Artificial Neural Networks - mathematical model with a future in building physics?

Artificial Neural Networks (ANNs) are a type of mathematical models which has only recently made its entry to the fields of civil engineering in general and building physics in particular Due to their pronounced flexibility ANNs enjoy a steadily growing number of applications to a wide (and widening) range of diverse tasks. After a description of the modelling technique and the properties of the ANN type in most widespread use in civil engineering contexts, the so-called Multilayer Perceptron (MLP) or feedforward network, previous experience in building physics applications is summarized briefly and potential fields of future application are indicated. At the present stage of development, ANNs appear as intelligent and flexible black box models, the practical potential of which has certainly not been fully exploited in the field of building physics yet.