Digital Engineering: Functional Virtual Prototyping, Part 2

The current bottleneck in globally competitive product design is the creation, instrumentation, testing and modification of system-level hardware prototypes. Traditional CAD/CAM/CAE methodologies do not provide a good means to break this bottleneck. New products in DMU, functional virtual prototyping and virtual factory simulation provide system-level counterparts to traditional component-focused CAD/CAM/CAE solutions and allow for breakthroughs in speed, cost and quality for new product design.

Validate

The importance of accurate validation of system-level models and modeling assumptions should not be under-emphasized. Functional virtual prototyping can yield a wealth of information to support rapid decision-making. It is critical to ensure that this information reflects the actual operating performance of the new product. The validation phase is not overly difficult, but often is not approached with as much rigor as is warranted. The companies with the very best records in making effective use of functional virtual prototyping have invested significant time and resources in building a validation library.

This library defines how models need to be constructed so that simulation performance results can be easily compared with test results. The library catalogues past validation work and summarizes modeling assumptions that have been validated. And the library is integrated with an internal product data management system so that data and information is readily available.

Good simulation tools and processes can greatly facilitate the validation process. For instance, a simulation software product that provides quick information on design sensitivity to various parameter changes can pinpoint areas of a model to be investigated to improve correlation between experiment and simulation. Also, as stated earlier, it is important to "simulate as you test," meaning that the same testing and instrumentation procedures should be used both in the physical and virtual test process.

In a typical validation process, the physical and virtual models are tested identically and baseline results are derived. The results are compared either manually or in automated computer-based fashion. Discrepancies are noted in specific performance results. Design sensitivity analyses are performed on the virtual model to identify design parameters or model areas that significantly contribute to the performance results that do not correlate. Then, a mixture of manual changes and computerized non-linear optimization techniques are employed to make changes to the model parameters identified or the test procedures until acceptable correlation is achieved and the model is validated across the different tests.

In a recent validation process, a virtual prototype of a Formula 1 race car was tested on a virtual representation of the Imola race course in Italy. A virtual driver model learned the course and was used to duplicate the behavior of a real race driver. Lap times were compared between the virtual car and a real car driven by a professional driver. The virtual prototype delivered a lap time that was within 0.1 seconds of the real driver. More importantly, a comparison of the vehicle lateral acceleration levels for the real and virtual vehicles showed outstanding correlation.

With experience, modeling assumptions can be correlated and catalogued. This allows for the automated creation of new product virtual prototypes that can be utilized with confidence - without the need for the construction and testing of an initial physical prototype. Physical prototypes are still needed downstream in the process to verify the design prior to production.

Refine

Refining a virtual prototype involves two aspects, refining the fidelity and breadth of the model, and refining the product design itself. As the design process progresses, the virtual prototype models will be relied upon to investigate more and more functionality. Initially, it may be enough to understand speed of operation, the space envelope of operation, the total power requirements, etc. This understanding can help drive component topology selection and overall design parameters.

Subsequently, as issues of comfort, noise, vibration, and durability need to be addressed; the virtual prototype model will need to be enhanced. It is important that the virtual prototype can access sub-system models of varying complexity and model fidelity. For investigations of more complex phenomenon, it will be important to enhance the model by replacing more and more of the rigid sub-system models with flexible counterparts. Models of the fluid power systems that interact with the mechanical and electrical components will need to be represented. Automatic control systems that alter the operating performance of the product will need to be accurately represented. These are all natural extensions of the initial virtual prototype. Component and sub-assembly models of varying complexity must be constructed in such a manner so as to be quickly interchangeable.

For instance, when investigating engine performance in a vehicle, it may be important to exercise a fairly detailed engine model that includes a flexible valve-train with cam-rocker contact. However, if vehicle dynamics is the main focus, the engine model can be effectively replaced with a much simpler representation. A template-based design system that allows for quick and easy exchange of various subsystem models is of paramount importance for effective design refinement.

Refining the actual product design is where functional virtual prototyping delivers real value. Once a validated, system-level virtual prototype has been created with interchangeable sub-system models of varying model fidelity, a very rigorous design refinement process is within reach. First, a complete battery of product functional tests are defined and finalized. These will be the virtual tests used to sign-off on the new product design. Next, potential design changes are identified in terms of component parameter changes, system topological changes, and potential manufacturing tolerances.

Performing the complete battery of selected tests with all combinations of parameters and tolerances is both impractical and unnecessary. Statistics-based, Design-Of-Experiment (DOE) methods are used to consider the entire universe of combinations of these changes and determine what combinations of these parameters must be simulated in the battery of virtual tests in order to give a statistically relevant prediction of the envelope of operating performance. The identified combinations are then simulated using both the virtual prototype and the battery of virtual tests, and the results are exported to a simple spreadsheet.

Curve fitting of these results allows for quick spreadsheet assessment of any potential design changes within the specified range. This approach facilitates rapid, knowledge-based decision-making in product design review meetings. Requested changes to system design points or parameters can be immediately assessed for their impact on performance, safety, durability, comfort and cost. Faster decisions and a better balance of competing functional performances result from this approach.

Automate

The aforementioned approach leads to significantly improved products at lower cost. To simultaneously reduce the overall development time, it is necessary to automate the virtual prototyping process.

This phase requires close cooperation between designers, development engineers, analysts and test engineers. Although this cooperation may not be easy to effect, the payback is significant. Ford Motor Company recently released results of applying an automated virtual prototyping process to three new vehicle programs. It demonstrated $40 million savings in engineering costs and more than $1 billion savings in manufacturing changes through reduction of late cycle changes.

Automating the process can be done very effectively in companies that make the same type of products year after year. It is much more difficult in organizations where radically different products are created over time.

Once the engineering analysts have worked through a few virtual prototyping cycles and helped create validated models that can be exercised through the parameter changes requested by the development engineers, the virtual prototyping environment can be automated through the use of a template-based design system. It works as follows. The engineering analysts catalogue:

• parametric topologies that are normally considered for new products.
• typical parameters that are varied in the design process.
• the range of validity of various modeling assumptions.
• the different levels of sub-assembly model representations required for various levels of fidelity.

An analyst then utilizes a template-based design system to create a series of design templates that can be used by the designers and development engineers to evaluate design changes. These templates automate the creation of the sub-assembly and system models. They allow input only within the range of the validated modeling assumptions. They hide the complexity of the model by only presenting the parameter changes that have traditionally been varied. And they automate the selection of sub-assembly representations in accordance with the type of test or performance output that is requested. If this is integrated with a product data management system, it allows for quick comparisons of new design performance with previous designs or competitive target designs. The analysts publish these design templates internally for use throughout the design process and even later in field troubleshooting.

This makes it possible to have an enterprise-wide virtual prototyping process where any engineer in a vehicle manufacturing organization can access a validated model of any previous vehicle or current new vehicle design. They can replace sub-systems, alter vehicle design parameters, add automatic control systems, and run the vehicles through standard test procedures to understand the effects of proposed changes. This is extremely powerful in stimulating creative input and capturing corporate design knowledge.

Realization of the Functional "Digital Car"

The confluence of technologies - such as DMU and functional virtual prototyping - are enabling the true realization of the functional digital car such that total vehicle performance can be evaluated and optimized on a computer. It is now possible to combine accurate mathematical model representations of chassis sub-systems, engine and driveline sub-systems, and body sub-systems to create a full virtual vehicle. The performance of this total vehicle can then be simulated in a virtual test lab environment or on a virtual test track and replicate real-world behavior. This digital car can then be integrated with hardware-in-the-loop simulations to investigate the real-time performance of the vehicle with real sub-systems such as ABS and TCS systems.

The functional digital car can be evaluated across multiple disciplines by replicating standard test suites and modifying typical vehicle design parameters throughout their acceptable ranges. Results can be shared globally among design teams using the Internet by means of Design-of-Experiment response surfaces, plotted results, and performance animations.

This, in essence, is the technology that will allow vehicle manufacturers to realize the projected dramatic reductions in cycle times while maintaining and increasing vehicle performance, safety and longevity. Effective implementation and automation of functional virtual prototyping can provide a significant competitive edge in the market.

Managing Risk More Effectively

The result of functional virtual prototyping is that manufacturers are much better equipped to manage the risks inherent in the product development cycle. Traditionally, the amount of information concerning the actual performance of a new vehicle was fairly low throughout the vehicle development process until the prototype and assembly stage. Then, behavioral information increased, and risk could be reduced through effective design changes. Unfortunately, late cycle changes are very expensive and prone to error. With functional virtual prototyping, behavioral performance predictions are obtained much earlier in the design cycle, thereby allowing more effective and cost efficient design changes and reducing overall risk substantially.

Reducing risk in this way has multiple benefits. It leads to vastly improved designs, limits warranty and liability issues, reduces late cycle changes and costs and helps to reduce overall development time.

Technology Enablers and Limiting Factors

A frequently asked question is "why haven't technologies such as DMU and functional virtual prototyping been applied extensively before now?" To understand this, it is important to look at factors that enable this technology and factors that inhibit it. Key enablers include the fact that 3-D solid models and component finite element models are now available for most system components, unlike in the past. Secondly, new technologies have been developed for simplifying the representation of component data so that it can be efficiently processed in large system simulations. Thirdly, fast graphic workstations that can quickly analyze and display system-level models have now become inexpensive and plentiful. Furthermore, PDM systems facilitate system-level design by making vast quantities of data available and current. These factors make it possible to effectively deploy DMU and functional virtual prototyping today.

A few limiting factors still exist which retard progress in applying these technologies. First, very few universities have instituted effective training in these technologies, thus limiting the number of knowledgeable candidates for deployment. Secondly, hardware testing is ingrained in most manufacturing organizations and this new technology is sometimes viewed as a threat rather than being synergistic. In addition, effective deployment requires some process change within these large organizations and that requires a significant amount of training and the passage of time for overall adoption.

Success Story

A number of major automotive OEMs and tier 1 suppliers have already made substantial progress in using virtual prototyping to reap cost, time and quality benefits. One of these is Volkswagen (VW). Starting with a clear set of design performance targets, Volkswagen set out to remake the venerable Beetle into a modern-day success, not only in its styling, but also its driveability. To achieve its goal, VW relied heavily on the use of a robust virtual prototyping process throughout their chassis and powertrain development groups. They made extensive use of virtual prototyping system simulation software* to evaluate thousands of design variations for vehicle ride and handling, vehicle durability, safety systems, as well as engine, clutch and transmission performance.

The results were outstanding. After thorough design in the virtual world, the vehicle behaved splendidly in the real world! It was released to wide acclaim in both North America and Europe.

David E. Davis Jr., a writer for Automobile Magazine¹, wrote: "The New Beetle is a landmark car." "The car is a blast to drive." "Steering, braking, shifting and clutch operation are, quite simply, a joy." "[It] is a very safe car..." "More important, for us, is first-rate dynamic performance." "The New Beetle is definitely a driver's car."

This is a rather stark contrast to the original Beetle that was very popular, but was never known for its performance. These great improvements were made possible in a cost-effective manner through the heavy reliance on virtual prototyping.

Conclusion

The current bottleneck in globally competitive product design is the creation, instrumentation, testing and modification of system-level hardware prototypes. Traditional CAD/CAM/CAE methodologies do not provide a good means to break this bottleneck. New products in the DMU area, functional virtual prototyping area, and virtual factory simulation provide system-level counterparts to traditional component-focused CAD/CAM/CAE solutions and allow for breakthroughs in speed, cost, and quality for new product design. Key enablers are present in the market to make these technologies practical today.

The aim of this article was to provide a brief overview of functional virtual prototyping and how it can be successfully implemented in major manufacturing industries. Clearly the need for this technology exists. Rapidly increasing product complexity coupled with declining development budgets and time-to-market pressures mandate an alternative to singular reliance on hardware prototype testing. New computer hardware and software have enabled cost-effective implementations of this functional virtual prototyping technology.

What remains is for manufacturers to adopt enterprise-wide processes that fully incorporate virtual prototyping as a mainstream practice and institutionalize the use of virtual prototyping software to improve design and development of new products. Critical success factors for functional virtual prototyping implementation include:

• A well-defined process.
• System-level focus.
• Effective target setting.
• Rapid simulation turnaround.
• High quality CAE infrastructure.

Implementation of functional virtual prototyping on an enterprise level requires a significant commitment of time and financial resources. The benefits of making this commitment are enormous in terms of return-on-investment and global competitiveness.

References

1. Davis, David, "Volkswagen New Beetle," Automobile Magazine, Vol. 13, 1999, pp. 92-95.

* ADAMS from Mechanical Dynamics.

Robert Ryan accepted the role of chief operating officer of worldwide operations for Mechanical Dynamics (AnnArbor, MI) in 1991 and was appointed president in 1997. Ryan began his career in software sales at SDRC, where he aided the company in the formation of an aerospace business unit. He worked as an independent consultant for NASA's Jet Propulsion Laboratories, Applied Information Memories, Failure Analysis Associates, General Motors Corporation and Chrysler Corporation before joining the faculty of the University of Michigan in 1986. Ryan received his bachelor's degree in engineering and business from the University of Cincinnati, and holds M.S. and PhD degrees in mechanical engineering from Stanford University.