sgonzalez90 Posted October 20, 2014 Posted October 20, 2014 Investigating an issue where two axes sit on the same seals. Excitation of one axis (Z-axis), displaces the other (Y-Axis) and vice versa. I have made attempts to tune the opposing with Integrator (with Integration mode = 1)...results got worse. Deadband Gain, had marginal improvements, but could be improved. My next approach is to use a notch filter in an attempt to use this as a low pass filter. For the cut-off frequency, being a component of the Velocity....with this FFT how would I determine the appropriate value, or by what process? Thank You
sgonzalez90 Posted October 20, 2014 Author Posted October 20, 2014 Investigating an issue where two axes sit on the same seals. Excitation of one axis (Z-axis), displaces the other (Y-Axis) and vice versa. I have made attempts to tune the opposing with Integrator (with Integration mode = 1)...results got worse. Deadband Gain, had marginal improvements, but could be improved. My next approach is to use a notch filter in an attempt to use this as a low pass filter. For the cut-off frequency, being a component of the Velocity....with this FFT how would I determine the appropriate value, or by what process? Thank YouY-Axis Velocity FFT.bmp
curtwilson Posted October 20, 2014 Posted October 20, 2014 Can you tell if there is a deterministic relationship between the action of one axis and the disturbance of the other? If so, the best strategy is to characterize this relationship and create your own cross-coupled feedforward algorithm (usually in a PLCC 0) to put a counteracting torque offset in the other motor before there is a chance for the error to build up. But the first step is to attempt to characterize the relationships. Are the errors proportional to the other motor's acceleration? Velocity? Both? I would look in time domain first, but both time and frequency plots can be valuable. P.S. You do have some pretty big spikes at the AC line frequency and harmonics. I can't tell if these are causing real problems for you, but you may want to investigate this further.
curtwilson Posted October 20, 2014 Posted October 20, 2014 Can you tell if there is a deterministic relationship between the action of one axis and the disturbance of the other? If so, the best strategy is to characterize this relationship and create your own cross-coupled feedforward algorithm (usually in a PLCC 0) to put a counteracting torque offset in the other motor before there is a chance for the error to build up. But the first step is to attempt to characterize the relationships. Are the errors proportional to the other motor's acceleration? Velocity? Both? I would look in time domain first, but both time and frequency plots can be valuable. P.S. You do have some pretty big spikes at the AC line frequency and harmonics. I can't tell if these are causing real problems for you, but you may want to investigate this further.
Recommended Posts