Term of Award
Master of Science in Applied Engineering (M.S.A.E.)
Document Type and Release Option
Thesis (restricted to Georgia Southern)
Department of Mechanical Engineering
Dr. Biswanath Samanta, Ph.D.
Committee Member 1
Dr. Frank Goforth, Ph.D.
Committee Member 2
Dr. Rocio Alba-Flores, Ph.D.
In this study, an approach is presented to nonlinear control of a magnetic levitation system using artificial neural networks (ANNs). Two neural networks, the multi-layer perceptron (MLP) and the single multiplicative neuron (SMN), are investigated in this work. A novel form of ANN, namely, single multiplicative neuron (SMN), is proposed in place of the more traditional multi-layer perceptron (MLP). SMN derives its name from the single neuron computation model in neuroscience. Both off-line training and on-line learning of SMN have been considered along with off-line training of MLP. The SMN model is first trained off-line, to estimate the network parameters (weights and biases), using a population based stochastic optimization technique, namely, particle swarm optimization (PSO). An on-line learning algorithm has been developed for updating the SMN model parameters using a gradient-descent method. The ANN based techniques have been compared with a feedback linearization approach. The control algorithms have been developed and implemented on a hardware-in-the-loop (HIL) system of magnetic levitation in LabVIEW environment. The ANN based controllers performed very well and much better than the feedback linearization controller. However, the SMN structure was much simpler than the MLP with similar performance. With a simpler structure and faster computation, the SMN has the potential to be preferred to conventional MLP type ANNs for implementation in real-life, complex, nonlinear magnetic levitation systems.
Hall, Daniel L., "Nonlinear Controller Design and Implementation for a Magnetic Levitation System" (2013). Electronic Theses and Dissertations. 884.