2003

Tammi, Kari

This work introduces two principles to compensate a deterministic excitation from a system: adaptive finite impulse-response-filter and convergent control. Both principles work as adaptive feedforward compensation algorithms by feeding a compensation signal into the system. The compensation signal is generated from a reference signal that is correlated with the excitation to be compensated. The difference between the principles is the algorithm to derive the compensation signal from the reference signal. These algorithms are reported in the work. The methods were compared with simulations by compensating a sinusoidal disturbance in a simple plant. The convergent control algorithm indicated smoother but slower convergence than the adaptive finite impulse-response-filter with least-mean-square algorithm. The convergent control was also tested in the rotor test environment. The displacement responses of the test rotor were measured when the convergent control was switched on. The measurements were carried out when the rotor was running 25 Hz, 40 Hz, and 65 Hz. The control force commands were also recorded at these speeds. The displacement response was also measured during a sweep from 11 Hz to 65 Hz. The convergent control was found working properly; the convergence of the algorithm was particularly fast. The performance of the algorithm may be improved by improving the quality of the compensation signal. The forces used for control were low: from 1 N to 3 N. The parameter update seemed to have an effect on response; the response was slightly amplified at the update frequency.