Computational Challenges In Modeling And Simulation

 Conceptual Modeling

Over the past decade, within the modeling and simulation community there has been a growing interest in, and concern about “conceptual modeling.” Generally accepted as crucial for any modeling and simulation project addressing a large and complex problem, conceptual modeling is not well-defined, nor is there a consensus on best practices.  “Important” and “not well understood” would seem to qualify conceptual modeling as a target for focused research. Some workshop participants defined conceptual models as “early stage” artifacts that integrate and provide requirements for a variety of more specialized models. In this view, conceptual models provide a foundation from which more formal and more detailed abstractions can be developed, and eventually elaborated into analysis models (e.g., for simulation). However, workshop discussion led us to recognize that “early” and “late” are relative terms that apply within each stage of development. 

For example, creating an analysis model might involve describing, (i.e., modeling) the analysis independently of software (“conceptually”) before implementation and execution. As a consequence, there might be multiple “early” models: conceptual models of reality and conceptual models of analysis; and there may be multiple versions of conceptual models as the understanding of the target system matures and the analysis design and implementation evolves. 

Conceptual Modeling Language/Formalism

An articulated conceptual model, whether describing the system of interest (the referent, in Robinson's terminology) or an analysis model of the system of interest, is expressed using some language, which may be formal or informal, graphical, textual, mathematical, or logical. Today, the situation is that most often, conceptual models are expressed using some combination of 26 sketches, flowcharts, data, and perhaps pseudo-code. Lack of general agreement on the implications of these techniques (i.e., ambiguity) limits the computational assistance that can be provided to engineers. Incorporating conceptual modeling into a modeling and simulation engineering discipline will require more explicit and formal conceptual modeling languages. 

However, conceptual modeling must be done in a manner accessible to domain engineers, who might not be trained in the necessary formalisms. This is addressed in the first subsection below.

In addition, formal conceptual modeling applies as much to analysis as to the referent systems, raising questions about the variety of approaches to simulation, as covered in the second subsection. Formality in model integration is discussed in section 3.3.

There are number of simulation software's for the FEA - Finite Element Analysis are available in the market like, Abaqus developed by Dassault System. Ansys Software products including Ansys Workbench, Ansys Fluent, Ansys APDL, CATIA, SolidWorks and Hype mesh are more common used by the industry. All this software used for their unique capabilities for pre processing and post processing work. To get trained on this simulation software many of the companies are offering research and training services to get trained on this software's also including PIGSO LEARNING LLP. They offered Abaqus Training.

Domain-Specific Formalisms

In mathematical logic, formalism is the application of model and proof theory to languages, to increase confidence in inferring new statements from existing ones (Bock, et al 2006). In practice, however, most mathematicians are more informal in their definitions and proofs, with peer review confirming results, or not. We expect conceptual modeling formalisms to be rigorous approaches to studying referent and analysis models, at least in the sense of mathematical practice. Formal approaches have fewer, more abstract categories and terms than less formal ones, facilitating integration across engineering domains and construction of analysis tools. 

However, by using more abstract language, formal approaches are often too far from the common language of applications to be easily understood by domain experts and too cumbersome to use in engineering practice e.g., in air traffic control, battlespace management, health care systems, logistics, etc. More specific formalisms would be useful not only to domain experts, for describing their systems, but also to technical or modeling experts who must translate the system description into analysis models and maintain them, and to other stakeholders who may need to participate in validation.

A Unified Theory for Simulation Formalisms

Conceptual modeling applies not only to the system of interest, but also to the analysis of that system. Our understanding of a system of interest evolves from our earliest concept of it as we gain deeper understanding through the development of system models. In the same way, our understanding of the analysis itself also may evolve as we better understand the system of interest and begin to elaborate our analysis model.

To support conceptual modeling of simulation analysis, it seems reasonable that we should first have the ontology, semantics and syntax to formally define a simulation. Unlike the case of other analyses, such as optimization, this requirement has not yet been satisfied for simulation. Several structures have been studied as simulation formalisms; however, there is little consensus on the best approach. In the same way that various models of computation provide a basis for theory within computer science, considering various simulation formalisms will further the development of a robust theory of simulation.

Conceptual Model Development Processes

Model development is a challenging and highly intricate process, with many questions needing to be answered, as discussed in this section. Currently, answering these questions in a systematic and informed manner is hampered by a lack of formalized knowledge in the modeling domains and in modeling and simulation in general. Providing these would constrain development decisions and the design of development processes themselves, reducing uncertainty in model lifecycle engineering. The first subsection below gives background on model development processes and analyzes questions about them. The next two subsections (effectiveness and maturity) describe complementary approaches to reducing model defects introduced during the modeling process. 

These help avoid difficult and high-cost amendments of the model after it is finished. It is impossible to reduce model defects to zero during development, leading to the need for validation after the model is built, the results of which are also useful during model development, as addressed in the last subsection. Taken together, progress in these areas can significantly enhance the credibility of models by improving the quality of processes that produce them.

Motivation and Research Approach

The purpose of modeling and simulation is to improve our understanding of the behavior of systems: an executable model M of a system S together with an experiment E allows the experiment E to be applied to the model M to answer questions about S (Collier 1991). Simulation is fundamentally an experiment on a model. A conceptual model C is the articulated description of S, upon which both M and E are developed. In science we seek to understand the behavior of natural systems; in engineering we seek to design systems that exhibit desired behavior. Because modeling and simulation facilities are themselves complex systems, it is seldom possible to go in one step from problem to solution.

The processes involved in modeling and simulation require different degrees of human interaction, different computer resources, are based on heterogeneous, partly uncertain knowledge defined more or less formally, and involve different types of expertise and users. Data, knowledge, processes, and orchestration vary depending on the system to be modeled, the questions to be answered, and the users. In these processes different versions of models and artifacts are generated, that need to be put into relation to each other.

Effectiveness Measures

In a model-based engineering (MBE) approach, the development team evolves a set of models throughout the system lifecycle to support design, analysis, and verification of the system under development. These models are intended to express aggregate knowledge about the system to enable communications and shared understanding among the development team and other program stakeholders. Program leadership must continue to determine what knowledge must be acquired at any given point in the lifecycle to maximize the likelihood of program success. The type of knowledge to be acquired can help identify the kind of design and analysis models that should be further developed and updated.

Maturity Models

The Capability Maturity Model (CMM) for software development has played a key role to guarantee the success of software projects (Paulk, et al 1993). CMM and CMM Integration (CMMI) originated in software engineering, but have been applied to many other areas over the years (CMMI 2016a). However, in M&S, there is no such standardized and systematic assessment methodology developed for M&S processes. Some related research and development results can be used as references to establish the maturity model of M&S:

Validation

As simulation models become more complex, validation of conceptual models and understanding their role in the broader process of validation will continue to be important research areas. Of course, understanding validation of conceptual models is dependent on a precise definition of the terms “conceptual model” and “validation.”  This section argues that a better consensus is needed on the first term, while a careful review of the validation literature will reveal the same for the second.

Conceptual Model Architecture and Services

Many modeling paradigms exist for most kinds of domain problems, applied to knowledge from many engineering disciplines. Understanding complex systems requires integrating these into a common composable reasoning scheme (NATO Research and Technology Organization 2014). The software and the system engineering communities have overcome similar challenges using architecture frameworks (e.g. OMG’s Unified Architecture Framework (OMG 2016)), but modeling and simulation does not have a similarly mature integration framework. The first subsection below concerns architectures for conceptual modeling, while the second outlines infrastructure services needed to support those architectures.

Model Architecture

At the foundation of a modeling architecture should be a fundamental theory of models, to enable reusability, composability, and extensibility. What theory of models could support the implementation of a model architecture? An epistemic study of existing modeling and integration paradigms is necessary to develop a theory of models. This should include a taxonomy of modeling paradigms, semantics, syntaxes and their decomposition into primitives that operate under common rules across paradigms, to integrate them as required by complex systems.

Services

The success of large-scale integration of knowledge required by complex systems fundamentally depends on modeling and simulation infrastructure services aggregated into platforms. These enable affordable solutions based on reusing domain-specific models and simulators, as well as integrating them into a multi-model co-simulation. For example, understanding vulnerabilities and resilience of complex engineered systems such as vehicles, manufacturing plants, or electric distribution networks requires the modeling and simulation-based analysis of not only the abstracted dynamics, but also some of the implementation details of networked embedded control systems.  Systems of such complexity are too expensive to model and analyze without reuse and synergies between projects