|
|
|
| We describe the Watt System Methodology (WattSys™), a systems engineering formalism and methodology used in developing models of products and processes in dynamically evolving, large-scale, complex systems (DELCS). Having identified fundamental problems in Traditional Systems Engineering’s (TSE) historically reductionist, hierarchical, machine-model approach, this methodology represents an alternative means of systems engineering. The methodology’s formalism imports static models in high level languages such as the Unified Modeling Language (UML™), its extension profile the Systems Modeling Language (SysML™), or mid-level languages such as Petri-nets and Harel state charts. The mathematical formalism allows us to map the underlying graph structure of these languages onto temporally extended, hybrid network graphs called models. The models represent hypotheses about the systems they represent and may be simulated, developed and evolved. Within models are classified three coupled groups, consisting of 1) graphs of the system under investigation (system graph), 2) the interface graphs, being the input, output and bias nodes to the system graphs, and 3) context graphs which represent meaningful, measurable context spaces and the topologies that define their boundaries. As more information is added during the modeling process, from high level models and internal refinements, it is mapped upon the graph constructs. Each of these three graphs are hybrids of network graphs, hybrid bond graphs and neural networks. The combinations of the couplings between system graphs, interface graphs and context graphs represent whole system models that may be processed by simulation, development, evolution, complexification, pattern abstraction and refinement. Data provides input values to the system interfaces, conditioned by forward calibrated Bayesian probabilities. These perturb activations of compositional pattern producing (neural) networks (CPPN), simulating growth and evolution of system models within their contexts. These models are then further refined by Inverse Bayesian methods. The methodology details the cyclical stages of growth and refinement in a fractal pattern until sufficiency of information is developed. As such, it resolves many problematic effects in the design, engineering and development of complex systems, including non-linearity, high-dimensional coupling, emergence, entropy and self-structuring over time. This paper is available upon e-mail application to WattSys Methodology UML and SysML are Trademarks of the Object Management Group (OMG)
|
|
The Unit of Power In Complex Systems™ Send mail to
webmaster with questions or comments
about this web site.
|