Manas Bajaj(1)

1. Georgia Institute of Technology
2. Akrometrix LLC (www.akrometrix.com)
3. InterCAX LLC (www.intercax.com)
4. Rockwell Collins
5. NIST

The realm of electronics product realization is marked by an extremely fast-paced market, stringent demands for product reliability and high importance to innovative design. Further, the time-to-market and the cost-to-realize play a critical role for product success. However, these attributes pose conflicting constraints on the realization process. While engineers need to converge quickly on a set of design alternatives, the high demand for reliability increases the breadth of behavioral simulations across the design space. Further, the multi-disciplinary nature poses integration challenges due to a disparate set of engineering tools, model representations, and simulation techniques.

In this presentation, we shall focus on the following three technical areas to alleviate these hurdles in knowledge management during electronics product realization:

(1) Design-Analysis Integration: In order to quickly converge on a set of feasible design alternatives while covering a wide range of behavioral simulations across the design space, it is essential for engineers to synthesize analysis and solution-specific product models for a given design alternative. This process is guided by ascertaining the context of analysis, identifying possible idealizations and mapping information from design specifications to the analysis specifications. Further, the modularity of this process is essential to explore all possible alternatives.

We leverage from over a decade's worth of experience spanning methodologies and tools (www.eislab.gatech.edu/research/dai/) and some recent advances in this area to demonstrate current and envisioned technologies for seamless design-analysis integration.

(2) Standards-based Knowledge Representation: In order to create high-fidelity design, analysis and manufacturable product models, it is essential to use a detailed and standard ontology for electronics product data specification. In this light, we employ STEP AP 210 (www.ap210.org) for electronic assembly packaging and design as the underlying representational structure for creating and archiving product models. Further, we use a harmonized set of STEP-based schemas for product model specifications across the design-analysis integration bridge. In this presentation, we shall focus on the ability of a standards-based knowledge representation scheme to support product and process related knowledge for electronics PLM.

(3) Experimental Validation of Simulation Techniques: In general, a simulation-based methodology needs to be validated against experimental results to justify reuse and instill confidence in decisions based on simulation results. In this presentation, we shall also demonstrate on algorithmic techniques for validating simulation results with experimental data and discuss some critical issues concerning the same.

Further, in this presentation, we will exemplify recent developments in the three technical areas using thermo-mechanical warpage analysis problem for printed circuit boards and assemblies, as part of the current collaborative effort between co-author organizations.

1. G.W. Woodruff School of Mechanical Engineering, GIT
2. Manufacturing Research Center GIT

In this presentation, we put forth the notion of a product view federation that embodies engineering processes related to creation, enrichment and usage of a particular product view. We also elucidate the interaction between product view federations to create the complete product specifications, represented as an open standards-based collective product model. Also, we outline a set of enabling technologies for this envisioned product model federation.

Further, to exemplify the ideas above, we leverage from our recent parallel efforts in the domain of knowledge management for federated simulations. In this scenario, we shall walk through a process of creating a federated simulation ontology (analogous to the collective product model for product specifications) from a set of given simulation ontologies (analogous to product views). We also demonstrate specific techniques for enabling this integration process and identify necessary extensions for utilizing these for the envisioned product model federation.

Product and system lifecycle management (PLM/SLM) helps enable creation of enterprise value. Here, value can be considered as a function of financial returns derived from such enterprise outputs as technologies, systems, products or services. It can also be a function of cost savings or less quantifiable benefits (e.g., improved enterprise capability). Integral to this notion of value creation are research and development, which serve as a front-end of the product/system lifecycle.

From this value-based perspective, R&D can be characterized as a multi-stage investment problem, in which resources are allocated to different lines of R&D, across multiple stages. The goal then involves some variant of maximizing total value created, where value is realized downstream from the R&D process. This investment problem involves considerable uncertainty, since R&D is characterized by technical risk (probability of unsuccessful results), and market/application risk (i.e., future value is unknown at time of investment decisions).

One key question involves identifying and quantifying R&D value levers - i.e., those factors that can be manipulated to increase value. For instance, a growing body of literature argues that valuating R&D investments using a real options (RO) framework is superior to a discounted cash flows (DCF) approach, since real options capture management's ability to fund initial work, and then discontinue funding in later stages, if value diminishes over time. Other process-oriented value levers are budget allocation among R&D stages, and even the number of stages.

This work studies strategic value levers in large-scale R&D organizations through use of organizational simulation. Organizational simulation is an emerging discipline that combines traditional discrete-event simulation with advanced decision logic, human decision-making, and ultimately immersive environments. Based on a model of R&D processes, initial results suggest that a real options framework facilitates more total value creation than DCF, while DCF emphasizes return on investment of R&D expenditure. Ongoing work is specifying a standardized data model for R&D "products" (i.e., results) and their value, to be integrated with the process model as a basis for strategic R&D value lifecycle management. This product model incorporates precedence relationships, future envisioned results and dynamic changes. Future work aims at integrating these models with strategic design.

Work funded by ESA at The University of Manchester has led to the development of an open source EXPRESS parser. The major feature of this parser is that the interface to the parsed model is an application programming interface (API) rather than an abstract syntax tree (AST); where possible, any processing that would be common to most applications based on an EXPRESS model are performed by the parser. Hence, the development time for applications that exploit this API is reduced.

Processing performed during the transformation of a model into the API data structures includes reference resolution, determining the data type domains of objects and the attributes of a complex entity instance, and the redeclarations that effect an attribute in a particular instance context. The Application Development Kit (ADK) supports multi-schema models and the concept of a viewpoint schema. During the population of the API, processing defined by Section 11, and Annexes B and C of the EXPRESS standard are performed to determine the independently and dependently instantiatable objects that are defined by the model and the various identifications that they have.

This talk will give an overview of the ADK API and describe some of the processing that is performed.

A brief history of CAD data translation issues and solutions, from users' point of view, will be presented. As CAD technology level moves higher, so does user expectation in the area of CAD data translation. We will present CAD data translation issues that users are facing today, tomorrow and beyond. This includes

• Model Based Definition. An appropriate level of interoperability is needed in order to support a paperless Model Based Definition environment. Feature-based translation and GD&T are part of the equation.
• Product Data Management (PDM) is playing an essential role in the CAD/CAM area. CAD is becoming a portion of the PDM system. Therefore the domain of CAD data translation is expanding from CAD to PDM.
• Translation Validation. There is always a need to validation translation quality. What do we need to validate? How to validate? How to communicate the results to the general user community?
• What should not be translated? Certain proprietary information should be kept in the company. How to prevent them from being translated out?
• Low End Viewer (LEV) is needed for various reasons. What technologies are available today? For what applications?
• Collaboration. What does it mean? How does CAD data translation help collaboration for designers?
• Long Term Data Retention. What are the current technologies for long term data retention? What can we expect out of it?

Over the last year we have been scrutinizing the STEP-TAS protocol and dictionary to become fit for the exchange of all space thermal analysis and test models. This has been done in an iterative cycle in which protocol and dictionary improvement have been verified continuously against real working interfaces for the main European thermal analysis tools ESARAD and THERMICA that have been implemented in the TASverter tool, that is distributed for free by ESA and is used by industry in production environments. Heavy use has been made of the Express-to-Python code generation capability (pyExpress) and a large regression test suite has been built up.

Some lessons learned will be presented on STEP interface development and verification in general as well as on the advantages of separating a data exchange standard into a basic robust protocol and an additional dictionary. Here a 'dictionary' is used in the same way a 'reference data library' in other STEP standards.

The problems this work addresses include:

• a product is made up of multiple STEP model schemas; thus product data access needs to access data in multiple schemas;
• repository user interfaces rely on SDAI or similar schema-dependent interfaces, which are extremely unfriendly for general users; a long term solution for multi-model access is needed;
• production quality user interfaces are expensive to develop, especially when they are broad in terms of the engineering domains covered.

We are working to develop an ontology-driven user interface capability, that we think can resolve these issues. It is based on:

• general ontology-driven models that can map to multiple ISO-STEP schemas;
• provides user interface capabilities that are schema-driven, allowing one interface to be self-tailoring to different domains;
• supports easy interfacing to different types of user interfaces: geometry, temporal relations, topological functional relations, tabular and scientific visualization methods.

The work described is being undertaken by two Ph.D. students at Georgia Tech in the College of Architecture. An NSF proposal has also been submitted for longer term funding of the work.

The advantages of DSL's is that they provided a more focused representation of design solutions, which can significantly increase the productivity of designers. The disadvantges are the problem of how to capture, manage, and exchange DSL data. By their own nature, DSLs typically have a bespoke representation, including the data relating to their domain concepts, their visual and textual representation, and even their dynamic semantics if they support dynamic features such as simulation rules, which need to be preserved across different platforms and tools.

This presentation will discuss experiences of working with DSL's on large projects and identify some potential approaches to the problem, some of which could be applied to general-purpose PDE.

INCOSE and the Object Management Group (OMG) are collaborating on an initiative to extend the Unified Modeling Language (UML) to provide a general-purpose systems modeling language to support specification, analysis, design, verification and validation of complex systems. The graphical modelling language is intended to be a key enabler for transitioning the practice of systems engineering from a document-centric to a model-centric approach. SysML is also be aligned with ISO AP233 to support exchange of systems engineering data and tool interoperability.

The UML for Systems Engineering (SE) Request for Proposal (RFP) was issued by the OMG in March 2003 and contains the requirements for the modeling language. The SysML Partners was established to respond to this RFP, and provide an extension to the recently adopted UML V2.0 to provide a robust systems modelling capability. This presentation will provide an advanced look of how UML 2 in conjunction with the SysML extension can address the needs of systems engineering. The adoption of SysML V1.0 is planned to begin in 2005. Additional information can be found on the SysML Partners web site at http://www.sysml.org.

Thanks principally to the advances in computer hardware, the execution of numerical and analytical simulations in the last few years no longer demands expensive investments in processor power or waiting time for results. Instead, the fact that simulation tools today sometimes tend to be used like "pocket calculators" leads almost inevitably to a situation where their application is restricted by the ability to manage the related data. The effort required for simulation data management is increasing with the amount and complexity of information that is manipulated and produced as part of the related processes. This seems diametrically opposed to scenarios requiring collaboration between partners in joint projects and processes that are getting more and more complex due to an increasing number of activities handled in parallel.

The application of Enterprise Application Integration (EAI) approaches for technical data today is mostly restricted to the area of product design (CAD) utilizing the comparably large overlap of information content maintained by existing CAD tools. An integration across different domains such as design and engineering where this overlap is rather low opens a completely new level of complexity not addressed today. This is made even more difficult by the fact that little Product Data Management (PDM) is used in the engineering domain and the variety of simulation tools tends to become complex with no common way existing to share CAE information across these tools. The key to address these challenges is simulation tool integration - opening ways to consistently manage CAE data as well as preparing this domain for integration with others.

The cost of changes in design and planning becomes dramatically higher with progressing maturity of a product's definition. The ability to make the right decisions as early as possible is becoming more and more crucial. Analysis and simulation can help to make the right decisions when used efficiently. The ability to manage the related information without concessions is crucial to this.

Developed in cooperation between CNES and CSTB (Centre Scientifique et Technique du Bâtiment - the French national research center for construction domain), the new version of Baghera View is able to load several STEP models from different schemas (STEP-TAS, AP203, AP214, IFC) in one single session for checking and comparing them.

Based on open source technologies like Open Cascade, Open SG, QT, PyExpress and new OSS Express toolkit developed by ESA and University of Manchester, Baghera View V3 propose standard viewing and reporting functions completed with innovative semantic comparison mechanisms, using EXPRESS models based analysis.

Developed on the basis of the CSTB Enriched Virtual Environment (EVE) platform and its portable kernel and GUI, Baghera View V3 can be used on Windows or UNIX/Linux systems or within the powerful CSTB Le Corbusier Immersion Space (http://salle-immersive.cstb.fr) for project reviews in virtual reality environments.

In this presentation, this new version of Baghera View and its mechanisms will be explained and demonstrated.

This presentation describes a method to (semi-)automatically derive a product model from product information collected from various processes. The method is called Georgia Tech Process to Product Modeling (GTPPM). GTPPM aims to empower domain experts to describe information required for their own processes through the requirements collection phase of product modeling and to allow product (data) modelers to automatically integrate and normalize the collected information through process modeling into a product model through the logical product modeling phase. It also aims to maintain the link between individual company data use-cases and a standard data model so that it can eventually support evolution and refinement of a model over time.

The integration of process models themselves is not our interest. Our interest is in the integration of information items collected through multiple process models. In GTPPM, the inconsistencies between information definitions collected from heterogeneous processes are handled at four levels. The first one is the semantic level consistency. In GTPPM, an information item can have only one meaning. No homonym or synonym is allowed. The second one is the consistency of the syntax for specifying product information. It was handled by the product information specification (PIS) framework. The third one is the consistency of the collected information: i.e., is the collected information logically coherent? This was handled by the dynamic consistency checking method. The last one is the consistency between information definitions collected through multiple processes. Such conflicts are resolved in the logical product modeling using the rules defined as design patterns. These methods
will be described with examples during the presentation.

GTPPM has been implemented and used by precast concrete companies in North America and several universities in the US, the UK, and Israel in IT-related research work.

Problem:
Design and analysis are two key aspects of the product development process. Designers determine product specifications based on required functions. Similarly, analysts simulate and evaluate the behavior of the resulting product. If the simulated behavior of the product is unacceptable, the design is sent back to the designers for modification. This process is repeated until the behavior of the product is within acceptable bounds. For example, in a beam design problem, this design-analysis iteration cycle may be repeated until the deflection or stress in a structural support beam is within the maximum allowable limits. It is evident from this scenario that the product-related information in a) design problems and b) analysis models is closely related and must be efficiently integrated. In this context, a design problem is comprised of the design variables, constraints, goals, preferences, etc. and an analysis model is used to predict the behavior of the product in terms of response variables.

The challenges associated with integrating design and analyses are twofold. First, the disparity in heterogeneous software applications and reliance on proprietary data formats limits sharing product knowledge across tools and organizations. Second, the product-related design information is often transformed between design and analysis domains. In this research, we address the second issue, specifically focusing on the manner by which different representations of the same physical product can be linked between designers and analysts. Thus, the research question addressed is - How can product-related information and the relationship between design problem and analysis models be captured?

Significance:
Exchanging product-related design information between heterogeneous design tools is an essential component in the product realization process. For example, standards-based product models facilitate the exchange of product information between heterogeneous design support tools. However, they do support capturing relationships between design problems and associated analysis models. Thus, there is a need for developing information representation aimed at capturing these relationships.

Example:
For demonstrating the information transformations, we present an example scenario for the design and analysis of the structural support beams. The structural support beam is simple, but clearly illustrates the linkages between design and analysis.

Currently, design processes for complex systems are formulated based on previous design experience. In simulation based multiscale design, the design processes represent the manner in which information generated by simulation models is utilized for satisfying design objectives. These processes are inherently complex because of the interdependencies between simulation models at various scales. Given this complexity of design processes, it is imperative that the design processes themselves be designed appropriately in a systematic manner. This is because inefficient design processes can lead to longer design time, thereby leading to higher costs [1]. The need for designing design processes is also emphasized by Herbert Simon [2] in his statement "design process strategies can affect not only the efficiency with which resources for designing are used, but also the nature of final design as well".

In order to design the design processes, two fundamental issues that need to be addressed are: 1) capturing design related information from past design scenarios and reusing it in new scenarios at a computational level, and 2) identifying appropriate design processes from a number of possible alternatives, depending on the design problem under consideration. In this paper, we address the first issue of capturing design related information by developing an information model that supports reusability and configurability of design processes. The second issue of metrics is addressed in a companion paper titled " Metrics for Analyzing Simulation Based Multiscale Design Processes" submitted to this conference.
The current practice in simulation-based design frameworks is to capture information about products and processes in a lumped fashion (i.e., the processes are captured in terms of the product information and the tools used), which restricts the utilization of processes for designing different products. There is very tight coupling between product and process information. Further, information about the design problem is not captured explicitly. Hence, it is not possible to capture the rationale behind selecting a design process for solving a design problem. This hinders reusability of previously utilized design processes and prevents design of design processes at a computational level. The 3-P information model proposed in this paper addresses this challenge.

Approach and Implementation
In order to address reusability and reconfigurability challenges associated with designing design processes, we present a 3-P information model, which is an object-oriented information model that captures design related information. The 3-P information model is built on the previous work presented by Panchal et al. in [3]. The three key components of the 3-P model include information models for a) design problem, b) design process, and c) product information. The design problem definition consists of the design variables, responses, constraints, goals, preferences, etc. In order to solve the design problem, a design process is laid out, which consists of a network of transformations on product information. These transformations have product information at State A as input and product information at State B as output.

In the 3-P information model, the information is captured as entity objects and relationship objects. For example, in the models for design processes, the entities refer to information transformations and relationships refer to the information flows between these transformations, whereas while modeling product information, the entities refer to components and the relationships refer to interfaces between components. The design problem information model is based on the compromise DSP construct developed as a part of the Decision Support Problem (DSP) Technique [4]. The processes are modeled as hierarchical systems in three levels - individual transformations, model interactions, and process compositions. Seven transformations and nine model interactions are identified specifically for simulation-based multiscale, multifunctional design. Each of these model interactions are associated with design processes and serve as standard reusable patterns for design processes.

The information model is instantiated as Java objects and integrated into FIPER. The reusability and reconfigurability aspects of the information model are shown via modeling design problem, processes and product information for datacenter cooling system.

== Overview ==

This presentation overviews a collaborative effort to infuse constrained object (COB) concepts within the emerging SysML standard. SysML is "a new visual modeling language designed by systems engineers for specifying systems of systems"(SoS) [1].

Georgia Tech has developed the COB knowledge representation over the past 12+ years to capture fine-grained relations within and among diverse models. Applications include analysis templates that facilitate interoperability among engineering design and analysis models.

In this presentation we show how SysML (and its emerging parametrics capabilities in particular) can represent the flap link analysis template tutorial described below. The SysML Parametric Diagram represents a network of relations among the properties of a system such as F=ma and Total Weight = sum (Part Weights). These diagrams are intended to capture inter-model associativity, including bridging design models with engineering analysis models. This concept-rich test case helps both to evaluate and demonstrate SysML capabilities (e.g., parametric diagram scalability) and to identify aspects needing further development.

Envisioned applications include a widely accepted unified representation of domain-specific models and their fine-grained associativity with system models, ultimately resulting in fundamental capabilities for next-generation SoS and product lifecycle management (PLM).

== Constrained Object (COB) and Analysis Template Background [2, 3] ==

The variety of engineering design and analysis contexts makes the generalized integration of computer-aided design and engineering (CAD/CAE) a challenging proposition. Transforming a detailed product design into an idealized analysis model can be a time-consuming and complicated process, which typically does not capture idealization and simplification knowledge explicitly. Georgia Tech has developed the multi-representation architecture (MRA) and analyzable product model (APM) techniques to bridge the CAD-CAE gap with stepping stone representations that support design-analysis diversity. These techniques employ constrained objects (COBs) as a generalized underlying representation.

The COB representation is based on object and constraint graph concepts to benefit from their modularity and multi-directional capabilities. Object techniques provide a semantically rich way to organize and reuse the complex relations and properties that naturally underlie engineering models. Representing relations as constraints makes COBs flexible because constraints can generally accept any combination of I/O information flows. This multi-directionality enables, for example, design sizing (synthesis) and design verification (analysis) using the same COB-based simulation model. Engineers perform such activities throughout the product lifecycle, with the former being characteristic of early design stages and vice versa.

Wilson et al. [2] present basic examples to illustrate COB concepts, including applications to analysis building blocks (ABBs) utilized later in a flap link tutorial example [2].

This flap link tutorial [3] demonstrates an MRA-based design-analysis panorama that supports these capabilities in a unified manner: multiple levels of abstraction and a diversity of physical behaviors, analysis fidelities, and CAD/CAE methods and tools.

To validate the COB representation, other work implemented electronic packaging and aerospace test cases in a COB-based toolkit called XaiTools™. In all, these test cases utilize some 260 different types of COBs with some 370 relations, including automated solving using commercial math and finite element analysis tools. Results show that the COB representation makes the MRA reusable, modular, and multi-directional, thus enhancing physical behavior modeling and knowledge capture for a wide variety of design models, analysis models, and engineering computing environments.

References:
1 - http://www.SysML.org/
2 - http://eislab.gatech.edu/pubs/conferences/2001-mit-cfsm-1-wilson-cobs/
3 - http://eislab.gatech.edu/pubs/conferences/2001-mit-cfsm-2-peak-xai-example/

Keywords:
SysML, UML, parametric diagram, constrained object (COB); constraint graph; constraint schematic, design-analysis integration; CAD-CAE interoperability; multi-representation architecture (MRA); simulation-based design (SBD); multi-fidelity; multi-directional; systems of systems (SoS); product lifecycle management (PLM).

The architecture of the PLM Services specification is intended to serve for a wide range of implementation platforms. Currently it is targeted at WebServices, XML Schema, SOAP, and Java.

PLM Services establishes an open and extensible set of functions based on the widely accepted STEP standards and bridges the gap between reliably and interoperable asynchronous data exchange scenarios and the online access methods of state of the art WebServices capabilities.

PLM Services are already implemented in several prototypes and products. The initiators are now seeking for more international dissemination and industry sectors. This presentation gives an introduction into the basic concepts of the PLM Services and demonstrates the functionality on the basis of the existing implementations.

The main emphasis of the presentation is to acquaint user and system developers, who are concerned with collaborative engineering scenarios, with the PLM Services and encourage them in adopting and implementing the standard.

One of the design goals of PGEF is to provide a set of built-in, modular capabilities and a few out-of-the-box, customizable applications. Target capabilities include importing and exporting engineering models, knowledge, and data using STEP (ISO 10303), UML, and OWL. Targets for out-of-the-box applications include a basic repository service with an underlying logical API that can be implemented in several network service interfaces, the initial one being XMLRPC over HTTP/HTTPS.

The PGEF architecture is built upon an eclectic set of open source software, including Python (a flexible object-oriented language, http://python.org), wxPython (a cross-platform GUI toolkit, http://wxpython.org), Twisted (a Python framework for asynchronous network protocols, http://twistedmatrix.com), and PostgreSQL (a robust relational database, http://www.postgresql.org), among others.

As PGEF evolves, it will integrate and interface with additional open source engineering and scientific applications, such as the Python Numeric and SciPy suites, which provide powerful computational, analytical, and visualization tools.

Further, we shall demonstrate IDA-STEP, an upcoming framework that embodies these technologies for rich knowledge representation and semantic continuity across the product lifecycle. More specifically, IDA-STEP (www.ida-step.net) is an open standards-based PLM framework that enables the integration of CAD, CAE, PDM and Manufacturing Process models in a multi-user, distributed environment. In its current generation, IDA-STEP uses ISO 10303 STEP-based content models for product and process descriptions in the following domains: electronics (AP210), automotive (AP214), mechanical (AP203), cabling (AP212), systems engineering (AP233). Many other STEP APs are supported at a basic viewing level for share aspects like geometry (e.g., AP209 for FEA and AP216 for ship molded forms). Additionally, IDA-STEP uses XML-based technologies for highly customizable and rich user interfaces. IDA-STEP supports models generated in many common ECAD, MCAD, Systems Engineering (SE), PDM, Manufacturing Process Planning (MPP) applications via STEP-based translators and interfaces available for them. In some cases, IDA-STEP has direct extensions available for these applications.

We shall also demonstrate usage scenarios typical of product and process-related knowledge management across PLM and outline our planned extensions to IDA-STEP and enabling open standards-based technologies.