TechArchive

Original article date: June 1998

JACQUI BLAKEMORE, servo product specialist at Mitsubishi shows how increased software power has taken the hard work out of setting up servo drives

It has been said that if experimentation involves more than seven variables it is an art not a science. Tuning servo drives has traditionally been seen as something of an art and in the days of analogue-controlled servo drives with typically 12 or more gain components and CR time constants to set up one can see why.

In more modern products digital technology has eliminated the soldering iron from the setting procedure. In the latest generations auto tuning has taken the place of manual successive approximation as a tuning technique. A typical auto-tuning system is initiated by a manual command – a push button for example. The response of the drive system to a predetermined disturbance is analysed by the system software and appropriate settings are calculated. In the more advanced systems options exist to allow the drive to be optimised for fastest response or minimum overshoot or other specific requirements.

So the problem is solved. Or is it? Unfortunately many servo drive applications do not have one stable set of load and inertia values to tune to. Take winders for example. As a roll increases in size the inertia also increases. Robots are another good example. When the arm is extended unloaded one set of characteristics apply. When the load is picked up another set apply. When the loaded arm moves close to the robot’s waist yet another set apply. Tuning for all three conditions involves a lot of compromise and far from optimum performance under all circumstances. As robots become smaller and lighter in response to commercial pressure for faster systems this particular problem gets worse since the load inertia represents an increasing proportion of the total.

The perfect solution would be for the drive to assess the tuning requirements continuously – to adapt to each new set of circumstances as they arise. How can this be accomplished? In Mitsubishi’s J2 real-time adaptive tuning servo-drive system, a 32bit RISC processor running at very high speeds executes the complex real-time adaptive algorithm with its two degrees of freedom one gain for the real system and one for the ideal model system. Both gains are continuously assessed and automatically updated while the drive is operating. The control system comprises three sections a mathematical model of the ideal system a disturbance suppression section and the real-time auto-tuning section. The ideal model represents the motor and load inertia with one value JA. This model assumes no disturbances or mechanical looseness. It has infinite rigidity allows maximum control gain setting and is configured to respond ideally to system commands. The outputs from this model are the torque commands T1 and speed command WA. Using the model the response of the actual load to a torque command T1 can be predicted when the model inertia is set to approximately the same value as the inertia of the motor and load.

The disturbance suppression section compensates for the differences between the model speed and the actual motor speed. If a torque disturbance decreases the motor speed the system outputs a torque command T2 to increase it. Thus the system suppresses or compensates for the disturbance. The section’s gain G2 is set by the auto-tuning system higher gain gives the section more effect lower gain less effect. The real-time auto-tuning section detects inertia fluctuations by monitoring model speed WA motor speed WM and model torque command T1 during operation. When it senses a change in these values it adjusts the model inertia JA model gain G1 and the disturbance suppression gain G2 in real time while the drive is running. By continuously detecting disturbances and adjusting the model parameters as above the system provides consistently stable operation even for machinery with large load fluctuations and backlash. The J2′s software also allows for manual setting of tuning parameters. Velocity and loop gain can specified as well as acceleration and deceleration times for most systems.

The benefits to the user are considerable. Apart from the obvious reduction in commissioning time and required degree of “black art” skill the system can deliver optimum performance in situations where a single tuning solution will simply not work. In addition machine vibration due to under-damped responses by the drive is reduced and sensitivity to vibration from other sources is all but eliminated.

• Mitsubishi Electric
• Jacqui Blakemore

June 1998