The growing complexity of electronic systems is prompting a major change in design processes. Whether they’re probing deeply into fuel combustion or integrating many control modules used in passive entry systems, engineers are doing more development in virtual environments.
At SAE's 2010 Convergence, engineers from Lear, Tata Technologies, and General Motors described their use of modeling and simulation tools. Though they have all been using the techniques for years, their approaches are constantly evolving.
“We’re moving towards more of a virtual desktop environment, with tools that make it appear to the engineer that he’s in a vehicle running tools used for calibration,” said Karla Wallace, Senior Electronic Engineering Manager for controls and calibration tools at GM. “Engineers can also use simulation in a batch mode, running many simulations and getting data back in an integrated format.”
Currently, GM researchers run batch simulations for powertrain designs on high-performance computers, not desktops, she said. When GM moves to desktops, those large systems will be used after the research team has a high degree of confidence in a design. These grid computing systems run thousands of test runs, setting the stage for real-world prototypes that can test critical functions and verify the simulations.
Models also helped Tata engineers design an entirely new vehicle, the Nano. One of the biggest benefits to the team designing the powertrain was the ability to use automatic code generation to trim time and reduce errors.
“Before we used model-based design, we hand coded in C, which took a lot of time and people and had a lot of bugs in the code,” said Debarsish Hazarika, Solutions Developer at Tata.
Wallace estimated that using automated code generation instead of hand coding gives GM a 30-35% efficiency gain. Lear provided similar figures for design, adding that the improvement rose to around 60% during validation.
All three companies used tools from The MathWorks to let engineers manage more complex designs in less time. In powertrains, the challenge is to eke more out of each drop of gasoline, while in body control the challenge is to get many controls to work together seamlessly.
“Passive entry may involve six ECUs, and all of them have to be validated,” said Jason Bauman, Engineering Manager for Lear’s Systems Integration Engineering group. “Almost every body feature is like that.”
These modules are usually developed by individual design teams, which can cause problems for those charged with integrating the functions. “When you have independent developments, there’s room for errors,” Bauman said. “We use encryption, so we have to be sure each side is developing it the same way.”
Another commonality for the three design teams is he need to respond to ever-changing requirements. Few programs remain static throughout the lengthy design cycle.
“Things go very fast; when you satisfy one function, another requirement comes up,” Sarkar said. “It was difficult to meet those changes with our earlier approach. Modeling makes that easier.”
Advances with design tools represent another major change in design processes. Tool providers are making it easier to integrate different types of equipment.
Moving from one segment of the virtual world to another is becoming easier. “There’s increasing compatibility with design tools and hardware-in-the-loop (HiL) equipment,” Bauman said.
The HiL tests for Lear’s most complex systems can take as long as eight weeks. But the company does most of its simulations on desktops. “We usually select a subset, managing the values we’re looking at so they can be run in eight to 10 hours,” said Jinming Yang, Principal Engineer for Systems Integration Engineering at Lear.
As design teams throughout the supply chain use more modeling and simulation, the availability of models is also growing. For example, libraries of models were not readily available at the outset of Tata’s design, but their availability is building rapidly.
“Things started gradually, with a few islands of automation, but now everything has been mushed into a stew,” said Prasanta Sarkar, Engineering Manager at Tata. That makes it simpler to combine autocoding and other new techniques with older design processes, he added.
Though models and autocoding now dominate the design process, programmers still have to step in. In the latest powertrain development for the Volt, nearly 100% of the software for many modules was generated automatically, Wallace said. “There’s always a bit of integration work that’s done by hand,” Wallace noted.