Interoperability tools for design and analysis, new elements, additional material models, faster solvers, and innovative training techniques make the latest finiteelement field worth a closer look.
Robert Williamson
Development Manager Algor Inc.
Pittsburgh, Pa.







Finite element technology is changing so fast that analysts who keep their noses to the grindstone may miss out on many of the recent, useful developments. For example, an actuator element can now simulate the motion of hydraulic, pneumatic, and electric cylinders and solenoids. This element lets designers analyze complex mechanical events that would have been nearly impossible even just a few months ago.
What's more, FEA companies are consistently improving their solvers to reduce solution times. They have to. The everincreasing memory and disk space capacities on today's computers encourage finiteelement analysts to work with larger and larger models. These trends and developments make it useful to step back from the daily grind for a better view of what's cropping up in the industry.
WHAT'S NEW IN PREPROCESSING?
Success using universal files, such as IGES, to transfer geometry from a CAD solid modeler to FEA software had been so intermittent that companies began integrating their FEA programs inside CAD software. This was intended to provide CAD/CAE interoperability with entrylevel FEA tools for engineers who typically need quick verification of their product design. Although this arrangement solved the problem of geometry transfers, extending the software for new analysis tasks is often difficult to impossible.
For example, limiting the number of available element types to simplify FEA programs also significantly limits what can be modeled. Engineers and analysts realized that the lack of other element types, such as beam, truss, or brick elements, in such a product keep them from analyzing even moderately complex parts or mechanisms.
Some FEA companies now recognize that a better approach is to reside just "one window away" from the CAD system while still directly transferring geometry through memory. Communication with the CAD solid modeler remains direct, but upgrades and expansions to the analysis program are easy.
It's also easier on analysts when they are presented with a consistent user interface regardless of the type of analysis performed. A consistent interface saves users from mastering a different GUI just because analysis needs change. In addition, the interface is the same no matter what the CAD package.
Another trend is the increasing level of expectations for engineering software. There was a time when engineers were comfortable working through a commandline prompt or an unfriendly technical terminal. Engineering software was almost expected to be difficult.
That is no longer the case. A modern graphic user interface with userfriendly dialogs and wizards are features analysts now expect in FEA software. For example, our company's effort to meet this demand has delivered additional easeofuse features such as contextsensitive help, which calls up relevant written information no matter where you are in the program. In addition, our software interfaces with Microsoft Office so users can, say, import loadcurve data from an Excel spreadsheet.
Furthermore, a single consistent interface across lowend, intermediate, and highend analysis means users aren't confused when they add other capabilities. So users work within the same FEA interface whether they work on problems involving heat transfer, electrostatic, linearstatic stress, linear vibration, fluid flow, or involving analysis of mechanical events.
Another recent development is the wider use of the 20node brick element in linearstatic stress problems. The element is modeled as a standard eightnode brick, but includes midside nodes. Although 20node brick elements are not new, they have been available mostly in more expensive nonlinear, FEA programs. Brick elements are generally regarded as more efficient than tetrahedral elements in that fewer of them obtain a similar level of accuracy in a particular model.
The additional nodes let the brick element capture bending more accurately than eightnode elements or tetrahedra. Another way to capture bending involves a large number of elements. The new element avoids this latter tactic. The 20node element works with isotropic and orthotropic material models, and can be temperaturedependent, making it applicable to a wide range of linear material models.
Additionally, a piezoelectric material model is available for the new brick element. In the same way that heat generates stress in a temperaturedependent element, voltage generates stress in the piezoelectric material model.
Composite elements are available for wider application than ever before. These elements are modeled in the same manner as plate/shell elements. The data input screen for a composite element lets users assign the number of plies that make up the laminate. Each ply is assigned a thickness, an orientation angle that positions it with respect to other plys, and a material. This happens in our program through a simple threecolumn spreadsheet that tracks the userdefined data.
Composite elements have been available for many engineering conditions, such as vibration analyses and linearstatic stress. More recently, they have been extended to model linear critical buckling. This type of analysis uses Euler approximations to find a buckling load multiplier. The composite element is also proving useful in simulations of mechanical events, including those involving largescale motion.
Kinematic elements, another recent invention, assist with large models by transmitting loads, motion, and displacements, but do not calculate stresses. This element is ideal for large models with small areas in which stress is of concern.
Communication links between different analysis modules are a more general development adding greater capabilities. Integrating different analyses together more directly lets the user model the multiple physical phenomena affecting parts in the real world. For instance, the link could let users perform an electrostatic analysis to determine unknown voltages and then automatically apply the calculated results to a linearstatic stress model using the piezoelectric material model to find the stress produced by the voltages.
WHAT'S NEW IN SOLVERS?
Analysts probably expect to see new elements and features in preprocessors since that is where they spend most of their time. But some of the developments they don't see, such as in solvers, can be equally impressive.
Recent solver developments have focused on sparse matrix and iterative technology. While aerospace companies frequently used this technology in the past, a growing number of analysts from all industries now take advantage of the solvers because they can process and analyze large models much faster. It wasn't long ago that users asked themselves: How can I simplify a model so it runs faster? The recent trend is to analyze the real thing — a more complete model. This has encouraged increased model size and the need for faster solvers, such as the sparse matrix and iterative versions.
Sparse solvers are based on the fact that zeroterms in the stiffness matrix need not be considered. The iterative solver, on the other hand, is often the fastest for analyzing large solid models. But there is a price to pay for its speed — it does not always converge. The iterative technology uses extrapolation techniques so it's not actually solving for every instance. Therefore, intermediate results in a linearstatic stress analysis are not meaningful.
Because it's important to tell whether or not a solution is converging or how it is converging, users can watch the progress of the analysis through a monitor program that generates realtime plots of convergence. This is also important when simulating moving mechanisms. For example, if a large mechanism is not converging due to its geometry, such as widely varying stiffness throughout the model, analysts can switch to the sparse solver, which will converge. In addition to the sparse solver, analysts can always choose reliable and standard skyline or banded solvers for more general scenarios. Both of these are stable, direct solvers that do not have convergence issues inherent in the iterative solver.
Iterative solvers, however, are perfect candidates for fluidflow problems, which traditionally involve large models where results do not significantly change from one time step to another. Expect to see big changes in this analysis arena in the next few months.
One of the most significant recent developments is the capability to analyze moving events in a finiteelement analysis using Mechanical Event Simulation software. The concept is simple: things move, they collide, and sometimes they break. But FEA technology for linearstatic analysis models only stationary objects, often requires the input of estimated loads, and is unreliable for large deformations. Truetolife simulations, on the other hand, should let users analyze more lifelike events, including those involving small and largescale motion.
An automatic timestepping feature in the Mechanical Event Simulation software makes it possible to simulate actual events. Consider a car headed toward a telephone pole, for example. Nothing happens during periods of relative inactivity so the software assigns a large time step to the event. But when contact is imminent, the software reduces the time step to fractions of a second to capture details of the impact.
This kind of automation in software is showing up in other places as well. Not long ago, engineers measuring the motion characteristics of a mechanism had to manually transfer loads from one program to another. More recent thinking says, should the engineer choose a staticstress analysis instead of the dynamic event, let the software perform the load transfers to an FEA program.
This idea has resulted in what's called an Inertial Load Transfer module. The capability expands upon recent geometrytransfer techniques that bring over an entire assembly and turn it into an FEA model comprised mainly of kinematic elements. After an analysis reveals the motion, the Inertial Load Transfer module automatically transfers loads to a linear or nonlinear stress analysis and performs an analysis on one or more parts to get the resulting stresses. Older systems require working in a kinematics package and on a separate kinematic model. Results would often be transferred manually, a process that invites error. In addition, the kinematic model is based on assumed joint stiffness and rigidbody motion, not flexiblebody motion.
WHAT'S COMING?
FEA companies are paying more attention to how they might simplify applying, modifying, and removing model loads and constraints. With recent technology, FEA users can rightclick on a model location to apply constraints or loads, or rightclick on an existing constraint to alter its magnitude. In addition, this technology is available for all analysis types. Because FEA is an iterative process with one analysis often leading to another, efforts in this area will increase a user's overall efficiency by enabling quicker, easier modification of model properties. That is, the user can quickly review results, change the model, and begin another analysis when necessary.
Rightclick applications also introduce more types of common loads, constraints, and ways of connecting parts. These are especially critical for motion studies, which require engineers to identify where and how links are jointed. For example, a piston and rod are connected at a joint that should not translate from side to side, but must rotate about a connection point. It's important to define the joint properly to capture the true motion of the mechanism.
To quickly model such connections, expect to see kinematic pivots, which let kinematic elements rotate in different relationships to each other. Kinematic pivots are the first of a series of joints that accurately model movement for analysis using Mechanical Event Simulation.
Also, look for more dynamic postprocessing capabilities, realtime monitoring, and presentation capabilities. For instance, users should be able to dynamically rotate, pan, or zoom in on a model to quickly examine von Mises stresses or other results regardless of where they reside on the model. In addition, analysts can take advantage of time saving monitoring capabilities, which make results available graphically during the analysis of timedependent analyses like transient heat transfer and Mechanical Event Simulation. Engineers need not wait for an analysis to finish before evaluating results to determine if modifications are necessary.
Once analysts have completed their work, the task of presenting the results remains. Report Wizards generate customizable HTML reports of the FEA model data that can include VRML, AVI, JPG, and TIF graphic files. The reports can be printed or distributed by an intranet or Internet.
Lastly, users should not be surprised to learn that most ideas for new features come from them. Many of the new features mentioned above stem from customer feedback. If you have an idea for a new feature, contact the developer of your FEA software. You could lighten your own workload.