CN1033473C - An improved self-tuning controller - Google Patents

Abstract

The present invention relates to a control parameter automatic-regulating and a method for an automatic tuning controller by adjusting a measured process variable signal procedure. A measured process variable signal is used to generate an error signal showing the closed loop response of a process to a disturbance condition, measure the partial extreme value of the error signal and select 3 continuous amplitude values in order to generate the measuring delay and the over-tuning characteristic of the error signal. The three continuous amplitude values are selected to make the measured attenuation characteristic of the error signal larger than the over-tuning characteristic. Based on the measured attenuation characteristic and the over-tuning characteristic, at least one controller control parameter is automatically adjusted to improve a difference value between a measured characteristic and a target characteristic.

Description

A kind of improved self tuning controller

The present invention relates to be used for the self-correcting adaptive controller that process is adjusted, more particularly, relate to pattern-recognition, self tuning controller.

Self-tuning mode identification method is unique.The direct performance feedback of the process variable of this method utilization monitoring is come the desired operation of deterministic process optimization.More particularly, pattern-recognizing self-correcting controller monitor continuously and automatically assess tested process variable for normal appearance and because the closed loop response pattern of the immeasurablel interference that variation caused of set point or load.After each the interference, closed loop error responses characteristic all can be identified such as crossing the mediation decay, and it is compared with desired characteristic.Difference between desired characteristic and the characteristic that measures is used to produce automatically new corrected value again, adjusts controller with the requirement according to process, so that reach the target waveform of error responses.The select target waveform makes the absolute error of integration be bordering on minimization.

Automatically first of the process response characteristic of identification and assessment closed loop response pattern overlaps successful auto-adaptive process control system, be H.Edgar H.Bristol, the II invention, and at United States Patent (USP) the 3rd, 798, open in No. 426 documents, according to the method for Bristol, self-reacting device is to exceed the closed loop response pattern that the interference in predetermined noise district causes by its amplitude and trigger by detecting.In case trigger, reach the predetermined identification retardation time needed time of percent value and the rise time of the extreme value of interference by response modes in the semiperiod before measuring.Suppose that retardation time is closely related with the most significant any process lag time in the early stage measurable variable response, the rise time then is assumed to be with the closed loop of process closely related free period.

The rise time of measuring is decided with scale, so that be the time interval that adaptive process is determined assessment.Selection makes the preceding semiperiod of resonance response form in an evaluation interval in order to determine the scale constant in the time interval, and first complete period then all forms in another interval.The assessment of response modes is that integral difference is carried out between the process control error measured by calculating and each evaluation time desired value at interval.Wherein survey the process control error, by the size normalization of first error responses peak value.Integral difference is used to the operating parameter of Correction and Control device, so that improve control action during next process is disturbed.

For the efficient adaptive Process Control System, several key parameters must be set by operating personnel.For example, the appropriate selection of scale constant is crucial for the rational evaluation time of definition at interval.General scale constant can not be used for dissimilar processes.This control system also needs operating personnel each evaluation time to select suitable desired value at interval.These desired values derive the experience of given process from operating personnel usually.

Thomas W.Kraus has proposed a kind of controller of improved pattern-recognizing self-correcting.According to the method for Kraus, when error surpasses the nominal noise threshold value, start adaptive process.In case start, just the closed loop response pattern is monitored, so that detect three continuous thresholds occurring at first or peak value (" peaks ") and their time of first peak value appearance relatively.

The over-damp control loop response of three peak values appears not having usually, so if the scheduled wait after checking out first peak value after the time second peak value do not occur, the Kraus adaptive process automatically is identified as overdamped response with this response.If second peak value exists, so, with first and second peak values between time cycle of being directly proportional time in, seek the 3rd peak value.If second or the 3rd peak value can not detect, just stop the peak value search, and specify pseudo-(" pseudo ") peak value.

Then, utilize the extreme value of the response modes measure to calculate closed loop response mode characteristic such as overtravel, damping and cycle.Characteristic that these are measured and the difference between the desired characteristic are used for calculating new control operation parameter again, so that the optimal control effect.This method has detailed description in the U.S. Patent number 4,602,326 of Thomas Kraus, the exercise question of this patent is " pattern-recognizing self-correcting controller ".As in conjunction with this paper awarding two United States Patent (USP)s that patent gives Bristol and Kraus describe as a reference, system comes correcting controller at last interference.If process is non-linear process, this can cause the unoptimizable of the controller that the next one is disturbed to proofread and correct.

Some key parameters of Kraus system also must be set by operating personnel.For example, in the system of Kraus, particularly process operation is when carrying out under range of condition widely, and it is critical selecting the scheduled wait cycle.This latent period why key is because before triggering peak value searching, and it has determined that also first-class treats the time.If latent period is provided with inappropriate, system just can not operate effectively.When not having second and the 3rd peak value in grass, this system also has a kind of trend of correction of fixation procedure.After several times disturbed, response may become overshoot oscillation, caused that system causes overdamped response to the exaggerated correction of process.Moreover, the threshold value of grass or selected or under open loop condition, determine during pre-adjustment by the user.This variation to condition in the closed loop procedure can not automated response.

The object of the present invention is to provide a kind of adaptive controller with correcting algorithm, described algorithm is to the mode characteristic sensitivity of the error responses under interference effect, and is then not too responsive with respect to the position that main procedure lags behind to the type disturbed or waveform and it.

Another purpose of the present invention is to provide a kind of method of comparatively fast restoring by utilizing the flex point detection scheme to find out accurate the second and the 3rd peak value from overdamping is proofreaied and correct.

Another object of the present invention is to be reduced to the number of the optimal control effect that reaches desirable, key parameter that operating personnel must set.

Another purpose of the present invention is to prevent the imbalance of disturbing with respect to a class when interference of second class or set point change occurring.

Another purpose of the present invention is when the departure mode characteristic of desired value and measurement can't be distinguished from noise, the prevention adaptively correcting.

Another purpose of the present invention is that definite adaptively first-class is treated the time after detecting a peak value, can not find out additional peak value so that guarantee before stopping searching peak value.

Another purpose of the present invention is to utilize when transient response begins successful correction in the past to proofread and correct non-linear process.

Another purpose of the present invention is to provide the method for revising the noise band in a kind of steady-state period between the error responses that separate.

Further aim of the present invention is to provide a kind of energy to determine the self tuning controller of process in the main procedural type parameter that distinguishes with master delay of lagging behind selected correction parameter so that help for the process that is suitable for specific type between the two.

In general, the self tuning controller that the present invention characterizes, it is characterized in that for development to the response performance that disturbs, the closed loop error responses that can determine suitable controlled variable variation.If the user only is concerned about a parameter of adjusting as the controller proportional band, then these error responses or error signal can directly be calculated by the process controlled variable.But in most preferred embodiment, in order automatically to adjust the more than one parameter of controller, error signal is to calculate as the difference between set point and the tested process variable signal.A kind of realization of the present invention is pattern feature by selecting carefully to obtain from error signal, makes the adjustment of controlled variable not too responsive to the waveform and the position of disturbing.

Therefore, the invention provides a kind of method that can automatically regulate the controlled variable that is used to adjust self tuning controller with a tested process variable signal.This method contains the error signal of the expression process of generation to the closed loop response of disturbed condition.For producing decay and the toning characteristic of measuring, select and three amplitudes in succession of measured error signal.These three in succession amplitude choose should make measurement attenuation characteristic greater than the toning characteristic.For reduce one or two the difference between the characteristic of surveying and target property, at least one controlled variable of controller can be adjusted automatically.

According to an aspect of the present invention, determine the type of controlled process, and the procedural type of determining is used for the variation of calculation control parameter to converge to target property rapidly.Type that it should be noted that process may perhaps be determined by information and existing controller parameter that error responses provided automatically by instruction manual.More particularly, be the characteristic of response based on the predetermined data of at least two various process, the procedural type parameter is used as the controller parameter relevant with interpolation or extrapolation factor.

The step of measuring amplitude comprises four or more continuous amplitude of measured error signal local extremum.If need the 4th amplitude, but detect less than, then can utilize the amplitude of before having measured to estimate.Then, three in the amplitude are combined in succession is used for determining that decay and toning characteristic are to adjust at least one controlled variable with four.The step of seeking and measure amplitude comprises that also searching is in first local extremum error signal flex point after a while.The time interval between first local extremum and this flex point is determined the time scale of error signal response.Then, this time interval is used to determine the time interval of seeking second local extremum.If this mistake at interval before finding out local extremum, then described flex point just is taken as amplitude, and stops the peak value search.

According to another aspect of the present invention, adjust the method for the controlled variable of self tuning controller automatically, can also comprise the step of the process variable sampling of when new disturbance response is detected, the user being selected.The process variable that the user selects can be subdivided into subrange, and each subrange is all corresponding to one group of previous controlled variable that adopts.The subrange of user's choice variable determines that the controller parameter of which group storage is the most suitable to firm detected interference.Classify in the controller parameter group that the symbol of initial error response also can be used for formerly storing and select.

For preventing the extreme error toning of the variation of response set point when having proofreaied and correct controller, also provide a lead-lag filter to filter configuration point signal at the load interference.Being provided with lags behind equals the integral time of controller.For reaching desired toning target, can adjust the ratio of lead-lag.

Above-mentioned and other purposes of the present invention, characteristics and advantage will be more obvious from the following detailed description to most preferred embodiment of the present invention, and as shown in drawings, wherein identical label relates to the same section among all different figure.

Fig. 1 is the block diagram according to auto-adaptive process control system of the present invention.

Fig. 2 is a kind of error signal of auto-adaptive process control system closed loop response shown in the presentation graphs 1.

Fig. 3 is the detailed diagram of auto-adaptive process control system shown in Figure 1.

Fig. 4 to Fig. 6 is the process flow diagram of explanation processor working state shown in Figure 3.

Fig. 7 illustrates has the error signal that an expression only has the flex point of the overdamped response that a peak value can be found out and confirm.

Fig. 8 and Fig. 9 are that expression processor deterministic process is proofreaied and correct the process flow diagram of controlled variable.

Figure 10 to Figure 12 is the performance graph that is produced by the experimental data to the PID control of the PI of pure delay process, integration deferring procedure control and integration deferring procedure.

For change the parameter of PID controller after each big departure response to the interference of normal appearance, the present invention has used performance feedback.That utilizes error responses comes measurement performance with irrelevant model shape feature, overtravel and the attenuation rate of amplitude and time scale.When response be vibration and decay when being not equal to overtravel, the error responses pattern feature of isolation provides the complete information that is used to upgrade controlled variable.When realizing target when suitably adjusting controller parameter such as ratio, integral time and derivative time, the selection of the target of error responses characteristic should make the time scale optimization of response.It should be noted, can utilize the knowledge sequencing of called procedure type derivative time to the ratio of integral time.The type of process can be by integral time of measuring carrying out interpolation to the class likelihood data of the known extreme procedural type of the ratio data in cycle and storage such as pure delay and integration delay or extrapolation is determined.The way of replacing is that the type of process can be by user or operating personnel's explanation.The knowledge of procedural type not only is used for the program adjustment of derivative time, but also the interpolation or the extrapolation that are used to the meter constant of data by the pertinence energy measurement and special process type are improved the rate of convergence of proportional band and integral time.

For oscillatory response, calculate overtravel according to two peak height ratios in succession.Attenuation rate is to comprise three semiperiod peak-peak value ratios of peak height in succession.Can also utilize derivative time comprise four in succession the complete period attenuation rate do as one likes of the peak of peak value-peak value ratio can feed back to determine.But, because the measurement of this performance is very responsive to noise, so be preferred modification of program derivative time of measuring based on procedural type.

Moreover the present invention utilizes the identification of response beginning, comes the optimal setting of selection from the pid correction of the previous employing of storage is provided with.In the end of response, this during the pid correction of storage is provided with is provided with and changed.Pid correction and error responses direction and/or relevant with the subrange of the variable of user's appointment.

Referring now to Fig. 1, auto-adaptive process control system 10 comprises that a self-aligning control system 12 and is by the controlled process 14 that characterizes such as the controlled process variable 16 of temperature, pressure, level or concentration.Self-aligning control system 12 is preferably formed by PID controller 18 with in order to the adaptive mechanism 20 of the controlled variable of automatic adjustment or correcting controller, so that make the closed-loop characteristic optimization.In general, tuner parameters 25 (below go through) is that poor (be defined as error signal 29 and produced by summing amplifier 23) between the numerical value that utilizes process controlled variable 16 and set point 26, these numerical value, output 22 and user's named variable 31 of controller calculate.Controlled variable can be the correction " constant " that comprises the lead-lag ratio etc. of proportional band (P), integral time (I), derivative time (D) and point compensation device.Utilize in these controlled variable certain several or whole, controller 18 produces one and can receive for the control module 24 as valve, in order to the control signal 22 of adjustment process 14.

The variation of process 14 response control units 24 causes when non-loaded interference 28, and the numerical value of controlled variable 16 is substantially equal to by the represented expectation value of set point 26.Like this, desired result is that the departure that Process Control System 10 can cause the variation or the load 28 on the process 14 of set point 26 responds.For example, if the big value that must be enough to 14 pairs of process controlled variables 16 of the process that causes of load variation amount is made suitable change, if its parameter is proofreaied and correct fully, controller 18 responds with the effect of proofreading and correct output 22, so that eliminate the influence of disturbing.If the error responses waveform is represented the interference of suboptimum and suppresses that 20 changes of adaptive mechanism controller parameter disturbs so that controller adapts to next time.

According to the present invention, the general description of auto-adaptive process control system 10 is preferably by shown in Figure 2 representing the characteristic of error signal 29 as the function of time.This error signal represents that 10 pairs of auto-adaptive process control systems cause the closed loop response that the disturbed conditions of difference between set point 26 and the process controlled variable 16 has been done.Described disturbed conditions is that load disturbs 28 (as the variations of throughput rate) to cause, and it upsets the numerical value of process controlled variable 16 significantly.But such as previously mentioned, perhaps disturbance is that the unexpected variation of set point 26 causes.The vertical coordinate of curve is represented the measured value (E) of error signal.Horizontal coordinate is the time (T) that has any base value.Isolating under the vibration situation, error signal 29 is to be characterized by four peaks (being also referred to as local extremum) that have respectively the peak E 1, E2, E3 and the E4 that occur at moment T1, T2, T3 and T4.At last, the integral action of controller 18 is determined the position of valve 24, makes the measured value of error signal be substantially equal to zero, and this has restored corresponding to process controlled variable 16 or has been condition for expectation value.One grass (NB), during the steady interval between the isolated disturbance, its value is the variation of desired error signal peak-peak, and is that center line illustrates with the horizontal time axis.In 10 operating periods of system, the numerical value of grass is selected to be discussed below.

Utilize the measuring amplitude of error signal and the time of appearance thereof, the closed loop response of auto-adaptive process control system 10 can characterize with decay, overtravel and time cycle, and the performance measurement of above-mentioned these description control loop qualities is that control engineering teacher is known.These performance measurement postpone (the main hysteresis) process or marginal process together with the type that controlled variable P, I and D can be used to determine process again such as deferring procedure, integration.Make the desired value of the process that reaches with overtravel for ordering about to decline, need to calculate new controlled variable.Measure these performance numbers and calculate suitable controlled variable, will carry out detailed discussion below.

Referring now to Fig. 3, shown in the figure is the more detailed block diagram of controller 18 of best interactive digital formula PID type, and it is by the analog controller unit of equivalence and help to explain adaptive control.PID controller 18 comprises four wave filters: 32, one differential filters 34 of 30, one integration filters of a scale filter and a set-point filter 38.Scale filter 30 comes down to the inverse of proportional band constant P, is used for determining the response of 22 pairs of errors 29 of controller output.Integration filter 32 receives an integral feedback input signal (can be control signal 22), and produces a positive feedback signal 33 that lags behind.Integration filter 32 is first-order lag wave filters, and its time constant is integration time constant I.Differential filter 34 is used in as shown in the figure to be measured on the backfeed loop, and operation is then as lead-lag filter.Differential filter 34 can comprise a second-order lag item, so that attenuate high frequency is measured noise, thereby prevents the activity that valve is unnecessary.Ratio, integration and differential in described these several wave filters is all to be that via line 25a, 25b, 25c are determined by adaptive mechanism 20 and the controlled variable of control, detailed discussion as after.In most preferred embodiment, controller 18 also comprises one for preventing to export the limiter 36 that 22 scopes that exceed control module 24 are provided with.

10 operating periods of system, set point 26 is preferably supplied with controllers 18 through wave filter 38.As shown in the figure, described wave filter comprises I and " constant " k integral time.As dotted line was indicated, wave filter 38 was optional in regulator applications, that is, and and the wave filter that links to each other with the constant set point.Yet, in the present embodiment,, preferably prevent the excessive toning state that set point change causes especially for main Delay Process.Similar with other correction parameters in the controller 18, these initially with pre-set, precorrection or default value mode determined that via line 25d is controlled by adaptive mechanism 20 then.

Afterwards, the output of wave filter 38 is combined with the output of differential filter 34 through summing amplifier 40.The consequential signal that produces multiply by the inverse of scale filter 30 proportionality constant P again, through summing amplifier 42, with the output addition again of integration filter 32.Then, if necessary with asking sum again by wave filter 36 amplitude limits, so that be that control valve 24 produces control signals 22.The output of integration filter 32 is positive feedback signals, when the output 22 of controller is between two limits values of limiter 36, this positive feedback signal provides the integral action to departure, but when controller output 22 is one of two limits values, then prevents integral windup.The output 22 of adaptive mechanism 20 monitor controllers is so that detect the restriction of output.

The people that control technology is known is clear that, is not with control signal 22 direct by-pass valve controls, but with the set point of signal 22 as the subcontrol (not shown).The controlled process variable of subcontrol is as the outside integral feedback signal integration filter 32 of feeding again.Similar with previously described auto-adaptive process control system, the measurement of main procedure or measured value should make it output (or integral feedback) as controller and set point and be applicable to differential filter 34 and adaptive mechanism 20.

Also show the detailed diagram of adaptive mechanism 20 among Fig. 3, receive and the processor 50 of the variable 31 that processing controller output 22 (or integral feedback), controlled variable 16, set point 26, error signal 29 and user select comprising one, produce correction parameter (P.I.D and k) according to the software that is stored in the ROM (read-only memory) (ROM) 52.These parameters via line 25a, 25b, 25c, 25d then impose on controller 18.The random-access memory (ram) 54 that provides contains in order to realize needed data-carrier store of software operation and register among the ROM by microprocessor.

When processor 50 initialization for the first time, will preset controlled variable (can be stored among the RAM 54 or the storer (not shown) of controller 18 in) and send processor 50 to.These correction parameters utilized describe with the front that similar precorrection characteristic obtains in the Kraus patent documentation or the person of being operated select.Other value of setting processor 50 of also feeding is comprising the minimum value W in order to stand-by period of setting up the Time Triggered pulse of searching peak value _MIN, maximal value W _MAX, initial grass NB item and the target capabilities value that can select by operating personnel similarly.Minimum latency W _MINBe used to confirm peak value, this is discussed below.Usually, this value is taken as 2 to 3 times of adaptive mechanism sampling time, and anti-locking system sorts out because the peak value that the excessive differential action causes.Going through of relevant other these values of setting carried out below.

Referring now to the process flow diagram of the processor 50 shown in Fig. 4-6, after be initialised (step 70), normally describe by six main modes of operation.Main mode of operation comprises that 56, four of stable states are used for the state 58,60,62 and 64 that processor is searched four continuous peak values, and a transient process state 68.

At stable state 56 or transient process state 68, processor is sought a new or continuous transient process.If during the transient process state, error signal is stayed in the grass one section appropriate time (half or complete period), then finishes transient process state 68 and processor enters stable state 56 (step 74 and 76).

Differentiation between these states is that if when processor is in the transient process state, error signal 29 surpasses the noise band and adds 1% of full scale measurement, and then error signal is predicated continuous oscillation (or stack response).If response starts from stable state, then be identified as to isolate and disturb.Then, processor 50 begins to search the mode characteristic of error signal 29 in step 78.

With reference to Fig. 2, the measured value of stable state and transient process state and error signal this condition between the upper and lower limit of grass is related.As long as the value of error signal 29 is retained between the upper and lower limit, just discerns the interference that does not make new advances and do not change controlled variable.If after the response of suitably damping, stable state continues one-period at least, step 82, then grass is with measured and change, step 84,86 and 88.

In the adaptive system in front, grass is the fixed value selected of user or is measuring in the pre-calibration phase of measuring as open loop.According to method of the present invention, grass is measured under the closed loop condition, and changes along with the variation of process condition.The user can insert a value at any time and replace and reinitialize this and be suitable for value.

As determined during the disturbance the last time, the change grass requires to extend steady-state period at least one closed loop free period (step 76,80 and 82).At first, cycle timer is set and equals the maximum latency W that the user selectes _MAXOr the stand-by period of the appointment of precorrection.If one of two following criterions are satisfied, then grass just is modified:

1. new grass value is less than previous grass value; Perhaps

2.a) previous determine cross tone pitch more than or equal to zero; With

B) error is not continuous oscillation (vibration that promptly observes at last starts from a stable state); With

C) the previous attenuation rate of measuring is less than or equal to the attenuation rate desired value and adds 0.1.

What note is to satisfy second criterion and allow the grass change to be big or smaller value.New peak-peak grass multiply by that any factor that preferably is made as six standard deviations calculates, step 84,86 and 88 by measuring noise (occurring in steady-state period) root-mean-square value.Adopt above-mentioned criterion to guarantee when allow increasing grass, control loop is neither overdamping underdamping too.

In case error signal 29 is passed the threshold value relevant with grass, step 74, processor enters first state of the four condition that is used to look for peak value or amplitude, and described amplitude will be taken as the error signal peak value and use.In most preferred embodiment, this threshold value is set equals the number percent that grass adds measurement range.First state 58 comprises an initialized cycle timer (counter increases progressively along with each new error sampling), determines the symbol of error signal and first peak value of searching error signal, step 90.For simplicity, negative if the error signal symbol is initially, then it be performed such symbol correction make with the deviation of the grass of symbol correction error for just.Moreover when error signal was at first detected, the variable 31 that not calibrated error symbol and user select was used for representing the process operation condition of new response.According to the value of initial error symbol and/or selected variable, the corrected value that has preferably utilized the past experience of this class process operating conditions and stored is inserted controller, step 78.

Appropriate users choice variable can comprise set point 26 or anticipate the sensing lead 28 of new final operating conditions.For example, set point can be used as the indicator of the following stationary value of process controlled variable that resembles user's choice variable.For example, if the characteristic of non-linear process mainly depends on controlled measured value, should select set-point (pre-service controlled variable).For another process, tested load variation (or output of controller) may be to select preferably.Optimal by selecting in the correction in the past of success, this system can better overcome the non-linear of process.This is that do as one likes can be fed back rather than based on determining or programme a kind of form of determined adaptive gain layout of the open loop of assignment procedure model.

In a preferred embodiment, previous adaptively correcting parameter group P.I.D and lead-lag ratio are to be stored and to retrieve according to the direction of error responses and the size of user's choice variable value.The scope of user's choice variable can be segmented and retrieve into the experimental process scope.For preferred embodiment, the scope of selected particular variables is divided into three subranges, makes each subrange all have two groups to proofread and correct pid parameter (depend on the initial symbol of error signal and the subrange of user's choice variable and use this setting).Select and change these corrected values, so that tackled new interference according to adopting and improve with the experience of living.Certainly, this notion can expand to a plurality of subrange situations of some variablees, still further, it is noted that the user can choose the mode of no choice variable.Proofread and correct the selection that is provided with and only depend on the initial error symbol.Such as, if process is linear, the user is choice variable not.

Existing amplitude by the error signal that will record is located first peak value that is referred to as to infer below the first peak value E1 with the comparison of previous measured value.If the current amplitude that records error signal is greater than first peak value of inferring, first peak value of Tui Ceing is just replaced by the amplitude of the up-to-date error signal that records so, and period counter is reset to zero.This new guess value is compared with the amplitude of the error signal that records in succession again.This circulation step proceeds to maximum latency W _MAXCrossed or the first honeybee value is confirmed to be till the real peak.If surpassed W from the time that the first peak of inferring begins before first peak value is identified _MAX, infer that so first honeybee just is identified, and the up-to-date error amount that records is by as the unacknowledged second honeybee value E2, step 92 and 94.When the controller correction was very blunt, this criterion was used to stop the peak value search.It is too late blunt that processor 50 utilizes this imperfect information that pid correction is not wanted again, and step 98 will be as below will discussing.

Reach W in peak counter _MAXBefore, always attempt to confirm that the peak value of inferring is first peak value E1 rather than the simulator and noise, step 96.Following table 1 has been listed the preferred criteria of confirming peak value.Select these criterions to be used to get rid of the local noise peak and reach the short period peak value that undue (imbalance) differentiate effect causes.Notice t _MINBe after the first peak, before the beginning concrete peak information gathering, period counter t ₁The minimum value that may reach.This " wait " time can be used to avoid detecting some higher-order of oscillation peak value that the differential action caused that may be added in the pure delay process by mistake.As indicating, if first peak value of inferring has surpassed grass with the difference between the current error, and existing error is less than 0.6 times supposition first peak value, and then first peak value just is identified.Above-mentioned location and confirm the method for first peak value will be used for the location equally and confirm other peak values.

Suppose that first peak value makes and confirm, processor 50 proceeds to usually by step 100 to the represented state 2 of 112 (Fig. 5).At first, cycle timer t ₁Be set to zero constantly at the first peak value E1, and be used for from the first peak value beginning Measuring Time.If this timer surpasses W _MAX, then second peak value of Tui Ceing will be taken as the second peak E2 and static controller and proofread and correct so that quicken this inertia quality, step 100 and 102.In this state, processor 50 is searched the amplitude of the flex point 114 that may be the second peak E2 or graph of errors (as shown in Figure 7), and available this amplitude is as unidentified second peak value.Flex point is the mode characteristic that the overdamped response of a peak value (E1) is only arranged.

When the flex point of inferring is positioned at by error responses 29 and t=0 on the drawn positive tangent 115 of the point of 0.6 times of the first peak E amplitude a bit on.With first peak value to 1.5 times of choosings inferring the flex point time as the semiperiod of inferring (step 106).2 times the time interval that equals this supposition semiperiod is used to search and confirm second peak value.This maximum search time is independent of previous response measurement or as disclosed presetting the stand-by period in the Kraus patent documentation.So, when having second peak value, set up the limited search time interval of seeking second peak value.The flex point of abandoning inferring if error responses 29 become on amplitude less than negative grass is because can find second peak thus.Then, infer that the semiperiod is corrected for first peak value and infers time between second peak value.But if after the semiperiod of two suppositions, second peak also is not identified, and then stops the search of peak value and enter PID calculating step 102.If the difference in magnitude on the current error responses that record 29 and the positive tangent 115 between the corresponding point is at W _MAXEqual the amplitude of noise band at interval, then flex point just is identified and is used as unacknowledged second peak value, step 108 and 112 in PID calculates.

Table 1:

Peak value is confirmed test

1. the current error responses that record are than 0.6 times more approaching zero of the peak value that is compared.

2. the current error responses that record are than the peak value that is identified near at least NB on approaching zero program.

3. in order to begin to confirm following peak value, since nearest peak value, over and done with time enough:

A) t _MIN=0, for first peak E1;

B) t _MIN=W _MIN, for the second peak E2; With

C) t _MIN=2 times of time intervals between E1 and E2 are for the 3rd peak E 3.

When second peak value has been identified, just begin to search the 3rd amplitude (state 3), this amplitude may be again peak value or flex point.Except the confirmed semiperiod no longer revised, the same process of above-mentioned search second peak value was performed.During searching, if the 3rd peak value of inferring has surpassed first peak value and second peak also is positive, then preceding two peaks are abandoned and are started the search of first peak, step 114 and 116 once more.If searching between the 3rd peak period, the process control variable has exceeded its scope, and termination is searched and the scope restriction of multilated is used as unacknowledged the 3rd peak estimated value in the new PID calculating, step 118 and 120.On the other hand, if the 3rd peak value of inferring greater than first peak value, but second peak value is born, and the product of the first and the 3rd peak value be not less than second peak value square, think that then response is unsettled, step 120, termination search peak value and the new PID that enters based on this information calculate.But, if starting from the transient process state rather than start from, response represents continuously or the stable state of overlapping response, then stop the search of the 3rd peak value and will utilize the pid correction parameter of this information calculations to input to controller, step 121.

Two semiperiods after second peak value are allowed to confirm the 3rd peak value.If during this period, be not identified, stop search and enter suitable substance P ID calculating.If the time behind second peak is no more than two semiperiods and error signal 29 remains negative, processor 50 utilizes the amplitude of second peak value with above-mentioned identical method searching inflection point, step 122,124 and 126.If the flex point value can be identified, it will be used as unacknowledged the 3rd peak value, step 128 and 130.

If the 3rd peak is identified, in the semiperiod after the 3rd peak value, will search the 4th peak value, step 134 and 136 (state 4), Fig. 6 with above-mentioned same method.In most preferred embodiment, if the 4th peak value of inferring is cancelled the affirmation to the 3rd peak so less than second peak value, and the pid correction new according to this information calculations, this shows that the process vibration is out of control.Do not plan to confirm the 4th peak value.

The amplitude that utilization records can be determined the mode characteristic value of error signal and be used for setting for the response just finished change pid control parameter, step 98 (Fig. 4).Semiperiod (T) and comprise overtravel (OVR) and the dimensionless mode characteristic that postpones is determined by following equation.

(1) toning=-E _(2+N)/ E _(1+N)

(3) T=T _(2+N)-T _(1+N)Variable N depends on and is selecting three continuous amplitudes so that be positioned before calculating above-mentioned mode characteristic and the number of the peak value skipped.The peak value of skipping will more go through below.If the 4th peak E 4 lost or the 5th peak value of needs, then can estimate peak value according to following formula in grass. $\left(4\right)E_{(3+N)}=\frac{E_{(1+N)}\×E_{(2+N)}}{E_{\left(N\right)}}$ This equation is based on such hypothesis, and promptly to present the real part of its three main closed loop limits are the accurate optimized attenuations that equate to error signal.

One group of peak value in succession that will be used to change controlled variable is according to determining step 140, Fig. 8 by first group three peak values (N=0) the comparison pad value and the overtravel of processor 50 location.If decay is less than overtravel, " peak moves " (Peak slip) will appear in processor 50 so, and promptly three peak values (N=1) based on next group calculate new pad value and overtravel, and determine whether second secondary peak moves (N=2) necessary.In other words, the peak moves and means that second peak will be by as first peak, and the 3rd peak will be by as second peak, or the like, its purpose is the controlled variable of proofreading and correct into determining, selects in order to calculate new decay and to cross the peak value in succession of tone pitch.In brief, if $\left(5\right)E_{(1+N)}\&times;E_{(3+N)}<E_{(2+N)}^2$

Then exist a secondary peak to move.

In case obtain making the N value of first product of equation (5) more than or equal to second product, then can test use three in succession first of peak value so that determine whether it triggers the peak value search.For calculating pid control parameter, it should be greater than the one percentage point plus noise band of full scale controlled variable, step 142.If this condition does not satisfy, then enough do not have the PID change greatly, and adaptive mechanism is converted into steady state (SS) because of response.But, if being disturbed by set point, error responses start, then the compensator factor k of set-point can move prior to the peak according to toning and be adjusted.Set point compensator factor k span from main Delay Process 0.2 to 1.0 of master delay process.This factor reduces one and surpasses the toning quantity that the target of moving prior to the peak is crossed tone pitch.Along with the inertia (negative toning) of isolating the set point response, the amount that this factor increases equals the toning deviation.After the peak moves but before P.I. and D are modified, decay and also suitable being adjusted of toning target (temporarily) step 144,146 and 148.If do this adjustment so that process is disturbed by the pulse of peak value movement requirement or the step of not having a movement requirement, in pid correction with no change.

Here, the ratio of the overtravel of error signal, decay and integration half period is determined, and can be used in now the PID adjustment process according to classifying from the integrality of error signal acquired information.Usually press following method process information by the first kind of step 150 and 152 expressions.If pad value exceeds measurement range greater than the 1.2 or second or the 3rd peak value, then control loop is identified as serious instability, step 150.For this situation, proportional is doubled, and integral is set equals the semiperiod, equals zero step 152 and differential term is set.The type of process (Ptype) is set up and equals 1.2.

Second class is handled expression only an information of confirming the overdamped response of peak value, by step 154 to 159 represented.If the semiperiod is more than or equal to W _MAX, proportional gets half and differential divisor is set to zero, step 155 and 156, otherwise proportional is by following recursion formula change, step 157.

(6) P=P * (1+1.5 (decay-decay target)) moreover, if overtravel then multiply by 0.8 integral time, step 158 and 159 less than negative 0.075.

Notice that for this situation, unacknowledged second peak E 2 is flex points 137 or is searching the minimum that finds during the 2nd E2.Moreover, because the 3rd peak E3 can not have the amplitude that surpasses peak-peak grass (PPNB) on unacknowledged second peak E, therefore calculate decay based on this extreme case (supposition postpones to be no more than its target).In equation (6), if decay is assumed that zero, this will make proportional band also bigger than the value that it should have.Its result can not exceedingly fix control when response is little when the second and the 3rd peak can not be differentiated with grass.

The third type usually by step 162 to 174 represented.When decay deducts overtravel less than 0.02 the time, this situation will occur, and this shows that closed loop response is bordering on quadratic form damping, step 162.Proportional and integral are calculated as follows.If procedural type is less than 0.2, this be a pure delay and as determine during early stage the employing or postpone measurement between (being bordering on pure delay) by the integration that the user is provided with, perhaps decay greater than 0.3, integral multiply by 1.4, step 164 and 166.At least going through of the judgement of procedural type will be carried out below.Impose on integral in order to the differential divisor of determining derivative time divided by 1.4, so that keep derivative time constant.If the decay be less than or equal to 0.3, then proportional multiply by 0.8, step 168 and 170, otherwise, if integral greater than 1.5 sesquialter cycles, proportional takes advantage of 1.2, step 169 and 171.If decay is greater than 0.9, then proportional multiply by 1.4, step 172 and 174; Otherwise proportional remains unchanged.

If can not be classified according to above-mentioned three types because of having the non-PID of information completely adjustment process, then processor determines whether response starts from a stable state, damping capacity subtract overtravel whether less than 0.6, whether the semiperiod less than W _MAX, step 176.If satisfy these criterions, then according to being predicated the type that fresh information completely comes change procedure, step 178.It is the higher value of process that procedural type can change to 1.0 and the unstable state that integration postpones from 0.0 of pure delay.According to most preferred embodiment, procedural type ptype by under establish an equation determine:

(7) ptype * ln (I/T) integration of ln (I/T)=record postpone+(1-ptype)

The ratio ln (I/T) that * ln (I/T) postpones equation (7) left side is known, because I value to be controller existing integral time and T are the semiperiods that records.Equation (7) the right is respectively the ratio ln (I/T) of integration delay and deferring procedure, is then provided through interpolation by the rate value of storing among Fig. 3 ROM.In a preferred embodiment, be total to total 27 rate values for each storage in three extremal process controller combinations shown in Figure 10 to Figure 12 curve.These rate values comprise about ln (P/Po), ln (I/Io) and ln (I/T) on the every bit of Figure 10 of the PI of pure delay and integration deferring procedure and PID control nine points that are marked with A to I printed words to the curve map shown in Figure 12.(because of best differential term D is zero, so made up the PI and the PID control of pure delay process among Figure 10).These rate values all pre-determine through experiment.Subzero mark refers to reference point A, chooses 0.1 overtravel and 0.2 delay arbitrarily.Proportional band P, integral time, I and D/I ratio all were for reaching the decay that records and crossing the needed numerical value of tone pitch.Should be noted in the discussion above that and utilize other experimental data that ratio produced that transmits similar information to be used.Moreover, though these data points in decay with subtract on the plane of overtravel and be spaced apart, are the purpose of interpolation or extrapolation, can select other interval and different number point of destination equally.

Process decay and overtravel that utilization records, for example the M point can be painted on Figure 10 to three figures shown in Figure 12 by point.The biquadratic interpolation of utilizing those skilled in the art all to know is determined and the relevant weight function of each storage data point, and is used for seeking ln (P/Po), ln (I/Io) and ln (I/T) the ratio group of each process controller combination.Why selecting the biquadratic interpolation, is because curve is produced by similar parabolical experimental data.If controller does not have the differential action, be labeled as stored ln (I/T) value data of representing at A-J point place among Figure 10 and Figure 11 and can be used to by top equation (7) deterministic process type (ptype).

Known integration postpones and existing integral time of I, proportional band P, procedural type and ratio ln (P/Po), the ln (I/Io) of deferring procedure, then processor can calculate proportional band Po and integral time Io, this is with existing decay and crosses tone pitch and move to the phase delay value at reference point A place respectively and cross tone pitch (0.1 overtravel and 0.2 delay) necessary.Notice that overtravel and pad value at reference point A place are selected as the selected objective target value.These new parameter values are to calculate the step 180 among Fig. 9 by Po and the Io that solves the equation (8) and equation (9) left side respectively.

(8) ln (P/Po)=ptype * ln (P/Po) integration delay+(1-ptype) * ln (P/

Po) postpone

(9) ln (I/Io)=ptype * ln (I/Io) integration delay+(1-ptype) * ln (I/

Io) postpone

If the target decay that the user selects is different with reference value with overtravel, proportional band P _TWith I integral time _TAccording to changing PI into P _TI _TEquation (8), (9) calculate.Then, Po, Io can resemble down establish an equation indicated by cancellation:

(10)ln(Pt/Po)-ln(P/Po)＝ln(P/P)＝r

(11)P＝pe ^r1；

(12) ln (I _T/ Io)-ln (I/Io)=ln (I _T/ I)=r; With

(13)Ir＝Ie ^r2；

More than calculate, suppose no differential term D.If controller has the differential action, in order to obtain the value that equation (7) is used to (9):

Ln (P/Po) integration postpones, and ln (I/Io) integration postpones and ln (I/T) integration postpones, and at first must utilize differential divisor (dfact) (will make short explanation) to make interpolation or extrapolation from Figure 11 and 12.

Utilize the rate value of these interpolations, to calculate the damping that to record and to cross the new controlled variable that tone pitch moves to desired value in order to following equation with above-mentioned similar method.For tested condition, differential divisor (dfact) is the ratio of D and I.

(14) ln (I/T) integration delay=dfact * ln (I/T) integration postpones, PID+

(1-dfact) * and the delay of ln (I/T) integration, PI

(15) ln (P/Po) integration delay=dfact * ln (P/Po) integration postpones, PID+

(1-dfact) * and the delay of ln (P/Po) integration, PI

(16) ln (I/Io) integration delay=dfact * ln (I/Io) integration postpones, PID+

(1-dfact) * and the delay of ln (I/Io) integration, PI

For goal condition, can utilize the relation of determining D/I and procedural type (ptype) by rule of thumb to arrange new differential divisor (dfact), step 178.If ptype is less than 0.03 (master delay process), differential divisor (dfactor) by permanent strangulation to zero till the user cancels this decision; Otherwise, if overtravel is greater than zero, then

(17)dfact＝2.2-1.25×ptype

If decay that records and overtravel both equal their desired value, even differential divisor changes, interpolating method can show proportional band or integral time no change.No matter the controller of integration deferring procedure has or not the differential action, makes correction according to the change of reference value Po and Io.This correction is proportional with amassing of procedural type with the variation of differential divisor.

If error signal starts from the transient process state or decay subtracts overtravel greater than 0.6, then the variation of parameter is removed by 2, step 182,184 and 186.

Be fast quick-recovery under the imbalance condition, adjust in the scope of the expectation that makes it to drop on the phase half period integral time.The criterion of determining this condition is pointed out to 198 in step 188.If overtravel greater than zero, does not then allow to surpass 1.5 times of the semiperiod integral time, step 187,188 and 190.If at least one peak value is identified, then be forced to get at least 0.25 sesquialter cycle, step 192 and 194 integral time.If three peak values are identified, then be not to arrange this response by quadratic term, and integral time less than between two, three peak values at a distance of 0.35 times (0.1+ptype) of time.Multiply by 1.4 factors integral time, step 196 and 198.When applying interference gradually, it is too little that these steps can prevent from integral time to become.Then, utilize differential divisor to multiply by that integrated value is surplus again comes computing differential, step 200 with 0.16.But, do not allow to surpass 0.25 times of the semiperiod derivative time.Then the controlled variable in the controller 18 on Fig. 3 is modified step 202.Then, processor returns to transient process state 68 shown in Figure 4.

Although with reference to preferred embodiment, the present invention concrete diagram and description have been done, but those skilled in the art can be understood that, are not violating under the condition of the spirit and scope of the invention that limits as appended claims, and any change from the form to the details all is possible.For example, similarly scheme can be applicable to nonreciprocal PID controller.

Claims (60)

1. method of automatically adjusting the controlled variable of self tuning controller, this self tuning controller is used to regulate the process with the process variable signal that measures, and it is characterized in that the step of this method includes:

Produce the error signal of an expression process to the closed loop response of disturbed conditions;

Find out and measure the amplitude that characterizes described error signal mode feature and show that decay is less than overtravel;

Selection is in order to the tested decay that produces described error signal and three continuous amplitudes of overtravel feature, and wherein said tested attenuation characteristic is greater than described toning characteristic; With

Automatically adjust at least one controlled variable of controller, so that improve the difference between in the described tested characteristic and target property.

2. the method for claim 1, it is characterized in that: described amplitude is the local extremum of error signal.

3. the method for claim 1 is characterized in that: described error signal equals poor between process variable signal and the set point signal.

4. the method for claim 1 is characterized in that: the described step of automatically adjusting controlled variable comprises the type of determining the process that is conditioned and determines variation as the controlled variable of procedural type function.

5. method as claimed in claim 4 is characterized in that: the described step of automatically adjusting controlled variable also comprises data that record and the interpolation factor that is conditioned procedural type at least by the comparison of two prior defined procedure interpolative datas with a definite expression.

6. method as claimed in claim 5 is characterized in that: described interpolative data comprises controlled variable.

7. method as claimed in claim 6 is characterized in that: described controller parameter comprise ratio and integral time data.

8. method as claimed in claim 4 is characterized in that: the described step of automatically adjusting also comprise the data that will record with by the data of at least two prior defined procedure interpolations relatively so that determine the suitable variation of controlled variable.

9. the method for claim 1 is characterized in that: the described step of automatically adjusting also comprise the data that will record with by the data of at least two prior defined procedure extrapolations relatively so that determine the suitable variation of controlled variable.

10. the method for claim 1, it is characterized in that: the described step of finding out and measure amplitude also comprises the amplitude of finding out and measure at least four continuous described error signal local extremums.

11. method as claimed in claim 10 is characterized in that: the described step of finding out and measure amplitude also comprises and estimates described the 4th amplitude if described the 4th amplitude can not be measured.

12. the method for claim 1 is characterized in that: the described step of finding out and measure amplitude, the first local extremum time that also was included in is found out the flex point of a described error signal after a while.

13. method as claimed in claim 10 is characterized in that: also comprise the cycle of determining between first local extremum and described flex point, so that determine the time scale of error signal response.

14. method as claimed in claim 13 is characterized in that: described first local extremum is selected as first in three continuous amplitudes, and described time scale is used for determining to seek the maximum latency of second amplitude.

15. method as claimed in claim 14 is characterized in that: the step of described selection amplitude comprises if seek before second local extremum through described maximum latency then use flex point as a described amplitude.

16. method as claimed in claim 13 is characterized in that: described time scale is used to calculate the maximum latency of finding out the 3rd amplitude.

17. method as claimed in claim 12 is characterized in that: the step of seeking and measure amplitude comprises the described flex point that is later than second local extremum on the hunting time.

18. method as claimed in claim 10 is characterized in that: the step that also comprises the flex point of confirming described error signal.

19. the method for claim 1, it is characterized in that: also be included in and find out and measure the process variable sampling that amplitude is selected the user before, from being provided with, previously stored controlled variable selects the previous controlled variable that adopts with the process variable sampled value of selecting according to the user, so that provide improved correction to the controlled condition of now.

20. method as claimed in claim 19 is characterized in that: the scope of described process variable is subdivided into the plurality of sub scope, and wherein each subrange is corresponding to one group of previous controlled variable that adopts.

21. the method for claim 1, it is characterized in that: be positive or negative and be the negative then value of described three amplitudes of reversing for step first amplitude of the described amplitude of simplified measurement in order to determine it if the step of three continuous amplitudes of described selection comprises the symbol that shows first amplitude.

22. the method for claim 1 is characterized in that: also comprise the step of the described set point signal filtering that has leading-lag filter, when load disturbance having been proofreaied and correct, prevent to respond the toning of set point change undue error with convenient controller.

23. method as claimed in claim 22 is characterized in that: also comprise adaptively correcting described leading-the leading and ratio that lags behind of lag filter cross tone pitch so that reach target.

24. the method for claim 1 is characterized in that: the step that also is included in the steady state period measurements error noise and the change grass of described error signal.

25. the method for claim 1 is characterized in that: the step that also comprises the local extremum amplitude that affirmation records.

26. a method of adjusting the controlled variable of the self tuning controller that is used to regulate the process that has the process variable signal of surveying automatically is characterized in that step comprises:

Produce the error signal of an expression process to the closed loop response of disturbed conditions;

Measure the amplitude of described error signal local extremum;

Select three continuous amplitudes (E1, E2 and E3) of described local extremum make second extreme value (E2) square to be less than or equal to first extreme value (E1) long-pending with the 3rd extreme value (E3);

At least one tested Performance Characteristics of combination results by described continuous and local extreme value; Automatically adjust at least one controller controlled variable, so that improve the difference between described at least one tested Performance Characteristics and the target capabilities characteristic.

27. method as claimed in claim 26 is characterized in that: described combination step comprises that producing one crosses the setting of tone pitch as a described Performance Characteristics.

28. method as claimed in claim 26 is characterized in that: described combination step comprises that generation one pad value is as a described tested Performance Characteristics.

29. method as claimed in claim 26 is characterized in that: the step of described measurement amplitude comprises the flex point of finding out and measure described error signal in time after a while in first local extremum.

30. method as claimed in claim 26 is characterized in that: the step that also is included in the steady state period measurements error noise and the change grass of described error signal.

31. method as claimed in claim 26 is characterized in that: also comprise the step of the set point signal filtering that has leading-lag filter, when load disturbance having been proofreaied and correct, prevent to respond the undue error toning of set point change with convenient controller.

32. a method of adjusting the self tuning controller controlled variable that is used to regulate the process that has a measurable process variable signal automatically is characterized in that step comprises:

Produce the error signal of an expression process to the disturbed conditions closed loop response;

Determine described error signal mode characteristic in order to the type of identification called procedure; With

According at least one controlled variable of the described definite automatic computing controller of mode characteristic that responds the called procedure type, and adjust the controlled variable of controller so that improve described definite characteristic and the difference between target property.

33. method as claimed in claim 32 is characterized in that: the step of described deterministic model characteristic comprises determines an overtravel ratio.

34. method as claimed in claim 32 is characterized in that: the step of described deterministic model characteristic comprises determines an attenuation rate.

35. a method of regulating the self tuning controller controlled variable that is used to adjust the process that has the process variable signal of surveying automatically is characterized in that step comprises:

Produce the error signal of an expression process to the disturbed conditions closed loop response;

Measure to characterize first and second amplitudes of described error signal mode characteristic, wherein said second flex point that amplitude is a described error signal is with the toning characteristic of the measurement that produces a described error signal;

Select one by the determined attenuation characteristic value of less target pad value or utilize the pad value of described magnitude determinations; With

Automatically adjust at least one controller controlled variable in order to the characteristic of improving a described measurement and the difference between the target property.

36. method as claimed in claim 35 is characterized in that: the grass that the attenuation characteristic value of described calculating equals to record subtracts flex point value amplitude divided by the first peak amplitude.

37. a method of adjusting the self tuning controller controlled variable that is used to regulate the process that has the process variable signal of surveying automatically is characterized in that step comprises:

Produce the error signal of an expression process to the disturbed conditions closed loop response;

Measure a noise component signal in order to the described error signal of setting up grass;

Measure first and second amplitudes of the mode characteristic of the described error signal of expression, wherein said first amplitude is one to surpass the local extremum of described grass to produce the toning characteristic that records of described error signal; With

Automatically adjust at least one controller controlled variable to improve the difference between a described toning characteristic and target toning characteristic, measure described error signal, and between the steady state period of described error signal, change described grass automatically with definite stable state.

38. method as claimed in claim 37 is characterized in that: also comprise and estimate the 3rd amplitude and select by the determined attenuation characteristic value of less target pad value or utilize the step of the pad value of described magnitude determinations.

39. method as claimed in claim 37 is characterized in that: the step of described measurement amplitude comprises the amplitude of at least four continuous described error signal local extremums of measurement.

40. method as claimed in claim 39 is characterized in that: the step of described measurement amplitude comprises estimates described the 4th amplitude if described the 4th amplitude can not surveyed.

41. method as claimed in claim 37 is characterized in that: also comprise the step of measuring the 3rd amplitude and calculating attenuation characteristic.

42. method as claimed in claim 41 is characterized in that: the step of described measurement amplitude also comprises selects three continuous amplitudes to produce the step of toning characteristic and lag characteristic, and the attenuation characteristic of wherein said measurement is greater than described toning characteristic.

43. method as claimed in claim 42 is characterized in that: described amplitude is the local extremum of described error signal.

44. method as claimed in claim 37 is characterized in that: the step of described measurement second amplitude is included in the flex point of finding out described error signal after a while in the time of described first amplitude.

45. method as claimed in claim 44 is characterized in that: comprise that also the cycle of determining between first amplitude and described flex point is to determine the step of the time scale that error signal responds.

46. method as claimed in claim 45 is characterized in that: described time scale is used for determining to search the time span of secondary extremal.

47. method as claimed in claim 45 is characterized in that: described time scale is in order to determine to find out and confirm the time range of the 3rd amplitude.

48. method as claimed in claim 44 is characterized in that: the step of described measurement amplitude comprises selects described flex point as described second amplitude.

49. method as claimed in claim 44 is characterized in that: the step of seeking and measure amplitude comprises the described flex point of the described error signal that is later than described second amplitude on the hunting time.

50. method as claimed in claim 37 is characterized in that: described automatic setting procedure comprise the type of determining called procedure and select to be suitable for decide the controlled variable of procedural type.

51. method as claimed in claim 50 is characterized in that: described automatic setting procedure also comprises the data that will measure and compares from the data of at least two prior defined procedure interpolations to determine that interpolation factor represents to adjust at least the called procedure type of a controlled variable.

52. method as claimed in claim 50 is characterized in that: described automatic setting procedure also comprises the data that will measure and compares from the data of at least two prior defined procedure extrapolations to determine that extrapolation factor represents the called procedure type.

53. method as claimed in claim 37 is characterized in that: also be included in and measure the user's selection course variable and make one's options so that improves step of sampling before the amplitude to the correction of active procedure condition from the controlled variable of before having stored.

54. method as claimed in claim 53 is characterized in that: described time-varying process variable is divided into the experimental process scope, and wherein each subrange is corresponding to one group of previous used controlled variable.

55. method as claimed in claim 37 is characterized in that: also comprise step, to prevent the undue error toning of response set point change when controller has been corrected load disturbance to the described set point signal filtering that has leading-lag filter.

56. method as claimed in claim 55 is characterized in that: also comprise adaptively correcting described leading-ratio of lag filter crosses tone pitch to reach target.

57. a corrective control is characterized in that comprising:

One has the controller of the controlled variable of the process that is used for process variable signal that accommodation zone is surveyed;

Be used to produce the device of expression process to the error signal of disturbed conditions closed loop response; With

The adaptive mechanism that is used for the auto--tuning controller controlled variable, described adaptive mechanism comprises:

Be used to measure the amplitude that characterizes described error signal mode characteristic and show the device of decay less than overtravel,

Be used to select three continuous amplitudes to measure the device of decay and toning characteristic to produce described error signal, described amplitude is selected to such an extent that make described measurement attenuation characteristic greater than described toning characteristic, and

Be used for adjusting automatically at least one controller controlled variable to improve the device of difference between a described measurement characteristics and target property.

58. a self-aligning control system is characterized in that comprising:

One has the controller of the process control parameters that is used for process variable that accommodation zone is surveyed;

Be used to produce the device of expression process to the error signal of disturbed conditions closed loop response; With

The adaptive mechanism of one auto--tuning controller controlled variable, described adaptive mechanism comprises:

Be used for determining that described error signal mode characteristic waits to transfer the device of the type of process with identification,

According to response wait to transfer procedural type really cover half formula characteristic calculate automatically at least one controller controlled variable device and

Be used to adjust the controller controlled variable to improve the device of determining the difference between characteristic and target property.

59. self-aligning control system as claimed in claim 58 is characterized in that: the device that is used for the deterministic model characteristic comprises the mediation attenuation rate excessively of determining described error signal.

60. self-aligning control system as claimed in claim 58 is characterized in that: described error signal equals poor between process variable signal and the set point signal.

Images