Harnessing the Power of Sim

Real cost savings could come from eliminating vehicle- and systems-level tests. Powerful simulation tools may be the only way to tackle the increasing complexity in mobility development.

According to experts, every ten years simulation will enjoy a 1,000X increase in comparative advantage over physical testing. But challenges remain. (Image credit: MATLAB)

Reliance on building and testing physical prototypes of systems and vehicles is giving way to more virtual testing using CAE simulations. It’s one of the most significant trends in the mobility industry, and details of how it is evolving was the focus of a conference, “Engineering Analysis & Simulation in the Automotive Industry,” held earlier this winter and attended by Automotive Engineering. The conference was sponsored by NAFEMS, the International Association for the Engineering Modeling, Analysis, and Simulation Community.

Engineers realize that while physical testing will always be needed, especially for materials properties and validating models, real cost savings could come from eliminating vehicle- and systems-level tests.

The complexity and compressed development time of next-gen mobility systems versus some iconic 20th century aerospace systems—F-104 Starfighter, Atlas missile, Apollo program — is staggering. (Image credit: DARPA/SMS THINKTANK)

To traditionalists, this notion may seem heretical. However, according to Keith Meintjes, Executive Consultant, Simulation for CIMdata, simple economics is giving the industry no choice. Following Moore’s Law, computers continue to grow more powerful and cheap.

“Every five years, computing has become ten times faster; in ten years 100 times faster,” he noted. “We can compute in one second today what it would have taken the Apollo program computers 1,700 years to do.”

At the same time, Meintjes cited the fact that the cost of physical testing is increasing about 7% a year. To compare it to simulation, testing’s cost will increase ten times in the same ten years that computing will decrease its cost 100 times. He asserted that this is an opportunity to be embraced – every ten years simulation will enjoy a 1,000X increase in comparative advantage over physical test.

In this analysis, like it or not, simulation will come to dominate the engineering development process. “With computing getting so cheap it is possible to brute force solutions to some of these difficult problems,” he said.

Challenges remain, however. Various participants in the conference detailed what they were doing to improve simulation, ranging from better physics models to how organizations will need to adapt to get best results.

Understanding models

System models of vehicles that use simplified parametric components exist that are useful early in the development process. Often, test data is converted into a statistical model and used in system models. Multibody products, such as MSC’s ADAMS or CarSim, provide vehicle dynamics simulations. Engines are modeled using system models and full 3D combustion models. But putting a highly detailed, 3D vehicle together is still a work in progress.

A panel discussion at the conference titled “Technology Gaps in Delivering Full Automotive Virtual Validation” attempted to identify critical shortfalls in developing a full 3D, virtual replacement for a physical prototype. While the list of gaps the panel of experts created was comprehensive and detailed, the following is a summary of the discussion:

The more difficult physics models still need to be understood or improved. These include high fidelity combustion models, tire models, battery chemistries, damping of running powertrains, accurate high-fidelity sensor models for ADAS, and vehicle-level durability analysis.

Material properties are fundamental to accurate mechanical simulations. Fortunately, CAE engineers have access to materials databases built up over decades of testing. Unfortunately, every year the industry continues to adopt new materials in the ever-lasting search for lighter, tougher and more durable materials. As one panelist commented: “We have not run out of materials to test. It is difficult to keep up.”

Another key point is that materials need to be tested “as manufactured,” the panelist noted. “It is difficult to couple manufacturing simulations with performance simulations,” he said—the material properties of a tested material may change after forming or stamping. Getting those properties right is still one of the bigger challenges, according to the panelist.

Frank Popielas of SMS ThinkTank: Simulation needs to be used by engineers even if they don’t fully understand the theory. (POPIELAS: BRUCE MOREY)

Modeling interactions between systems at their interfaces. Simulation experts have, quite rightly, developed accurate individual system models such as chassis, engines and transmissions. Each may have the right physics and numerical models to get answers within each domain. Putting those component models together and have them interact with each other in a simulation gets quite involved. Modeling the interfaces is important. Further development is needed to ensure accuracy and to understand how variations propagate.

A culture for implementation

It may be more than technology. “I recommend looking at organization, culture and processes first before you start to talk about technology,” recommended panelist Frank Popielas, Managing Partner for SMS ThinkTank. Adding to that are business challenges that include a global, interconnected supply chain and new competitors. Potential technology solutions such as IoT, Big Data and Cognitive Engineering techniques are advancing rapidly. These factors will challenge a comprehensive simulation model, requiring inputs from diverse sources and keeping track of simulated results.

It’s a tougher problem than with experimental data, the experts noted, since simulations can crank out much more data.

Popielas pointed to new concepts that can help this, such as Digital Twin and Digital Thread. “But the key question,” he asked rhetorically, “is do you know what you need out of those [technologies] and do you have the culture to implement it?”

Another key element in making simulation more broadly useful is getting it into the hands of more users. “It needs to be used by engineers who are not simulation experts, even if they don’t fully understand the theory,” Popielas explained. Such simulation tools need to be as simple, perhaps, as an app on a smart phone.

An ideal use of simulation in a flexible system, illustrating its power, is in developing a virtual world for testing of autonomous driving, according Dr. Ashley Micks, Technical Specialist at Ford. Ford created its aDRIVE Simulation Framework to validate vehicle features at higher levels of autonomy. It is a prototype simulation workflow that flexibly meets the needs of specific projects through the use of open source gaming engine technology.

In a clever piece of systems engineering, Ford’s team exploited an existing gaming engine, Unreal Engine 4 from Epic Games, connecting to it various sensor and vehicle dynamics models to create a framework to test autonomous driving algorithms. [See SAE Technical Paper 2017-01-0107: “Creating 3D Virtual Driving Environments for Simulation Aided Development of Autonomous Driving and Active Safety.”]