The paper presents a novel, object-oriented (O-O) approach to structural modeling and analysis to enable interactive, virtual prototyping in the next generation of the design environment. The interactivity of analysis of build-up structures can only be achieved by re-using results of component analyses, therefore, the applicability of data based analysis to the O-O approach is investigated in this paper. An object in the O-O approach is a neurocomputing model of a structural component developed for a given set of parameterized design variables, loading, and "smart" connectors. Such connectors define how the given component is connected to the assembly, and what is the validity of such a connection. Since component models can be subject to affine transformations, a family of similar structural components is available for modeling. Neurocomputing models are used to enable data based analysis, because they can relate any design variables and load conditions to the selected system response. The applicability of feed-forward and feature-sensitive neural nets is examined in this context. Object-oriented modeling is illustrated with a design of an assembly of beams using a single-point (SP) connector and an assembly of plates subject to a thermal load using a multi-point (MP) connector. (C) 2000 Elsevier Science Ltd. All rights reserved.