SE442921B - SELF-ADJUSTING REGULATOR - Google Patents

Description

7802097-1 2 ning. The purpose is thus to find a suitable sampling time for a stable control based on a different maturity-k, ie the number of sampling intervals until the system responds. This avoids the risk of an unstable system where k is too small. The invention is further illustrated in the attached figure 2, which shows a block diagram of a self-adjusting regulator according to the invention. Figure 1 is an illustration of the invention: back gum (see above).

The process, which is intended to be controlled by the self-adjusting regulator according to the invention, is shown at 1 (figure 2). A control value is added to the process 2. | o The device includes a linear, digital controller 3 of, for example, PID character, to which, among other things, a reference value is supplied. is it connected to a calculation and setting means 5 for optimal setting of the control parameters, adaptation of the control parameters. A further control and setting means 6 for setting the appropriate sampling time (T) is also connected to the controller at a sampling device 7 for switching on and off the control signals (via 8) (adaptation of the sampling time T).

The sequencing sequence is that the controller 3 is switched on at each sampling time point, while the adaptation can take place according to one of the following three alternatives: 110 1) the same frequency as the controller 3 2) slower frequency than the controller 3 3) during installation or only in the event.

The controller according to the invention is thus completely self-adjusting, ie no parameters need to be adjusted manually when connecting the controller 3 to the process 1. The controller with its adapters 5-6 functions as a time mist system where the measurement signals from the process are sampled (T) and the sampling values are adapted to a model for the process, suitably built into the controller. Based on the model, an optimal steering signal is selected.

Unlike previously known self-adjusting regulators, in the case of the invention, the sampling time T / dislccet-running time does not need to be adapted to a model fox: Proe- 0985921- 10 15 20 25 7802097-1 According to an embodiment of the invention, for example. fix the discrete maturity in the process (= - the number of sampling times T / continuous maturity) to 2 or more (see figure 1). The sampling times the controller chooses between can be illustrated in the following formula (recursive): = ---- T. _ (1) Ti are thus the different sampling times. You start with To and go up until you find the appropriate T, which of course happens automatically in the controller.

It can be shown that there is exactly one sampling time of those generated by said formula (1), which suits the process. In addition, a few sampling times are generated to be tested. For k = 2, TO = 20 msec Only 13 different sampling times are needed to cover. range 20 msec - 82 sec.

To find the correct sampling time of the generated sampling times according to; formula (1) during the izx adjustment, the controller evaluates a loss G, consisting of a weighted square sum of measurement sigzals at consecutive sampling times. . 2 2 g = yzp fi) + Å. [y (t-mk fl z + X2 Ef (t-zæk fl +} \ 5 Ef (t-smk fl + (2) The correct sampling time Tk is the one for which the following applies: (A) ß has local mmm in sk, section G ( ik) <min ((š (wk_1), G (TkHD where Tk is generated by the formula (1).

(B) G is stationary, ie 1 GN (Tk) / 603196 1 P) 9 where p is selected as the appropriate constant (p = P1 _ {(N, 00) if hypothesis testing of _ 2 "regulator optimal" with rice level OC 'is desired ). f; T) means the loss evaluated during the N last sampling times.

N k G shall thus. meet these two. conditions for the system to be stable.

If, for example, the ratio according to B falls outside (š-, p) during the hypothesis test, the system becomes unstable. Üvsozoav-1 10 15 4 According to another utföz-Lngg 'form is fixed on. same 'as above the discrete duration k to any integer k Z 2.

The sampling time is allowed to change at disk times tN according to the formulas below. alm = tN + m - TOGN), where m is a fixed arbitrary integer Z1 and T, the sampling time is determined in the time tN. now; filtered measured value züN) times as a function of measured quantity: in the sampling fi- points tN_1, tH_1 + T (tn), ..., tN as weighted sum. of the following quantities: 1121; sšyïtN) = ä) y (tN_1 + i * m (tN_1)) y (tN_1 + (in) m (tN_1)) 11z = 1 = s; f <«N> = iåå y <fl N; 1 + i»> uT > 111 syßN) = _5_Oy (tN_1 + ieeæ (tN_1)) 1, _ The adaptation device 6 in Figure 2 is designed as a self-adjusting controller with the same calculation methodology as in 3.95, Figure 2, with the only difference that z (tN) is considered as "measured value" and T (tN) (sampling time at time tN) as "control value".

The device as above can be varied in many ways within the scope of the following patent claims.

Claims (2)

1. 7802097-1 UI 1- "» - 1. Device for self-adjusting controllers, which controller comprises a linear, digital regent (5), a; e: possibly PI or Pin powers, wherein 1 the device also includes a: meteumngedrg (5) for optimal setting of the control parameters, characterized in that the device includes a sampler (7) for control axles, and that another regulating and adjusting means (6) is connected to the controller. for controlling the sampling device (7), for the purpose of adapting the sampling time for optimal control, wherein new sampling time is arranged to be calculated for each setting. 2. A device according to claim 1, characterized in that the calculation-Benet is metaubem a diekret laptid (k), after which since this inteente a for the system etdbilieet 'Suitable eempligget time (m) is arranged to be selected automatically. Device according to claim 1, characterized in that the sampling time (es) is arranged in a selection according to the following gene: the following formula 'where' Bi is the different sampling times and where appropriate T is arranged to be selected in the controller.