Several types of cranes are widely applied for short-distance load transportation. The so-called overhead cranes are often used in construction sites, ship yards and storage areas. The overhead cranes mostly move on rails, and carry the load underneath their cart, hanging on a wire.
Similarly to the other types of cranes, the swinging of the load poses a problem in case of the overhead cranes as well. The swinging is not only dangerous, but can delay the transportation time. In practice this problem is usually prevented by experienced operators. Yet there exist control algorithms to handle this problem, however they are based on the value of the swinging angle. In most cases this is unknown, there are no sensors available to measure the angle. Hence other methods are required for the determination of it's value. One possible method to achieve this is using state observation.
The state observation consists of the determination of the states of a dynamical system, based on the information collected. In general, this information is obtained through measurement of the inputs and outputs. In some cases though, the inputs of the system are considered unknown, and hence they must also be estimated.
The reason behind may be the desired fault detection of the actuator, but another similar task could be the determination of an unknown disturbance. In case of mechanical systems, the disturbance is typically the friction changing the force produced by the actuator.
In this thesis, sorted state observers are presented for the state observation of friction loaded mechatronical systems in general, and of the overhead cranes in particular. Simulation results are shown for the study and evaluation of these observers' performance. Models of the overhead crane system are used in different complexity (with nonlinear or linearized dynamics; with varying or fixed rope length).
The computations implemented in embedded controllers operate on sampled signals, hence the realization of the observer is also done in discrete-time. Therefore specific attention is given to discrete-time UIO-s and in particular to the so called algebraic observers. Observers with such structure, as opposed to the classical ones, do not have states and they solve a linear or nonlinear set of equations during each sampling period. This implies that algebraic observers can be used without checking the separation principle to prove the stability of the closed loop system.
As a result of this study two methods are proposed for the determination of the swinging angle. The first one is a specially designed Kálmán-filter, with calculates the angle quite accurately. The second one is a nonlinear algebraic observer. While it is less precise in the determination of the state variables, its further development could be more promising. This method could be applicable for the observation of a larger class of nonlinear systems.