Modeling elevator dynamics using neural networks

A new neural network model of a commercial SCD elevator is proposed. The main goal of the research project is to improve elevator ride comfort via speed profile design. The main objective in modeling is to obtain a good and reliable tool for process analysis and control system development. The work consists of measurement and filter planning as well as actual model identification. Much emphasis is put on designing and preprocessing measurements without forgetting practical engineering aspects. The model combines nonlinear and linear networks into a gray-box model instead of the common black-box model. Also physical knowledge is embedded into network construction. The results show that the empirical model implemented within neural network framework is able to represent the real process up to small details.