1 Intelligent flight controllers for helicopter control1)A methodology for design of an intelligent helicopter flight controller is presented in this paper. 2)The methodology combines artificial neural network, genetic algorithms, conventional PID-controllers, and fuzzy logic algorithms in the design procedure.3) In this approach, the design of an intelligent controller is based on experimental data collected from actual helicopter flight and not based on analytical models.4) A neural network is trained to learn the dynamic characteristics of the helicopter. 5)Based on the neural model, the coefficients of a PID-controller used for blade angle control are searched for using genetic algorithms. 6)The main rotor speed controller is designed using fuzzy logic algorithm based on knowledge generated from understanding the aerodynamic theory and analyzing the helicopter experimental flight data. 7)The intelligent helicopter flight controller is formed by combining the blade angle PID-controller and the rotor speed fuzzy controller. 8)Simulation results showed that for desired altitude input H d, the intelligent controller was able to generate proper control signals for both the blade angle control and the rotor speed control. 9）The helicopter model altitude output H t follows the desired altitude input H d. 10）The design procedures, theoretical background, results and conclusions are presented in this paper2. EDFA gain modeling using Pspice and neural networks1）A new methodology in modeling the performance of an Erbium Doped Fiber Amplifier (EDFA) Gain is presented in this dissertation. 2）It is based on the gain characteristics of an EDFA under different input signal power spectra and pump levels. 3）The novelty of this dissertation is based on the training and use of a neural network applied to the performance characteristics of an EDFA.;The modeling is divided in two sections. 4）The first model is an electrical equivalent circuit of an EDFA that can be simulated and analyzed by Pspice, a commercial electrical simulation program from OrCad. 5）The second section is the development and implementation of a Neural Network to model the gain response of the EDFA for a given number of input power spectra.;6）In order to provide data and verify the performance of both models, a simulation program was developed. 7）The simulation program's output performance was verified using an EDFA analysis program called Oasix.;8）The results from both models were accurate with respect to the theoretical and experimental data. 9）The electrical equivalent model's output spectrum was proven to match the output of the simulation program. 10）The neural network was trained successfully to identify the EDFA gain spectra to its correspondent input spectrum and pump power level with an accuracy of 98%.4、Characterization and Modeling of the Power Insulated Gate Bipolar T ransistor.The power Insulated Gate Bipolar Transistor (IGBT) is a new switching device designed to overcome the high on-state loss of the power MOSFET. The IGBT behaves as a bipolar transistor which is supplied base current by a MOSFET. The bipolar transistor of the IGBT has a wide base with the base contact at the collector edge of the base and is operated with its base in high-level injection. Because of this, the traditional bipolar transistor models are not adequate for the IGBT and the new model developed in this dissertation must be used. The new model is developedusing ambipolar transport and does not assume the quasi-static condition for the transient analysis.;The new IGBT model is used to describe measurements for extracting the essential physical device parameters of the model. With these extracted parameters, the new IGBT model consistently describes the measured electrical characteristics of IGBTs with different base lifetimes. The important electrical characteristics of the IGBT are the on-state I-V characteristics, the steady-state saturation current, and the switching transient current and voltage waveforms. The transient waveforms are examined in detail for constant anode voltage switching, clamped inductive load switching, and series resistor, inductor load switching.;The disadvantage of the IGBT is its slow turn-off speed relative to that of the power MOSFET. The two methods which have been proposed to reduce the turn-off time of the IGBT are base lifetime reduction and buffer layer inclusion. Both methods have the disadvantage, though, of also increasing the on-state voltage. The buffer layer is a high-doped portion of the bipolar transistor base at the base-emitter junction. The new IGBT model discussed above is extended to include the buffer layer. Using the extended model, it is shown that the buffer layer IGBT can be made to have a faster switching time for a given on-state voltage than that of the nonbuffer layer (lifetime reduction) IGBT.;In summary, a new model is developed for the IGBT. It is shown that the new model must be used to accurately describe the steady-state and transient characteristics of the IGBT. The model is used to compare the effects of lifetime reduction and buffer layer inclusion. The results of the comparison show that a better on-state voltage, switching speed trade-off is obtained using the buffer layer than is obtained using lifetime reduction.5、Developing modeling and simulation methodology for virtual prototype power supply systemThis dissertation develops a modeling and simulation methodology for design, verification, and testing (DVT) power supply system using a virtual prototype. The virtual prototype is implemented before the hardware prototyping to detect most of the design errors and circuit deficiencies that occur in the later stage of a standard hardware design verification and testing procedure. The design iterations and product cost are reduced significantly by using this approach. ^ The proposed modeling and simulation methodology consists of four major parts: system partitioning, multi-level modeling of device/function block, hierarchical test sequence, and multi-level simulation. By applying the proposed methodology, the designer can use the virtual prototype effectively by keeping a short simulation CPU time as well as catching most of the design problems. ^ The proposed virtual prototype DVT procedure is demonstrated by simulating a 5 V power supply system with a main power supply, a bias power supply, and other protection, monitoring circuitry. The total CPU time is about 8 hours for 780 tests that include the basic function test, steady stage analysis, small-signal stability analysis, large-signal transient analysis, subsystem interaction test, and system interaction test. By comparing the simulation results with the measurements, it shows that the virtual prototype can represent the important behavior of the power supply system accurately. Since the proposed virtual prototype DVT procedure verifies the circuit design with different types of the tests over different line and load conditions, many circuit problems that are not obvious in the original circuit design can be detected by the simulation. ^ The developed virtual prototype DVT procedure is not only capable of detecting most of the design errors, but also plays an important role in design modifications. This dissertation also demonstrates how to analyze the anomalies of the forward converter with active-clamp reset circuit extensively and facilitate the design and improve the circuitperformances by utilizing the virtual prototype. With the help of the virtual prototype, it is the first time that the designer is able to analyze the dynamic behavior of the active-clamp forward converter during large-signal transient and optimize the design correspondingly.7、Adaptive nonlinear control of spacecraftThe problem of spacecraft attitude control and momentum management is addressed using nonlinear controllers based on feedback linearization. A chief limitation of the feedback linearization technique is that it requires an exact cancellation of nonlinear terms in order to obtain linear input-output behavior. Adaptive nonlinear controllers for linearizable systems are investigated to overcome this restriction and to achieve asymptotic linear behavior.The spacecraft attitude control and momentum mangement model is characterized by having uncertain parameters appearing linearly on both sides of the dynamicl equation. Therefore，two adaptive controf laws are developed forgeneral models in which the parameters appear linearly on both sides of the dynamical equations.The spacecraft attitude control and momentum mangement problem is then addressed with the adaptive nonlineu control laws utilized and uncertainties in the spacecraft principal intertias assumed. The adaptive nonlinear approach is shown to effectively control the Space Station attitude and effector momentum, while providing accurate estimates of inertias. The combined tracking prediction-error based tracking is shown to have improvement over the performance of the tracking-error based adaptive controller.8、Discrete-Time Control of A Spacecraft with Retargetable Flexible Antennas.This dissertation considers the discrete-time control of a spacecraft consisting of a rigid-platform. with retargetable flexible antennas. The mission consists of independent minimum-time maneuvers of each antenna to coincide with predetermined lines of sight, while the platform. is stabilized in an inertial space and elastic vibration of the antennas is suppressed. The system is governed by a set of linearized, time-varying equations of motion. A discrete-time approach permits consideration of the time-varying nature of the system in designing the control law.;Both global and decentralized controls are proposed for a noise-free system with full-state feedback. Initially, a time-varying linear-quadratic regulator (LQR) is implemented, followed by two types of decentralized controllers. First, a collocated control law is devised in which actuator forces are based on the position and velocity at the actuator locations. Next, a new metho d called Substructure-Decentralized Control is proposed, where each flexible substructure is controlled based on state measurements associated with the substructure modes of the separately modeled appendages.;In both global and decentralized cases, a linear control law is first implemented coupled with an open-loop disturbance-accommodating control based on the known inertial disturbances caused by the maneuver. Elastic motion is next controlled using nonlinear (on-off) antenna controllers for each decentralized case. For Substructure-Decentralized Control, the controls translate into quantized actual controls. Lastly, nonlinear (on-off) control laws are also used to control the rigid-body motion for each case.;Next, the problem of controlling the time-varying system in the presence of noisy actuators and sensors is examined. It is assumed that only displacements, not velocities, are sensed for both rigid-body and elastic motion, making state reconstruction also necessary. A discrete-time, full-order Kalman filter is constructed for the time-varying system. A pseudo-decentralized control is proposed whereby feedback controls are based on system state estimates. As before, both linear and nonlinear controls are implemented. For each case mentioned, a numerical example is presented involving a spacecraft with a singleflexible maneuvering antenna.。