In this thesis a class of robust non-linear controllers for a gantry crane system are discussed. The gantry crane has three degrees of freedom, all of which are interrelated. These are the horizontal traverse of the cart, the vertical motion of the goods (i.e. rope length) and the swing angle made by the goods during the movement of the cart. The objective is to control all three degrees of freedom. This means achieving setpoint control for the cart and the rope length and cancellation of the swing oscillations. A mathematical model of the gantry crane system is developed using Lagrangian dynamics. In this thesis it is shown that a model of the gantry crane system can be represented as two sub models which are coupled by a term which includes the rope length as a parameter. The first system will consist of the cart and swing dynamics and the other system is the hoist dynamics. The mathematical model of these two systems will be derived independent of the other system. The model that is comprised of the two sub models is verified as an accurate model of a gantry crane system and it will be used to simulate the performance of the controllers using Matlab. For completeness a fully coupled mathematical model of the gantry crane system is also developed. A detailed design of a gain scheduled sliding mode controller is presented. This will guarantee the controller's robustness in the presence of uncertainties and bounded matched disturbances. This controller is developed to achieve cart setpoint and swing control while achieving rope length setpoint control. A non gain scheduled sliding mode controller is also developed to determine if the more complex gain scheduled sliding mode controller gives any significant improvement in performance. In the implementation of both sliding mode controllers, all system states must be available. In the real-time gantry crane system used in this thesis, the cart velocity and the swing angle velocity are not directly available from the system. They will be estimated using an alpha-beta state estimator. To overcome this limitation and provide a more practical solution an optimal output feedback model following controller is designed. It is demonstrated that by expressing the system and the model for which the system is to follow in a non-minimal state space representation, LQR techniques can be used to design the controller. This produces a dynamic controller that has a proper transfer function, and negates the need for the availability of all system states. This thesis presents an alternative method of solving the LQR problem by using a generic eigenvalue solution to solve the Riccati equation and thus determine the optimal feedback gains. In this thesis it is shown that by using a combination of sliding mode and H??? control techniques, a non-linear controller is achieved which is robust in the presence of a wide variety of uncertainties and disturbances. A supervisory controller is also described in this thesis. The supervisory control is made up of a feedforward and a feedback component. It is shown that the feedforward component is the crane operator's action, and the feedback component is a sliding mode controller which compensates as the system's output deviates from the desired trajectory because of the operator's inappropriate actions or external disturbances such as wind gusts and noise. All controllers are simulated using Matlab and implemented in real-time on a scale model of the gantry crane system using the program RTShell. The real-time results are compared against simulated results to determine the controller's performance in a real-time environment.

In this thesis a class of robust non-linear controllers for a gantry crane system are discussed. The gantry crane has three degrees of freedom, all of which are interrelated. These are the horizontal traverse of the cart, the vertical motion of the goods (i.e. rope length) and the swing angle made by the goods during the movement of the cart. The objective is to control all three degrees of freedom. This means achieving setpoint control for the cart and the rope length and cancellation of the swing oscillations. A mathematical model of the gantry crane system is developed using Lagrangian dynamics. In this thesis it is shown that a model of the gantry crane system can be represented as two sub models which are coupled by a term which includes the rope length as a parameter. The first system will consist of the cart and swing dynamics and the other system is the hoist dynamics. The mathematical model of these two systems will be derived independent of the other system. The model that is comprised of the two sub models is verified as an accurate model of a gantry crane system and it will be used to simulate the performance of the controllers using Matlab. For completeness a fully coupled mathematical model of the gantry crane system is also developed. A detailed design of a gain scheduled sliding mode controller is presented. This will guarantee the controller's robustness in the presence of uncertainties and bounded matched disturbances. This controller is developed to achieve cart setpoint and swing control while achieving rope length setpoint control. A non gain scheduled sliding mode controller is also developed to determine if the more complex gain scheduled sliding mode controller gives any significant improvement in performance. In the implementation of both sliding mode controllers, all system states must be available. In the real-time gantry crane system used in this thesis, the cart velocity and the swing angle velocity are not directly available from the system. They will be estimated using an alpha-beta state estimator. To overcome this limitation and provide a more practical solution an optimal output feedback model following controller is designed. It is demonstrated that by expressing the system and the model for which the system is to follow in a non-minimal state space representation, LQR techniques can be used to design the controller. This produces a dynamic controller that has a proper transfer function, and negates the need for the availability of all system states. This thesis presents an alternative method of solving the LQR problem by using a generic eigenvalue solution to solve the Riccati equation and thus determine the optimal feedback gains. In this thesis it is shown that by using a combination of sliding mode and H??? control techniques, a non-linear controller is achieved which is robust in the presence of a wide variety of uncertainties and disturbances. A supervisory controller is also described in this thesis. The supervisory control is made up of a feedforward and a feedback component. It is shown that the feedforward component is the crane operator's action, and the feedback component is a sliding mode controller which compensates as the system's output deviates from the desired trajectory because of the operator's inappropriate actions or external disturbances such as wind gusts and noise. All controllers are simulated using Matlab and implemented in real-time on a scale model of the gantry crane system using the program RTShell. The real-time results are compared against simulated results to determine the controller's performance in a real-time environment.

O problema de transferir uma carga ao se movimentar num plano em tempo mínimo e sem oscilação no ponto de descarga, num guindaste portuário tipo pórtico é investigado neste trabalho. Assume-se que a carga esteja inicialmente em repouso na posição vertical no ponto de carga acima do navio e igualmente em repouso no ponto de descarga na moega de alimentação no porto. Assume-se também que o carro do guindaste esteja em repouso em ambos os pontos. Um modelo completo é apresentado para o sistema do guindaste onde as equações dinâmicas não-lineares são linearizadas para ângulos de oscilação pequenos o suficiente e reescritas para a forma adimensional. A solução de tempo mínimo é buscada considerando como variáveis de controle as funções do tempo que descrevem tanto a força aplicada no carro para produzir seu deslocamento horizontal, como a velocidade de içamento da carga. Um método iterativo preditor-corretor usando a Programação Linear (PL) é proposto, baseado no modelo do sistema de tempo discreto onde as variáveis de controle são tomadas constantes por trechos. Na etapa corretora, assume-se que o movimento de içamento é dado e uma solução de tempo mínimo é obtida resolvendo-se uma seqüência de problemas de PL de tempo fixo e máximo deslocamento. Na etapa preditora, um modelo linearizado é empregado para obter-se uma correção ótima do movimento de içamento usando a PL. O problema de controle de tempo mínimo é formulado levando-se em consideração restrições práticas na velocidade do carro do guindaste, velocidade máxima de içamento, assim como na máxima força que pode ser aplicada ao carro. Resultados numéricos são apresentados e mostram a efetividade do método. / The problem of minimum-time anti-swing transfer of a load in a ship-to-pier gantry crane is investigated in this work. The load is assumed to be initially at rest at the vertical position at the loading point above the ship and equally at rest at the unloading point above the hopper. The trolley is also assumed to be at rest at both points. A complete model is presented for the crane system where the nonlinear dynamic equations are linearized for sufficiently small swing angles and then rewritten in dimensionless form. The minimum-time solution is sought by considering as control variables both the force applied on the trolley that produces its horizontal motion and the hoisting speed of the load as functions of time. A predictor-corrector iterative method using Linear Programming (LP) is proposed based on a discretetime model of the system where the control variables are taken as stepwise constants. At the corrector step, the hoisting motion is assumed given and a minimum-time solution is obtained by solving a sequence of LP problems representing fixed-time maximum-range problems. At the predictor step, a linearized model is employed to obtain an optimal correction of the hoisting motion using LP. The minimum-time control problem is formulated by taking into account practical constraints on the maximum speeds of both the trolley and the load hoisting, as well as on the maximum force that can be applied to the trolley. Numerical results are presented and show the effectiveness of the method.

O problema de transferir uma carga ao se movimentar num plano em tempo mínimo e sem oscilação no ponto de descarga, num guindaste portuário tipo pórtico é investigado neste trabalho. Assume-se que a carga esteja inicialmente em repouso na posição vertical no ponto de carga acima do navio e igualmente em repouso no ponto de descarga na moega de alimentação no porto. Assume-se também que o carro do guindaste esteja em repouso em ambos os pontos. Um modelo completo é apresentado para o sistema do guindaste onde as equações dinâmicas não-lineares são linearizadas para ângulos de oscilação pequenos o suficiente e reescritas para a forma adimensional. A solução de tempo mínimo é buscada considerando como variáveis de controle as funções do tempo que descrevem tanto a força aplicada no carro para produzir seu deslocamento horizontal, como a velocidade de içamento da carga. Um método iterativo preditor-corretor usando a Programação Linear (PL) é proposto, baseado no modelo do sistema de tempo discreto onde as variáveis de controle são tomadas constantes por trechos. Na etapa corretora, assume-se que o movimento de içamento é dado e uma solução de tempo mínimo é obtida resolvendo-se uma seqüência de problemas de PL de tempo fixo e máximo deslocamento. Na etapa preditora, um modelo linearizado é empregado para obter-se uma correção ótima do movimento de içamento usando a PL. O problema de controle de tempo mínimo é formulado levando-se em consideração restrições práticas na velocidade do carro do guindaste, velocidade máxima de içamento, assim como na máxima força que pode ser aplicada ao carro. Resultados numéricos são apresentados e mostram a efetividade do método. / The problem of minimum-time anti-swing transfer of a load in a ship-to-pier gantry crane is investigated in this work. The load is assumed to be initially at rest at the vertical position at the loading point above the ship and equally at rest at the unloading point above the hopper. The trolley is also assumed to be at rest at both points. A complete model is presented for the crane system where the nonlinear dynamic equations are linearized for sufficiently small swing angles and then rewritten in dimensionless form. The minimum-time solution is sought by considering as control variables both the force applied on the trolley that produces its horizontal motion and the hoisting speed of the load as functions of time. A predictor-corrector iterative method using Linear Programming (LP) is proposed based on a discretetime model of the system where the control variables are taken as stepwise constants. At the corrector step, the hoisting motion is assumed given and a minimum-time solution is obtained by solving a sequence of LP problems representing fixed-time maximum-range problems. At the predictor step, a linearized model is employed to obtain an optimal correction of the hoisting motion using LP. The minimum-time control problem is formulated by taking into account practical constraints on the maximum speeds of both the trolley and the load hoisting, as well as on the maximum force that can be applied to the trolley. Numerical results are presented and show the effectiveness of the method.