Deep learning-based applications of control engineering

Within the applications of deep learning, control is a frequently visited subject. Especially system parameter identification is a research field that diverges to many different potential improvements. In this thesis, system parameter identification is performed for the damping coefficient and the natural frequency of a second order system, using deep recurrent neural networks. The network trained has been implemented into a closed loop configuration, furthermore in the loop training is actualized. The integration of deep learning into the domain of parameter identification aims to enhance the accuracy of parameter identification. The parameters predicted are utilized to tune a PID controller for the investigated system, therefore an adaptive controller concept is realized.