CAE Data Management: A Critical Success Factor for Virtual Product Design and Subsystem Outsourcing
By Gahl Berkooz, EVP PLM (berkooz@dhbrown.com) and Don Brown, Chairman, D. H. Brown Associates, Inc. (dhbrown@dhbrown.com)

Conclusion
The inexorable acceleration in technological innovation creates a distinct and major opportunity - and rising challenge - for a major breakthrough in simulation across the supply chain backed by CAE data management. The trend also creates a significant opening for vendors. So far, however, we have not seen any vendors that comprehend the trends and tradeoffs. Indeed, many organizations mistakenly rely on their existing data management system (typically, PDM) to achieve their CAE data management goals. A new range of offerings targeting emerging requirements for performance and CAE data management is essential to the continuing evolution of virtual product development.

Pressure and Pain Points Driving Requirements
An emerging set of coherent business requirements is evolving that will make a major contribution down the line when architecting, designing, and selecting implementation strategies for performance and CAE data management. Short term, the pressures to compress the design cycle, and to employ analysis early in the design cycle, also drive the rationalization of data management strategies. Unfortunately, the CAE analysts and designers today often manage the data in a very unstructured way, making it hard to reuse data, to locate data, and to create archives that can facilitate product decisions for future generations of products.

Upfront CAE early in the design cycle and virtual product development in general, require quality assurance mechanisms to support the management of performance information for making decisions. That in turn creates a critical need for traceability. In a virtual development environment, the source of information that contributed to a decision must be identified. The process itself, and the data management implementation supporting that process become essential in tracking the quality of a design.

CAE analysts are only now beginning to appreciate how test data is managed to support lifecycle requirements. A product's lifecycle stretches all the way out to satisfy regulatory and maintenance concerns. The CAE side has not even begun to develop a parallel methodology for CAE information that would replace the test data.

Another pressure point relates to the new tools being introduced into our design processes: Stochastic analysis, robust design, design for Six Sigma, and CAE workflow automation - all of these approaches shape the trends.

When the virtual product development environment includes suppliers collaborating on the design, a host of new requirements emerge. For example, how do we segregate supplier information to maintain confidentiality? Moreover, warranty responsibilities will ultimately trace back the decision processes based on these digital records across the full supply chain. Considering these requirements, performance and CAE data management will touch on many constituents and business processes through design analysis, purchasing, and test. It becomes an enterprise and supply chain issue encompassing the full range of lifecycle activities.

Key Features
Given that broad charter, many of the key required features may be grouped in several categories. Tiered data management controls the CAE model revisions with respect to the CAD models, and maintains consistency. A host of issues arise related to product definition, including effectivity.

Another specific CAE issue relates to the whole notion of simplified geometries. The new versions create the need to maintain consistency and integrity of the simplified geometries for analysis in connection with CAD models for the parts or systems. Once the part or system changes, one must make sure that simplified geometry updates as well.

Consistency of CAE becomes a concern with respect to the product "as manufactured." CAE happens relatively early in the design cycle and does not capture a lot of smaller changes that may happen when the product shifts to actual production.

Tier One suppliers in the automotive industry are particularly sensitive to demands for CAE data interoperability and neutral storage. They are also concerned with the need for segregation, abstraction, and generalization of CAE data to preserve supplier confidentiality.

Productivity Issues
Up to now, these summary conclusions cover the point of view of data management. The productivity of analysts raises different concerns. Analysts may work individually or as a group in CAE. In all cases, however, their efforts involve others working on the same product in other disciplines which must share many assumptions and a common understanding of the product. The assumptions and understanding often needs to be made explicit, updated regularly, and published to non-analysts. Otherwise, the risk remains that the data and the models are used out of context.

An analysis history tree presents a particular peculiarity of analysis. As the analysis progresses, the process evolves the context in terms of the underlying assumptions. For example, the analysis may narrow or zero in on the ranges of parameters. By including these updated parameters, the full understanding of the specific piece of analysis and the context of the evolution in the analysis becomes paramount. Otherwise, the simulation results may be nonsense.

Other productivity features may involve quick starts for follow on analyses, or the modification of existing analysis. This becomes especially important in applying analysis in the early phase of design. Web-based post processing may increase productivity and also reduce the cost of software licensing. Features such as the automatic tuning of empirical parameters may improve the match with test results. Indeed, all too often the reconciliation for CAE and test is done ad hoc. Tiered data management must meet the productivity requirement of indexing for future reuse.

Integrity and Systems Engineering
The need for integrity in systems engineering involves another prioritization of key features required by yet another constituency and point of view. A breakdown of a methodology to cover product performance verification must assess the tradeoffs and use of CAE or test. The methodology should cover the stage of the design that needs validation as well as the compliance checking against specifications. In a sense, a road map for a product needs to lay out the steps and process for performance verification with CAE.

From a different perspective that entails subtle tweaks, the need to reconcile and compare CAE results with physical testing comes up again in considering the infrastructure. Interoperability requirements extend to consider the interface with proprietary internal databases that capture the properties of materials, reliability, fatigue, and standards. Various testing approaches need to cover wind tunnels and crash dummies - indeed the full range employed.

Given the goal of shortening the time-to-market, the pressure keeps building in the automotive area to either do without test, or minimize the test to the point where it is employed for confirmation purposes only. This is often quite difficult: Most companies are struggling because their test and analysis groups are usually not under the same management. That separation causes many problems. Despite the extreme desire for keeping track of all of that data through one stream, or through one management system, a bureaucratic roadblock for combining CAE and test stops it cold. Success has been achieved in cases where the test and analysis were actually used together. That provides better consistency for the CAE models with test on design decisions based on the CAE models. Effectively, that shortcuts much of the test work that might be done otherwise.

Aerospace may be significantly ahead of automotive in converging test with CAE. Aerodynamic design, for example, is exclusively done in a digital environment. Results from the wind tunnels calibrate the CFD models. There are different levels of maturity in different industries on this convergence.

The behavioral simulation of the full system becomes important. The abstraction of CAE to support system-level simulation also ties into a key product definition issue in assuring that the CAE data actually reflects the configuration that the analysts think they are simulating.

Common Pitfall
The natural instinct of people is to use their existing data management system, namely their PDM system, to achieve their objectives for CAE. PDM offerings, however, introduce a fundamental mismatch because the PDM systems are really designed for configuration control and release, and for dissemination to all interested parties. Performance information management presents fundamental differences with the emphasis on the processes for support of ongoing decisions, on the documentation, and on the traceability of the information that led to those decisions. It is painful to try to fit one into the other. Because of the mismatch in data sets, the approach does not seem like a productive avenue.

For more information, visit the D.H. Brown Associates, Inc. at www.dhbrown.com