There are plans for the next two calls with the topics of CATIA V4/V5 Interoperability; and CATBench Hardware Evaluation. Dates are November 13, and January date TBD. Be watching for more information to come.
As a reminder, the abstract for the call today was as follows:
One puzzle has been why does computing seem so hard to do. Despite the best efforts of many, there seems to be no consistent process to ensure successful implementation. John M. Switlik will talk about a few computational issues related to advanced computing, such as CAD/CAM/CAE and KBE (Knowledge Based Engineering). Switlik will show how these involve trades that are related to representation, methods, and quality.
Representation deals with technical concerns, such as structure and parameter definitions. Choices related to representation have a large influence. One trade involves choice between explicit and implicit, the latter being required for smarter processing.
Methods are of many types involving both human and computational expertise. In the computational era, methods imply management of external and internal capability. Capability involves an effective mix of procedures, algorithms, and heuristics that know how to handle inherent problems.
Quality varies by context and is affected by both representation and methods. Classic trades exist between some quality criteria. An example would be trades for a surface (fit, continuity, and footprint). Criteria can differ by whether the use is interactive or automated as human judgment (assuming it is well trained) affords a broad evaluative set that can be difficult to duplicate via automation.
We can have progress by appropriate handling of the trades and other inherent issues. Like the weather, these cannot be managed away; we can and ought to make provisions for controlling their effects. Brief mention is made of one concept called undecidability that will be of growing importance to KBE.
An example KBE project handles types of data requirements by integrating methods from the Boeing mathematics library with rules expressed using ICAD. The project provides a mixture of algorithms and heuristics to implement knowledge related to geometry (spline) creation and evaluation. One offering is the application of optimization, using Sparse Optimal Control Software - SOCS, to surface and curve fitting; this approach has demonstrated the ability to handle problematic data (minimal, sparse, noisy, or originating from multiple/conflicting sources).
The presentation looks at recent work involving physical modeling, discusses possible improvements in defining quality measurements for models, and offers a list of open issues that will require continued and joint effort within the industry.
John Switlik's interests have an integration/quality focus for life-cycle control (Conceptual, Define, Produce, Maintain). The involved methodology includes applied/computational mathematics (FEA/FEM/Statics, simulation/dynamics, NURBS-based numeric processes), knowledge modeling/management, and advanced techniques such as adaptive approaches, statistical learning, data mining, and categorical modeling/processing. With a Professor Emeritus, Mathematics, Switlik has been exploring the basis for modern mathematical analysis as applied with the computational metaphor. Switlik received his M.A. and B.A from the University of Arizona in economics with focus on quantitative analysis.
Author: Michelle Brozell