Date of Award
Doctor of Philosophy
Electrical and Computer Engineering
A spline network, that is an alternative to artificial neural networks, is introduced in this dissertation. This network has an input layer, a single hidden layer, and an output layer. Spline basis functions, with small support, are used as the activation functions. The network is used to model the steady state operation of a complex Heating Ventilation and Air-conditioning (HVAC) system. Real data was used to train the spline network. A neural network was also trained on the same set of data. Based on the training process, it is possible to conclude that when compared to artificial neural networks, the spline network is much faster to train, needed fewer input-output pairs, and had no convergence problems. The weights of the spline network are obtained by solving a set of linear equations;The spline network model of the HVAC system is used to detect faulty operation of the actual system. Once abnormal operation of the system is monitored, a fuzzy neural network is used to locate the faulty component. The fuzzy neural network is trained on data obtained by simulating fault scenarios. This network minimizes ambiguities at decision boundaries. The results of fault classification are presented in the dissertation.
Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/
Mathew Scaria Chackalackal
Chackalackal, Mathew Scaria, "Spline network modeling and fault classification of a heating ventilation and air-conditioning system " (1994). Retrospective Theses and Dissertations. 10544.