Modular Analysis of Complex Systems with Numerically Described Multidimensional Probability Distributions

Abstract

To analyze complex systems, many teams have to collaborate, often from different disciplines. The presented approach allows a systematic modular description of such systems (esp. systems, that are influenced by uncertain values, e.g. from human behavior or from mechanical/material uncertainty). The model of the system is made up from a network of modules, which are connected by interfaces. Each interface is a set of values, each described by Numerically Described Multidimensional Probability Distributions (NDMPD). This allows a description even of scattered values. The behavior of each module can be described by

• formulas (the related computer tool will evaluate it for parameters given as NDMPD and provide the result as NDMPD)

• NDMPD, gathered by

  • surveys,

  • evaluations of the module as a subsystem,

  • many other methods, e.g. Monte Carlo,

  • appropriate assumptions.

If a module is considered as a subsystem, it can be hierarchically modeled by sub-modules. The behavior of the subsystem can be described with NDMPD, which are derived from the behavior of its (sub-)modules. As the system behavior is described numerically, computers can evaluate it. This allows automatic reevaluation, e.g. after a change of a single module (after a refinement, for new findings, for other hypothesis etc.).