‘Smart data’ instead of ‘big data’

Technical questions in research and industry are often addressed using classic machine learning methods, which, however, have the disadvantage that good results require large amounts of training data – ‘big data’. Stefan Posch and his team at the new CD Laboratory for Physics-driven Machine Learning in lndustrial Applications at TU Graz are therefore working on an innovative extension of this principle, supported by commercial partners BRP-Rotax GmbH & Co KG, Andritz Hydro GmbH, Palfinger Europe GmbH, Engineering Center Steyr GmbH & Co KG and MAN Energy Solutions SE: The aim is to combine traditional machine learning with physics-based methods in order to obtain ‘physics-driven machine learning’ models that ‘know’ the physical rules to be observed and thus enable greater accuracy with far less training data required - meaning its is worked with 'Smart Data' and an important contribution to significantly accelerate numerical simulations in the future (and thus to faster, more efficient product development in industry) is made!