Mathematics-based computational mechanics involves idealization in going from the observed behaviour of a system into mathematical equations representing the underlying mechanics of that behaviour. Idealization may lead mathematical models that exclude certain aspects of the complex behaviour that may be significant. An alternative approach is data-centric modelling that constitutes a fundamental shift from mathematical equations to data that contain the required information about the underlying mechanics. However, purely data-centric methods often fail for infrequent events and large state changes. In this article, a new hybrid modelling framework is proposed to improve accuracy in simulation of real-world systems. In the hybrid framework, a mathematical model is complemented by information-based components. The role of informational components is to model aspects which the mathematical model leaves out. The missing aspects are extracted and identified through Autoprogressive Algorithms. The proposed hybrid modelling framework has a wide range of potential applications for natural and engineered systems. The potential of the hybrid methodology is illustrated through modelling highly pinched hysteretic behaviour of beam-to-column connections in steel frames.
|Original language||English (US)|
|Journal||IOP Conference Series: Materials Science and Engineering|
|State||Published - Jan 1 2014|
|Event||9th World Congress on Computational Mechanics, WCCM 2010, Held in Conjuction with the 4th Asian Pacific Congress on Computational Mechanics, APCOM 2010 - Sydney, Australia|
Duration: Jul 19 2010 → Jul 23 2010
All Science Journal Classification (ASJC) codes
- Materials Science(all)