Do it right the first time--not necessarily good advice

To learn more about the EcoCAR competition and trends in model-based design, I spoke with Chad Conway, a sophomore Rose-Hulman; Zac Chambers, an associate professor at Rose-Hulman; and Paul Smith, director of consulting services at The Mathworks. (This article is part 2 of 2. Read part 1, “Blowing up thousands of $5000 engines-virtually.”)

Zac Chambers mentioned that he’s blown up thousands of $5000 engines, virtually. What’s the value in that approach?

Smith: Before joining the Mathworks I worked at one of the big three automakers, and we were introducing model-based design at the time. I remember having a debate with my manager. At the time mantra was, “build it right the first time,” and I said, no, we want to build it wrong 1000 times first, because that what engineers are good at - they are good at trying things, learning from that, and then doing it differently and refining it. And that’s the power of the virtual environment-it lets you build it wrong a thousand times. Then when you go to real hardware–when it really matters, when you are going to break real metal or when you risk hurting somebody-it just works.

Chambers: We have exactly one [real] engine, and we can’t afford to break it.

Smith: The aerospace industry has been forced to do this for years and years because airplanes aren’t good test beds. Rose-Hulman is leading the pack in integrating into its curriculum model-based design as a field of study.

What’s the makeup of the Rose-Hulman EcoCAR team?

Conway: The official team leader is double major like myself. He is senior this year; he was at the tail end of Challenge X and has lot of knowledge. The total team has from 25 to 30 members, and right now I’m in the process of recruiting more freshmen. EcoCAR is a really good way to recruit electrical guys who got into [electrical engineering] just for the sake of computers. It’s been interesting recruiting new people and finding new team members, and we’ve got a really good response from the freshmen this year. And especially in year 2 there is a lot of fun stuff to work on, so our team will probably grow quickly.

Chambers: One of the big challenges you face is people leaving the project after three years. They come in their sophomore year and get this incredible reserve of knowledge, and then they graduate. We’ve been working real hard as new people come onto the team to get them up to speed, and fortunately our alumni are very good at staying in contact with us, taking some time during evening to answer e-mails and help the new guys along.

Are all the team members undergrads?

Conway: Yes. When looking for colleges, one of things I looked at is that Rose is primarily an undergraduate school [with only about 50 graduate students of 1800 total] so I would have a chance to do a lot of hands-on work as an undergraduate. That had great appeal, to jump right in and learn a lot before finishing four years of college.

Do other departments, such as marketing, participate in the EcoCAR program?

Chambers: We are a 100% engineering and science school. Our outreach coordinator, Ashley Erffmeyer, has her undergraduate degree from Rose-Hulman in mechanical engineering, and she is pursuing her master’s degree in engineering management at Rose-Hulman, so that’s the closest connection we have to business and marketing. But we are forging some very good new relationships with Indiana State University in Terre Haute, and it looks like we will be able to grab a hold of one of their up and coming marketing managers from their Motorsports division to be our marketing person for year 3.

What you gain here in these sessions at the Mathworks?

Conway: An increased knowledge of the software and how to use it to our advantage.

Smith: We offer a multitrack training program here on our campus this week. We recognize there are some students that have been involved for multiple years, and their needs are different from somebody who is brand new to the program or the software. So there some intro courses Matlab and Simulink and some more advance topics. The most senior consultant I have was a powertrain calibration specialist before he joined the Mathworks, and he is teaching some cutting-edge powertrain calibration and optimization techniques using one of the toolboxes we have called Model-Based Calibration toolbox.

One of the courses was centered around the physical modeling tools we have for electrical systems and power systems. It goes back to a comment Zac made earlier. Here you’ve got undergraduate students with some basic calculus under their belts-how do you teach them the advanced controls that are needed in a hybrid? It seems like maybe there’s a mismatch. But what the physical modeling tools do, like Simpower or SimElectronics or SimMechanics, is they embody the calculus under the hood, so, and I’m simplifying, but you draw pictures of the topology of the system, and the tools solve the equations for you so you can focus on engineering rather than the mathematics. We see this as the next layer of abstraction-you’ve gone from machine code to assembly code to modeling tools where you use blocks to model equations. Now you are dropping an electric motor model into a larger model and just wiring it up and away it goes. You don’t have to put the equations of motion underneath–that’s the next layer of abstraction that’s coming.

Chambers: I fully support this next level of abstraction. In the Challenge X competition we elected to do a power split architecture like the Toyota Prius, and I worked for about six weeks trying to get all the equations of motion put together to make this thing respond dynamically and just accelerate from zero to 60, and it didn’t work right. Then our Mathworks mentor Sameer Prabhu came by and he said, let’s see what you are doing. And he said, let me show you SimDriveline, and honestly in 20 minutes he did what I could not do in six weeks, and we had a working vehicle model by the end of the day.

I have been hearing a lot about early verification…

Chambers: There are a couple of curves you can draw in terms of development vs cost, and the fastest you can find a problem and the earlier you can verify that something is very right or, more important, very wrong, the less it costs you to fix it at that point. And it’s very cheap to fix things in software as opposed to getting to the very end. You get to production and go, oh my gosh that doesn’t work. Early verification is very important to be cost-competitive for a project.

Smith: You have to break it a thousand times before you build it right. And that’s what early verification is really about. It’s building virtual environments to break it a thousand times.

Follow me on Twitter: www.twitter.com/Rick_editor.