A novel approach to model unsteady fluid dynamics in a compressor network by using a bond graph is presented. The model is intended in particular for compressor control system development. First, we develop a bond graph model of a single compression system. Bond graph modeling offers a different perspective to previous work by modeling the compression system based on energy flow instead of fluid dynamics. Analyzing the bond graph model explains the energy flow during compressor surge. Two principal solutions for compressor surge problem are identified: upstream energy injection and downstream energy dissipation. Both principal solutions are verified in bond graph modelings of single compression system equipped with a surge avoidance system (SAS) and single compression system equipped with an active control system. Moreover, the bond graph model of single compressor equipped with SAS is able to show the effect of recycling flow to the compressor upstream states which improves the current available model. The bond graph model of a single compression system is then used as the base model and combined to build compressor network models. Two compressor networks are modeled: serial compressors and parallel compressors. Simulation results show the surge conditions in both compressor networks.

A novel approach to active surge control in compressors using piston actuation is presented. Two control laws are compared in order to evaluate the feasibility of implementing the concept. The first control law is a nonlinear feedback control derived by using backstepping and the second one is a linear feedback control derived by analyzing the eigenvalues of the linearized system around the operating point. The nonlinear feedback control law makes the closed loop system globally asymptotically stable (GAS) and uses full states feedback. The linear feedback control is only using feedback from plenum pressure and piston velocity and the removal of the mass flow feedback is advantageous for implementation. The closed loop system with the linear feedback control is locally asymptotically stable around the operating point. Simulations show that both controllers are capable of stabilizing surge.

A Lyapunov-based control system design of a twowheeled robot is presented. A two-wheeled robot (TWR) is a statically unstable system such that a control system is required to stabilize it. In order to get dynamics equations of the robot which will be applied in the control system design, modeling the robot is done through analyzing free body diagram and applying the Newton's second law. The modeling results in non-linear dynamics equations. Lyapunov-based control method is applied in the control system design of the TWR. It results in a global asymptotic stability (GAS) as proven mathematically. The GAS makes the robot's closed loop system is stable for the whole operating area. Performance of the designed control system for stabilizing the robot is evaluated through computer simulations.

A control design for two-wheeled robot (TWR) stabilization using linear quadratic regulator (LQR) method and states estimation is presented. A two-wheeled robot is an unstable system such that a control system is required to stabilize it. LQR is a control design method where the control gain is calculated by minimizing a performance index. The LQR method results in a full states feedback control law. Applying the LQR method for TWR stabilization will require feedback from pitching angle and pitching rate angle measurements. However, the common available sensor in the TWR is only a rate gyro which measures the pitching rate angle. Theoretically, time integration of the pitching rate angle results in pitching angle but it is not practically applicable because the time integration of measurement data may result in an accumulation of error and noise. Instead of the time integration, a Luenberger observer is applied to estimate the pitching angle based on the pitching rate angle measurement. Simulations of the TWR stabilization using the designed controller are presented through two scenarios: 1) both required state feedback are assumed to be available, and 2) both required state feedback are obtained through estimation using Luenberger observer. The first scenario is representing an ideal condition, while the second scenario is representing a practical condition. Simulation results show that the TWR is stabilized for the both scenarios, where the control system performance in the second scenario is slightly less than the first scenario but still acceptable.