Performing substructuring on a motorized linear shaking table, but hindered by dead-zone of motor. Need new solutions, any help much appreciated.

I am using Simulink, interfaced to a linear motorized shaking table by Dspace. Substructuring requires great control accuracy but the DC motor has dead-zone of +/- 3.77 V when loaded - a big problem when the motor continually has to traverse zero V. Tried overcoming the dead-zone in Simulink by inverse dead-zones, relay blocks, Look-up tables, fuzzy interference (badly!), all with PI control. Whilst some work well in Simulink, it causes the table to vibrate at high frequency, which would destabalise the substructure. Anyone think of any other solutions, or ever accurately overcome such a problem in practice before? No external components if possible please. Many thanks for any help or comments.

 

If the dead zone is caused by the "shoot through" delay in the PWM servo amplifier, the dead zone is difficult to overcome externally. Some amplifiers are much better in this regard than others. Older PWM amplifiers for brush type motors used "Hard Switching" which virtually eliminated the problem.

 

Nick,
you need real servo amp in current or velocity mode as inner loop. But better way for practical application is using real motion controller instead simulation tool.

 

one method to reduce the effect of dead zone is to increase gain of encoder input by some factor for example by 2, 5, 10 factor (multiply input with 2, 5, 10) is one method.

 

It seems that your dead zone problem is related to mechanical friction. To overcome this problem with simulation you may use for example load feedforward feedback compensating such friction.