How many times has the first design iteration submitted to FEA modeling passed the design criteria?
The answer is close to zero, but even if it does happen by stroke of fortune, the design is not the optimal design – which means that although design requirements are met and validated by FEA, there is always scope of improvement either in terms of cost or in terms of performance. In general, it is not unusual to reach the optimal design in 15 to 20 iterations.
An analyst know the pain of creating a detailed finite element simulation model. Most of the steps involved, such as geometry cleaning and meshing, are very time-consuming, and they are primarily driven by geometry. Let’s look at the workflow in more detail:
An analyst in automotive industry often performs finite element modeling work in Hypermesh, stress analysis in Abaqus, optimization in Optistruct, and durability in Fe-Safe or N-code. An analyst in the aerospace industry often performs CAD composites work in CATIA, finite element modeling in Abaqus CAE, stress analysis in Abaqus or Nastran, and durability in Fe-Safe. An analyst working in other industries has his own suite of FEA tools to work with. The entire process requires data flow from one simulation code to the other. This means output from one code serves as an input to the other. Quite often this work is also done manually by the analyst.
This means that in situations where optimal design is obtained in 20 iterations as mentioned above, an analyst has to perform geometry cleaning 20 times, create FE meshes manually 20 times, and also transfer the simulation data from one piece of code to the other 20 times. By the time these design iterations are over, the analyst’s face and computer looks somewhat like this:
Let analysts remain as analysts and let simulation robot do the rest!
The traditional job of finite element analyst is to build robust high fidelity simulation models that gives correct results under real life load applications. The analyst is not an FE robot who can perform repetitive tasks with ease. In situations like one mentioned above, it makes perfect sense to let FE analyst create a robust FE model only once per FE code involved. Subsequently introduce a simulation robot that can capture hidden steps and workflow, create a script and execute that script multiple times. This simulation robot is called ISight.
Dassault Systemes’ SIMULIA offers ISight: an open-source, code-neutral process automation and design optimization tool that can create robust simulation workflows with ease. It is suitable to automate frequently-used applications that are often multi-step and conditional in nature. It comes with its own graphical user interface and component libraries. Isight provides designers, engineers, and researchers with an open system for integrating design and simulation models—created with various CAD, CAE, and other software applications—to automate the execution of hundreds or thousands of simulations.
Just like any FEA code, ISight has a user interface for pre- and post-processing, as well as a solver. The nomenclature is, however, a bit different. The user interface is called Design and Runtime gateway and the solver is called Simflow. The component library of ISight is very exhaustive, which means that most of the tools – CAD, CAE, as well as general computational and desktop tools – can be incorporated in ISight Simflow. The usage is as easy as dragging and dropping the components into the viewport, configuring each component, and connecting them logically as per the simulation workflow definition. A few examples of supported components include Abaqus, Dymola, CATIA, Solidworks, Matlab, Mathcad, Calculator, E-Mail, Data Matching and many more.
Once the process automation task is completed, it’s easy to execute this automated workflow multiple times without the manual intervention of the FE analyst. But each individual execution may be time-consuming, as well as storage-intensive -especially if it is associated with a bulky FE model. So it makes perfect sense to execute the Simflow with a certain objective, which could be design of experiments, parametric optimization, Monte Carlo simulation, Six Sigma, etc. ISight offers value-added intelligent algorithms such that any of these objectives can be accomplished with as few iterations as possible.
Remember: Mr. Analyst got frustrated with 20 design iterations. How about achieving the same optimal design in 10 iterations instead?
Let’s look at few of these value added techniques in more detail:
Design of Experiments (DOE)
The DOE component enables engineers to quickly assess the impact of various design variables based on a set of objectives and identify significant interactions. Access a full suite of methods including Central Composite, Data File, Full Factorial, Fractional-Factorial, Box-Behnken, Latin Hypercube, Optimal Latin Hypercube, Orthogonal Array, Dependent Variable Sampling and Parameter Study with appropriate post-processing options.
Isight provides a comprehensive selection of parallelized optimization techniques that can be applied to a variety of problems. It also includes techniques that can handle multi-objective optimization problems. Define your optimization problem in terms of variables and multiple weighted and scaled objectives and constraints using the following algorithms: Gradient: NLPQL, MMFD, LSGRG2; Pattern: Hooke-Jeeves, Downhill Simplex, Adaptive Simulated Annealing; Mixed Integer/Real: MISQP, MOST; Genetic Algorithms: Evolution, Multi-Island Genetic Algorithm.
Use probabilistic analysis to measure the quality of a design given uncertainty or randomness of a product or process. Perform reliability analysis with the mean value method, FORM and SORM reliability method, importance sampling, sobol sampling, DOE sample, or Monte Carlo Analysis. This component can be used in combination with optimization and approximation techniques to perform fast Six Sigma optimization.
The amount of output data generated from Simflow can be overwhelming, in terms of both storage and visualization. This is especially true in cases where Simflow containing big FE models execute hundreds of FE simulations. ISight offers advanced simulation data management and visualization options to make the best use of output data. The Runtime Gateway enables desktop and distributed execution of engineering process flows. It also enables the creation of graphs and tables to visualize results. All job results are saved automatically to a locally managed database. The user interface supports the creation of visual tools for real-time post-processing of data such as tables, 2D and 3D plots, and statistical analysis.
Latest posts by Ankur Kumar (see all)
- Multiple ways of using FEA: From point solutions to simulation democratization - November 16, 2017
- Laser Projection methods in CATIA Composites - November 7, 2017
- Implicit vs. explicit schemes in Abaqus - September 18, 2017