PID setting for seismic table


Thread Starter


I am working on a project: Structural Vibration Control using Tuned Liquid Damper at IIT Bombay, India. We simulate earthquake in lab using seismic table. We give earthquake time histories to servo controlled single directional actuator (5 ton capacity having two servo valves 40 lpm each) to which, shake table is connected. We have NI-DAQ, VB Instruments and LabView for data acquisition and it's post processing.

While carrying out preliminary experiments on seismic table, I found that PID settings of the servo hydrolic system need to be set as per frequency content of the input time waveform. Our system being very old (18 years), we do not
have records such as process variables, tuning parameters etc with us. Moreover, I speculate that after many years of use, system parameters may have changed.

Can anybody please write me how to form equation which can be used to monitor PID control settings? I can not approach to Servo Hydraulic people as it has now merged with Instron.

Any advise/comments/hints will be highly appreciated.


I don't really think that PID-control is a good option in the case you describe. The reason I tend to think so is the function of your system. You are not controlling a process, you are designing a servo system. The I-component of the PID-control wouldn't be useful because you don^Òt need to eliminate a static error of a certain process variable in a steady state. The input that your hydraulic system must track is pretty dynamic and the system may not have steady states at all. It would only reduce the stability margins. The D-component of the PID-control wouldn't help either. Vice versa, it may affect the system performance at the input signals you have making the input very noisy. As a result, with the only component (P-component) of the PID
control you wouldn't be able to build a really good system.
A much better approach, which totally agrees with the practice of such system design, would be application of lag-lead compensation with transfer function K*(T1*s+1)/(T2*s+1) where T1>T1 or transfer functions of higher order or even
more sophisticated control methods. The principle is that you try to raise the open-loop gain of the system as high as possible within the spectral range of the input signals you have simultaneously insuring stability (with certain
margins) of the system. This type of compensation is described in many textbooks on control.
Hope this would help.

Dr. Igor Boiko
Consulting in Control is available
(including modeling, simulations and control design)
[email protected]
Tel: 1-403-294-2745