▼ Closed loop control of linear stochastic control system with unknown parameters is studied by using a dual control approach. For control problems with either imperfect state observations or parameter uncertainties, the control signal to the system can actively influence the knowledge about the state of the system or about the parameters of the system. The dual controller is designed to regulate the state or output of the system to reach a certain goal and actively probe the system to obtain more information about the unknown system parameters. At each stage, the cost functional associated with the system objective is decomposed into a certainty equivalence cost and a dual cost. The active learning feature of the (dual) control explicitly accounts for the reduction of the dual cost, which measures the uncertainty of the system. An approximation technique is used to obtain a recursive closed form solution of the equations of functional dynamic programming. The dual control algorithm developed reduces the overall cost. Moreover, the algorithm is shown to be the best affine dual controller and it is very simple to implement. In certain situations, using an appropriate dual cost approximation, the computational overhead for the algorithm is minimum, and real-time control applications are practical. Advisors/Committee Members: Loparo, Kenneth A.

▼ The purpose of this investigation was to design and build a test rig to investigate the heat generated by a high speed sprag clutch assembly. The parameters used to design the test rig were to apply the cyclic torsion load at a level of 756 inch-pounds at 5000 rpm. The test rig was designed using a four square arrangement, which utilizes the circulation of power principle, a well known method commonly used in load testing power transmissions and has the advantage that only lost power must be supplied to the test rig. The sprag clutch test unit was placed between the high speed pinions of the circulating power loop. The test unit was designed to have replaceable inner and outer races, which contain the instrumentation to monitor the sprag clutch, to allow for cost effective replacement and versatility to test different sprag clutch designs. The oil supply system for the sprag clutch was developed to have the ability to lubricate the clutch from the inner race or through the sides of the sprag clutch and the oil flow rate could be changed by changing the size of the supplying orifice. The torque loading device was chosen to be a water cooled magnetic clutch, which is controlled either manually or through a computer. A DC electric motor, which will be used to supply the lost power to the four square loop, was chosen with a speed range of 0 to 2800 rpm, which allows testing at various speeds. The electric motor was coupled to the four square test loop by an increaser belt system, which will allow the high speed gear shafts to have a maximum speed of 5000 rpm. The data obtained from this investigation will be used to verify the results from a computer model and provide the ability to test present and future sprag clutch designs. Advisors/Committee Members: Kicher, Thomas P.