While engineering groups have steadily increased their appetite for simulation, the simulation process itself remains a challenge. Simulation generally involves multiple parties across various engineering disciplines working independently on what oftentimes are conflicting strategies. Each is armed with a different set of tools and, most likely, in hot pursuit of their own objectives. To complicate what by nature is a lengthy and intricate process, simulation efforts are also still deeply mired in manual steps that prevent the various silos from sharing results until late in the game when it is difficult to make changes without having a negative impact on product delivery schedules and budgets.
Perhaps the biggest hurdle in the use of simulation for complex product design is the growing requirement for multi-objective optimization—or, in simple terms, the ability to achieve multiple goals simultaneously as a means of working towards the greater product design. Consider, for example, one engineering group charged with making a product more aerodynamically efficient, while another engineering team is tasked with ensuring safety of the same product. The group focused on performance might leverage Computational Fluid Dynamics (CFD) and other analysis tools to zero in on the optimal design in terms of aerodynamic features and air flow. Conversely, the group charged with safety might have a completely different set of goals and consequently employ alternative simulation methods like Finite Element Analysis (FEA) to find a design with optimal structural integrity. More likely than not, the two simulation exercises will produce radically different interpretations of the “best” design.
This, of course, is a relatively straight-forward example of trying to reconcile two potentially conflicting design objectives. But what happens when multiple disciplines are introduced, demanding additional optimization scenarios and trade-offs? Without the proper workflows, process integration, and communications and analysis capabilities in place, siloed simulation efforts are more likely to produce results that are at cross-purposes, impeding that universal quest for the overall optimal design.
Moreover, as the different disciplines advance the cause of their individual objectives, the onus shifts to engineering management to hash out a balanced design. Instead of yielding the best design based on true multi-objective optimization, this approach tends to produce an outcome borne out of emotional gut instincts and compromise rather than something grounded on proven scientific principles.
An optimized simulation process
So what’s the answer for companies seeking out a process that brings balance and rigor to multidisciplinary, multi-objective optimization, while allowing them to quickly see the greater design picture? It comes in the form of modeFRONTIER, a process integration and simulation platform from ESTECO that has been built from the ground up to solve this specific simulation challenge. modeFRONTIER helps companies identify the set of best possible solutions available while eliminating guesswork and introducing rigor and automation in simulation processes.
modeFRONTIER achieves this through three primary areas of functionality: Process integration and automation, optimization, and data visualization and analysis. Let’s start with process integration and automation—capabilities designed to address the problem of myriad product engineers and designers employing a wide variety of CAE, CAD, and other application tools in the course of the development effort in a disconnected, trial-and-error fashion.
Using modeFRONTIER, an automated process can be easily constructed using a workflow editor that couples the different computational steps and data formats into an automated, streamlined and repeatable process. Moreover, modeFRONTIER integrates seamlessly with dozens of third-party tools, simplifying the utilization of CAE parametric models by recognizing their predefined parameters and mapping them to input and output variables.
By automating both repetitive and concurrent simulations, engineering teams are freed from the painstaking process of manually combining output from different applications, reducing the time and cost associated with simulation, while also eliminating the guesswork when trying to balance multiple design objectives.
On the optimization front, modeFRONTIER comes with a wide range of algorithms, including those tuned to tackle discrete or continuous variables. It has the ability to determine the set of best possible solutions combining opposite objectives, while respecting any user-defined constraints. After preliminary exploration of the design space via the Design of Experiments module, modeFRONTIER’s optimization algorithms work to pinpoint the optimal solutions, displaying the results on the trade-off curve of designs, known as the Pareto Frontier, where each identified point can be considered an optimum.
modeFRONTIER’s third area of focus is statistical analysis and data visualization features for analyzing DOEs, simulation results, and Response Surface Models (RSMs). RSMs are an interpolation and regression methodology used to approximate, analyze, and simulate complex real-world systems with the aim of saving time and maximizing limited resources. As part of this area of functionality, modeFRONTIER includes a variety of multi-dimensional and interactive charts that assist in the screening of the number of variables and in controlling how the initial points are distributed--both tasks associated with DOE analysis. In addition, the advanced visualization capabilities help clarify the relationship between the initial conditions and the simulation results, and to better leverage the analysis, modeFRONTIER offers additional tools and methodologies to guide users in the decision process.
modeFRONTIER at Work
Consider BMW, which leveraged modeFRONTIER to optimize the fatigue life of a diesel engine crankcase and gasket. The objectives were to maximize safety factors at the inter-cylinder walls, while maximizing the minimum bead pressure during the low temperature phases of combustion.
Challenge: BMW’s traditional optimization techniques tended to “over optimize,” producing solutions that performed well at the design point, but showed different results due to off-design characteristics. To address this issue, the team needed to take into account the uncertainty of certain input parameters since they couldn’t always be precisely determined under real manufacturing and operating conditions. Moreover, a product designed for one specific environmental scenario was not especially suited in other environments; thus the team needed a way to come up with designs which had lower variability of performance.
Solution: modeFRONTIER’s robust design algorithms helped address the multi-objective design optimization challenge by allowing one variable and three constants to be defined as stochastic. In that way, the platform automatically created a set of sample designs during the optimization with a user-specified distribution and variance for each stochastic variable, centered at the initial value point. modeFRONTIER’s workflow editor, together with several direct integration nodes, helped the team create an automated pipeline connecting the different solvers required to solve the optimization challenge, including Paramesh to modify the geometry, Abaqus for the thermo-structural computations, and Femfat, to compute the safety factors.
Results: By using the RSM approach to run a virtual, robust, optimization with thousands of computations, in addition to running ABAQUS to validate the virtual results, BMW was able to improve the Fatigue Safety Factor by 15%, a solution which also constrained the variation of the measured output to less than 1%.
The future Frontier
The current desktop-centric approach will progressively give way to a new enterprise-wide, collaborative design and development process in which different people, belonging to different disciplines and possessing different skill levels, will perform complementary tasks, all while sharing data, simulation results, and knowledge. Existing simulation systems will evolve to support new usage paradigms that are tailored to the task at hand as well as the related level of expertise, guiding users through simulation steps via customized processes that orchestrate the activities and hide complexities. At the same time, transparent access to on-demand computing power will drastically reduce wait times for simulation results, and will open up new possibilities around designing more sophisticated, reliable, and safe products.l