A new speed observer with guaranteed bounds using interval arithmetic

In this paper a new speed and disturbance torque observer, applied to the induction machine with encoder, is proposed. In practical situations, when designing the control loop, the noise statistics needed for tuning a stochastic observer are generally unknown. However, the bounds of the process and measurement noises can be clearly defined, for example, if they are generated by a quantizer. The increased DSP computational power and new mathematical tools, like the interval arithmetic, allow us to calculate both the bounds of the estimation error and the observed states. A nonlinear optimization algorithm is used here to obtain the gain of the observer that minimizes the estimation error. Some experimental tests with an AC motor have been performed in order to compare the proposed observer with a state bounding algorithm.