JPH0425563B2 - - Google Patents

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a method and device for diagnosing a control system,
In particular, the present invention relates to a control system diagnostic method and device suitable for use in diagnosing whether the state of an automatic control device or the like is normal or abnormal.

[Conventional technology]

In recent years, manufacturing equipment has become larger and more sophisticated.
The health of each piece of equipment tends to affect not only the productivity of each piece of equipment alone, but also the productivity of the entire manufacturing facility, which is made up of a huge number of pieces of equipment. Therefore, it is necessary to confirm the health of the entire manufacturing facility and each component, to quickly identify signs of abnormality, and to take appropriate measures. For this purpose, there is a method to examine the so-called step response, in which the setting signal of the control system is changed stepwise and the corresponding change in the controlled variable is observed, or a sine wave is applied as the setting signal and its frequency is sequentially changed. and observe the corresponding change in the control amount.
A method of examining so-called frequency response has been put into practical use.

[Problem to be solved by the invention]

However, all of these methods of examining the step response or frequency response require the use of step signals or sine wave signals that are completely different from the signals used during operation; therefore, they cannot be carried out during operation. It is possible. (2) Since these input signals differ in shape, amplitude, etc. from signals during operation, they are insufficient as a means for diagnosing the state of the control system during operation. (3) In a control system consisting of multiple control loops, it is impossible to accurately diagnose the state of the observed loop due to interference with control loops other than the observed loop. It had the following drawbacks. That is, for example, as shown in FIG. 1, a main loop regulator A, a minor loop regulator B,
Controlled object C of the minor loop, controlled objects D and E of the main loop, feedback element F of the minor loop, feedback element G of the main loop, element H influencing the parallel loop, and regulator I of the parallel loop. , a control system having a parallel loop controlled object J, an influence element K from the parallel loop, a parallel loop controlled object L, and a parallel loop feedback element M, a minor consisting of elements B, C, and F is considered. The loop is itself a closed control loop, and at the same time
The block B.C/(1+B.C.F) is one of the control targets of the main loop. Therefore, each element B, C, F of the minor loop
The characteristics of not only affect the controllability of the minor loop but also have a significant effect on the controllability of the main loop. In FIG. 1, a is the target value of the main loop, b is the control deviation of the main loop (=an),
c is the operation amount of the main loop (=target value of the minor loop), d is the disturbance of the main loop, and e is the
Minor loop control deviation (=c+d-m), f
is the operating amount of the minor loop, g is the disturbance of the minor loop, h=f+g, i is the controlled amount of the minor loop, j=i・D, k=j・E, 1 is the controlled amount of the main loop, m is the amount of feedback of the minor loop, n is the amount of feedback of the main loop, o is the interference disturbance to the loop that is parallel to the main loop and interferes with each other (hereinafter referred to as parallel loop), p is the amount of feedback of the parallel loop. The target value, q is the control deviation of the parallel loop (=p-w), r is the manipulated variable of the parallel loop, s=
rJ, t is interference disturbance from parallel loop, u=
s·L, v is the control amount of the parallel loop, and w is the feedback amount of the parallel loop. Here, in order to diagnose the controllability of the main loop, when the target value a of the main loop is changed, the response to it is not only A, D, E, and G, which are the main elements of the main loop, but also It is also affected by minor loop elements B, C, and F. Furthermore, in order to diagnose the controllability of the minor loop, when changing the target value c of the minor loop and observing the corresponding feedback amount m of the minor loop, the control amount i of the minor loop is the same as that of the main loop. element D,
The target value c of the minor loop changes via E, G, and A. Furthermore, the control amount 1 of the main loop includes, in addition to the internal state amount k of the main loop,
The internal state quantity j of the main loop is H・I・J・K・
Signal t multiplied by M/(1+I・J・L・M)
, it is also affected by the parallel loop consisting of I, J, L, and M, and cannot be determined only by the main elements of the main loop, A, D, E, and G. On the other hand, as another method for diagnosing the state of a control system, a white noise signal or a variable frequency sine wave signal is applied to the input of a certain element, and changes in the amplitude and phase angle of the corresponding output are used as a transfer function. There is a way to measure it. However, this method can only be applied when the target element is linear and has one input and one output.
When this method is applied to an actual control system, errors occur due to approximation to a linear one-input, one-output system. That is, for example, in a two-input, two-output system as shown in FIG.
as well as the target value (other input) of the parallel loop p
It is also affected by disturbances d and g. Similarly, the control amount (output) v of the parallel loop is
It is affected by two inputs a and p and two disturbances d and g. Therefore, in such a system, even if an attempt is made to extract only the relationship between a certain specific input and output, it is impossible to extract it because the influence of other systems cannot be excluded. Furthermore, as is well known, the transfer function method is based on the Labras transformation, and therefore cannot be applied to nonlinear systems. When analyzing changes in the transfer function of an actual control system, it is important to check whether it is due to changes in the characteristics of linear elements.
It is necessary to distinguish whether this is due to a change in nonlinearity, but this is not possible with the conventional transfer function method. As described above, all of the conventional methods have serious drawbacks, and it has been difficult to accurately diagnose the normality of the control system. The present invention was made in order to eliminate the above-mentioned conventional drawbacks, and it is possible to easily maintain the normality of the control system.
It is an object of the present invention to provide a control system diagnostic method and device that can accurately diagnose a control system.

[Means to solve the problem]

The present invention provides a method for diagnosing a control system, first of all,
When diagnosing the normality of the control system, the square integral value of the control deviation of the control system is calculated, and the square integral value is
When the value is larger than a preset reference value, it is determined that the controllability of the control loop is poor, and for each element of the control system, a coherence function representing the contribution of the input of the element to the output is calculated, and the coherence function is calculated. If the value of is larger than a preset reference value, the relationship between the input and output of the element is determined as a transfer function, and the gain margin and phase margin of the transfer function are determined according to each preset reference value. When the value of the coherence function is outside the range, it is determined that the control characteristics of the element are poor; on the other hand, when the value of the coherence function is less than or equal to the reference value, the control characteristics of the element are not determined to be good or bad; The above objective has been achieved. Further, the present invention provides an apparatus for achieving the above object, comprising: a calculation means for calculating a square integral value of a control deviation of a control system; and a first calculating means for calculating a square integral value of a control deviation of a control system; a comparison means, a frequency analysis means for determining, for each element of the control system, a coherence function representing the degree of contribution of the input of the element to the output and a transfer function representing the relationship between the input and the output of the element; , a second comparison means for comparing with a preset reference value; a third comparison means for comparing the transfer function with each preset reference value to obtain a gain margin and a phase margin; According to the output of each comparison means, when the square integral value is larger than the reference value, it is determined that the controllability of the control loop is poor, and when the value of the coherence function is larger than the reference value. When the values of the gain margin and phase margin of the transfer function are outside the respective preset reference ranges, it is determined that the control characteristics of the element is poor, and the value of the coherence function is determined to be less than or equal to the reference value. When this is the case, it is constructed using a logic circuit that does not judge whether the control characteristics of the element are good or bad.

[Effect]

The present invention has the following features: (1) Regardless of linearity or number of inputs and outputs, the control deviation is
This was done by focusing on the following points: (2) the reliability of the transfer function can be evaluated using the coherence relationship; and (2) the reliability of the transfer function can be evaluated using the coherence relationship. The method of the present invention will be explained below by taking as an example a control loop that includes a minor loop in the main loop and further has parallel loops, as shown in FIG. 1 above. The main loop to be diagnosed has a target value a, a controlled variable 1, a control deviation b, and a manipulated variable c, and the transfer function of the controlled variable 1 to the manipulated variable c is N.
Then, as shown in FIG. 2, it can be expressed as a control loop consisting of a regulator A, an equivalent controlled object N, and a feedback element G. Here, the equivalent controlled object N is not a simple element with one input and one output, but
In addition to the input c, it can be considered that disturbances d, g, and p are added. If this entire control loop is defined as a closed loop 10, then
The input/output a, l of this closed loop 10, the state quantity b,
c, n, disturbances d, g, p, as shown in Figure 2,
input to the diagnostic device 12; This diagnostic device 12 is
For example, as shown in FIG.
A first comparator 16 that compares the product integral value ISE with the reference ISE ₀ , and a coherence function CF of each element of the control system.
and a two-dimensional frequency analyzer 18 for determining the transfer function.
a second comparator 20 that compares the value of the coherence function CF with a reference value CF ₀ ; a third comparator 22 that compares the transfer function with the reference value to obtain a gain margin and a phase margin; Comparators 16, 20,
22, and a logic circuit 24 for determining the normality of the control system based on the output of the control system 22. Diagnosis of the normality of the control system in the diagnostic device 12 as described above is performed as follows. That is, first, the controllability as a control loop is evaluated as the square integral value ISE of the control deviation b. ISE=∫〓 ₀ b ² dt...(1) Here, the integration time τ is set corresponding to the time constant of the control loop. The absolute value of the control deviation b is large in both cases of slow and sensitive control, and furthermore, due to the degree of influence on the operation, the control deviation b
It is appropriate to evaluate using the square of . Note that when evaluating a transient response, it is appropriate to integrate and evaluate the square value of the control deviation b over the elapsed time of the transient phenomenon. If the square integral value ISE is smaller than the reference value ISE ₀ of the loop, there is no problem with the controllability of the control loop, but if it becomes greater than the reference value ISE ₀ ,
Understand the characteristics of each element and make necessary adjustments, including repairs and replacements, to restore normal conditions. Even if the controllability of the control loop can be understood, it is necessary to accurately understand the characteristics of each element within it for correct operation. For this purpose, evaluation is performed using the transfer function shown below. At this time, in order to increase the reliability of the transfer function, it is also possible to perform evaluation using the transfer function in an angular frequency range where the coherence function is reliable. Specifically, first, the target value a and the control amount 1 are set to 2.
A two-dimensional frequency analyzer 18 performs two-dimensional frequency analysis to obtain a coherence function CFal between the two. This coherence function CFal is a function representing the degree of contribution of input a to output 1, and is mathematically determined by the following equation. CFa1=｜C(ω)｜ ² /{I(ω)・0(ω) ² }...
(2) Here, I(ω) and 0(ω) are the auto power spectra of input signal a and output signal 1, respectively, and C
(ω) is the input and output cross power spectrum,
ω is the angular frequency. Coherence function determined by equation (2) above
The larger CFal is, the more the controlled variable 1 is influenced by the target value a. On the other hand, when the coherence function CFal is small, the controlled variable 1 is influenced by inputs (or disturbances) other than the target value a. This indicates that the control amount is strongly influenced, or that the elements between the target value a and the control amount 1 have strong nonlinearity. Therefore, the coherence function CFal is the reference value CFal ₀
If the value is larger, that is, if the controlled variable 1 strongly corresponds to the target value a, the relationship between the controlled variable l (output) and the target value a (input) is determined as a transfer function. On the other hand, the coherence function CFal is the reference value
If CFal is less than ₀ , the control amount 1 is the target value a
Since the transfer function is dominated by other factors, the reliability of the transfer function is low. Therefore, without determining whether the control characteristics of the element (target value a) are good or bad, it is possible to investigate, for example, inputs from others, disturbances, or control characteristics of the element that have a strong influence on the control amount l. When the coherence function CFal is large and the transfer function is determined, controllability is evaluated using the margin of the control loop. For example, when diagnosing after stopping operations,
White noise as a simulated signal at the target value application point, or
Applying a variable frequency sine wave signal and connecting to one input terminal of the two-dimensional frequency analyzer 18,
Furthermore, the control amount 1 corresponding to the simulated signal is connected to the other input terminal of the two-dimensional frequency analyzer 18, and a transfer function between the two inputs is mathematically expressed by the following equation,
a first comparing means for comparing the square integral value with a preset reference value; and a coherence function representing the degree of contribution of the input of the element to the output for each element of the control system, and an output for the input of the element. frequency analysis means for determining a transfer function representing the relationship between the two; second comparison means for comparing the value of the coherence function with a preset reference value; and a second comparison means for comparing the value of the coherence function with a preset reference value. A third comparison means for comparing the gain margin and a phase margin, and the output of each of the comparison means, when the square integral value is larger than the reference value, the controllability as a control loop is determined. When the value of the coherence function is determined to be defective and the value of the coherence function is larger than its reference value, and the value of the gain margin and phase margin of the transfer function are outside the respective preset reference ranges, the element When the control characteristic of the element is determined to be poor and the value of the coherence function is less than or equal to the reference value, the control characteristic of the element is determined using a logic circuit that does not determine whether the control characteristic of the element is good or bad. G=C(ω)/I(ω)……(3) On the other hand, when measuring the transfer function without causing any disturbance to the operation during operation, do not apply a simulated signal and input the element to be measured. and the output are connected to the two input terminals of the two-dimensional frequency analyzer 18. There are two methods for diagnosing controllability from the transfer function obtained in this way. (1) The margin for the phase θf ₁ of -180° at the frequency f ₁ where the gain |G| becomes 0 dB, that is, the phase margin Me = θ _f1 +180°, and the phase θ is -180°.
The margin for the gain |G|f ₂ dB at the frequency f ₂ that becomes, that is, the gain margin M _G = |G|f ₂ is
A method of judging based on how far away it is from the reference values M〓 _O and M _GO , respectively. Although this method is effective for accurately diagnosing the characteristics of a control system, it is inappropriate to use it during operation because it requires obtaining significant signals over a wide frequency range. (2) Gain |G | is the value in the low frequency range |G ₀
The frequency f-3dB when the frequency has decreased _by 3dB from
3dB) ₀ , (f90°) How to diagnose how far it deviates from ₀ . Although this method is somewhat unstable in terms of accuracy,
It can be sufficiently evaluated in a relatively narrow frequency range. By the method described above, the characteristics of each element can be quantitatively evaluated and the normality of the device can be diagnosed. 4A to 4C show examples of measured waveforms of a transfer function, coherence function, cross power spectrum, and auto power spectrum according to the present invention. The coherence function and transfer function can be displayed on a cathode ray tube 26 for observation, as shown in FIG. 2, or can be copied using a hard copy machine 28 such as a dot printer or a digital block. . In addition, the square integral value ISE can be displayed numerically or copied, or can be indicated or recorded in an analog manner using a digital/analog converter and an indicator recorder (not shown). is also possible. Furthermore, by utilizing a computer, it is also possible to sequentially investigate a large number of measurement points and automatically measure the characteristics of each element.

【Example】

FIG. 5 shows a flowchart of an embodiment of the diagnostic method according to the present invention. In this embodiment, an example is shown in which the margin M is used as a method for evaluating the transfer function, but it is of course also possible to evaluate using f-3 dB and f90°. Further, FIG. 5 also describes adjustments, etc. that should be performed when each determination is determined to be abnormal. The flowchart shown in FIG. 5 will be explained in detail below. First, in step 101, the closed loop 10
input/output a, l, state quantities b, c, n, disturbance d,
Read g and p. Next, proceed to step 102,
All counters α, β, and γ representing the diagnostic process are reset. Next, the process proceeds to step 103, where the square integral value of the control deviation b calculated using the above equation (1) is calculated.
It is determined whether ISE is less than the reference value _ISE0 . If the determination result is positive, it is determined that the control system is normal, and the diagnosis is terminated. On the other hand, if the determination result in step 103 is negative, the process proceeds to step 104, where it is determined whether the count value of counter α is 1 or less. If the determination result is negative, proceed to step 105.
Investigate other factors to finalize the diagnosis. On the other hand, if the determination result in step 104 is positive, the process proceeds to step 106, where it is determined whether the coherence function CFal calculated by equation (2) is greater than its reference value CFalo. do.
If the determination result is positive, the process proceeds to step 107, where it is determined whether the margin Mal is within the reference value Malo±ΔMal. If the judgment result is positive, proceed to step 108 and set the reference values ISE ₀ , CFalo,
Review Malo, △Mal. Next, the process proceeds to step 109 to investigate other factors and complete the diagnosis. On the other hand, if the determination result in step 107 is negative, the process proceeds to step 110, where it is determined whether the count value of counter γ is less than 1. If the judgment result is positive, proceed to step 111.
Coherence function CFLn between control amount 1 and feedback amount n
is larger than the reference value CFLno, and whether or not the margin Mln between the two is within the reference value Mlno±ΔMln. If the determination result is negative, the process proceeds to step 112 and the feedback element G is adjusted.
After step 112 is completed, or if the judgment result in step 111 is positive, step
Proceeding to step 113, the coherence function CFcl between the manipulated variable c and the controlled variable 1 is larger than its reference value CFclo,
Moreover, the margin Mcl between the two is the reference value Mcl±△
Determine whether it is within Mcl. If the result of the determination is negative, the process proceeds to step 114, where the transfer function of the equivalent controlled object N is adjusted. After step 114, or if the judgment result in step 113 is positive, the process proceeds to step 115, where the coherence function CFbc between the control deviation b and the manipulated variable c is
It is determined whether the reference value CFbco is greater than the reference value CFbco and the margin Mbc between the two is within the reference value Mbco±ΔMbc. If the judgment result is negative,
Proceed to step 116 and adjust regulator A. After step 116, or if the judgment result in step 115 is positive, proceed to step 117, increment the counter γ by 1, and
Return to step 107 above. On the other hand, if the judgment result in step 110 is negative, the process proceeds to step 118 and the reference value is
CFlno, Mlno, △Mln, CFclo, Mclo, △Mcl,
Review CFbco, Mbco, and △Mbc, and further investigate other factors in step 119, and repeat step 106 above.
Return to If the judgment result in step 106 is negative, the process proceeds to step 120, where the counter β is
It is determined whether the count value of is less than 1. If the determination result is negative, proceed to step 121.
After investigating other factors, in step 122, count up the counter α by 1 and repeat the previous step.
Return to 103. On the other hand, if the judgment result in step 120 is positive, the process proceeds to step 123, where the coherence function CFdl between the disturbance d and the control amount 1, the coherence function CFgl between the disturbance g and the control amount 1, the coherence function CFgl between the disturbance p and the control It is determined whether the coherence function CFpl between the quantities 1 is larger than the respective reference values CFdlo, CFglo, and CFplo. If the determination result is positive, the process proceeds to step 124 and the disturbances d, g, and p are reduced. On the other hand, if the determination result in step 123 is negative, the process proceeds to step 125, where the feedback element G, equivalent control object N, and regulator A are adjusted. After step 125 or step 124 described above is completed, the process proceeds to step 126, increments the counter β by 1, and returns to step 106 described above. Although the above embodiment describes the case of diagnosing the main loop, the main loop can also be used as part of diagnosing the main loop or when diagnosing minor loops or parallel loops for completely different purposes. The diagnosis can be made using the same procedure as explained above. Next, an example of diagnosis when the control state becomes oversensitive in a hydraulic automatic plate thickness control system (hereinafter referred to as a hydraulic AGC system) of a thick plate rolling mill will be described. The hydraulic AGC system consists of an analog control system.
The hydraulic cylinder position control system includes a hydraulic servo valve opening control system, and is further provided with an automatic plate thickness control system in parallel with the hydraulic cylinder position control system. In this control system, the hydraulic cylinder position control system interferes with the automatic plate thickness control system, so the coherence function is small. The hydraulic servo valve opening control system is
Since it is a single control system and can be regarded as an almost linear system, it has a large coherence function. Therefore, in the diagnosis, first, the control deviation of the control system was periodically measured, the measured values were subjected to parallel calculations using a square calculator, and the results were added using an adder. The adder is reset every set number of additions (50 times in this case). This added value was repeatedly printed using a digital printer. Then, this added value (ISE) becomes
The result was 7857/50 times, which was larger than the reference value of 1000/50 times, so when the coherence function CFal between the target value a of the control system and the controlled variable 1 was determined using a two-dimensional frequency analyzer, it was found that the commonly used frequency range was 0. At ~50Hz, the coherence function CFal was above the reference value of 0.9. Therefore, when the transfer function between the target value a, the controlled variable 1, and the manipulated variable c was obtained using a two-dimensional frequency analyzer, it was found that the gain of the regulator was insufficient (70% of the reference value). In contrast, the gain of the regulator is set to the reference value of 102
When adjusted to %, controllability improved dramatically,
ISE returned to normal value at 0.048/50 times.

【Effect of the invention】

As described above, the present invention has the excellent effect of being able to easily and accurately diagnose abnormalities in the control system.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of a complicated control loop to which the control system diagnosis method according to the present invention is applied, and FIG. A block diagram showing the state in which the main loop of the controlled object as shown in Figure 1 is being diagnosed,
FIG. 3 is a block diagram showing the configuration of the diagnostic device, and FIGS. 4A to 4C show transfer functions according to the present invention,
FIG. 5 is a diagram showing examples of measurement waveforms of a coherence function, a cross power spectrum, and an auto power spectrum. FIG. 5 is a flowchart showing an embodiment of the diagnostic method according to the present invention. 10...Closed loop, A...Adjuster, N...Equivalent control object, G...Feedback element, 12...Diagnostic device,
14... Arithmetic unit, 16, 20, 22... Comparator,
18...Two-dimensional frequency analyzer, 24...Logic circuit.

Claims (1)

[Claims] 1. When diagnosing the normality of the control system, calculate the square integral value of the control deviation of the control system, and when the square integral value is larger than a preset reference value, It is determined that the controllability of the control loop is poor, and for each element of the control system, a coherence function representing the contribution of the input of the element to the output is calculated, and if the value of the coherence function is greater than a preset reference value. If so, the relationship between the input and output of the element is determined as a transfer function, and when the gain margin and phase margin of the transfer function are outside the respective preset reference ranges, it is determined that the control characteristics of the element are defective. A method for diagnosing a control system, characterized in that, on the other hand, when the value of the coherence function is less than or equal to the reference value, the control characteristics of the element are not determined. 2. Calculating means for calculating the square integral value of the control deviation of the control system; first comparing means for comparing the square integral value with a preset reference value; and for each element of the control system, the element frequency analysis means for determining a coherence function representing the contribution of the input to the output of the element and a transfer function representing the relationship between the input and the output of the element; and a second frequency analysis means for comparing the value of the coherence function with a preset reference value. a comparison means; a third comparison means for determining a gain margin and a phase margin by comparing the transfer function with respective preset reference values; When the value is larger than the reference value, it is determined that the controllability as a control loop is poor, and when the value of the coherence function is larger than the reference value, the values of the gain margin and phase margin of the transfer function are determined. is outside the respective preset reference ranges, it is determined that the control characteristics of the element are poor, and when the value of the coherence function is below the reference value, the control characteristics of the element are determined to be good or bad. A control system diagnostic device comprising: a logic circuit that does not perform judgment;