Optimize Automotive Designs Without Optimization or Programming Expertise with pSeven

3D plot of optimization data. (Image courtesy of DATADVANCE.)

3D plot of optimization data. (Image courtesy of DATADVANCE.)

As products become more complex, the number of parameters used to define them increases. By varying these parameters, engineers can navigate the design space to discover product configurations with optimum performance.

However, with design cycles from various industries increasingly coming under time pressures, especially in the automotive sector, it can be a challenge to navigate the entire design space manually.

Engineers can employ computer-aided engineering (CAE) tools, specifically design space exploration software, predictive modeling and data model analysis to discover an optimum configuration for their product.

Design space exploration tools, such as pSeven from DATADVANCE, can connect to various multidisciplinary CAD/CAE tools to guide the software through an optimization iteration cycle. Numerous real-world applications from various industries exist, but few are as representative of the software’s capabilities as the optimization of an automotive side mirror.

Engineers designing a side mirror will need to ensure that its shape has a minimal effect on the car’s aerodynamics. As a result, the mirror would need to become smaller and sleeker with respect to the car’s frame.

However, sleeker and smaller designs might negatively affect the visibility the mirror provides for the driver. These smaller mirror designs might also be flimsier and might not be able to withstand an accidental run-in with the edge of a garage door.

Let’s not even mention the rattling and whistling the mirror on last year’s model made as soon as the car hit 50 mph. Structural, fluid dynamic, and noise, vibration and harshness (NVH) simulations might all be included within one optimization of the mirror. Engineers can also create a predictive model based on the mirror’s wind tunnel data and add that into the optimization. Talk about a grab-bag of disciplines.

pSeven performs multidisciplinary optimizations using mathematical algorithms wrapped up into several tools. Fortunately, a thorough understanding of optimization theory is not a prerequisite to use pSeven. DATADVANCE strives to have its software seen—and used—by a wide audience, not just optimization gurus. They report that pSeven will pick the proper optimization algorithm for you. It also uses simple drag-and-drop block diagrams so you don’t have to become a programmer.

How pSeven connects to Other CAD/CAE Tools to Optimize a Product

So you have a side mirror design, you’ve run it through the appropriate computational fluid dynamics (CFD) and finite element analysis (FEA) CAE tools and now you want to optimize it. What do you do?

pSeven will link your chosen CAD and other CAE software when iterating through design space explorations and optimizations. It will vary the series of parameters set up within the CAD/CAE software to map the design space and look for optima. All the user needs to do is drag and drop the CAD/CAE connections, instructions, commands and the order they should be performed in using block diagrams or integration blocks. These blocks can be specific to a particular CAD/CAE tool or general purpose.

The workflow is as follows:

  1. Create an optimization project.
  2. Add blocks from the library to create the workflow.
  3. Configure each block with the appropriate links to control the target CAD/CAE tool.
  4. Connect the blocks to complete the workflow.
  5. Start the optimization or design space exploration.
  6. Post-process the data to find correlations, trade-offs, local optima and more.
The pSeven graphical interface gives engineers the ability to drag and drop blocks and macros that connect and control third-party CAD/CAE software and arrange them into a workflow. The software will determine the best way to optimize the output of the workflow. (Image courtesy of DATADVANCE.)

The pSeven graphical interface gives engineers the ability to drag and drop blocks and macros that connect and control third-party CAD/CAE software and arrange them into a workflow. The software will determine the best way to optimize the output of the workflow. (Image courtesy of DATADVANCE.)

Alexander Prokhorov, head of software development at DATADVANCE, notes that the workflow of pSeven is designed to be intuitive for non-optimization experts performing general-purpose optimizations. He said, “Once the user understands some basic principles, it doesn’t matter what type of problem needs to be solved. The workflow is constructed with blocks and connections that are easy to understand and implement.”

Since engineers won’t have to focus on programming the optimization cycle, they can spend more time on optimizing an actual side mirror, or any other component. This also serves as a means to make the software easier so more users can operate it correctly.

Another way pSeven democratizes optimization software is with the automatic selection of a design space exploration algorithm. “Optimization software users typically need to choose an appropriate algorithm from a long list of complex names and track its effectiveness,” explained Prokhorov. “This is covered by our SmartSelection technology that allows user to focus on the problem itself.”

But when DATADVANCE puts such a sophisticated CAE tool in the hands of those who aren’t necessarily experts in optimization, isn’t there a risk of a user error?

Prokhorov insists that the SmartSelection technology is actually smart enough to protect the optimization newbie. pSeven not only chooses the optimization algorithm, it also assesses how the optimization is behaving after each iteration. If SmartSelection notices that something in the optimization process is going wrong—for example, if the data is not converging properly—it changes the chosen algorithm or makes changes to the current one. By taking one of the most common sources of error in the optimization workflow out of the hands of the user, DATADVANCE suggests its software is safe to use—even by amateurs.

Some of the algorithms under the hood of pSeven include:

  • Single or multi-objective quasi-Newton
  • Quadratic programming
  • Sequential quadratic programming (SQP) with adaptive filter
  • Quadratically constrained SQP
  • Robust optimization
  • Surrogate-based optimization
  • Surrogate-based robust optimization
  • Multi-objective surrogate-based optimization (SBO)

As previously mentioned, pSeven is able to integrate CAD/CAE tools from various big players in the engineering world out of the box. For example, it can connect to:

  • PTC Creo
  • Siemens NX
  • Dassault Systèmes SOLIDWORKS
  • ANSYS Workbench
  • Dassault Systèmes CATIA
  • Kompas 3D

Naturally, there are still a lot of major CAD/CAE tools not on that list that your organization might use. AutoCAD, for example, is conspicuously absent. However, DATAVANCE mentions that Inventor will be added shortly.

To connect to these third-party software options, engineers must manually set it up using a general-purpose integration block. Some of the CAD/CAE software that this block can connect to includes:

  • ANSYS CFX/Fluent
  • SIMULIA Abaqus
  • FloEFD
  • Nastran
  • MATLAB
  • OpenFOAM
  • LMS Image.Lab AMESim
  • MSC Nastran
  • Altair HyperWorks
  •  Altair RADIOSS
  •  and many others

With this wide list of CAD/CAE tools, many automotive design engineers will find their in-house programs compatible—which is good news for those engineers creating a side mirror.

How to Simplify Optimization in pSeven with Surrogate Models

Surrogate model (in green) becomes more accurate to the true model (in blue) as it is based on more and more points. This process is governed by a “design of experiments” algorithm. (image courtesy of DATADVANCE.)

Surrogate model (in green) becomes more accurate to the true model (in blue) as it is based on more and more points. This process is governed by a “design of experiments” algorithm. (image courtesy of DATADVANCE.)

Since pSeven iteratively cycles through a large number of parameters in order to attempt optimization, you can imagine how this could be a significantly large computational task.

This is particularly true when dealing with large and computationally demanding CAD/CAE tools. This is of concern to our side mirror design team that wishes to include the computationally demanding CFD, structural FEA and NVH simulations within the optimization process.

To help reduce the computational demand of each iteration, engineers can use surrogate modeling.

A surrogate model is a simplification of another model. It is pre-calculated to produce similar results to the original model while reducing the computational demand.

In other words, instead of pSeven running the same 3D FEA to ensure the mirror won’t crack, the program can create a 1D formula that will mimic the output of the model.

“Basically, surrogate modeling involves creating a simulation— or any other— model imitation. It is often less computationally expensive while presuming a high accuracy level of input-output behavior,” said Prokhorov. “A surrogate model can also be called a behavioral model or a black box and can be used in many engineering applications or design processes like optimization and system-level modeling.”

The surrogate model is built using a design of experiments that varies the parameters of the true model in an attempt to map it. As the design of experiments iterates through the parameters, it will map a more accurate surrogate model.

Processing Optimization Data with pSeven

 Plot scores the performance of each optimization iteration against design criteria. The idea is to simplify the identification of an optimal design. Plot will automatically update with each iteration of the optimization. (Image courtesy of DATADVANCE.)

Plot scores the performance of each optimization iteration against design criteria. The idea is to simplify the identification of an optimal design. Plot will automatically update with each iteration of the optimization. (Image courtesy of DATADVANCE.)

As pSeven runs an optimization algorithm, it passes the captured information into a project database. Engineers can access and post-process this information with pSeven’s analysis tools.

The graphs, tables, plots and diagrams created by the engineer during the post-processing step will automatically update as each iteration adds more information into the database.

Real-time data capture will let engineers look at the data immediately and start crunching the numbers using the post processing tools as they come in. This parallel workflow ensures that time isn’t wasted as the optimization is being processed.

Additionally, the engineer will be able to stop the optimization iteration cycle or change parameters if something fishy comes through the numbers. If push comes to shove, he or she can also stop and start the cycle over again to implement significant changes.

In other words, the design team doesn’t have to wait for the optimization iteration cycle to finish. This is good news, since it could take days or even weeks to complete. This will save time for that side mirror development team, which is hoping to get the final CAD drawing ready for production in time for the first full-scale prototype.

Some post-processing tools available in pSeven include:

  • Tables
  • 2D & 3D plots
  •  Scatter plots
  • Structured parallel coordinates
  • Histograms
  • Correlations
  • Statistics
  • Sensitivity & dependency analysis
  • 3D model displays

To learn more about DATADVANCE and its optimization technology, check out the company website. You can also request a free 30-day trial license to get acquainted with pSeven.

DATAVANCE has sponsored ENGINEERING.com to write this article. It has provided no editorial input. All opinions are mine. —Shawn Wasserman



from Department of Space Exploration