Relay feedback experiment

To initiate an ATV test, the process should be at steady-state or near steady-state conditions, Cq and Jo- Next, the controller output is set io Cq + h (or Cg - h) until y deviates significantly from Jq. At that point, the controller output is set to Cq - / (or Cq + h), which will turn the process back toward jo- Then, each time y crosses yo. the controller output is switched from Cq + h to Cq - h or from Cq - h to Cg + h. The process is also referred to as a relay feedback experiment. A standing wave is established after 3 to 4 cycles therefore, the values of a and the ultimate period, P , can be measured directly, and the ATV test is concluded. The ultimate gain, K, is calculated by... 

The relay feedback experiment was made popular in the field of process control by Astrom and Hagglund (1984). This experiment was suggested as a means to automate the Ziegler-Nichols scheme for determining ultimate gain and frequency information about a process. Their approach followed directly from a describing function approximation (DFA) to the nonlinear relay element. The objective was to use the obtained process information for automatic tuning of PID controllers. 

Recursive Estimation from Relay Feedback Experiments... 

Portions of this chapter have been reprinted from Automatica 35, L. Wang, M.L. Desarmo and W.R. Cluett, Real-time estimation of process frequency response and step response from relay feedback experiments , pp. 1427-1436, 1999, with permission from Elsevier Science. 

The next part involves controller tuning. We must determine the tuning constants for the controllers in the plant. While this task is often performed by using heuristics and experience, it can sometimes be a nontrivial exercise for certain loops. We recommend using a relay-feedback test that determines the ultimate gain and period for the control loop, from which controller settings can be calculated (Luyben and Luyben, 1997). 

Astrom and Hagglund s work (1984) has prompted research in several different directions. One of these directions, and the focus of Chapter 8, is in the area of process identification, where the objective is to obtain a more complete and accurate model of the process firom data generated under relay feedback. Fitting a more complete process model (i.e. a transfer function model) normally requires knowledge of several points on the process Nyquist curve. Given that the standard relay experiment combined with the DFA identification technique is able to identify only a single point, fitting such a model either requires the availability of some prior process information (e.g. Luyben, 1987) or requires the user to conduct a series of relay experiments in... 

In the circuit of Fig. 12.7, there is an action that "feeds back" and makes the operation of the relay weaker, rather than stronger. This action is simple and direct — it is the disconnection of power to the coil. In the first experiment, the capacitor is not attached. A high pitched buzz is heard, because the armature vibrates (oscillates) at a fairly high frequency. The negative feedback occurs too soon for the armature to get all the way down to the NO contacts. 

As discussed in the previous section, a standard relay experiment typically produces a limit cycle that is dominated by a single frequency. However, this information is not sufficient for the estimation of an accmate process step response model. This raises the issue of how to generate the appropriate information. One approach is to inject a dither signal while the process is under some sort of feedback control, either additively to the controller output or via the setpoint. However, this requires design of the dither signal, i.e. decisions must be made concerning its power spectrum. Another alternative is to make use of multiple relay experiments to generate frequency response information at several frequencies. 

Past experience and feedback on electro-mechanical safety relays confirm this reasoning a relay that remains in a raised position, despite an interruption of power, occurring, on average, several billion times more often than not. And it is man, in this case a maintenance agent, who preventively or remediaUy forces the automation, constituted by a signaling installation, to stay within its area of operation. 

---