig. 30: Microstructure fractions th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 10 The initial microstructure conditions are 90% ferrite, 10% pearlite. The resulting distribution of the phase fractions shows a martensitic microstructure with remaining local austenite spots. The resulting distributions of stresses SX (perpendicular to the weld) and SZ (longitudinal) are shown in Fig. 31. These stresses are used to consider the influence of residual stresses on service stress limits. Fig. 31: Leblond kinetics approach 3.3 Test welding Laboratory test were carried out to gain temperature and stress data for verification of analytical weld process simulation. The geometry indicates similar geometry as the real structure. Position of stress measurement (at the top and at the bottom) 35° 5 4 3 2 1 20 60 3 100 100 0 2 00 Fig.32: Test configuration 5th layer 300 250 A 60 40 40 B 4th layer 20 D C 150 3rd layer Process: GMAW (CO2), PA-position 2nd layer position A position B position C position D 1st layer temperatur (°C) 200 Welding parameters: 1st layer 22V 160A 20cm/min E=1,1KJ/mm 2nd layer 26V 220A 25cm/min E=1,4KJ/mm 3rd layer 30V 350A 25cm/min E=2,5KJ/mm 4th layer 31V 380A 20cm/min E=3,5KJ/mm 5th layer 31V 400A 20cm/min E=3,7KJ/mm 100 50 pre - heating 14 min 18 min 19 min 20 min 0 0 1000 2000 3000 4000 5000 time (min) Fig.33: Measured temperatures th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 11 6000 7000 8000 9000 10000 after welding (room temperature) Restraint 190 (-170) top side (bottom side) N/mm2 loosen 20 (0) N/mm2 -10 (-130) N/mm2 -70 (-70) N/mm2 210 (-150) N/mm2 40 (30) N/mm2 10 (-140) N/mm2 -60 (-80) N/mm2 Fig.34: Measured stresses Transverse stresses: The results of the stress measurement shows, as expected, heavy bending of the plate due to the 2 irregularity of the welding volume. The transverse stresses of about ±180 N/mm have nearly completely relaxed after the restraints are removed. Longitudinal stresses: 2 The stresses in longitudinal direction of about - 70 N/mm after loosening the clamps are plausible. These compressive stresses in the plate indicate high tension in and near the weld. During restraint condition 2 there is an additional bending in longitudinal direction of about ± 60 N/mm . This may be an effect of the plate dimension and the clamping arrangement and is not likely to appear on the real component. 3.4 Effect of preheating on residual stress Preheating prior to welding is a simple measure to reduce the shrinkage stresses. The main reason to utilize preheat is to decrease the cooling rate in the weld and the base metal. Circular welds Longitudinal welds Less deformation after welding in longitudinal direction Elongation of inner part by reason of preheating Elongation of plate by reason of preheating restraint outer part (prevents free transverse shrinkage) thin plate not restraint (transverse shrinkage possible) thick plate inner part Preheating reduces longitudinal shrinkage stresses Preheating increases transverse shrinkage stresses transverse shrinkage Fig.35: Preheating of high joint restraint longitudinal shrinkage Fig. 35 shows schematically representation the effects of preheating. Whether they are positive or negative depends on the weld joint application. The single bevel butt weld in question corresponds to the left application. This means that preheating is advantageous, since it reduces longitudinal shrinkage stresses that are produced by the present boundary conditions. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 12 4. Fatigue strength analyses 4.1 Stress evaluation The welding seams of the present structure are subjected to multi-axial stresses. Not only the magnitude but also the direction of the principle stress changes during one load cycle, i.e. one revolution of the crank shaft. A variety of different hypothesis for fatigue under such conditions have been published. Wärtsilä uses the von Mises criterion (GH) with signed principle stress now for many years. This approach results in high stress values and is often a very conservative interpretation of the load. The normal stress criterion (NH) enables less conservative predictions of the fatigue strength to be made. Cracks are formed preferably perpendicular to the maximum tensile stress. According to fracture mechanics the stress component normal to a possible crack-plane is the crucial one. The normal stress criterion takes into consideration possible root imperfections in correlation to the critical load component. Surface: Root: 20 A B 12 C 10 angle (0° = TDC Cyl.1) B C 4 60° 120° 180° 240° -10 -20 300° 360° stress sx (transverse) stress sy (longitudinal) 60° 120° -8 -20 G6 A cyl.6 - + B F5 cyl.5 F 5 240° 300° 360° C F6 cyl.6 + - 180° -4 -16 -40 F5 0° -12 stress sxy (shear) -30 F5 angle (0° = TDC Cyl.1) 0 2 ƒÐ N = }14 N/mm 0° dyn.stress (N/mm2) 0 2 ƒÐ N = }27 N/mm dyn.stress (N/mm2) A 8 cyl.6 - cyl.5 F6 cyl.5 G5 F6 A A Fig.36 : stress condition on root and surface side (geometric stress concept) The crucial stress component due to manufacturing deficiencies (imperfections perpendicular to the stress direction) is the transverse stress σx. According to Fig. 36, the maximum stress range occurs between position A and B. Normal stress criterion (NH): σN = σ x +σ y 2 2 A-A ⎛ σ x −σ y ⎞ + ⎜⎜ ⎝ 2 ⎟⎟ +τ xy2 ⎠ Shear stresses are very low at the relevant positions and can be neglected. If the normal stresses σx and σy always act proportional or synchronous in phase the values are to be inserted in the formula with the same (positive) signs /3/. Regarding stresses on welding surface the transverse and longitudinal stresses act synchronous in phase. This means that there is no additional influence from the longitudinal stress component. Therefore normal stress criterion will be generated only by the transverse stress: σN = σx 2 The decisive root area of the weld is loaded by σx = ± 14 N/mm . Fatigue resitance is usually derived from constant or variable amplitude tests. The fatigue assessment of classified structural details (FAT-classes) is based on the nominal stress range. This fatigue values are based on representative experimental tests and includes the effects of normal fabriction standard. 2 Comparison of σx with the allowable stress range for FAT 71, σAK = ± 29 N/mm , reveals that endurance is guaranteed with a considerable margin, provided the quality of the weld corresponds to FAT71 (full penetration without any imperfections). Nevertheless, long term experiences shows that normally a sytematic defect in the root must be taken into account. In this case the safety drops down dramatically. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 13 4.2 Fatigue resitance stress concentration factor kt As the geometrical stress concept does not apply to root defects, it is necessary to evaluate the fatigue resistance by use of more refined techniques such as effective notch stress concept or linear elastic fracture mechanics. Possible root defects of single bevel butt welds were classified with: a) stress concentration factor kt for lack of penetration failure (partial penetration) b) stress intensity factor KI for root crack or lack of fusion failure 3.0 F root-side (load tension) root-side (load bending) M surface-side (load tension) surface-side (load bending) 2.5 2.0 1.5 1.0 8% 18% 28% lack of penetration 38% Fig.37: stress concentration factor kt for bending and tension/compression Lack of penetration up to 4 mm (20% in case of a plate thickness of 20 mm) could easily occur if the fit-up is not done properly. Corresponingly, the stress in the critical area will be increased by a factor 1.6 -1.8 depend on load distribution (see Fig. 37). Therewith, the safety margin of the above example drops just to zero. More critical defects such as root cracks or lack of fusion must be assesed with linear elastic fracture mechanics. The resitance of a material against cyclic crack propagation is characterized by the material parameters of the “Paris” power law of crack propagation /3/: da da =0 = Co ∗ ΔK m ΔKth =190−144∗ R Δ K < Δ K th if then dN dN The treshold value ΔKth is influenced by the stress ratio R as idicated in the estimation formula above. Compressive residual stresses cause a lower R and, therewith, an increased ΔKth, which means that a higher range ΔK is required to let a crack-like defect grow. This means that the damage tolerance of the weld (allowance of possible or hypothetical crack-like imperfections, which could not be found by Non Destructive Testing) is improved by the residual stresses, which is a big advantage concerning the inspection concept. 4.3 Safety It is shown theoretically that the residual stress has an important influence on the fatigue performance of this welded joint. Depending on the defect size, location and orientation on one hand, and the distribution of the residual stresses on the other, the effect on the fatigue strength can be positive or negative. To decide about the most beneficial weld procedure and eventual PWHT the critical locations and the corresponding residual stress distribution should be known. orientation of imperfections orientation of significant load F Utilisation of residual stresses + To fully benefit from the positive effect of residual stress restriction and detailed information are necessary for production: - Optimised welding sequence - Procedure for root repair required - Flame straightening only in exceptional cases Only in this special case together with a correct application the decision to not apply PWHT will be successful. th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 14 5. Preliminary conclusions and further investigations Residual stresses and weld defects play a major role in the fatigue behaviour of welded structures. Their effects should be taken into account in a theoretical analysis. Compressive residual stresses can increase the damage tolerance of a weld significantly. Therefore, a PWHT is not necessarily beneficial. To decide about the most adequate procedures one has to be sure of all influencing factors. This only gives the possibility to restrict the production of welded components regarding residual stress sensitivity. With the SST “Schweißsimulationstool” is it now possible to apply analytical investigations for different applications. However, regarding the variety of required parameters, the results have to be validated by benchmark measurements. Examples for further calculations: A) Variation of heat input: number of layers, excessive parameters B) Influence of preheating, flame straightening C) Variation of restraint: circular weld, restricted shrinkage D) Repair welding 6. References /1/ SST Schweißsimulationstool – Werkzeug zur vollständigen numerischen Simulation des Schmelzschweißens, Schlussbericht zum BMBF-Verbundprojekt 02PD1051, 2005 /2/ J.B.Leblond, J.Deveaux, Acta metall., 32 (1984), 137. /3/ FKM-Guideline, Analytical strength assessment of components in mechanical engineering th 5 , revised edition, 2003, Forschungskuratorium Maschinenbau /4/ Schindler, H.J., Martens, H.J., Sönnichsen, S., “A Fracture Mechanics Approach to Estimate the Fatigue Endurance of Welded T-Joints including Residual Stress Effects”, to be published in Fatigue and Fracture of Engineeing Materials and Structures, 2006 /5/ Schindler, H.J.,”Experimental determination of crack closure by the cut compliance method, “ASTM STP 1343, R. McClung and J.C. Newman, Eds., American Soc. For Testing and Materials, West Conshohocken, PA. (1999), 175-187 th 24 CADFEM Users’ Meeting 2006 International Congress on FEM Technology with 2006 German ANSYS Conference October 25-27, 2006 Schwabenlandhalle Stuttgart/Fellbach, Germany 15