CAD to Crash in 24 Hours

Engineering analysts are continually being asked to get more done in less time. In addition, they are overwhelmed with learning to use many individual tools that claim the ability to perform a specific task or component of a job. This further drives the need for one integrated suite of tools that can use the same licensing system to complete a process from start to finish. To address this need, Altair has implemented many complete processes that are flexible and configurable with the Altair HyperWorks Suite.

The CAD2Crash24 Challenge:

In 2009, Altair demonstrated the power, speed and accuracy of RADIOSS, the finite element analysis (FEA) solver for linear and non-linear simulations within the HyperWorks suite, by running a 1 million- element full car crash model in less than five minutes. With this being only one part of a larger process for crash analysis, Altair decided to take the next step and utilized the full HyperWorks suite of products to cover the comprehensive process of setting up and simulating a full car crash event, the CAD2CRASH24 Challenge. Altair set an aggressive goal: to run through all the necessary steps of the crash process in just 24 hours. This challenge demanded analysts and HyperWorks to be able to manage thousands of parts, assemblies, and connections (spot welds, adhesives, bolts, hemmings, etc) in one model.

The products from within the HyperWorks suite that were used for this project were:

  • HyperMesh
  • BatchMesher
  • HyperView
  • HyperCrash
  • HyperWorks API for customizations
  • BatchAssembler
  • PBS Pro for Work load management
  • RADIOSS
  • Process Manager
  • Altair’s new Compute Cloud

The project was split into major sections, milestones were established, and analysis specialists were assigned to manage the sections. The team for the project consisted of 4 specialists and, at any point in time, only one person was working on the project.

The team reviewed a few existing crash programs, processes that were followed, the time needed to complete various tasks, and any issues experienced by the engineers. The team noted that the time needed for meshing, assembly, connections and model set-up ranged from three to eight weeks. If the processes could be reduced to 24 hours, a large amount of time will become available for additional product engineering and tuning. Engineers would have more time for analysis of the effects of different parts, optimization or morphing of parts and other design of experiments that were not done because of time constraints. In order to make the 24 hour goal, all parts of the processes were evaluated and only the ones that truly added value were kept.

Current Challenges:

As with any challenge, it is important to understand why it was not attempted or completed before. This review leads many to see that while each part of the process can be done by one tool or another, no other company has a comprehensive suite of tools under one licensing system to fulfill the necessary steps of the entire process. Today, most analysts use multiple products to complete their processes, and this fact tends to result in file transfer and conversions issues, additional time spent to understand and be trained in different tools, and money being wasted to map the various applications to a specific process.

The involvement of global teams, presents even greater challenges in defining consistent processes and agreement on which tools to use. Frequently, each team is left to decide on the tools they want to use, resulting in incompatible processes and costly customizations specific to each team’s usage. Additionally, it continues the practice of developing processes that cannot easily be interchanged among teams and consequently yields inconsistent data management for global teams.

Accessing high performance compute resources is another challenge, not so much from an investment perspective but more so from a maintenance and overhead perspective.

The need for continuous improvement in FEA has led to great advancements in this process resulting in a compressed timeline from months to weeks. With this project, Altair aligned its technology, tools and expertise within the HyperWorks Platform to accomplish something this grand in a very short time.

The project’s keys to success were centered on four main areas that needed to work harmoniously together.

1) HyperWorks process automation provides a comprehensive process automation framework for computer-aided engineering (CAE) automation, process guidance and process integration. This enabled the team to:

  • Capture best practices
  • Leverage domain expertise
  • Improve reliability and repeatability
  • Improve productivity

2) Altair’s experience and support has a simulation driven vision for product development which combines expertise and technology in a creative environment.

  • Altair PD chosen as a development partner based on their success in numerous projects including the validation of crashworthiness and safety of the “Smart Two“ for both Euro-NCAP standards and U.S. standards
  • Achieve compliance with regulations and ratings for two vehicle variants (Coupê and Convertible).

3) PBS job management for the best in class solutions for job scheduling, job monitoring, batchmeshing and assembly.

  • HPC workload optimization solutions from PBS Works
  • Shorten design times, lower infrastructure costs and maximize the utilization of high performance computing (HPC) systems
  • Process simulation studies

4) RADIOSS solver power is a proven leader in scalability, reliability and accuracy. Combined with hybrid MPP, multi domain and advance mass scaling, RADIOSS has the power to handle complex problems that other solvers cannot.

Project and Technical Details:

As the project entered the conceptualization stage, Altair contacted multiple OEMs to obtain real world computer-aided design (CAD) data that Altair could utilize in-house for this project. Ford Motor Company quickly agreed to participate in the project and Altair started the process of acquiring a legacy CAD model.

  • Ford provided a body in white (BIW) CAD data for a generic sedan (see figure 3). Due to confidentially agreements with their vendors and internal departments the instrument panel, powertrain, front bumper, seat assemblies, tires and suspension components were omitted from the CAD data provided. Altair completed the model by obtaining and utilizing part meshes, materials and connection information from a public domain National Highway Traffic Safety Administration (NHTSA) model. These parts were then morphed to fit this generic model.
  • All parts and assemblies were batch meshed in parallel on the Altair Compute Cloud and key parts were checked and inspected manually.
  • Pre-meshed parts (IP, front bumper, etc.) and vendor sub assemblies were attached using Batch Assembler on the Altair Compute Cloud.  All connections were verified and fixed manually.
  • Crash setup and model validation were completed with HyperMesh, HyperCrash and Process Manager for streamlining.
  • Job was submitted on the Altair Compute Cloud with 64 CPU’s.
  • Automated report generation was completed with HyperView.
  • Final Model:
    • Approximately 1 million elements (990,000)
    • Over 9,500 connections

Loadcases

  • NCAP frontal crash was simulated for 100 milliseconds
  • Used FTSS 50th percentile Hybrid III dummy
  • Used IMM airbag

Materials

  • Employed Ford material properties for the BIW and sub-assemblies
  • Generic materials used for all other parts not supplied by Ford

Connections

  • BIW and closures were connected using spotwelds
  • Bolts were connected with rigid spiders

Process Overview:

Altair developed a process to capture the main attributes of a frontal crash simulation. Every element of the process was questioned and any sections that were of little or no value were removed. The overall process and its parts were reviewed to determine if it could be automated and how to achieve the fastest possible turnaround time.

Batch Mesher:

Parallel batch meshing was achieved by using the Altair Compute Cloud. First, the job submission environment was linked to the CAD repository holding all the parts to be batch meshed using the HyperWorks Batch Mesher (see figure 5 below). Once the job was submitted, the parts were processed in parallel based on the number of available CPU’s. As each Job was completed, an H3D file was also generated, which could be dragged and dropped on to the dashboard of the Cloud environment to get a preview of the meshed part or assembly. If the part needed modifications or if it failed, the user could resubmit an updated or corrected part to the queue.

Batch Assembly:

Batch Assembler was used to assemble the meshed parts from the Batch Mesher and the vendor sub-assemblies. The Batch Assembler widget on Altair Compute Cloud requires the user to point to:

  • The mesh repository from batch meshing
  • Mesh repository for vendor sub-assemblies which were provided as assembled meshes
  • Bill of materials (BOM) xml file to extract mass trimming information
  • Material creation and property assignments
  • Weld xml for defining all connections

By working with this information, the Batch Assembler was able to provide a fully assembled model complete with materials and properties assigned.

Model Verification and Fixing:

The model was checked for any weld issues. During this model review of more than 9,500 connections, it was discovered that for a variety of reasons approximately 500 of them failed. The failed connections were mainly hems and some incorrectly positioned welds that were able to be fixed manually.

Crash Setup:

The fully assembled model was set-up for a crash load case. The set-up process was guided using an optimized HyperWorks Process Manager automation template (see figure 7 below). The process started with the mass trimming. The masses recorded from the BOM were compared with the CAE mass at an assembly level as well as a part level. The differential between the masses was applied in an automated fashion to adjust for the mass difference throughout the model. The template also allowed for manual adjustment of mass where the user could select a region to distribute the masses.

A dummy model, seat belts, airbags, contacts and an initial velocity load case with an infinite rigid wall were added to the model. Again, the process was guided by a Process Manager template.

Model Validation:

The model then was validated using HyperCrash to leverage its excellent model checking, seat deformation, penetration checker and model validation capabilities. All errors were checked and fixed manually before submitting the job to the Altair Compute Cloud.

Job Submission:

Before submitting the final job, two to five millisecond iterations were performed as a final error check for fine tuning the airbag firing times, contact parameters, timesteps, and so on.

The job was submitted on the Altair Compute Cloud, using 64 CPU’s and a SPMD executable.

Reporting:

Once the submitted job was completed, a HyperView based automation generated a report of the simulation. This report contained multiple views, section cuts, plots and contours of the analysis.

Results

To identify the critical areas that would define the timing of the complete process, the main tasks were split into nine bins (see figure 9 below). For each area, Altair assigned what it termed the “target” time. In working on the tools for this project, Altair discovered that it could perform even better than the targets in most areas, and modified them to be more aggressive. The team’s new time goals were called the “budget”. The idea was to achieve the budget, but not exceed the target time. Overall, the target was 24 hours and the budget was approximately 19 hours. In Altair’s real world test, the project was completed in 21.

Task

Target (Hours)

Budget (hours)

Actual (Hours)

BIW BatchMesh

2

1

1.5

Sub Assembly BatchMesh

2

1

1

Assembly – BIW (welding)

2

2

1.5

Assembly – Subassembly

4

4

2

Mass Trimming

2

2

2.75

Crash Set up

2

2

3.25

Model Validation

2

2

1.5

Solutions (64 CPU’s) (final iteration)

6

4

6.5

Reporting

2

1

1

Total

24

19

21

Figure 9:  Timing report

Conclusion:

Altair showed that a complex simulation problem such as a full frontal crash that today takes weeks to set-up, solve and post process can be accomplished in less than 24 hours. This achievement was only possible by utilizing the various elements of the HyperWorks integrated platform, version 10.0 with all updates and the Altair Compute Cloud. Similar automations can be easily extended to other CAE processes if well defined specifications are put into place and rigorously followed.

Altair truly provides a unique platform for innovation by seamlessly integrating a complete suite of products with tailored processes and a powerful compute cloud.

► Read the press release

Learn more about the products used in this article:

► HyperMesh / BatchMesher► HyperCrash► RADIOSS► HyperView► Process Manager► PBS Professional

Darius Fadanelli

Director - Customer Service and Process at Altair
Darius Fadanelli

Latest posts by Darius Fadanelli (see all)