A model predictive protection system for actuators placed in hostile environments

In most applications where motors control several degrees of freedom of a mechanical system there is the necessity to implement a protection which is able to stop the motors when the mechanical system reaches the extreme positions of its working space. When the mechanical system is located in an hostile environment and/or when the limit positions define a complex surface this simple protection strategy is not always viable, since it's not possible to place sensors and limit switches in the proximity of the limit positions. We will present in this paper a solution for the actuator protection problem based on a model predictive algorithm capable to estimate in real time the actuator state variables (positions and speeds). The idea of the model predictive protection scheme is to implement a model of the mechanical system forced by the emergency braking actions in order to predict, using as initial condition the current observed state of the system, the final rest position together with its uncertainty. If the probability that the predicted rest position lays outside the physical boundaries of the working space exceeds a predefined threshold the protection trips. With this approach we will also reach the maximum exploitation of the system's working space, since it is adapted to the current system state. We will show the application of this concept to the protection of a set of steering microwave antennas placed in an hostile environment (hard vacuum condition, strong neutron and gamma radiation, strong magnetic fields) inside a nuclear fusion reactor