DE10010712A1 - Deviation stabilization in a control system, involves establishing a final setting time for a regulating unit - Google Patents

Abstract

A controller establishes a final setting time for a regulating unit, stabilizing deviations.

Description

The invention relates to a method for correcting a control deviation in a Control path, which is characterized in that that was influenced by the controller Actuator has a finite positioning time.

The reason for the invention of this control method was the control of a fill level in a container. The regulation is to take place via the inlet to the tank by means of a control valve with a control time of 120 s. The flow of the container represents the disturbance variable. ( Fig. 1)

Attempting to control the controlled system with a classic PID resulted in following difficulties.

• - The control value output of the controller changed significantly faster than the valve position could change. As a result of the asynchrony between the controller output and the manipulated variable ( Fig. 2), the control loop tends to be unstable. Furthermore, a bumpless changeover from automatic to manual mode is not always guaranteed.
• - The I component can become disproportionately large in the event of a long-lasting control deviation and thus lead to instability, especially if the actuator reaches a stop.
  If you reduce AI to minimize this effect, it takes longer until a control deviation is corrected.
• - Due to the above problems with the stability of the control loop, it is very difficult to find the optimal parameters for KP, KI, KD, the following claims are enough
• - Stable control loop
• - Fast correction of faults
• - Fast setting of new setpoints
• - No permanent deviation

This is particularly difficult because the three parameters mentioned differ influence each other, and thus not be optimized independently can.

The aim of the invention is to demonstrate a method with which it is possible to Control deviation of a controlled system, which is characterized in that by the Regulator-influenced actuator has a finite actuating time. The The control loop should be absolutely stable, control deviations should be corrected quickly The controller should be easy to parameterize and the control process should be kept simple be that it is e.g. B. is executable on a standard PLC.

Furthermore, a bumpless changeover from automatic to manual mode should be ensured at all times to be guaranteed.

The following recursive calculation rule with general coefficients is given in the literature [1] for all digital P, PI and PID controllers:

y _k = a ₁ . y _k-1 + b ₀ . x _{d, k} + b ₁ . x _{d, k-1} + b ₂ . x _{d, k-2}

y _k : control output of the controller in the current cycle
y _k-1 : control output of the controller in the previous cycle
x _{d, k} : control deviation in the current cycle
x _{d, k-1} : control deviation in the previous cycle
x _{d, k-2} : control deviation in the previous cycle
T: cycle time of the digital control
Tv: retention time
Tn: reset time
Ts: actuating time (time required for a manipulated variable jump from 0 to 100% or vice versa)

Here are the factors for the PD controller

a ₁ = 0 b ₀ = Kr (1 + Tv / T) b ₁ = -Kr (Tv / T) b ₂ = 0

and the factors for the PID controller

a ₁ = 1 b ₀ = Kr (1 + Tv / T) b ₁ = -Kr (1 - T / Tn + 2Tv / T) b ₂ = Kr. TV / T

called.

The method newly developed by us according to the invention is based on the same general calculation rule with the coefficients

a ₁ = 1 b ₀ = 50 / T (1 + Tv / T) b ₁ = -50 / T (Tv / T) b ₂ = 0

calculated.

It becomes clear that this is a modification of the PD algorithm, in which the following parameters have been changed:

• 1. The coefficient a ₁ = 1 was set. This eliminates a permanent control deviation without the need for the reset time (Tn) required for a PID controller.
• 2. In the coefficients b ₀ and b ₁ , the controller gain Kr has been replaced by the expression 50 / T.
  The factor 50 was developed experimentally. The fact that this factor is still divided by the cycle time shows that it is not directly comparable to Kr.
  The size of this factor is, however, only of minor importance for the control quality, since the controller output is normally determined by the maximum possible change in the control variable Δy _k (see explanation below) if there is a clear control deviation.
  The terms b ₀ and b ₁ are only effective for small control deviations or in the steady state.
  While the conventional PD controller calculates y _k from the terms containing the coefficients b ₀ and b ₁ in each cycle, in the newly developed method only the change in y _{k is} calculated from these terms and becomes y _{k- 1} added.

The result is:

y _k = y _k-1 + 50 / T (1 + Tv / T). x _{d, k} - 50 / T (Tv / T). x _{d, k-1}

converted to y _k = x _k-1 + 50 / T (x _{d, k} + Tv / T. (x _{d, k} - x _{d, k-1} ))

with Δy _k = 50 / T (x _{d, k} + Tv / T. (x _{d, k} - x _{d, k-1} ))

one obtains: y _k = y _k-1 + Δy _k

Here Δy _{k represents} the change in the control output per computing cycle.

According to the invention, we limit this change to the maximum change that the actuator allows due to its finite actuating time ( Fig. 3). The actuating time (time required for a manipulated variable jump from 0 to 100% or vice versa) can be found in the documents of the actuator supplier and is communicated to the controller as parameter Ts.

This results in: Δy _{k max} = T / Ts. The following calculation instructions for the new controller are thus obtained.

Δy _k = 50 / T (x _{d, k} + Tv / T. (X _{d, k} - x _{d, k-1} ))

(in the event of setpoint changes, the term (x _{d, k} - x _{d, k-1} ) is set to zero to increase the stability)
if Δy _k <Δy _{k max} then Δy _k : = Δy _{k max}

if Δy _k <-Δy _{k max} then Δy _k : = -Δy _{k max}

y _k = y _k-1 + Δy _k

Here Tv (lead time) is the only parameter to be optimized by the user With more precise knowledge of the controlled system, methods are conceivable in which the Retention time Tv depending on the setpoint, disturbance variable, actuating time, manipulated variable and Control deviation is automatically optimized.

Since a control algorithm is processed cyclically in a digital controller, the control value is changed once per cycle. To make a To determine the actuating speed, knowledge of the cycle time is required. Since the Cycle times of digital controls are partly freely configurable or from Computing effort of the entire program and the performance of the processor depend, knowledge of the cycle time is required. It must be ensured that the cycle time is the same every cycle and not z. B. by hard disk access in some cycles is significantly extended. This cycle time can also be adjusted by suitable Calculations can be determined automatically.

Since the parameterization of the controller essentially by the actuator (or its Positioning speed), we call the controller "AOC" (actuator optimized controller).

Bibliography

[1] Lutz / Wendt Paperback of control engineering Verlag Harm Deutsch Thun and Frankfurt am Main, 1995 Page 419-421

Claims (5)

1. A method for correcting a control deviation in a controlled system, which is characterized in that the actuator influenced by the controller has a finite actuating time. 2. The procedure described here achieves a complete regulation of the Control deviation without using a reset time usual for PID controllers, which significantly increases the stability of the control loop. 3. The procedure described here can be adjusted by adapting only one Parameters (lead time) can be optimized. 4. The process described here is not very computationally intensive and therefore low Requirements for the performance of a digital control. In most Cases is z. B. a standard hardware PLC is sufficient. For extreme requirements to the speed (highly dynamic drives) z. B. a soft PLC in Connection with a powerful processor can be used. 5. The change in the manipulated variable output y _k is limited to the maximum actuating speed of the actuator, which significantly increases the stability of the controlled system. In the case of fast actuators, the actuating speed can be limited to a maximum value desired by the user to protect the material.