FIELD OF THE INVENTION The present invention relates to a method of performing control of a closed loop control system and a computer-readable medium having stored therein instructions for causing a processing unit to execute such a method and a control unit, e.g. an actuator, adapted to be used as part of a system controlled in a control system.

BACKGROUND OF THE INVENTION

The PID regulator is very common for the control of industrial processes and provides for proportional, integrating and derivative control. A process of larger scope employs a large number of such regulators. PID regulators are manufactured in large series as standard products. It is more and more common that the regulators are based on microcomputers, whereby more complicated control functions can be used.

Either because of inappropriate settings of the parameters in the PID regulator or because the characteristics of the process change, oscillations can occur in the control signal from the PID regulator to the controlled elements. These oscillations are undesired since they introduce wear out of the controlled elements, that are actuators and valves, in the system. Further, these oscillations introduce a control of the process, which is not optimal.

OBJECT AND SUMMARY OF THE INVENTION

The object of the invention is to solve the above mentioned problems.

This is obtained by a method of performing control of a control system controlled by a controller, wherein said controller delivers the control input to at least one control unit in said system e.g. a valve, said method comprises the steps of:

- detecting the oscillation level in said control input to said control unit, - amplifying said control input by a multiplication factor based on said detected oscillation level.

Thereby characteristics of the control unit are changed as oscillations are detected. Thereby oscillations are minimized, and e.g. the wear of the control units is reduced and service intervals are increased. Further by minimizing oscillations of the control unit the energy consumption of the control unit are lowered and an improved control is obtained.

The oscillation level is an indication of how much the input signal to the control unit oscillates. The oscillation level is an indication of control loop stability. A high oscillation level indicates a nearly unstable control system or that th system is very close to instability.

The control unit could e.g. be an actuator for a valve or a similar device.

The controller could e.g. be a PID controller, a Pl controller or a P controller or another controller.

The control system could be a closed loop control system, but also other control systems are relevant.

In an embodiment said oscillation level is detected by calculating a decay ratio, wherein said decay ratio is used for determining said multiplication factor. Calculation of decay ratio is a special simple way of determining oscillation level.

In an embodiment a range of decay ratios is chosen as the optimal decay ratio resulting in a multiplication factor of 1.

In an embodiment a decay ratio lower than said optimal decay ratio results in a multiplication factor more than 1. This can improve control performance if e.g. a closed-loop response is too slow In an embodiment a decay ratio higher than said optimal decay ratio results in a multiplication factor less than 1. This can reduce overshoots and oscillatory behavior of the control unit in the closed-loop configuration.

In an embodiment the method further comprises the step of determining the amplification of the control input for the entire working range of said control unit based on said determined multiplication factor.

In an embodiment said method of determining the amplification of the entire working range comprises the steps of:

- determining the amplification value for at least three predefined working points representing said entire working range,

- interpolating said amplification values.

In an embodiment the amplification of said control input has to be between a maximum and a minimum preset amplification. Otherwise permanent periodic disturbance might decrease control input amplification toward zero.

The invention further relates to a computer-readable medium having stored therein instructions for causing a processing unit to execute a method according to the above.

The invention further relates to an actuator such as a valve adapted to be used as part of a system controlled in a control system by a controller, wherein said controller delivers the control input to said control unit, said unit comprises:

- means for detecting the oscillation level in said control input from said controller,

- means for amplifying said control input by a multiplication factor based on said detected oscillation level.

The invention further relates to a control system comprising a control unit according to any of the claims 10-12. BRIEF DESCRIPTION OF THE DRAWINGS

In the following, preferred embodiments of the invention will be described referring to the figures, where figure 1 illustrates a control loop and the steps comprised in the method of controlling according to the present invention,

figure 2 illustrates an example of dependency between measured decay ratio and multiplication factor,

figure 3 illustrates a control loop and an embodiment of further steps comprised in the method of controlling according to the present invention,

figure 4a illustrates the amplification of a control unit being a valve interpolated between three working points,

figure 4b illustrates a relationship between gain reduction and valveinput,

figure 5 illustrates the change of the amplification in the three working points illustrated above after oscillations have been detected,

figure 6 illustrates a method of detecting oscillations in the control input.

DESCRIPTION OF EMBODIMENTS In figure 1 a closed loop control system 100 is illustrated. The system comprises a PID controller 101 receiving the control error e as input, where the control error has been calculated by subtracting the process output from a reference r. Based on the control error, the PID controller calculates a controller output given to the system 103 as control input. The control input is given to a control unit (CU) 105 from which the system can be influenced, e.g. an actuator for a valve or the like. The control unit responds to the control input and influences the process Gp(s). If the control unit is an actuator for a valve, it responds by changing how open or closed the valve is. According to the present invention, instead of using the output of the controller 101 as control input to the control unit 105, the control input is being processed as illustrated in the flow diagram 109. Here the control input is low pass filtered in 111 (F_CI), then in 113 (OSC) it is detected whether there are oscillations in the control input. If oscillations are present, then in 115 the control input is amplified (A_CI) with a multiplication factor typically less than 1 in order to dampen the control input to the control unit. If no oscillations are detected, the algorithm is ended in 117 (E). The algorithm as described could be performed each time the general control program for controlling the process is executed, e.g. each time, whereby oscillations are detected as they occur as an integrated part of determining the control input to the control unit.

Further to the presence of oscillations being detected, also the degree of oscillations is determined and used for determining the value of the multiplication factor.

The decay ratio of the oscillations is calculated as:

Decay ratio (DR) = (peaki - peak2) / (peak3 - peak4),

where peaki is the last peak and peak 4 is the first detected peak.

High values of decay ratio means that the system is very oscillating and close to instability. The values more than 1 denote an unstable system.

In figure 2 an example of dependency between calculated decay ratio DR and multiplication factor MF is illustrated. For optimal value of decay ratio the range from 0.15 to 0.25 is chosen 201 , and if the measured decay ratio DR is in this range, the multiplication factor is 1 and the control input will not be changed.

If a smaller decay ratio is determined, the damping of the control input is too high, and the multiplication factor is more than 1 in order to increase control input to the control unit. As the decay ratio is getting smaller than 0.15, the multiplication factor is linearly increasing.

Further, if a higher decay ratio is determined, the damping of the control input is too low, and the multiplication factor is less than 1 in order to decrease control input to the control unit. As the decay ratio is getting larger than 0.25, the multiplication factor is linearly decreasing.

Note that the interval 0.15 - 0.25 is just an example where tests have shown successful results (good tracking and disturbance rejection properties), but other intervals having another size and position could be chosen depending on the control system and the characteristics of the control unit. Further, the multiplication factor needs not be linearly decreasing and increasing around the interval - the main importance is that oscillations are damped via a multiplication factor less than 1 , and too much damping is amplified via a multiplication factor more than 1.

In figure 3 a control loop is illustrated 301 similar to the one in figure 1 and where further steps are comprised in the method of controlling. The oscillation detection is performed in 304 and consists of the control input being low pass filtered in 303 (F_CI) in order to remove high frequency noise. Further in 305 (OSC) it is detected whether there are oscillations in the control input. If oscillations are present, then in 307 the decay ratio is calculated (C_DR) as described above.

In 309, based on the calculated decay ratio, the corresponding multiplication factor is determined (D_MF), and the control input is amplified with the multiplication factor. Finally, in 311 the amplification of the control input for the entire working range (AJEVM) is calculated and used. If no oscillations are detected, the algorithm is ended in 313 (E). The algorithm as described could be performed each time the general control program for controlling the process is executed, e.g. each time, whereby oscillations are detected as they occur as an integrated part of determining the control input to the control unit. In figure 4a an example is given of the step 311 in figure 3 where the control unit is a valve. The working range of the valve is represented with three points 401, 403, 405 - these points represent amplification of the valve at 0%, 50% and 100% openness of the valve. Amplification can never be higher or lower than preset value (parameter gainjnax (e.g. value 1) and gain_min (eg. value 0.2)). Further, the value of amplification in the range 0% to 50% and in the range 50% to 100% is determined by interpolating the three points representing the working range.

The reduction of gain can e.g. be calculated in selected points (e.g. o%, 50% and 100%) by using the following function:

g=1 - (1-k)./(1 + 10*(v-c).*(v-c)),

where g is gain reduction at different values of valve input signal (v), c is a particular valve input signal at which oscillations occur, and k is the gain reduction for oscillations in the point c.

In figure 4b the relationship between v and g is illustrated in an example where oscillations have been detected near 40%. Here v is the valve input in the range 0 to 1 and g is the gain reduction for oscillation detected in c = 0.4 and with central reduction k=0.5.

An example of control according to the present invention is that if calculations show that a reduction of the amplification of a factor 0.5 is needed, then amplification is reduced in all three points, but not in the same degree for all points. Amplification is reduced most in the point nearest the point where the oscillations appeared and least in the point most distant to the point of the oscillation. Algorithms for calculating the reductions for each point could in an embodiment be algorithms similar to the ones used in fuzzy control systems. In figure 5 the change of the amplification in the three working points is illustrated, where after oscillations have been detected near 40% of openness of the valve. Amplification is reduced by suitably changing the three points representing the working range. Most changed is the point near the range where oscillations appeared (50%) and the least changed point is the most distant point (100%). The working range after this change is then represented with three points 501 , 503, 505. These points represent amplification of the valve at 0%, 50% and 100% openness of valve. Again the value of amplification in the range 0% to 50% and in the range 50% to 100% is determined by interpolating the three points representing the working range.

In figure 6 the oscillation detector, illustrated as 304 in figure 3, is further described Initially, in 601 the algorithm is started. In 603 the input signal is filtered (F_CI) and in 604 it is determined whether the routine shall search for a negative or a positive peak (S_P_P), this depends on which peak where detected previously and if it is the first iteration the algorithm starts searching for a positive peak. In an alternative embodiment the routine could also start by searching for a negative peak.

In 605, if a positive peak is confirmed (P_P?) (a positive peak is confirmed if the signal is less than a maximum detected value minus hysteresis), the peak is counted in 607 by adding to a counter a new positive peak (A_N_P_P).

In 609, if a negative peak is confirmed (N_P?) (a negative peak is confirmed if the signal is more than a minimum detected value plus hysteresis), the negative peak is counted in 611 by adding to a counter a new negative peak (A_N_N_P).

In 613, if more than 2 peaks have been detected (P > 2), then in 615 the first peak is disregarded and detection of other peeks (oscillation detection) is initiated, the first peak is deleted in order to avoid oscillations invoked by changing set point, if this is not deleted the later calculated decay ratio might be inaccurate. In 617 if a sufficient number of peaks (e.g. 4 (P > 4)) is detected the decay ratio is calculated in 619 (C_DR) e.g. as described previosly. If in 621 it is determined that oscillations have already been detected, then in 627 a flag is set that a new decay ratio is calculated (SF_N__DR_C), and in 629 all parameters are initialized and a new search is started (lnit_S).

If in 621 oscillations have not been detected yet, then in 623 it is determined whether the decay ratio is very high (1 or more) (D>=1) meaning that serious oscillations is present and this might be because of an instable loop or disturbance occuring when searching for peaks, and if this is the case then in 625 it is determined that oscillations are detected and one peak is discarded (OSC_D_DP) this is in order to ensure that the decay ratio really is above 1 , whereby one more peak will be searched for.

In 631 the algorithm is ended (E).