Dr. Marc Halpern: I want to preface my message today with some comments on collaboration. When we talk about collaboration, we talk about communication. We talk about defining common objectives. We talk about roles that are aligned to meet those objectives. And of course, it always helps to have the right set of software and hardware in your infrastructure to support the type of process you want to enable.

In the course of our ongoing research on CAE effectiveness, we’ve uncovered four key guidelines that I’d like to talk about that I think are the most important to enable collaborative engineering across CAE and design. Number 1 is that engineers must manage and monitor CAE simulation use. If you want collaboration between CAE simulation and design, never forget that these tools were written for engineers and they speak the engineer’s language. It is up to the engineers to provide the interpretation that a designer can get value out of their use and perhaps even use these tools under the right circumstances.

Over the past decade, there’s been a lot of talk about designers doing CAE directly. Yet, in our ongoing monitoring of our clients, manufacturing companies, we continue to see that the mandate really is for collaboration between engineers and designers regarding these tools, as opposed to the designers using them directly. Even if engineers are coming in to the company with the right type of background with these tools, notably they do know something about CAE simulation technology. They do have grounding in engineering principles; it still takes significant amounts of time for them to become effective with these tools because they must understand something about the behavior and the functions of the products they are designing. So if this is a barrier for engineers, you can imagine what kind of barrier this is for designers.

Even if designers can set up the problems and the simulation correctly, they still have to know how to interpret the results. And that interpretation really does require engineering expertise. So if engineers have to be in the driver’s seat, what is the real challenge? There is a scarcity of qualified personnel. One out of every four engineers is qualified to use these tools. The experts from the aerospace industry indicate that there are not enough experts to go around. So what do you do? How can you leverage their expertise across all of the work that needs to be done?

When you’re involved with a simulation problem, or you’re employing simulation to solve a problem for the first time, you’re in a learning mode, and very clearly the engineer has to be involved in that learning mode. Until they have their arms around the problem – they scope out the details of how to model the problem, how to apply information to make your design decision – not just one, but perhaps many times they have to go through that process. After a while, if you’ve done the proper experience capture, the experience can be reused in what I call the rapid mode. If this type of problem is recurring time and time again, you should be able to reuse that experience.

At the earliest phases in the learning mode, you might be building a CAD model and what would happen is you go through the experience of doing your modeling with FEA, and you learn several things, and then you make some design modifications, and you validate those modifications, and finally you come up with a viable concept. When you’ve done this enough, and if you can, to some degree, automate that process, you can do what we call rapid mode analysis. If your problem is well-grounded, the decision criteria can be embedded into software. Certainly then a designer can use these types of tools. With enough experience with the tools, you might end up developing heuristic rules, where you can embed those directly into CAD software. Perhaps with parametric systems you could define the rules, or even with software that has a knowledge-based capability of setting up making criteria as software logic.

Imagine a company has in a design flanges on parts or flanges on beams. They’re discovering that the flanges continue to fail. It is up to the engineer to decide what type of simulation needs to be done on that flange – perhaps a nonlinear analysis or a fatigue analysis. When they have their arms around that problem and they know that problem occurs again and again, perhaps they can automate this with some type of process-specific customization tools so that a designer can literally step through the process, and they know that the analysis and interpretation of the results is a similar process going on from time to time. Then, the designer, under the mentoring of the analysis expert, can use these tools. Finally, when the designer has had enough experience, and they’ve collected data from the field on these flanges, they may come up with sets of design rules based on this experience. For example, perhaps if the wall thickness of the flanges is between 5mm and 10mm, it satisfies performance and manufacturing requirements. And of course, the fillet radii should be no less than 80% of this wall thickness, or you might have unacceptable stress concentrations.

I believe that this process is achievable. However, I’d like to emphasize that the CAE expert has to be central to this process. They have to define the criteria, the process by which the tools are being used, and they also should have an approval process, by which they can review the types of work that the designers are doing at any given time to make sure that the tools are being applied correctly.

Number 2: Role definitions must encourage collaboration. If people who must work together don’t have a common set of objectives, collaboration efforts fail. I’d like to depict a scenario that I’ve seen in a lot of companies we’ve worked with. All of the companies know that CAE should be up front. So they’ve got a new product that they’re planning to produce. They get started on the right foot, and they will do CAE up front for the conceptual design. But all of a sudden, they have their production schedules. So they get involved with the designers because the designers are primarily responsible for getting packaging out – perhaps they have manufacturing responsibilities – and they’re working because they know they’re being measured on how quickly they can get that prototype out. So they finally build that first prototype and they test it. And what they discover is that the test results do not correlate the CAE simulation that was done up front. They bring this back to their CAE expert, who says "that’s not really what I analyzed." Since the designer is not really familiar with all of the engineering principles, when they make these changes, they’ve actually modified mechanical performance design variables related to what they were working on. The CAE specialist is left in the cold, and the real added value they can provide isn’t delivered because they’re not involved in the process.

Part of this has to do with the performance metrics by which these operational people are being measured. For example, designers are very often measured by how quickly they produce their CAD models, and CAE specialists simply by how quickly they perform a CAE simulation, which is strictly validation. What I maintain is that you have to have new role definitions and new measures of performance in order to enable collaboration. The key ones I’d like to point are that a designer should not be measured simply on how quickly they can build a solid model. A designer should be responsible for the performance of the product and components they design throughout the life cycle of that product or component. By the same token, the CAE specialist should not be responsible just for validation simulation. They should be responsible for optimizing the design for robustness – under sets of unpredictable or uncontrollable performance or manufacturing conditions, the performance of those parts is going to be reliable.

This means that the CAE specialist is actually contributing to the design throughout the life cycle of those parts or products. And if the performance reviews involving the success of the parts and components throughout the product life cycle, and the designers and CAE specialists know that they’re being evaluated according to the same measures, then I can guarantee you that the designers and CAE specialists will be working together much more closely.

Number 3: CAE effectiveness must align with the business objectives. Collaboration will never work unless it has support of management at the highest level. And in order to get that support, you must always be ensured that your goals are aligned with what their business objectives are for the growth of the company.

Key CAE Effectiveness Measures Identified

Cost differences between prototype testing and CAE -- Requires cost tracking and estimation CAE costs and time trends Reduction in the amount of prototype testing -- Can foster unhealthy competition between test and CAE personnel without revised role definitions Correlation of product team performance and the level of CAE deployed -- Requires monitoring of CAE use across product design teams through a central organization -- Correlate CAE use to prototype failures, development costs, development time, and product acceptance -- Requires systematic analysis of data Treatment of CAE Services as a business unit -- Monitor growth or decline in demand for services as internal billings from design groups as a measure of effectiveness Meeting the Challenge: Strategies for Enabling CAE Copyright 1999 by D.H. Brown Associates, Inc.

These are the types of performance measures they’re going to be concerned about. [See chart above.] Within these, I have examples of these measures; CAE impacts all of these. In defining the CAE metrics, if you map those CAE metrics into overall business goals, senior executives and senior management will be sympathetic and supportive of those types of efforts.

And finally, in this competitive environment, you have to have the right infrastructure to enable collaboration. Most collaboration does require the right organization, the right culture, but in today’s type of electronic environment, you really need the infrastructure to be competitive. We’ve been monitoring, for the last 10 years, the performance of these tools out in the field – how well companies do with these tools. And what we found was, from 1991, for an individual CAE expert, it took about 60 hours to perform an analysis. By 1996, because of the types of software and hardware enhancements that have been made, that time was cut by 50% to a little over 30 hours. And based on our reviews of these technologies that we’ve been conducting over the past year or so, what we’ve found from some of the new innovations, we expect the time to reduce by 50% again by the year 2001, to an average time of 16 hours.

This is the type of infrastructure that enhances the productivity of the individual engineer. But how about when you want to enable collaboration? Now you want to get the leverage of these experts across a broader enterprise, or you really need to enable collaboration across your OEMs and suppliers. You need more than the tools delivered today. On the software requirement side, in order to be able to understand what’s being designed, and the types of simulation models being used to simulate those designs, you need remote visualization, and a web browser could be very useful for tracking those models. Since change management is a critical component of this whole effort, you really do need change management tools, data management tools. If you want to do the type of customization I was talking about, you need process templates, creation capabilities.

The key issue on the hardware issue is always networking. In our studies of the CAD industry, for example, in taking a look at assemblies, collaboration of packaging of large assemblies, what we found was that the network bandwidth was always the critical variable, and it continues to be. So high-bandwidth networks are a key requirement to enable collaboration in this space. And finally, since you want widespread access to these models, you want reliable and affordable repository type servers for storing and retrieving these types of models.

I want to emphasize that collaboration does require process practices, or else all of the software and hardware infrastructure means absolutely nothing. And what you really need to do is document your experiences, lessons learned, in order that the information can be reused. This can be done with engineering notebooks. And also, focus groups. Sharing experiences with designers or other CAE experts within the enterprise.

My recommendations for companies interested in collaboration across CAE experts and designers are:

1. Establish mechanisms and practices to capture and reuse experiences.
2. Evaluate and redefine role definitions and job performance measurements so that the experts you want to have work together feel they have a common goal
3. Implement CAE effectiveness measures that can report back to senior management on how well you’re doing and also in the process of developing these measures, the operational managers responsible for these tasks and the senior managers develop a common language of communication
4. Audit and upgrade your software and hardware infrastructure.

 

Top

Rich Smith: My experience is unique in the role I play at my company, in the sense that our network services area works very closely with the engineering and manufacturing departments, as well as the business applications. Because of my past experience, I find that the thing that is most rewarding, maintaining that close link with manufacturing and engineering.

Crane is a leading manufacturer of snack and drink vending machines in hotels, airports, and schools. We’re in St. Louis, MO, and we’ve been in business over 70 years, with about 155 million dollars in sales in 1998. Our machines typically have over 400 parts per machine, so we need to take these machines that live on 120 volts and 20 amps, and put them in a hostile environment, such as underneath school stairwells, or wherever they may live. They need to support themselves with minimum support for extended periods of time and deliver the products reliably.

The systems included in those machines are refrigeration systems, heating for product delivery, anti-pilfering systems, and inventory, which allow us to maintain records of what we sell. There are a lot of electronics that control these electromechanical devices.

We faced design challenges with this new product we’re coming out with to expand our market presence – a new coffee machine. It is a combination of sheet metal and molded plastic components. The emphasis in this design was styling, particularly the molded plastic front. Other design challenges are very common to manufacturing companies, including time to market, and everyone is ambitious about getting from the design phase to production as rapidly as possible. There is an ambitious schedule to bring this line to market.

Another issue is the use of an industrial design firm located in California. We’re in St. Louis. We’ve used design firms in the past, but this is the first collaboration with them. They are separated by distance, obviously, so communicating design concepts with outside firms typically involve travel, making mockup, and going through those iterations of the industrial part of the design. The other challenge was to make the design changes early in the process. In order to meet the schedule and reduce the cost, if you can make the design changes early at the solid model level, and you don’t cut any tools, you save the cost of tooling changes. These concepts are not new, but I want to emphasize that it’s very important.

The strategy and opportunity that came about to develop the product line came about with an offer from Co-Create to be a part of their Fast Track program, which was the evaluation of their software called OneSpace. The software is a web-collaboration tool that allows you to work with many web clients at one time on one single model of data, and interact with that data in terms of viewing it, pointing out features and forms, the functionality of it, actually measuring things on the solid model, and making changes to the model as a collaboration over the web. So we had that opportunity that coincided with the timing of the design of the new coffee machine. Our design firm also is a Co-Create customer.

We started in the first quarter of this year with the whole process, and we’re very early in the product development. The challenges of the strategy were to develop a unique-looking machine for upscale locations. We wanted to make use of solid modeling. We made a commitment to solid modeling in 1996, and to do that on an Intel NT platform, we felt that the technology on the hardware end with the Intel NT platform was mature to the point that it would support the software necessary to do the work, and that the software was there and practically useful and cost-effective to us to be able to go into solid modeling. For a company of our size, there’s no definitive time of when to do that.

So we looked at this opportunity to evaluate the software with this real-world application. In the past, we had worked with solid models on similar design projects, and it very much was typical: We’d take the design process, whether it was internally or externally developed, and take a water-color picture or sketch or rendering, and develop that into mockups in our modeling area and make evaluations on that. And then take that into the actual product design, and go from that and make hard designs, and then transfer that over to our tool makers, and then go through the iterations of what is the most valid way of designing this from the toolmaker’s standpoint.

We were looking at making this a more collaborative process, at least from the industrial design and engineering staff, making that a more parallel effort instead of a serial effort.

The hardware client was a Kayak Pentium II 450 Xeon, with 384 MB RAM with a Fx6 video card, a pretty robust platform. Installing OneSpace is about as easy as installing a web browser. We installed the software with a secure connection, configured all the packet filtering into configuration for two collaboration web servers in less than one hour.

The general theme of our early collaborative sessions was training, and they involved our design firm, myself, and the application engineer. The servers we were conversing with were in Vancouver, BC, and Germany, and the performance was slow. We attribute that to being a temporary configuration for those servers we were dealing with, and I don’t think they had a very robust connection. At times of the day, you’re at the mercy of the Internet traffic. The models we were dealing with were basically a plastic model, which was small (less than 100K) and it loaded rapidly. The graphical interface was kind of a mix of Windows standard and a UNIX style user interface that was easy to use. What was most important to me is that we’re always trying to get faster video cards. Engineers are always trying to get bigger assemblies on their screens and spin them around and dive into them and do fancy things graphically. You need a fire-breathing graphics card. Doing this over the web, I was very surprised with how the performance of the graphics was.

Generally we spent about two hours training total, and we covered the major topics of manipulating the geometry. After that, we gave a presentation to engineering management and a few engineers, and told them that this was the direction we were taking.

Sketches and drawings were used initially to evaluate and create the concepts. It included a trip to California so our design firm could get the design intent. The first sessions were more of a presentation format, where we were the audience, and the designers were the presenters. We always had the AE, the third person in the loop, help us get through the learning curve issues we had with the tool.

The objective for these first sessions was to review and refine down to the final two concept design selections that were actually modeled, particularly issues of styling, the look, the general aesthetics of the design. The screen captures that were made were useful for internal distribution to management.

Those two concept designs were selected with no prototypes and models. It was definitely a deviation on that part from past methodology.

Our vice president of engineering and product development commented that he was impressed with the involvement of all team members in the design phase. It wasn’t just a select group of people that got to travel to California and be part of the design process. We had engineers, designers, technicians, everybody huddle around the speakerphone and talk about collaborating on this design. They were involved early in the process.

There will always be a lag when you’re transferring large amounts of information across the Internet. You’re moving things on your screen, and you’re interacting with what you see, but your audience is two seconds behind. That can be hard to deal with.

Our later sessions were more technical detail sessions, more component-level form, location, and function of the parts, and it looked for fit, form, and function type issues with the engineers. It was more of a true collaboration. Our engineers were involved talking with the designers, and trying to make sure that the design intent didn’t violate some manufacturing parameters that we understood. The server location was in a different location, and it was a much better Internet day, so performance was significantly better. This was a bigger than 5 MG model, and that’s a compressed model, which expanded out to twice that size, so the performance was surprising. The biggest thing is that the large number of issues that were covered rapidly by our engineers and the designers in California, the number of issues that were covered was an overload of information.

Our engineers were impressed with the number and variety of design issues that could be covered in one session. There was still a problem with the images lagging behind the conversation. We’re too early in the project to come up with some hard, objective figures that validate how much faster this tool allows us to go through the process. In general, the engineers agree that it enhances the communication and decision-making process tremendously. From this part of the design phase, the tool maker-design engineer-industrial designer collaboration would take the first part of the design phase and put that in parallel. That also includes internal collaboration with manufacturing and design engineering.

Top

 

Gregory Roth: The trends in the product development process arena basically cover reducing time to design, develop, and prove out; to reduce costs, waste, and warranty returns; and improve quality. While these are trends we’ve seen before, what have changed are the motivation and the urgency to implement them. In the past, this is usually done as a bean-counter exercise to increase return on assets and investments and to improve and increase the process. Now the main focus is to get the products out faster to meet the customer’s needs and wants. Traditionally, much market research is done, and that goes into the product development process. It takes years for a product to be developed, and often, by the time the product is developed, it’s no longer needed or the needs have changed, so it doesn’t match the customer’s needs.

In the automotive industry, comparing product development cycles, it takes about 48 to 37 months to develop a platform. There’s a lot that goes into that, and there are some approximations, but generally what happens. There is now an effort to reduce that in half to 24 months. That is a major compression of time. How do you do that? Streamline the product development process. One technique to do this is to embed and integrate CAD/CAE functions earlier in the design cycle. In the past, typically these tools like CAD and CAE are taken and thrown into the product development process mix and everybody stands back and hopes a miracle occurs and things improve. In fact, in a lot of cases, it slows down the process. Translation problems from CAD to CAE; models must be built more robustly, and people think because these fancy tools are being implemented, things must be improving when they aren’t.

The basic requirements to achieve this include a well-defined product development process. The point is that you have to have a very good structure to ensure that certain protocols are followed and not bypassed. We as engineers know how to get around problems. What we’re trying to do is install a very well-defined process. You must be able to do basic level CAE analysis of concepts early in the design cycle to filter out performance concerns. The point isn’t to fully prove out concepts – it’s just to weed out bad performance features early before they get embedded.

You must have CAE codes that can be used by designers and engineers. The key to this is if you’re going to use this type of capability, it has to be easy to use, learn, and remember. Another thing making much headway is robust interface between CAD and CAE. The ultimate goal of all this is to achieve first pass success. What this entails is to be able to prove the part, design it, have it go through its first testing, and pass all performance criteria. This is the Holy Grail of the industry. It would be great if we could always achieve that.

In the traditional product development timeline, most of the work is done just prior to release. That’s when most of the problems are encountered in manufacturing, product prove-out tests, and there are a lot of fires. That’s typically how CAE got started, to help with these last-minute plant closure problems when everybody is trying to solve a problem at whatever cost. The trick now is to push that work up front. If you do your homework up front, you’ll have less overall work.

This requires a paradigm shift, because the way we typically operate is to get the first prototypes out the door fast because we know we’re going to encounter those problems.

There are three major areas in the design process that implementing CAE up front can help address. It can help reduce the need for numerous redesign prototype-building task cycles, help identify design issues up front, and reduce prototype and test costs.

In the proposed process, the design process runs in parallel and interactively with CAE activity. In this process there’s an interactive feedback loop where the CAD engineer is working almost directly with the CAE engineer and engineering in general, and things are being analyzed on the fly, kind of like a virtual prove-out mechanism. What you’re trying to do is get these ideas proven out before they even go downstream. What happens is it then goes to tooling, and in the best of worlds, it passes the tests.

One thing you should notice is that in the early phases with the CAE design process up front, the design phase will take longer because your models must be developed, and they’re changed, and you’re getting information back that this feature is no longer good.

As you reduce the iteration, you’re actually saving significant time and money, and this is quite obvious if you think about it. Every time you do design iteration, it requires failure analysis and when the part fails, it has to be redesigned and has to be rebuilt with new prototypes, maybe analysis done, and then retested. All these steps add up to huge chunks of time and money. The major concern is that you’re saving engineering and prototype time, and testing time.

In the past, much of the work in this part of the industry has been in the pre-processor and CAD arena. Up to a 55% reduction has been achieved in reducing the time to get to the point of CAE, getting the model ready. Now with the improved meshable CAD solid modelers, these tools can mesh and get everything ready to go. Now the focus is on CAE. Newer tools allow a quicker turnaround to a viable concept.

In the traditional sense, in "In Series" CAD/CAE process models, the design is handed off to the CAE group, and all the effort stops on the design side. The analyst does his work, passes it back, and all effort begins again. They actually can run in parallel. Some non-critical areas can actually be developed in parallel with the analyst. Likewise, when the analyst has an idea of generally what’s going on with this part, he can pass it back and give a heads-up to the designer. This isn’t rocket science.

In using CAD at the end of a typical design process, in other words, the designer has done all his work, he’s handed it off to the CAE person to see if it makes any sense. If CAE is done up front, especially by the engineers, now you’re getting economy of scale, where more people are saving time across the community, rather than just discreet locations.

Product designs traditionally are funneled into a CAE analysis group. There are often insufficient resources to do the work, analysis time can take quite a bit of work, and also there is a general perception of fear of a bad outcome. You have people who don’t want to hear bad news. So there’s a tradition to bypass the group, and then these products just go on their merry way the traditional way.

Product designs come in and the basic CAE analysis is done by engineers and designers. The intent here is to filter out the bad ideas early. Most of them would then be evaluated and some of them would go to the CAE analyst for further study, and some would just be checked by a monitoring process for quality, and that there’s no major risks being taken. More problems are being evaluated and filtered, so can bad ideas and concepts get through? Yes, of course. But is there a better chance that you’ve made your best try to eliminate them? Yes, very much so.

The key issue is engineers or designers doing analysis work versus analysts. That’s a hot topic these days. While I believe the best scenario would be to have an analysis core of thousands of people all in engineering work that could be done in a company, being in the industry, I know this is not very likely. There are resource problems; there is cannibalism in the company.

So what’s needed make the process flow work? First, an easy-to-use CAE type code that will allow for quick "what if" studies at the engineering and designer levels, and that is intuitive and consistent with other codes. It must be easy to remember, not allow for serious pitfalls, and should provide robust solutions. It must be compatible with CAD solid modeling codes; it must fit in with the standard engineering/design processes and cultures; be portable to different hardware environments, and be documentable.

Top

 

Doug Johnson: I think we make no mistake that the collaboration theme is booming in popularity. It might become the cell phone of the next decade. It’s not going to replace CAD or FEA or PDA, just as the cell phone hasn’t replaced the traditional wired phone. In fact, for all the people you see with cell phones, only 8% of the traffic last year was carried wireless – the other 92% was still in the traditional phones.

So what you might see in the collaboration world is a real complementing with collaboration capabilities to the tools that exist out there today. You might think that a leading finite element opportunity and a leading collaboration provider might be working together.

The technology for collaboration continues to move forward in a variety of ways. One of the most key elements that will continue to move forward is network capacity. It’s been a bottleneck for years and has slowed down the adoption of multimedia and digital video that were pioneered in the 80s but never really got off the ground because the network capacity wasn’t there. Today, smart programs like compression algorithms for digital video or the algorithms that are used in collaboration products are allowing orders of magnitude improvement over the available network bandwidth that’s there today. These algorithms are what are making the current wave of collaborative products and digital video products to become practical for engineers to use. We see a revolution occurring in network capacity. It’s the same kind of revolution phone companies and primary carriers are spending enormous sums as a percentage not like we’ve seen since the 1960s and 70s when we got into the long-distance telephone craze. Before then, you had to have an operator make a call for you.

Today, digital traffic is bigger than analog traffic, for the long-distance carriers. Today, the demand is growing at 60% per year. When that happens, technology and high-tech companies respond, and put in place the kinds of capacities that are necessary. We’re also going to see networks much easier to use. About 10 years ago, the first time I got into a collaborative session between two personal computers located about 1500 miles apart, we scheduled a one-hour meeting and spent 59 minutes trying to get our network connection working. We spent the last one minute conducting our business without the network connection.

There is continued work in directories and automatic routing mechanisms. Even today, I’ll bet most of you think it’s easier to address an e-mail message than it is to write out long-hand the address on an envelope, put a stamp on it, and take it down to the mailbox. That’s a long way for a networking and ease-of-use in networking over where we were 10 or 15 years ago.

The next interesting revolution will have to do something with data. The first computer I had had a 5 MB disk in it. The laptop I just got has a 6 GB disk in it. That’s a 1000-fold improvement in 20 years. We’ll see that kind of explosion continue, which will mean that 10 years from now, you’ll have a half-trillion bytes in your computer, maybe even in your laptop computer. You’ll be able to save everything imaginable. You’ll have more memory in your laptop than all the mainframes in the world had in 1980. This explosion of data will mean that we’ll have more things on which to collaborate, lots more information associated in the enterprise, and more to talk to our partners about.

I see this being a very promising and exploding future, and companies like those here would like to continue to bring those capabilities to all of you.

Top

 

Paul Bemis: The longer you wait to find the problem, the more it costs. It is not a linear relationship, and in fact, it’s more exponential in scope. Tools that are easy to learn, use, and remember are important. There is a new class of user now coming to the forefront of the analysis engineering simulation community, and that is someone that we generally refer to as a periodic user. Up to this point, analysis has been something that’s been done by very sophisticated, well-educated people who typically have a master's, if not a Ph.D., in engineering simulation studies. These people typically go to class for a quite a period of time to get to know the products, and then use the products daily so they know the product. They are rather complex products.

There is a new class of user coming along, however, that does not have that class of experience, nor does it have the time to get up that steep learning curve. This is the periodic user – someone who needs to be able to get to know a tool without really going through much training, if at all, or maybe online. They also need to be able to reintroduce the tool after a period of time and be able to still use it.

ANSYS 5.5 aims at the educated user who knows about singularities, who knows about stresses and strains and large deformations, and who knows about these issues, but is in fact using the tool quite a bit. DesignSpace 4.0 is for the periodic user, someone who is more of a generalist. Not a specialist. Someone who is only going to periodically touch this piece of software. In the design engineering community you have people who are not all that educated in the academics or theory associated with the tools. These are degreed engineers, but they are not advanced users.

We believe it is not enough just to develop the tools for people, but also the collaboration must exist so these teams can work together over both organizational and geographic boundaries.

The advanced engineering community may be a consumer or supplier of technology. In the supply chain analogy, the analyst might be the consumer. The design engineering community might, in fact, be the supplier. What we’re trying to do is put collaborative processes in place so that the advanced analyst community can help the design community determine what tools they are using, and configure those tools. Much of what we do in the design community is generally referred to as reanalysis. We’ve done this design 10, 15, 100 times, now all we’re going to do is vary the base design. The issue is where are the limits? How much can I change wall thickness with respect to radius before it starts to fail? This is knowledge that can be put into the DesignSpace product and ultimately that knowledge is configured and delivered by the advanced community. It’s the concept of capturing knowledge or best practices that have been laid out in the advanced and engineering analysis community, and encapsulating them into a product that can be used by a broader audience, without the sophisticated training.

We do this to some extent today in DesignSpace, using concepts of wizards. With a wizard, you put the floppy disk in, hit the go button, and it steps you through a sequence of events. When it wants input, it prompts you. It says, "Now I want you to tell me where the directory is that I want this product to go into." With the wizards in the DesignSpace environment, we step you along and you get to a point and we say, "What kind of analysis do you want, what kind of materials do you want to choose? Where are the loads that you’re going to put on?" And we see this in the future where the boundary conditions and the loads, and everything that is sensitive has been pre-set and what the design person is able to do is work within the constraints of variability that the analysis community has laid in place for them. This is what the concept of advanced controls means. It means capturing best practices in the analysis community and delivering them in a product to the design community that is constrained by the physics that have been pre-calculated, and manifesting it in a very easy-to-use manner.

These collaborative products cannot succeed without high-level management support, and they must be delivered in a staged fashion with training. The people I am aware of who are implementing this are doing it in a staged fashion – they’ve actually named their engineering community or put their analysis community into some level of gradient: Level three is an advanced engineer; two is a mid-range analysis engineer; and level one is a design engineer. They bring them back for periodic training, and as they train, they deliver more simulation technology to them, but only after training.

Engineering simulation is moving out of research, where it has been, and into production. This is a fact. This is happening whether we like it or not. The fact that it’s happening is a direct result of the fact that hardware costs are dropping and now you can get done what you couldn’t get done before. When you give a design person a 450-MHz desktop Pentium III or II, they will now have the capacity to do things they couldn’t do before. We have the opportunity of getting out in front of that and becoming part of it and helping to bring it along, or dealing with the consequences of playing with those tools without our help.

Mechanisms are now available to help guide this process. We are putting in place mechanisms such as wizards and other tools to allow guidance to be provided to the design community during this process under the control of the more advanced audience.

Top

Q & A:
Panelists field questions from the audience...

Question: I see how this web-based collaboration would work when the context is basically aesthetics and physical design look and feel. Would you comment on how that can be expanded into the analysis side where on the one hand once you make a change to perform the analysis it is time-consuming, and when you get the data, you also have millions of data points that you have to analyze and discuss with the team. How can that kind of magnitude of processing be handled on a web-based system?

Paul Bemis: When the design engineer gets done with their design and generates a report, and the report says, by the way, you failed, it could be introduced by something that the design engineer does not understand. But today, for them to get the high-end analysis person involved, they’d have to send an e-mail or phone, and say, could you please look at www dot whatever, at my report, or here it is attached as an e-mail A much better way would be to use OneSpace, as an example, to share with the analysis person the actual geometry of what they’re working on and get some advice about it, either after the analyst has seen the report, or prior, and clearly after the analyst has done a little more evaluation. They can both get on the phone over boundaries, and say, here’s what I think you ought to do.

Question: What do you feel is a good ratio for designers to CAE engineers, and do you feel that CAE engineers should be the gurus of the 3D CAD modeling system?

Doug Johnson: Last count, there are at least a half-dozen CAD systems that one would have to become an expert on. Today, with all of these CAD systems being so popular, and with supply chain integration being so tight, you now have to be an expert on multiple CAD systems. I think we’ll see a trend more that people are experts in their areas of expertise, and we’ll see collaboration and people working together in order to bridge those two worlds.

Gregory Roth: I think that the CAE analyst needs to be quite conversant with the CAD tool of choice of the company. I agree that there are a lot of different CAD tools out there, and it’s difficult to take all of them on, but there should be some standardized tool within that segment of the analysis process. Everything will be analyzed in a particular tool of choice. The CAE analyst, to be effective, also has to be able to propose changes and often do some of those changes quickly to see if they’re valid. You don’t want to bog down the CAD designer’s time with a couple hundred what-ifs. If you can do them fairly quickly and churn them out, as an analyst, you can then go back to the designer and say "here’s your best direction to go."

I think there’s a difference between "want" and "what is." The ratios out there can be anywhere from 10 to 1, to 30 to 1. The key issue is how many analyses are being requested. I can see one analyst supporting five designers, depending on if there’s actual analysis work in each of those teams.

Marc Halpern: If you remember when I was talking about the ratios of experts to practitioners, and it ranged anywhere from 4 to 1, to 20 to 1, that’s not an accident. The reason those ratios exist is because in the aerospace industry, structural analysis is essential to design. In the automotive industry, there are certain issues related to crash-worthiness, noise vibration, harshness, and other related analysis issues. Yes, you need your CAE experts, but there’s a lot of design going on where CAE really doesn’t add value, such as interior design, styling, etc. So the ratios are based on the need. And what I would say is yes, although all industries require more CAE being performed, it’s really a function of the industry and the type of design being performed.

I don’t think that there’s going to be much of a shift in the ratio of analysts to designers, simply because of the learning curve required to become an analyst. Right now, industry would like there to be more analysts to designers, but what I suspect will happen, as these collaborative tools become more effective, we’ll be able to leverage analysis among a greater number of designers, so it may be that the ratio will decrease.

Question: How are processes truly being integrated into CAE software?

Paul Bemis: In the design area, there are opportunities to do a few things. For one, geometry has become relatively standardized. We now have STEP and IGES. Geometry itself went from a proprietary environment where CAD people had control of it and were unwilling to release it, to more of a standards format today, where you have STEP, IGES, and other mechanisms, and the CAD people are willing to release it. This has improved the transfer of geometry into analysis, and has made that much more highly automated than it ever was before. Today, the opportunity to attach and read with a native API the CAD graphical data or geometry data is in fact quite good.

Furthermore, there are other areas of automation in the design area that offer the opportunity to take the manual work out of it, and automate it without giving up accuracy. What we have done with the base is add things like wizards, that allow the opportunity for the design engineer, who may not remember how to go through the setup and solve a relatively fundamental problem, to step through it using a wizard or other procedure.

Marc Halpern: I do concur with Paul on the ease of sharing CAD data across the CAE simulation tools and the CAD system itself. I’d like to emphasize a little more some of the customization capabilities within ANSYS. Paul alluded to the capabilities of DesignSpace to introduce wizards. I’m seeing a trend across the industry, where, through APIs and other mechanisms, there are open architectures that enable customization such that a designer could step through the process. I think there is going to be a set of generic tools. I believe there is a tremendous opportunity for new cottage industries taking these framework or infrastructure tools, and applying them towards industries - automotive, electronics, aerospace, consumer products – where experts from those industries will be able to take these tools and set up process-centric CAE capabilities. So literally, they’re taking the frameworks and building tools that they would then sell within their specific industries. It’s going to be a whole new generation of software within what I would call an emerging cottage industry.

Gregory Roth: In the industry, the CAE tools, CAD tools, and a lot of software tools have matured and become quite robust. At this point, it’s not a matter of additional functionality. The real issue is applying it – getting these tools in the hands of the people that are going to use them. Too many times these tools are actually just thrown into the group, and they expect to see all these miracles, and they don’t achieve the value add. The trick is to get them actually used in the process, and use them wisely. That’s what’s missing in all the situations I’ve seen. The tools are there. Even bad tools can be used effectively to improve the process and reduce iterations. You have to put it in the process – embed it.

You may have a designer on a tight timeline, and you say, "I need to do a study to make sure this part will meet your criteria." They say, "I don’t have time. I have to get this part out. I’ve got to get it prototyped and out the door." And that’s got to change. These other tools will expedite that. Collaborative tools and all these other things are just to support. But they have to be implemented. That’s where the last problem is in this whole process.

Bill Neill: At Hewlett-Packard, our Design Chain Engineering approach to collaborative environments includes a great deal of integration of the tools with the design process in a consultative fashion. We recognize that no matter how good the tools are, if you don’t know how to apply them, it’s too frustrating. Our clients really want to know how to implement, not just install, tools into their processes.

I encourage you to go to our web center, because that is where we’ll be scheduling and announcing our future collaborative design audioconferences, as well as our design managers’ conferences in the future. We’ll be sharing with you new ideas, and allow you to ask questions, ask for information, and help you to implement and integrate your collaborative environment.

For more information on collaborative engineering, and for the dates and places of upcoming interactive audioconferences, visit the HP Design Chain Engineering Web site.

---