Methodical Approach
Partial models in fields of analysis and synthesis in the engineer technical scientific context are predominantly represented by physical mathematical descriptions in the theoretical area. An effective processing is increasingly made possible by numeric methods. The method of the finite elements is used in rapidly rising measures in all fields of engineering, including the building and construction industry. This relates to both the research and the practice. This method shall therefore play a central role in processing and incorporate methods and procedures of stochastic modelling, multi-scales and adaptive modelling on a necessary scale.
The quality of models can be evaluated over the definition of used parameters and flow injection analysis systems (see table). This makes it possible, through different standards, to define general concepts and to delimit these standards from each other for the used partial models. Partial models in this context are defined as sub models, which depict partial aspects during the project processing of a building e.g. global load-capacity, constructional-detailing, subsoil behaviour. To date, the use of a complete global model depicting all views of a structure is not technically possible.
With reference to the different partial models, three abstraction levels can be defined, which reflect different qualities in the procedures. The number of standards, described as levels, is fixed and thus follows practical points of view by analogy with corresponding definitions in the reliability theory.
The development of theoretical fundamentals of civil engineering and the available analysis techniques shows at present a steady transition of essentially deterministic, geometrically and physically linear considerations to nonlinear considerations and semi probabilistic assumptions for loads, material properties etc.. These transitions are partly carried out independently of each other in individual engineering disciplines and thereby with the connected partial models. Due to the fact that the norm development is strongly subdivided into individual branches like soil mechanics, solid construction, steel-construction etc., very frequently as a consequence, different (Partial-) model standards are then coupled insufficiently arising from tradition or ignorance (figure 2).
Transitions between different levels of partial models are usually difficult to model because they are often considered independent of each other and the incorporated input parameters are not representable, thus depicting results different in nature. Furthermore, the applied analysis procedures of the partial models follow different methodical approaches. Statements regarding the entire model are finally derived from the integrated analysis of all partial models. Interactions between separate partial models are depicted in many cases over simple, highly abstract interfaces. An example for this is the ground-structure interaction which is often simulated by floor springs.
In the theoretical analysis it is to be investigated, how a transition between partial models can take place, which can be assigned different levels of abstraction.The aim is to increase the quality of the complete statement and to provide scientific based methods for a coupling.
From an engineering point of view, the problem of the independent development of partial models at higher abstraction levels and the impact of these developments on those partial models coupled on other partial models must be investigated.
Here, methods have to be developed, which permit an adaptive customization of the coupled partial models and thereby adopt the quality requirements consistently from one to the other partial model (figure 2).
A supplementary procedure for the assessment of the model quality according to the concept introduced above is the development of a methodology, for the quantitative assessment of the quality of partial models and model couplings, which is based on energy statements.
Within the solution space of the mechanical / physical problem, limited by the abstraction process, expanding and reducing processes can be employed on the dimensions of the solution space, following well known procedures to assess discretization errors (in space and time) as well as dimension and model errors. Their impacts on various energy levels (e.g. global and local distortion energies, dissipative and damage energies etc.) are to be quantitatively evaluated. With this methodology, different influences / causes for the deviation model of partial models, such as dimension and scale effects, influences of different levels for abstraction of material laws, consideration or negligence of transient effects or impacts as a result of different linearization approaches within the context of nonlinear problems, will be quantitatively determined.
To investigate the model sensitivity as a measure of the quantitative and possibly qualitative changes in the statements of a model by changes in parameters and input variables, customized procedures from the reliability analysis can be employed. Therefore, for example, in areas of continuous solution functions with few results, response functions as a buffer area in the foreseen parameter space can be generated. The derivations of the response surfaces can be understood as a measurement for sensitivity of response by fluctuations in input parameters and therefore, enable a quantitative assessment.
To evaluate the model robustness for feasible evidence, as calculation for the capability of a model for parameters and input values within a, for the construction industry typically wide, utility area, the size of the response area can be found in the adequate smooth response surfaces and no singularities. Methods of the perturbation theory and interval analysis can be used as alternative procedures to assess the robustness.
Based on deviations of energy quantities between disturbed solutions and undisturbed reference solutions; this assessment also requires calculations adhering to the demands on the utility area.
The error propagation for complex tasks of engineering, which is required for the coupling of partial models is to date not documented in the literature.
Further implementations of the proposed methodology for quantitative evaluation of partial models and model couplings are found in the description of the main research I.
For the conception and realization of the necessary theoretical information of cooperation platforms, approaches are examined both, with versioned object models and with operative models. This shall develop computer tools, which support the planning processes, using coupled partial models and allow a synchronous manipulation and timely calculation and visualization of model responses. Existing single user applications in a distributed surrounding will be employed as far as possible. Current technologies like object-oriented methods as well as the agent technology will be employed.