WO1993009481A1 - Unite de controle amelioree a reglage automatique - Google Patents

Unite de controle amelioree a reglage automatique Download PDF

Info

Publication number
WO1993009481A1
WO1993009481A1 PCT/US1991/007957 US9107957W WO9309481A1 WO 1993009481 A1 WO1993009481 A1 WO 1993009481A1 US 9107957 W US9107957 W US 9107957W WO 9309481 A1 WO9309481 A1 WO 9309481A1
Authority
WO
WIPO (PCT)
Prior art keywords
error signal
measured
control parameters
controller
overshoot
Prior art date
Application number
PCT/US1991/007957
Other languages
English (en)
Inventor
Peter D. Hansen
Edgar H. Bristol
Original Assignee
The Foxboro Company
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by The Foxboro Company filed Critical The Foxboro Company
Priority to JP50223392A priority Critical patent/JP3470197B2/ja
Priority to AU91210/91A priority patent/AU664980B2/en
Priority to PCT/US1991/007957 priority patent/WO1993009481A1/fr
Priority to DE69129241T priority patent/DE69129241T2/de
Priority to CA002122472A priority patent/CA2122472C/fr
Priority to EP92901622A priority patent/EP0610189B1/fr
Priority to KR1019940701411A priority patent/KR100194451B1/ko
Priority claimed from CA002122472A external-priority patent/CA2122472C/fr
Publication of WO1993009481A1 publication Critical patent/WO1993009481A1/fr

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0205—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system
    • G05B13/024—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system in which a parameter or coefficient is automatically adjusted to optimise the performance

Definitions

  • This invention relates to self-tuning adaptive controllers for controlling a process and more particularly to pattern-recognition, self-tuning controllers.
  • the pattern recognition approach to self-tuning is unique. It uses direct-performance feedback of a monitored process variable to determine the required action for optimizing the process. More specifically, a pattern-recognition, self-tuning controller continuously monitors and automatically evaluates a closed-loop response pattern of a measured process variable to naturally occurring, unmeasured upsets caused by a change in set point or load. After each upset, closed-loop error response characteristics such as overshoot and decay are identified and compared with desired characteristics. Differences between the desired characteristics and the measured characteristics are then used to automatically generate new tuning values for adjusting the controller according to the requirements of the process in order to achieve an error response target shape. The target shape is chosen to approximately minimize integrated absolute error.
  • the first successful adaptive process control system to automatically identify and evaluate process response characteristics of the closed-loop response pattern was invented by Edgar H. Bristol, II and is disclosed in U.S. Patent No. 3,798,426.
  • the adaptation system is triggered by detecting the closed-loop response pattern resulting from an upset having a magnitude which exceeds a preselected noise band.
  • dead and rise times are identified by measuring the time required for the response pattern to reach predetermined percentages of the extremum value of the upset during its first half cycle. The dead time is assumed to bear a significant relationship to any process dead time which is most apparent early in the measured variable response, and rise time is assumed to bear a significant relationship to process closed-loop natural period.
  • the measured rise time is scaled to establish evaluation intervals for the adaptation process. Scaling constants, used to determine these intervals, are selected so that the first half cycle in a resonant response is developed during one evaluation interval and the first full cycle is fully developed during another.
  • the response pattern is evaluated by calculating the integrated difference between the measured process control error, normalized by the magnitude of the first error response peak, and a target value for each of the evaluation intervals. The integrated differences are used to adapt the operating parameters of the controller to improve control action during the next process upset.
  • Kraus' adaptive process automatically recognizes the response as overdamped if, after a pre-specified wait period after verifying the first peak, the second peak is not found. If a second peak is found, the third peak is sought for a time period proportional to the time between the first and second peaks. If the second or third peak is not detected, the search for peaks is terminated and "pseudo" peak values are assigned. Characteristics of the closed-loop response pattern, such as overshoot, damping, and period, are then calculated using the measured extremum values of the response pattern. Differences between these measured characteristics and desired characteristics are then used to calculate new control operating parameters to optimize the control action.
  • Kraus' system also has critical parameters which must be specified by an operator. For example, choosing the pre-specified wait period in Kraus' system is critical, particularly when the process operates over a wide range of conditions. This wait period is critical because it also establishes a wait time before the peak search is activated. If it is set improperly, the system may not operate efficiently. The system also has a tendency to tighten the tuning of the process when the second and third peaks are lost in the noise band. After several disturbances, the response can become excessively oscillatory, causing the system to overcorrect the process which results in overdamping the response. Further, thresholds for the noise band are either user-selected or determined during a pretuning operation for an open loop condition. It does not adaptively respond to condition changes in a closed-loop ' process.
  • the invention features a self-tuning
  • SUB TITUTESHEET controller that characterizes a closed-loop error response for determining appropriate control parameter changes to improve the performance of the process in response to a disturbance.
  • This error response or error signal can be calculated directly from a process controlled variable if the user is only interested in adapting only one parameter such as the proportional band of the controller.
  • the error signal is calculated as the difference between a set point and the measured process variable signal in order to automatically adjust more than one parameter of the controller.
  • a realization of the present invention is that, by a judicious selection of pattern.features available from the error signal, the adaptation of control parameters can be made less sensitive to the disturbance shape and location.
  • the present invention provides a method for automatically adjusting the control parameters of the self-tuning controller used to regulate a process having a measured process variable signal.
  • the method includes generating an error signal representing a closed-loop response of the process to an upset condition. Three successive amplitude values of the error signal are selected and measured to produce measured decay and overshoot characteristics. The three successive amplitudes are selected such that the measured decay characteristic is greater than the overshoot characteristic. At least one of the control parameters of the controller is then automatically adjusted to reduce the difference between one or both of the measured characteristics and target characteristics.
  • the type of process to be regulated is determined and the determined process type is used to calculate changes in the control parameters for achieving rapid convergence to the target characteristic.
  • the type of process may be declared by the user or determined automatically from information provided by the error response and the current controller parameters. More specifically, the process type parameter is used as an interpolating or extrapolating factor relating controller parameters to response characteristics based on predetermined data from at least two different processes.
  • the step of measuring amplitude values can include measuring four or more successive amplitude values of local extrema of the error signal. If the fourth amplitude value is needed but cannot be detected, it can be estimated using the previously measured amplitudes. Three of the four successive amplitude values are then combined to determine the decay and overshoot characteristic values to adjust at least one of the control parameters.
  • the step of locating and measuring the amplitudes may further include locating a knee of the error signal later in time to a first local extremum.
  • the method for automatically adjusting the control parameters of the self-tuning controller may further include the step of sampling a user-selected process variable when a new disturbance response is detected.
  • the user-selected process variable can be subdivided into subranges in which each subrange corresponds to a set of previously adapted control parameters.
  • the subrange of the user-selected variable determines which set of stored controller parameters is most appropriate for the just detected disturbance.
  • the sign of the initial error response can also be used to catalog and select among previous stored sets of controller parameters.
  • a lead-lag filter can also be provided to filter the set-point signal to prevent an excessive error overshoot in response to a set-point change when the controller is well tuned for a load upset.
  • the lag is set equal to the controller integral time.
  • the ratio of lead to lag can be adapted to achieve the desired overshoot target.
  • Fig. 1 is a block diagram of an adaptive process control system according to the present invention.
  • Fig. 2 is an error signal representing a closed-loop response of the adaptive process control system of Fig. 1.
  • Fig. 3 is a detailed block diagram of the adaptive process control system of Fig. 1.
  • Figs. 4-6 are flow diagrams indicating the operating states of the processor shown in Fig. 3.
  • Fig.- 7 shows an error signal having a knee which indicates an overdamped response that has only one peak that can be located and confirmed.
  • Figs. 8 and 9 are flow diagrams indicating the operation of the processor for determining corrective control parameters for the process.
  • Figs. 10-12 are performance maps generated from experimental data for PI control of a pure delay process, an integral delay process and PID control of an integral delay process.
  • the present invention uses performance feedback to update the parameters of a PID controller following each significant control error response to naturally occurring disturbances. Performance is measured using the pattern shape features of the error response, overshoot and decay ratio, which are independent of the amplitude and time scales.
  • the pattern features of an isolated error response provide complete information for updating the control parameters when the response is oscillatory and decay is not equal to overshoot.
  • Targets for the error response features are chosen so that the time scale of the response is optimized when the targets are realized by proper adjustment of the controller parameters such as proportional, integral time and derivative time. It should be noted that the derivative time to integral time ratio may be programmed using the knowledge of type of process to be regulated.
  • the process type can be determined through interpolation or extrapolation of measured integral-time-to-period ratio data with similar data stored for known extreme process types such as pure delay and integral delay.
  • the process type can be declared by the user or operator.
  • Knowledge of the process type is not only used for programmed adaptation of the derivative time but is also used to improve the convergence rate of the proportional band and integral time through interpolation or extrapolation of data relating performance measures to tuning constants for specific process types.
  • overshoot is calculated from the ratio of two successive peak heights.
  • the decay ratio is the half period peak-to-peak ratio involving three successive peak heights.
  • the derivative time can also be determined fa y performance feedback using a full period decay ratio involving a peak-to-peak ratio of four successive peaks.
  • this performance measure is very sensitive to noise, programmed adaptation for derivative time based on a measure of process type is preferred.
  • the present invention makes use of the recognition of a beginning of • a response to choose the most appropriate set from stored sets of previously adapted PID tunings. At the end of the response, this set of stored sets of PID tunings is updated.
  • the PID tunings are correlated with error response direction and/or with subranges of a user-specified variable.
  • an adaptive process control system 10 includes a self-tuning control system 12 and a process 14 which is characterized by a controlled process variable 16, such as temperature, pressure, level, or concentration.
  • Self-tuning control system 12 preferably consists of a PID controller 18 and an adaptor 20 for automatically adjusting or tuning the control parameters of the controller to optimize the closed-loop performance.
  • tuning parameters 25 are calculated using the values of process controlled variable 16 and a set point 26, the difference between these values (defined as an error signal 29 and produced by a summing amplifier 23), the controller output 22, and- a user-specified variable 31.
  • control parameters are tuning "constants," which may include proportional band (P) , integral time (I), derivative time (D) and the lead/lag ratio of a set-point compensator.
  • controller 18 uses some or all of these control parameters, controller 18 produces a control signal 22, which is received by a control element 24, such as a valve, for regulating process 14.
  • Process 14 responds to changes in control element 24 such that the value of controlled variable 16 becomes substantially equal to the desired value represented by set point 26 in the absence of load disturbances 28.
  • the desired result is that process control system 10 be responsive to control error caused by changes in set point 26 or loads 28 on process 14.
  • the controller 18, providing its parameters have been we ' ll tuned will respond accordingly with corrective output 22 action to remove the effects of the disturbance.
  • the adaptor 20 updates the controller parameters to condition the controller for the next disturbance.
  • a general description of the adaptive process control system 10 according to the present invention can be best presented in terms of the behavior of error signal 29 as a function of time shown in Fig. 2.
  • This error signal represents the closed-loop response of adaptive process control system 10 to an upset condition which causes a difference between set point 26 and the value of process controlled variable 16.
  • the upset condition depicted is caused by a load disturbance 28 (such as a change in production rate) which significantly disturbs the value of the process controlled variable 16.
  • the vertical axis of the plot indicates the measured value (E) of the error signal.
  • the horizontal axis is time (T) with an arbitrary base value.
  • the error signal 29 is characterized by as many as four peaks (also known as local extrema) having peak values El, E2, E3, and E4 which occur at times Tl, T2, T3, and T4, respectively.
  • the integral action of controller 18 positions the valve 24 so that the measured value of the error signal becomes substantially equal to zero, which corresponds to the condition where the process controlled variable 16 has returned or changed to its desired value.
  • a noise band (NB) whose value is the expected peak-to-peak variation of the error signal (PPNB) during quiet intervals between isolated upsets and is shown centered on the horizontal time axis.
  • the noise band value is adapted as will be discussed below.
  • the closed-loop response of adaptive process control system 10 can be characterized in terms of decay, overshoot, and time period, which are performance measures that are well known to control engineers for describing the behavior of the control loop. These performance measures together with the control parameters, P, I, and D, may then be used to determine the process type, such as a delay process, integral delay (dominant lag) process, or something in between. New control parameters are calculated for driving the decay and overshoot values to target values for that process. Specific details for measuring these performance values and for calculating the proper control parameters will be discussed below.
  • a more detailed block diagram of a preferred interacting digital type PID controller 18 is shown in terms of an equivalent analog controller element to aid in the explanation of the adaptive control.
  • the PID controller 18 includes four filters: a proportional filter 30, an integral filter 32, a derivative filter 34, and a set-point filter 38.
  • Proportional filter 30 is basically the inverse of the proportional band constant P and is used to determine the responsiveness of controller output 22 to error 29.
  • Integral filter 32 receives an integral feedback input signal which may be the .control signal 22 and generates a lagged positive feedback signal 33.
  • Integral filter .32 is a first-order lag filter whose time constant is the integral time constant I.
  • Derivative filter 34 is used in the measurement feedback loop as shown and operates as a lead-lag filter.
  • Derivative filter 34 may include a second order lag term to attenuate high frequency measurement noise and thereby prevent unnecessary valve activity.
  • the proportional, integral, and derivative terms in these filters are the control parameters determined and controlled, via lines 25a-c, by adaptor 20 and are discussed in greater detail below.
  • controller 18 further includes a limiter 36 which is set to prevent the output 22 from exceeding the range of the control element 24.
  • set point 26 is preferably supplied to controller 18 via filter 38.
  • This filter includes the integral time I, and a "constant" k as shown.
  • filter 38 is not required in a regulator application; i.e., one with a constant set point. However, it is preferred in the present embodiment to prevent a set-point change from causing an excessive overshoot condition, particularly with a dominant lag process. Similar to the other tuning parameters in controller 18, these terms are initially determined by preset, pretuned, or default values and thereafter controlled by adaptor 20 via line 25d.
  • the output of filter 38 is thereafter combined, via a summing amplifier 40, with the output of derivative filter 34.
  • the resulting signal is then multiplied by the inverse of the proportional constant P of proportional filter 30 and summed, via a summing amplifier 42, with the output of integral filter 32.
  • the sum is then limited if necessary by filter 36 to produce control signal 22 for regulating valve 24.
  • the output of integration filter 32 is a positive feedback signal that provides integral action on the control error when controller output 22 is between the limits of the limiter 36, but prevents integral windup when controller output 22 is at one of the limits.
  • Adaptor 20 monitors controller output 22 to detect output limiting.
  • signal 22 could be used as a set point of a secondary controller (not shown).
  • the controlled process variable of that secondary controller would then be fed to integral filter 32 as an external integral feedback signal.
  • the primary process measurement, or measured value is made available to derivative filter 34 and to adaptor 20 as are the controller's output (or -integral feedback) and set point.
  • adaptor 20 includes a processor 50 which receives and processes the controller output 22 (or integral feedback), controlled variable 16, set point 26, error signal 29, and the user-selected variable 31, in accordance with software stored in Read Only Memory (ROM) 52 to produce tuning parameters (P,I,D, and k) . These parameters are subsequently applied to controller 18 via lines 25a-d.
  • ROM Read Only Memory
  • a Random Access Memory (RAM) 54 is provided to contain the data memory and registers required by the microprocessor for implementing the software operations in the ROM.
  • preset control parameters (which may be stored in RAM 54 or in memory, not shown, of controller 18) are transferred to processor 50. These tuning parameters may have been achieved using a pretuning feature similar to that described in the Kraus patent referenced above or may have been selected by an operator.
  • Other settings are also fed to processor 50 and include a minimum, W MIN , and maximum, W j ⁇ , wait times " for establishing time windows for -searching for peak values, an initial noise band NB term and target performance values which can similarly be selected by the operator.
  • the minimum wait time W nicotine_ blend is used in the confirmation of peaks, which will be discussed below. Typically this value is set to 2 or 3 times the sampling time of the adapter and prevents the system from picking out a peak caused by excessive derivative action.
  • the primary operating states include a quiet state 56, four states (States 1-4) in which the processor searches for four consecutive peaks, 58, 60, 62, and 64, and a settling state 68.
  • the processor is looking for a new or continuing transient. If, during the settling state, the error 5 signal remains within the noise band for an adequate time (a half or full period), the settling state 68 is terminated and the processor enters the quiet state 56, Steps 74 and 76.
  • the quiet and settling o states are associated with the condition where the measured value of the error signal is between the upper and lower levels of the noise band. So long as the value of the error signal 29 remains between these levels, no new disturbance is recognized and 5 no changes in the control parameters are made. If quiet state lasts at least one period following an adequately damped response, Step 82, the noise band is measured and updated, Steps 84, 86, and 88.
  • the noise 0 band was a fixed value selected by a user or measured during a pretuning phase as an open loop measurement. According to the present method, the noise band is measured under closed-loop conditions and is updated with changing process conditions. The user at any time can insert a value which will override and reinitialize this adapted value.
  • Updating the noise band requires that the quiet period extend for at least one.natural period of the closed-loop as determined during the last upset, Steps 76, 80, and 82.
  • the period timer is set equal to the maximum wait time, W-. TM , selected by the user or a ⁇ retune specified wait time.
  • the noise band is updated if one of the two following criteria is satisfied: 1. The new noise band value is less than the previous noise band value; or
  • the previously determined overshoot value is greater than or equal to zero; and b) the error is not a continuing oscillation (that is the last observed oscillation started from a quiet condition) ; and c) the previous measured decay ratio is less than or equal to the decay ratio target plus 0.1.
  • the new peak-to-peak noise band is calculated as the root-mean-square (RMS) value of the measured noise (that occurred during the quiet period) times an arbitrary factor, which is preferably set to six standard deviations. Steps 84, 86, and 88. These criteria are imposed to assure that the loop was neither overdamped nor too underdamped when the noise band is allowed to increase.
  • the processor enters the first of four states for locating peaks or amplitudes which will be used as if they were peaks of the error signal.
  • this threshold is set equal to the noise band plus one percent of the measurement range.
  • the first state 58 includes initializing a period timer (a counter incremented with each new error sample), determining the sign of the error signal, and searching for the first peak of the error signal, Step 90. For convenience, if the sign of the error signal is initially negative, it is sign corrected such that the initial deviation from the noise band of the sign corrected error is positive.
  • the uncorrected error sign and the user-selected variable 31 are used to indicate the likely process operating condition for the new response. Based on the value of initial error sign and/or the selected variable, stored tuning values which best exploit past experience in coping with this process operating condition are inserted into the controller, Step 78.
  • An appropriate user-selected variable can include the set point 26 or a measured load 28 which anticipates the new final operating condition.
  • the set point can be used as the user-selected variable as an indicator of the future settled value of the process controlled variable. For example, if the nonlinear process behavior is most dependent on the value of the controlled measurement, the set point (which anticipates the controlled variable) should be selected. For another process, a measured load variable (or the controller output) may be a better choice. By selecting the most appropriate among successful past tunings, the system is better able to cope with process nonlinearity. This is a form of adaptive gain scheduling determined by performance feedback rather than open loop programming based on an identified or specified process model.
  • previous sets of adaptive tuning parameters, P, I, D and lead-lag ratio are stored and indexed according to the error response direction and magnitude values of the user-selected variable.
  • the range of the user-selected variable may be subdivided and indexed into several subranges.
  • the range of the selected specified variable is divided into three subranges, such that for each subrange there are two sets of tuning PID parameters (the set to be used depends on the initial sign of the error signal and the subrange of the user-selected variable) .
  • tuning values are selected and updated to exploit and improve upon past experiences in coping with new disturbances.
  • this concept could be extended to multiple subranges of several variables.
  • the user may elect to select no variable. Then the selected set of tunings would depend on initial error sign alone. For example, the user would not select a variable if the process is linear.
  • the first peak, El is located by comparing the current magnitude of the measured error signal to a previously measured value, hereinafter referred to as a tentative first peak. If the current magnitude of the measured error signal ' is greater than the tentative first peak, then the value of the tentative first peak is replaced with the magnitude of the currently measured error signal and the period timer is reset to zero. This new tentative value is then compared with successively measured magnitudes of the error signal. This recursive step continues until the maximum wait time, W, M..A.X,., expires or when the tentative first peak is confirmed as a bona fide peak.
  • Step 96 Before the peak timer reaches W j - , an attempt is made to confirm that the tentative peak is the first peak, El, and not a noise artifact, Step 96.
  • Table 1 set forth below, lists the preferred criteria for confirming peaks. These criteria are selected to reject local noise peaks and short-period peaks caused- by excessive (mistuned) derivative action. Note that t MIN is the minimum value that the period timer t, may achieve after the first peak before starting information gathering for a particular peak. This "wait" time may be used to avoid detecting some of the high frequency oscillation peaks that may be the result of derivative action misapplied to a pure delay process.
  • the first peak is- confirmed.
  • the method described above for locating and confirming the first peak will also be used in locating and confirming other peaks.
  • processor 50 enters state 2, generally indicated by Steps 100 to 112 (Fig. 5). Initially, the period timer, t, is set to zero at the first peak El and is used for measuring the time since the first peak. If this timer exceeds W Mft ⁇ , the tentative second peak value will be used as the second peak E2 and the controller tuning is tightened to speed up this sluggish behavior. Steps 100 and 102. In this state, processor 50 is searching for an amplitude which may be a second peak E2 or a knee 114 in the error curve (as shown in Fig. 7), which may be used as if it were an unconfirmed second peak. A knee is a pattern feature of an overdamped response that has only one peak (El) .
  • a tentative half period is chosen as 1.5 times the time from the first peak to the tentative knee, Step 106.
  • a time interval equal to two times this tentative half period is used to search for and confirm the second peak.
  • This maximum search time is independent of a previous response measurement or a preset wait time as disclosed in Kraus.
  • a finite search time interval for locating the second peak is established even when a second peak exists.
  • the tentative knee is abandoned if the error response 29 becomes less in magnitude than the negative noise band because a second peak can then be located.
  • the tentative half period is then revised to be the time between the first peak and the tentative second peak. If, however, after two tentative half periods a second peak has not been confirmed, the search for peaks is terminated and the PID calculations are entered, Step 102. The knee is confirmed and used in the PID calculations as the unconfirmed second peak, if during the WiA interval the amplitude difference between the currently measured error response 29 and a corresponding point on the tangent 115 equals the magnitude of the noise band, Steps 108 and 112.
  • the currently measured error response is closer to zero than 0.6 * peak being compared.
  • the currently measured error response is at least NB closer to zero than the peak being confirmed.
  • the search for a third amplitude which again may be a peak or a knee, is begun (state 3).
  • the same procedure used for locating the second peak, described above, is followed, except that the confirmed half period is 5 not further revised. If, during the search, the third tentative peak exceeds the first peak and the second peak is also positive, the first two peaks are abandoned and the search for a first peak is again initiated, Steps 114 and 116. If the process
  • Step 121 ⁇ and PID tuning parameters calculated using this information are entered into the controller, Step 121.
  • processor 50 searches for a knee in the same manner described above, Steps 122, 124, and 126, using the magnitude of the second peak. If the knee value can then be confirmed, it will be used as the unconfirmed third peak, Steps 128 and 130. Providing the third peak is confirmed, a fourth peak will be sought in the same manner described above for a half period after the third peak, Steps 134 and 136 (state 4), Fig. 6..
  • the tentative fourth peak value is less than the second peak value
  • confirmation of the third peak is withdrawn and new PID tunings are calculated based on this information, which indicates that the process is oscillating out of control. There is no attempt to confirm the fourth peak.
  • values for pattern features of the error signal can be determined and used to update the set of PID control parameters for the just-completed responses, Step 98 (Fig. 4).
  • the half-period (T) and dimensionless pattern features which include overshoot (OVR) and decay are determined according to the equation set forth below.
  • variable N depends on the number of peaks that were located and skipped before selecting three successive amplitudes to calculate the above pattern features. Skipping peaks will be discussed in greater detail below. If the fourth peak E4 is lost in the noise band or a fifth peak is needed, a peak can be estimated according to the formula set forth below.-
  • This equation is based on the assumption that the error signal is exhibiting a near optimal decay where the real parts of the three dominant closed-loop poles are identical.
  • peak slipping implies that the second peak will be declared the first peak, the third peak will be declared the second peak, and so on, for the purposes of selecting successive peaks that will be used to calculate new decay and overshoot values for determining corrective control parameters. More simply, if:
  • the first of the three successive peaks used may be tested to determine if it would trigger a peak search. It should be greater than one percent of the full-scale controlled variable plus the noise band in order to calculate the PID control parameters. Step 142. If this condition is not met, there is no PID update because the response is not large enough and the adaptor is switched to the settling state. However the set-point compensator factor k, may. be adjusted, based on the overshoot prior to peak slipping, provided that the error response was initiated by a set-point disturbance. The set-point compensator factor k ranges from 0.2 for a dominant-lag process to 1.0 for a dominant delay process.
  • This factor is decreased by the amount of overshoot which exceeds the target overshoot prior to peak slipping. This factor is increased following a sluggish isolated set-point response (negative overshoot) by an amount equal to the overshoot deviation.
  • the decay and overshoot targets are also appropriately adjusted
  • Steps 144, 146 and 148 (temporarily) after a peak slip, but before P, I, and D are updated, Steps 144, 146 and 148.
  • the adjustment is made so that there will be no change in the PID tuning if the process is disturbed either by an impulse requiring peak slipping or a step requiring no slipping.
  • a first category generally indicated by Steps 150 and 152, deals with information in the following way. If the value of decay is greater than 1.2 or the second or third peak is beyond the measurement range, the control loop is recognized as grossly unstable, Step 150. For this case, the proportional term is doubled, the integral term is set equal to the half period, and the derivative term is set equal to zero. Step 152.
  • the process type (ptype) is also set equal to 1.2.
  • a second category deals with information which indicates an overdamped response with only one confirmed peak and is generally indicated by Steps 154 to 159. If the half-period is greater than or equal to MA ⁇ , the proportional term is halved and the derivative factor is set to zero, Steps 155 and 156. Otherwise, the proportional term is updated according to- the following recursive formula. Step 157.
  • the integral time is multiplied by 0.8, if the overshoot is less than minus 0.075, Steps 158 and 159.
  • the unconfirmed second peak E2 was either the knee 137 or the lowest value found during the search for the second peak E2.
  • the third peak E3 cannot have an amplitude more than the peak-to-peak noise band (PPNB) above the unconfirmed- second peak E2
  • decay is calculated based on this extreme case (provided decay does not exceed its target) . This will cause the proportional band to be larger than it would have been had the decay been assumed zero in equation 6.
  • the control is not overly tightened when the response is so small that the second and third peaks cannot be distinguished from noise.
  • a third category is generally indicated by Steps 162 to 174. This case occurs when the decay minus overshoot is less than 0.02, which indicates that the closed-loop response is close to being that of a damped quadratic, Step 162.
  • the proportional and integral terms are calculated as follows. If the process type is less than 0.2, which is a measure that is between a pure delay and an integral delay (i.e.. close to a pure delay) as determined during an earlier adaptation or set by the user, or decay is greater than 0.3, the integral term is multiplied by 1.4, Steps 164 and 166. The determination of the process type will be discussed in greater detail below.
  • the derivative factor a factor applied to the integral term to determine the derivative time, is divided by 1.4 to keep the derivative time unchanged.
  • the proportional term is multiplied by 0.8, Steps 168 and 170. Otherwise, if the integral term is greater than one and one-half times the half period, the proportional term is multiplied by 1.2, Steps 169 and 171. If decay is greater than 0.9, then the proportional term is multiplied by 1.4, Steps 172 and 174; otherwise, the proportional term remains unchanged.
  • the processor determines whether the response started from a quiet state, decay minus overshoot is less than 0.6, and the half period is less than W j ⁇ , Step 176. If these criteria are satisfied, then the process type is updated based on the new information judged to be complete. Step 178.
  • the process type can change from 0.0 for a pure delay to 1.0 for an integral delay and even larger for a statically unstable process.
  • the process type, ptype is determined from the following equation:
  • the ratio ln(I/T) on the left hand side of the equation (7) is known because I is the current integral time controller value and T is the measured half-period.
  • the ratios, ln(I/T), on the right-hand side of equation (7) for the integral-delay and delay processes respectively, are interpolated from ratio values stored in ROM 52, Fig. 3. In the preferred embodiment, a total of 27 ratio values are stored for each of the three extreme process controller combinations shown in the plots of Figs. 10 through 12. These ratio values, include ln(P/Po), ln(I/Io), and ln(I/T) at each of nine points labeled A-I, on the plots shown in Figs.
  • a point M for example, can be plotted on the three graphs shown in Figs. 10 through 12.
  • the weighting function associated with each of the stored data points is determined and used to find the set of ratios ln(P/P ), ln(I/I ) and ln(I/T) for each process controller combination.
  • Biquadratic interpolation was chosen because the curves generated by the experimental data resembled parabolas. If the controller is to have no derivative action, the stored data values of ln(I/T) at the indicated points labeled A-J in Figs. 10 and 11 can be used to determine the process type (ptype) using equation (7) above.
  • the proportional band P ⁇ and integral time I_ can be calculated according to the following with P and I of equations 8 and 9 replaced by P disregard and I_.
  • dfact is the ratio of D to I.
  • the new derivative factor, dfact may be scheduled using an empirically determined relationship for D/I with ptype. Step 178. If ptype is less than 0.03 (a dominant delay process) the dfactor is permanently clamped to zero until the user overrides this decision; otherwise, if the overshoot is greater than zero:
  • Steps 182, 184 and 186 If the error signal started from the settling state or decay minus overshoot is greater than 0.6, then the parameter changes are divided by 2, Steps 182, 184 and 186.
  • the integral time is adjusted to fall in an expected range relative to the half period.
  • the criteria for determining this condition are set forth in Steps 188 to 198. If overshoot is greater than zero, the integral time is not allowed to exceed 1.5 times the half-period. Steps 187, 188, and 190. If at least one peak is confirmed, the integral time is forced to be at least 0.25 times the half-period, Steps 192 and 194. Also if three peaks have been confirmed, the response is not dominated by the quadratic term, and the integral time is less than 0.35 times (1.0 + ptype) times the time between peaks two and three.
  • the integral time is multiplied by a factor of 1.4, Steps 196 and 198. These steps prevent the integral time from becoming too small when the disturbance is applied gradually.
  • the derivative is thereafter calculated by multiplying the derivative factor times the integral value times 0.16, Step 200. However, the derivative time is not allowed to exceed 0.25 times the half-period.
  • the control parameters are then updated in controller 18, Fig. 3, Step 202.
  • the processor thereafter returns to the settling state 68, Fig. 4.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)

Abstract

Dispositif et procédé servant à ajuster automatiquement les paramètres de commande d'une unité de contrôle à réglage automatique utilisée pour réguler un processus comportant un signal variable de processus mesuré. Un signal d'erreur représentant une réaction en boucle fermée du processus à une condition de dérèglement est produit par l'intermédiaire du signal variable de processus mesuré. On mesure les extrêmes locaux du signal d'erreur et on sélectionne trois valeurs d'amplitude successives, de façon à produire des caractéristiques mesurées de décroissance et de surdépassement du signal d'erreur. On sélectionne les trois valeurs d'amplitude successives, de telle façon que la caractéristique de décroissance mesurée est supérieure à la caractéristique de surdépassement. En se basant sur les caractéristiques mesurées de décroissance et de surdépassement, au moins un des paramètres de commande de l'unité de contrôle s'ajuste automatiquement, de façon à améliorer la différence entre l'une des caractéristiques mesurées et une caractéristique cible.
PCT/US1991/007957 1991-10-28 1991-10-28 Unite de controle amelioree a reglage automatique WO1993009481A1 (fr)

Priority Applications (7)

Application Number Priority Date Filing Date Title
JP50223392A JP3470197B2 (ja) 1991-10-28 1991-10-28 自己調節型コントローラー
AU91210/91A AU664980B2 (en) 1991-10-28 1991-10-28 An improved self-tuning controller
PCT/US1991/007957 WO1993009481A1 (fr) 1991-10-28 1991-10-28 Unite de controle amelioree a reglage automatique
DE69129241T DE69129241T2 (de) 1991-10-28 1991-10-28 Selbsteinstellender regler
CA002122472A CA2122472C (fr) 1991-10-28 1991-10-28 Controleur a auto-reglage ameliore
EP92901622A EP0610189B1 (fr) 1991-10-28 1991-10-28 Unite de controle amelioree a reglage automatique
KR1019940701411A KR100194451B1 (ko) 1991-10-28 1991-10-28 자기동조 제어기

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
PCT/US1991/007957 WO1993009481A1 (fr) 1991-10-28 1991-10-28 Unite de controle amelioree a reglage automatique
CA002122472A CA2122472C (fr) 1991-10-28 1991-10-28 Controleur a auto-reglage ameliore

Publications (1)

Publication Number Publication Date
WO1993009481A1 true WO1993009481A1 (fr) 1993-05-13

Family

ID=25677225

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US1991/007957 WO1993009481A1 (fr) 1991-10-28 1991-10-28 Unite de controle amelioree a reglage automatique

Country Status (2)

Country Link
JP (1) JP3470197B2 (fr)
WO (1) WO1993009481A1 (fr)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0626627A1 (fr) * 1993-05-26 1994-11-30 The Foxboro Company Régulateur de processus auto-adaptatif à temps mort
EP0732636A2 (fr) * 1995-03-14 1996-09-18 ASCOMETAL (Société anonyme) Procédé de régulation d'une installation à comportement non linéaire, notamment d'un four à arc électrique à l'aide d'un régulateur de type proportionnel-intégral adaptatif
WO1997007444A2 (fr) * 1995-08-15 1997-02-27 Rosemount Inc. Systeme de commande de processus a reglage automatique adapte au site
EP1810104A2 (fr) * 2004-10-20 2007-07-25 Fisher Controls International LLC Filtre d'entree avance-retard pour des boucles de regulation electro-pneumatiques
EP2725434A1 (fr) * 2012-10-29 2014-04-30 Omron Corporation Procédé, dispositif et programme de commande de rétroaction

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4309326B2 (ja) * 2004-10-06 2009-08-05 本田技研工業株式会社 プラントの制御装置
US7349745B2 (en) * 2004-10-20 2008-03-25 Fisher Controls International Llc. Lead-lag input filter arrangement with adjustable initial conditions for electro-pneumatic control loops

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4602326A (en) * 1983-12-12 1986-07-22 The Foxboro Company Pattern-recognizing self-tuning controller

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4602326A (en) * 1983-12-12 1986-07-22 The Foxboro Company Pattern-recognizing self-tuning controller

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
ISA TRANSACTIONS. vol. 27, no. 2, 1988, PITTSBURGH US pages 43 - 50; F. RADKE: 'MICROPROCESSOR-BASED ADAPTIVE PID-CONTROLLERS' *
PATENT ABSTRACTS OF JAPAN vol. 12, no. 259 (P-733)21 July 1988 & JP,A,63 046 502 ( YOKOKAWA ELECTRIC CORP ) 27 February 1988 *
PATENT ABSTRACTS OF JAPAN vol. 13, no. 572 (P-978)18 December 1989 & JP,A,1 239 603 ( TOSHIBA CORP ) 25 September 1989 *
PROCEEDINGS OF THE 27TH IEEE CONFERENCE ON DECISION AND CONTROL vol. 3/3, December 1988, AUSTIN US pages 1947 - 1951; MAURICIO ALVES DA SILVA ET AL.: 'A RULE BASED PROCEDURE FOR SELFTUNING PID CONTROLLERS' *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0626627A1 (fr) * 1993-05-26 1994-11-30 The Foxboro Company Régulateur de processus auto-adaptatif à temps mort
US5420785A (en) * 1993-05-26 1995-05-30 The Foxboro Company Self-tuning deadtime process controller
EP0732636A2 (fr) * 1995-03-14 1996-09-18 ASCOMETAL (Société anonyme) Procédé de régulation d'une installation à comportement non linéaire, notamment d'un four à arc électrique à l'aide d'un régulateur de type proportionnel-intégral adaptatif
FR2731810A1 (fr) * 1995-03-14 1996-09-20 Ascometal Sa Procede de regulation d'une installation a comportement non lineaire, notamment d'un four a arc electrique, a l'aide d'un regulateur de type proportionnel-integral adaptatif
EP0732636A3 (fr) * 1995-03-14 1997-07-30 Ascometal Sa Procédé de régulation d'une installation à comportement non linéaire, notamment d'un four à arc électrique à l'aide d'un régulateur de type proportionnel-intégral adaptatif
WO1997007444A2 (fr) * 1995-08-15 1997-02-27 Rosemount Inc. Systeme de commande de processus a reglage automatique adapte au site
US5691896A (en) * 1995-08-15 1997-11-25 Rosemount, Inc. Field based process control system with auto-tuning
WO1997007444A3 (fr) * 1995-08-15 1998-02-05 Rosemount Inc Systeme de commande de processus a reglage automatique adapte au site
EP1810104A2 (fr) * 2004-10-20 2007-07-25 Fisher Controls International LLC Filtre d'entree avance-retard pour des boucles de regulation electro-pneumatiques
EP1810104A4 (fr) * 2004-10-20 2010-07-28 Fisher Controls Int Filtre d'entree avance-retard pour des boucles de regulation electro-pneumatiques
US7917233B2 (en) 2004-10-20 2011-03-29 Fisher Controls International Llc Lead-lag filter arrangement for electro-pneumatic control loops
EP2725434A1 (fr) * 2012-10-29 2014-04-30 Omron Corporation Procédé, dispositif et programme de commande de rétroaction

Also Published As

Publication number Publication date
JPH07503082A (ja) 1995-03-30
JP3470197B2 (ja) 2003-11-25

Similar Documents

Publication Publication Date Title
US5587896A (en) Self-tuning controller
US5406474A (en) Self-tuning controller
US5153807A (en) Self-tuning controller apparatus and process control system
US4602326A (en) Pattern-recognizing self-tuning controller
US5355305A (en) Pattern recognition adaptive controller
US5166873A (en) Process control device
US6937909B2 (en) Pattern recognition adaptive controller
EP0462815A2 (fr) Commande adaptative à horizon variable avec des moyens pour minimiser les coûts de fonctionnement
US5541833A (en) Multivariable feedforward adaptive controller
US5170341A (en) Adaptive controller in a process control system and a method therefor
Zervos et al. On PID controller tuning using orthonormal series identification
USRE33267E (en) Pattern-recognizing self-tuning controller
EP0139243A2 (fr) Appareil de commande de processus
US6959219B2 (en) Control apparatus having a limit cycle auto-tuning function
WO1995004956A1 (fr) Procede et dispositif de commande de logique floue a reglage automatique
AU664980B2 (en) An improved self-tuning controller
EP0626627B1 (fr) Régulateur de processus auto-adaptatif à temps mort
WO1993009481A1 (fr) Unite de controle amelioree a reglage automatique
Tavakoli et al. Robust PI controller for load disturbance rejection and setpoint regulation
KR100194451B1 (ko) 자기동조 제어기
CN1033473C (zh) 一种改进的自校正控制器
Seki Adaptive IMC-PI controllers for process applications
JPS63116204A (ja) 適応制御装置
EP0554442B1 (fr) Commande a action directe multivariable et adaptative
GB2173019A (en) Pattern-recognizing self-tuning controller

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AU CA JP KR SU

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IT LU NL SE

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
WWE Wipo information: entry into national phase

Ref document number: 2122472

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: 1992901622

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1992901622

Country of ref document: EP

WWG Wipo information: grant in national office

Ref document number: 1992901622

Country of ref document: EP