Académique Documents
Professionnel Documents
Culture Documents
Element Based Fatigue Analysis Authors: Dr NWM Bishop, MSC Frimley and Alan Caserio. MSC Costa Mesa
Abstract
Fatigue analysis procedures for the design of modern structures rely on techniques, which have been developed over the last 100 years or so. The first accepted technique was the S-N or stress-life method generally given credit to the German August Woehler for his systematic tests done on railway axles in the 1870s. Initially these techniques were relatively simple procedures, which compared measured constant amplitude stresses (from prototype tests) with material data from test coupons. These techniques have become progressively more sophisticated with the introduction of strain based techniques to deal with local plasticity effects. Nowadays, variable stress responses can be dealt with. Furthermore, techniques exist to predict how fast a crack will grow through a component, instead of the more limited capability to simply predict the time to failure. Even more recently techniques have been introduced to deal with the occurrence of stresses in more than one principal direction (multi-axial fatigue) and to deal with vibrating structures where responses are predicted as PSDs (Power Spectral Densitys) of stress. Even more recently researchers have addressed the requirements for the design of specific components such as spot welds. All of these techniques were developed outside of the Finite Element environment. However, they have now been implanted into many FE based analysis programs, the best known of which is MSCFATIGUE. The FE environment introduces additional considerations relating to how input data is processed and how fatigue life, or damage, results are post processed. This paper will deal with the issues associated with how fatigue techniques can be incorporated into the FE environment. Modern examples of FE based fatigue design will be included.
Results
Fatigue analysis has traditionally been a test-based activity. Components or models are tested with service loads, which are as close to the in-service signals as possible. In a test situation loading is usually a stress signal measured remotely from a critical location. Geometry is usually a stress concentration factor to account for the separation of the critical location and measurement point and materials are the cyclic fatigue properties. The biggest drawback with testing is that it can not be undertaken until a prototype exists. If a design problem then occurs it is usually very difficult to rectify. It is also very expensive to perform fatigue tests. For these reasons FEA based fatigue analysis has been perceived as an excellent enhancement to the testing process. The FEA model effectively replaces the geometry box in Figure 3. Loading signals are now forces, displacements or some other driving function. Material properties still have to be obtained through test, however empirical approximations can be made based solely on the UTS and Youngs Modulus of the material. The correctness and accuracy of each of these inputs is important in that an error with any of these will be magnified through the fatigue analysis procedure (the fourth box,) since this process is logarithmic. A 10% error in loading magnitude could result in a 100%, or more, error in the predicted fatigue life. The fifth box is the post-processing or results evaluation. This can take on the form of color contours on a finite element model or a tabular listing, but also quite often leads back into the three inputs to see what effect variations of these inputs will have on the life prediction. This is referred to as a sensitivity or a what if study. This is extremely useful at times when you are not quite sure about the accuracy of one of the inputs.
MSC/FATIGUE
1E9
Life Contours
Cross Plot of Data : SAETRN
5
Loading Histories
Strain Life Plot
MANTEN Sf': 917 b: -0.095 Ef': 0.26 c: -0.47
Stress (total) Life Strain (initiation) Life Crack Propagation Vibration Fatigue Multi-axial Fatigue Spot Weld Analyzer Software Strain Gauge Utilities
) ( r o t c a F e l a c S
4
1E2
1E3
1E4
Life(Repeats)
nCode nSoft
Sensitivity Analysis
DAMAGE HISTOGRAM DISTRIBUTION FOR : SAETRN.DHH Maximum height : 6.8848E-5 Z Units :
1E-1
6.8848E-5
Damage
1E-2
3496.5 0 0
Range uE
1E-3
Mean uE
5281.3-1732.5
nCode nSoft
1E-4 1E0 1E1 1E2 1E3 1E4 1E5 1E6 1E7 1E8
Life (Reversals)
Materials
Damage Distributions
Infor
Total Life
Nf
FEA based stress analysis options
Ni
Np
There are several FEA based methods for obtaining the stress information that is required to perform a fatigue life calculation [1] Static structural (and fatigue) analysis can be undertaken utilising MSC/FATIGUEs superposition capabilities for combining multi load application inputs. Unit inputs of load are applied to all desired load application points. The resultant stresses (caused by the unit load cases) are then factored by the actual time history of loading for that load application point. This process is repeated for all load application points and the results are linearly superimposed. Fatigue life calculations are then performed using these combined stress histories. This method ignores dynamic influences such as mass effects. [2]. Dynamic transient analysis. If this approach is used, the stress histories are produced at each point of interest using a FE transient analysis method. These stress histories are also superimposed to obtain the required combined stress histories, but the FE solver handles this. Fatigue life calculations are then performed on these stress time histories. This method accounts for all dynamic effects but is less versatile in that all loads must be combined in a single FE analysis. [3]. Frequency Response analysis. In this approach the transfer functions are produced using the desired solver. These transfer functions are then resolved onto the desired stress axis system (usually principal stress). The response caused by multiple random loading inputs is then obtained using standard random process techniques. The effect of correlation between inputs can be dealt with by including Cross Power Spectral Density functions in the input loading data. This method accounts for all dynamic effects and is quite versatile. [4]. Random Vibration analysis. In this approach the response Power Spectral Density function is determined directly from the FE solver. Effects due to multiple load inputs must be dealt with in the FE analysis as with a transient analysis approach. All dynamic effects are accounted for but this method has the limitation that fatigue life can only be computed for a single component direction. Stress response results are not resolved onto a desired stress axis system by the FE analysis.
Design Philosophies
There are three main fatigue design philosophies. Each centers around one of the fatigue life estimation methodologies. To illustrate the three consider the design of a stool.
Safe Life. The safe life philosophy is a philosophy adopted by many. Products are designed to survive a specific design life. Full scale tests are usually carried out with margins of safety applied. In general, this philosophy results in fairly optimized structures such as a stool with three legs. Any less than three legs and it would fall over.
Fail Safe . On the other end of the spectrum of design philosophies is that of fail safe. This is where a failure must be avoided at all costs. And if the structure were to fail it would fall into a state such that it would survive until repairs could be made. This is illustrated with our stool now having six legs. If one were to fail the stool would remain standing until repairs could be made. This philosophy is heavily used in safety critical items such as in the aerospace or offshore industries.
Damage Tolerant. The middle ground philosophy is that of damage tolerance. This philosophy, adopted heavily in the aerospace community and nuclear power generation, relies on the assumption that a flaw already exists and that a periodic inspection schedule will be set up to ensure that the crack does not propagate to a critical state between inspection periods. As implied, this philosophy adopts the crack growth method. This is illustrated using our stool (now with four legs) but with someone inspecting it. This particular design philosophy is generally used in conjunction with the fail safe philosophy first to try and design such that no failure is expected but then to assume that, for whatever reason, a flaw does exist and must be monitored.
Fatigue Testing
The first fatigue tests were carried out on full-scale components to establish their safe working stress. Later, the more complete relationship between cyclic stress or strain and fatigue life was established. Small-scale specimens were tested to study component life and also fatigue mechanisms. In more recent times, as tests had to become increasingly realistic, special test techniques were developed such as Remote Parameter Control. Today, testing is still the most common way of confirming the fatigue life of a product prior to releasing it onto the market. However, testing often reveals weaknesses, which necessitate re-design. Assessing the suitability of particular design modifications using fatigue testing alone can be time consuming and cost far more than just a delayed product.
Fatigue Modeling
The estimation of fatigue life using mathematical modeling techniques was developed to assist the engineer in solving fatigue problems without always having to physically test all the options. For this reason, techniques such as local strain or crack initiation modeling have become widely used. Improvements in the power of computers have enabled the effective use of these techniques. Today, most major companies designing mechanical structures will use a fatigue life estimation tool such as MSC/FATIGUE in conjunction with testing. The late 1980s had establish the use of finite element analysis (FEA) as a tool for stress analysis. At the same time the integration of FEA and fatigue life estimation through the MSC/FATIGUE product began to provide new benefits by assessing fatigue earlier in the development process.
Sophisticated crack initiation or strain-life (-N) modeling provides a method for estimating life to the initiation of an engineering crack. Crack initiation specific features include: Neuber elastic-plastic correction. Advanced elastic-plastic correction based on Mertens-Dittman or Seeger-Beste methods Cyclic stress-strain tracking using Massings hypothesis and material memory modeling Smith-Watson-Topper and Morrow mean stress correction Advanced biaxial corrections (proportional loading) based on Parameter Modification or Hoffman-Seeger
When a notch becomes a crack, the stress field becomes a singularity (in theoretical elastic terms) and the stress concentration, K t, is no longer a useful way of describing the feature. Rather we need something that describes the intensity of the stress field around the singularity. This concept is well illustrated by the diagram below where a hole is introduced into an infinite plate. As the hole becomes an ellipse and the ratio of the length to width of the ellipse becomes greater and greater, tending towards infinity, so does the stress concentration.
m da ------- = C ( K ) dN
Equation 1. Paris Law
K = Y ( a )
Equation 2. Stress Intensity K versus compliance factor Y
Y is known as the compliance function and describes the geometry in which the crack exists. It relates crack length to geometric features of the part or component. Perhaps one way to describe a compliance function in physical terms is the change in stiffness or flexibility (compliance) as the crack grows, i.e., the structure becomes more compliant as the crack gets longer. The dictionary defines compliant as ready or disposed to comply, and compliance as the act or process of complying to a desire, demand, or proposal or to coercion. In engineering terms it is the ability of an object to yield when a force is applied. Material response is modeled by measuring crack growth rates versus stress intensity (K) in constant amplitude tests. From these tests are derived the da/dN curve and the threshold characteristics and fracture toughness of the material. In fatigue we are concerned with stable crack growth occurring below a catastrophic level. When you plot crack growth rates against K on log scales, you get sigmoidal shaped curves like these, which have three distinct regions. There is a linear region in the middle of these curves, which is described by the Paris Equation. At the bottom end of the curves there is a threshold below which no crack growth occurs (very similar to a fatigue limit). This is caused by crack closure and the interaction of the crack with the micro-structure. If the mean stress is raised the threshold decreases because the cracks are held open for more of the time. At the other end of the curve, crack growth rates increase as the maximum stress of each cycle gets close to the fracture toughness of the material. This curve is called the apparent K curve. However there are many effects that this equation does not take into account, such as crack closure, corrosive environments, the influences of a notch, and static fracture mode contributions to name a few. MSC/FATIGUE models these by using an Effective K curve which has the effect of linearizing the entire Apparent K curve through all three of its distinct regions. It is this Effective K that is the actual (effective) driving force that is then used in the Paris Equation to determine crack growth. As discussed earlier, the Fracture Mechanics Triangle relates stress intensity, stress, and crack length. When speaking in terms of crack growth and overall life, a rectangular rather than a triangular representation is used. In Crack Growth there is a relationship between stress range and life just as with the Total Life (S-N) method except it is extended to include the initial and final crack lengths (and all crack sizes in-between these two limits). So in a similar way to solving the triangle, the fatigue crack propagation rectangle can be solved by knowing any three of the four corners to derive the fourth.
da/dN
Stress Range
Such systems are currently designed, and analyzed, predominantly through the use of expensive and time consuming test based procedures. MSC/FATIGUE vibration allows designers to identify and deal with such damage at a much earlier stage in the design process, thus reducing or eliminating the need for expensive prototype tests. As well as a fatigue analyzer this module also contains a state of the art analysis tool which provides a complete solution path for multiple load case frequency domain based analysis. It includes new advances in stress tensor mobility and biaxiality checking. The simplest method of obtaining stress based PSD information is to read the PSDs directly from the FE solver and this is supported. A more flexible and sophisticated approach is to read transfer functions from the solver database. These transfer functions are rotated onto any one of a number of user defined stress systems (maximum principal being the most relevant). Results for these axis systems are computed for each incoming load case (each frequency and each load application point).
PSD Force
frequency
PSD Force
G11()
G12()
frequency
PSD Force
frequency
PSD Force
G21()
G22()
frequency
Reaction
The software strain gauges are defined as finite element groups, each containing between 1 to 3 elements. All standard strain gauge definitions are supported in both planar and stacked formulations. User defined gauges may also be created, with definitions stored in a gauge definition file. The virtual strain gauges are positioned on the finite element model surface, with the gauge aligned in any orientation, and the gauge covering multiple finite elements. The results obtained from the Software Strain Gauge are averaged results from the underlying finite elements, modeling the same geometric averaging obtained with actual instrumentation. Results are transformed to the coordinate system and alignment of the software strain gauge. The Software Strain Gauge has the following features: Multiple Gauge Geometries Uniaxial Gauges T Gauges Delta & Rectangular Gauges Stacked & Planar Gauges User Specified Gauge Definitions Gauge Definition Files (user definable gauges) Up to 200 simultaneous Software Strain Gauges
The Software Strain Gauge is also of benefit to the analyst performing MSC/FATIGUE weld durability calculations in accordance with British Standard 7608. The Gauge tool allows ready access to strain time histories at the weld toe, providing important information for weld durability calculations.
10
Analysis of welds joining two metal sheets S-N (total life) technology The number of Spot Welds within the model is limited only by the FEA analyzer Up to 20 different groups of similarly defined Spot Welds may be simultaneously analyzed in the FEA model Unlimited number of Spot Welds per definition group The analyzer simultaneously calculates weld nugget and sheet fatigue life 108 sets of fatigue calculations are performed for each spot weld
The local strain methods allow global and single node/element fatigue calculations with a variety of postprocessing options including fringe plots, histogram plots, time correlated damage plots and polar damage plots. The method includes a new multiaxial nonproportional notch correction procedure, incorporating an energy-based notch rule based on Neuber's rule and a Mroz-Garud cyclic plasticity model.
Figure 13. Fatigue life contour plot for steering knuckle under the application of 12 correlated loading inputs
11
12
Result access features include FE results data from: MSC/PATRAN database in the following form: Linear static (stresses & strains) Linear transient (stresses & strains) Includes results from MSC/NASTRAN, ANSYS, ABAQUS, MARC & any other PATRAN supported analysis codes MSC/ACCESS (MSC/NASTRAN xdb files) Linear static & transient stresses SDRC Universal files Linear static stresses & strains External result files PATRAN nodal & elemental result files MSC/PATRAN FEA result files
USER
Materials
MSC / FATIGUE
Model Definition Analysis Setup Post-Processing
Time History
PATRAN
MSC / NASTRAN ABAQUS ANSYS MARC
13
A materials database manager stores and manipulates a library of cyclic material properties. Features include: Approximately 200 materials (steels) supplied Add, create or modify your own or supplied materials data (Imperial & SI units supported) Generate materials data from UTS & E Weld classifier based on BS7608 Graphical display of: Component & material S-N curves Cyclic & monotonic stress/strain curves Strain-life curves Elastic-plastic lines Fatigue limits (endurance limits) Graphical display, hardcopies & tabular comparison of materials
1E-1
) e l c y c / m ( N d / a d
1E-6
1E3
1E-7
1E-2
1E-8
1E2 1E-3
1E-9
1E-10
1E1 1E0 1E1 1E2 1E3 1E4 1E5 1E6 1E7 1E8 1E9
1E-4 1E0 1E1 1E2 1E3 1E4 1E5 1E6 1E7 1E8
Life (Cycles)
nCode nSoft
Life (Reversals)
nCode nSoft
S-N
E-N_ LEFM Figure 15. The 3 alternative material curves used in fatigue analysis
A variety of results data is reported both in tabular form and graphically. A single location analyzer can be used for what-if studies after a global analysis has identified hot spots. Results data include: Damage/life (reported in linear & log form) Multiaxial assessment parameters: Factor of safety Contour Plotting of: Life Estimates Log of Life Damage Component Specific Life Units (Flights, Miles, etc.) X-Y Plots of Sensitivity Studies Histogram plotting
14
A number of what if scenarios can be investigated including back calculations based on design life of: Scale factor (stress concentration) Residual stress Probability of failure (design criterion) Sensitivity studies of: Multiple scale factors (stress concentrations) Multiple residual stress values Multiple probabilities of failures (design criteria) Surface finish/treatment Mean stress correction methods Graphical display and hardcopy of sensitivity plots Change materials or surface finish/treatment Material searches based design life
Future Development
Future plans for development include the following Thermal-Mechanical Fatigue. This involves a strategic relationship with a major US car company and automobile producer. Fatigue of Cast Iron. This involves a strategic relationship with a major US agricultural equipment company. Fatigue Editing. This involves a major ground vehicle company in the US. Fatigue Analysis of Rotating Structures (wheels). This involves a major aerospace company. Other MSC/FATIGUE future technology projects. Elastic-plastic non-linear FE results, load step analysis, families of S-N curves and an open architecture
15