Term of Award

Spring 2011

Degree Name

Master of Science in Applied Engineering (M.S.A.E.)

Document Type and Release Option

Thesis (open access)

Department

Department of Mechanical Engineering

Committee Chair

Fernando Rios-Gutierrez

Committee Member 1

M. Rocio Alba-Flores

Committee Member 2

Robert Cook

Abstract

DC motor systems have played an important role in the improvement and development of the industrial revolution, making them the heart of different applications beside AC motor systems. Therefore, the development of a more efficient control strategy that can be used for the control of a DC servomotor system, and a well defined mathematical model that can be used for off line simulation are essential for this type of systems Servomotor systems are known to have nonlinear parameters and dynamic factors, such as backlash, dead zone and Coulomb friction that make the systems hard to control using conventional control methods such as PID controllers. Also, the dynamics of the servomotor and outside factors add more complexity to the analysis of the system, for example when the load attached to the control system changes. Due to these parameters and factors new intelligent control techniques such as Neural Networks, genetic algorithms and Fuzzy logic methods are under research consideration in order to solve the complex problems related to the control of these nonlinear systems. In this research we are using a combination of two multilayer neural networks to implement the control system: a) The first network is used to build a model that mimics the function of DC servomotor system, and b) a second network is used to implement the controller that controls the operation of the model network using backpropagation learning technique. The proposed combination of the two neural networks will be able to deal with the nonlinear parameters and dynamic factors involved in the original servomotor system and hence generate the proper control of the output speed and position. Off line simulation using MATLAB Neural Network toolbox is used to show final results, and to compare them with a conventional PID controller results for the same model.

Share

COinS