WO1999044105A9 - Procede et dispositif de commande - Google Patents

Procede et dispositif de commande

Info

Publication number
WO1999044105A9
WO1999044105A9 PCT/JP1999/000837 JP9900837W WO9944105A9 WO 1999044105 A9 WO1999044105 A9 WO 1999044105A9 JP 9900837 W JP9900837 W JP 9900837W WO 9944105 A9 WO9944105 A9 WO 9944105A9
Authority
WO
WIPO (PCT)
Prior art keywords
value
response function
control
noise
signal
Prior art date
Legal status (The legal status 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 status listed.)
Ceased
Application number
PCT/JP1999/000837
Other languages
English (en)
Japanese (ja)
Other versions
WO1999044105A1 (fr
Inventor
Takehiko Futatsugi
Mitsuharu Chiba
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Adtex Inc
Original Assignee
Adtex Inc
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 Adtex Inc filed Critical Adtex Inc
Priority to JP54029799A priority Critical patent/JP3309142B2/ja
Priority to AU26394/99A priority patent/AU2639499A/en
Publication of WO1999044105A1 publication Critical patent/WO1999044105A1/fr
Publication of WO1999044105A9 publication Critical patent/WO1999044105A9/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

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
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
    • 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
    • G05B21/00—Systems involving sampling of the variable controlled
    • G05B21/02—Systems involving sampling of the variable controlled electric

Definitions

  • the present invention amplifies sudden failures (falling into an uncontrollable state or an oscillating state) and noise that occur in digital control using modern control theory that identifies a response function (hereinafter, including correction) at the same time as control. It relates to a control method that does not fall into various control states and its control device. If there is a disturbance that is measurable or can be known according to a schedule (expressed as an observable disturbance D), a feedforward that cancels out the adverse effects of D can also be realized.
  • D observable disturbance
  • the model system is required for calculation. Usually, we consider a less complex system, which can be represented by a differential equation for a continuous system and a difference equation for a discrete system. The system is then assumed to have the principles of energy theorem, linearity, and superposition. A system without energy production capability is called a passive system. If the control system is a passive system, if the operating value C is kept at a constant value, the control device R will eventually approach a constant value. It is said that the energy theorem holds when R keeps constant when C is held. An example of energy production is when a gunpowder is ignited or an inverted pendulum. Normally, explosives cannot control the subsequent explosion condition once ignited.
  • An energy-producing system is called an active system, but a controllable active system must have the energy theorem or be able to change R at least freely by changing C.
  • the energy theorem expresses the fact that in a system that does not generate energy, the energy changes to a large entropy state and an equilibrium state is realized.
  • the result R approaches a constant value (energy theorem) in numerical terms, so it is not the raw value (called the actual value) that is measured or set, but the amount of change ( (Differential value over time or difference of a sequence representing a time series).
  • the energy theorem can be expressed as "If the causes c and d become 0, the result r eventually becomes 0."
  • the relationship is represented by a linear differential equation as in (3).
  • u (d) r (t) v (3) c (t) + w (a) d (t) (3A) u (9) ⁇ uo + ui d + U2 d 2 + - ⁇ - + u "q d ⁇ uq
  • the derivative is made a difference by (3A), and 1 ", £ ;
  • Laplacian transformation after replacing with (3C ') sampled by, r (t) ⁇ rn ⁇ (t-nT), rn ⁇ r (nT)
  • ⁇ (t) ⁇ gn ⁇ (t-nT), gn ⁇ g (nT) (5D ")
  • (4D) (5D) is the basic equation of control theory using Z-transform.
  • a subscript for example, k is added to a symbol (for example, ⁇ ) or an expression (for example, A + ⁇ ) that represents a sequence, and the sequence or the sequence
  • the components are represented as " ⁇ '," (A + B).
  • the general term (general term) of a sequence is the nth term, and it is expressed as A ⁇ ⁇ AJ, or by arranging terms, or by writing a condition. If you write only one term in U, it means the nth term.
  • the control start time is set to the 0th term and the control cycle is set to T
  • the values of the subsequent T, 2 ⁇ , 3 ⁇ ,... Are referred to as the 1st, 2nd, 3rd,.
  • Ayo ⁇ a. , A 1, a 2, ... ⁇ , A ' ⁇ ⁇ ..., 0, 0, a. , A,, a 2,... ⁇ ⁇ ' A (11) It is expressed by "(n,” that all the terms less than n are 0. "(- ⁇ ,” is replaced by "(, It is abbreviated as “(n,” and the n-th term is not 0 by "[n,”, and the existence of such n is abbreviated as "[,”. , The n-th term is called the first term.
  • a sequence or an expression is given as “ct A”, “a (A + BC)” after “ ⁇ ”, and is expressed as a sequence A or an expression A + BC Similarly, the first order of the sequence is expressed in the same way as “ ⁇ ,” and “ ⁇ )” when all the terms in n-th order are all 0 and “ ⁇ ]” when the ⁇ -th order is not 0. , "+ ⁇ )” is abbreviated to ")” and " ⁇ ]” is abbreviated to "]”.
  • ⁇ is the final term
  • the ⁇ term is the final term
  • the final position is expressed using “ ⁇ .”
  • the terms from less than X and more than Y in the sequence ⁇ are set to 0, and the terms from X to Y are calculated. , Observing If, for convenience to the X ⁇ ⁇ , the ⁇ and ⁇ ⁇ , and expressed as ⁇ € ( ⁇ ⁇ ' ⁇ ⁇ ). Refers to the number of terms is not limited to 0 and the number of terms.
  • (12A) the set of sequences in which the first place is X and the last place is ⁇ is defined as [ ⁇ , ⁇ ].
  • (12B) the set of sequences in which terms are less than X and more than ⁇ are ⁇ is defined as (X, ⁇ ). The same applies to (12C) and (12D).
  • a sequence with the first term [,) is called a left regular sequence and is denoted by [], and a sequence with a final term (,] is called a right regular sequence (].)
  • a finite sequence with one or more terms [,] Is called a finite sequence and is abbreviated as a mouth
  • the intersection of (] and [) is []
  • the union of 0 in (8G) and [), (], [] Sets) are called left-regular sequence, right-regular sequence, and finite sequence, and are represented by [) +0, (] +0, [] +0.
  • the set that can be freely calculated is [) +0, and the set that can be freely calculated by adding, subtracting, multiplying, and dividing by expanding [:] is called a rational sequence.
  • (20D) is a dynamic connection between cause c and result r.
  • (4F) (5F) is (4 ⁇ ') (5 ⁇ ').
  • the relational expression that expresses these causal relationships is the basic control equation called the transfer equation.
  • sequence ⁇ For the sequence ⁇ , define the sequences " ⁇ " and "' ⁇ " excluding the first and last terms.
  • A ⁇ An ⁇ S [) ⁇ "A ⁇ A-Airri ⁇ ⁇ " ⁇ (23A)
  • A ⁇ An ⁇ £ (] ⁇ A ⁇ A- ⁇ country, ⁇ ⁇ " , ⁇ (23B)
  • ⁇ " and " ⁇ " are symbols that define a sequence excluding the first term and the last term.
  • ⁇ ⁇ ⁇ is the first and last term of ⁇ , and the power of ⁇ is a sequence defined in (8C).
  • the quotient of a left regular sequence of zeros is defined as zero.
  • the left regular sequence is always divisible except for 0 / B ⁇ 0 B ⁇ 0 (141-1) division by 0, and its quotient is a left regular sequence. The following relationships hold.
  • the right quotient of a right regular sequence by a right regular sequence is a right regular sequence, but not necessarily a left regular sequence. Only when the right quotient becomes a finite sequence, the quotient of the left regular sequence (when distinguishing it is called the left quotient) matches. Therefore, right division is a tool for finding a quotient that becomes a finite sequence from the last term side.
  • ⁇ m works by shifting the arbitrary sequence A by m terms to the right.
  • a m A means A before m time points.
  • ⁇ can be expressed as 1 1 ⁇ by ⁇ , and the product of the sequence A and any sequence is a difference.
  • Room 1 ⁇ (8A ')
  • ⁇ ⁇ can be expressed as ⁇ - 1 in ⁇ , and sums the product with any sequence ⁇ .
  • ⁇ ⁇ -1 (8L)
  • R f C (20D ")
  • C are actual values obtained by adjusting the origin of the unit to satisfy R £ [) and C £ [). (20D ")
  • R f.
  • the pulse response function and the step response function are related to the sum of the differences.
  • control measures are usually expressed ignoring time variations.
  • the control valve is so on can be adjusted in flow rate 5 0 mm 3 / SeC per degree opening. This means that the flow rate does not change by 5 Ommsec at the moment when the opening is changed by 1 degree, but changes over time.
  • the static characteristics of the operating means are nothing more than the extreme values of the step response function.
  • the eigenvalue of the denominator 1-Q of the rational expression of the response function is called the pole. If any of the poles has an absolute value of 1 or more, the force that does not stop changing forever will continue to increase. This violates the energy theorem. In a control system where the energy theorem holds, the absolute values of all poles are less than 1 and decrease exponentially.
  • the poles When there is an oscillating element, the poles are represented by more than one pair of complex numbers, and it is considered that energy exchange occurs between these poles. Sound pressure and kinetic energy, pendulum position energy and kinetic energy, electromagnetic wave magnetic field
  • M k M u (32A)
  • the diagonal component of M-M is 1 or 0, and 0 above the diagonal component of the diagonal sequence whose diagonal component is 0 (the part with *) Not necessarily. Rows with diagonal components of 0 have all columns 0.
  • M—If the diagonal component of the i-th row of M is 0, is dependent on -th order and cannot be determined independently. The solution when the k component ki of the row where the diagonal component becomes 0 is set to 0 is given by M ⁇ u. If M is regular, the diagonal components of M—M are all k M u (32B) 1 and the four lines marked with * are not executed. If M is not regular, ignoring * and performing operations when the diagonal component is not 0 will result in a divide-by-zero error. If M is nonsingular, M_ is the inverse matrix M— 1 .
  • MM + w ⁇ 'XX; u ⁇ u + w ⁇ 1 X (30C) is called weighting, and the method of (30C) is called weighting type. If w, in (30C) indicates the effective number of sets, M, u before update contains n — 1 data sets, and M, u after update contains n data sets. Expressing one set of M and u gives (30D), and if 1 / n is p r 'w, the updated type (30E) is obtained.
  • Increasing data set number in (30B) is nM- (n- 1) M + 1 XX; nu- (n- 1) u + y ⁇ 1 x (30D) M- (1- 1 / n) M + (1 / n) ⁇ 1 xx; u — (1-1 / n) u + (1 / n) ⁇ y ⁇ 1 x
  • the frequency of occurrence of the data X in the primary equation (28A) is extremely different, using (30C) and (30D) will cause the less frequent parts to lose digit and lose information.
  • the data may be divided into groups with different frequency of occurrence and have different observation formulas. As an example of two cases, the observation equation is set up and solved for each of kl and k2 as shown in (28D).
  • X, y in both the there is error, true of k is the answer between the k x and k y.
  • the basic form (35A) is to use the correction amount as it is (33A), and the update type (update rate (35B), which is dynamically weighted to reduce the effect of noise Is the weight type (weight Wl ) (35C).
  • (34A) calculates the sum of squares where temperature, height, power, etc. are mixed It will be. Ki corresponding to Xi, which has become a small numerical value in terms of units, is hardly identified, and kj, which corresponds to Xj, which has become large, is conversely ⁇ due to the inaccuracy of other ki ⁇ j. Continue to receive larger fixes.
  • a value for setting a certain value (called a control value, denoted by R) to match a target value (also called a set value, denoted by S) (called settling)
  • the control device is at least an input device of length and S, an output device of C, a timer for periodic processing, and a device for determining C according to the difference between R and S (computing device) And represented by U).
  • U requires storage device M.
  • a display device, an alarm device, a safety device input device, etc. are attached as necessary.
  • the present invention requires an input device (including the detection device if it can be detected at the location of the device) to know when an abnormality occurs, and an input device for available intelligent disturbance D.
  • You. M contains programs and initial value data required for control. These program / data or timer signals can also be obtained externally using communication means. Externally obtained communication means are considered to be attached to the equipment.
  • Tightened paper (Rule 91) In a digital controller, the computation between input and output values performed by U plays the role of a machine cam or gear.
  • the object to be controlled (control system) has a causal relationship with C being output from the control device and D being observed, with R being the result.
  • the equation describing the causal relationship is called the transfer equation, and the coefficient when the transfer equation is described by a linear equation is called the response function.
  • the modern control method is characterized in that a response function is obtained and an output value C is determined based on the response function.
  • the theory was complicated in the conventional method, and when controlling while obtaining the response function, it suddenly fell into an uncontrollable state or amplified noise.
  • high-speed calculations are required and the number of calculations is limited, if the response function is identified at the end of control, the rest time, the start of the next control, etc. after saving the data, Sometimes became impossible. While analyzing the actual behavior using the theory using left-regular sequences, we discovered these causes.
  • R is obtained from the input device, and C is output from the output device.
  • C is output from the output device.
  • R is obtained from the input device
  • C is output from the output device.
  • these numerical values are discrete integer values or converted values of integer values.
  • the response function describes the relationship between these discretized numbers, and R is obtained by measurement.
  • a person with a height of 1.65 m is measured with a standard error of 0.05 m and rounded to the nearest lm.
  • about 135 out of 100,000 times is expected to be lm, and the remaining 99,865 times is expected to be 2 m. This average is 1.99865m, not 1.65m.
  • One billion or one trillion measurements will not help. Few people will do such rounding when measuring and recording their height.
  • the learning effect occurs only by identification while controlling, and the burden of preliminary measurement and equipment startup is reduced. Considering the quantum phenomena, since the signal-to-noise ratio S / N is 5 or more, it cannot be expected that increasing the number of measurements will further improve the S / N.
  • the identification of the response function must be avoided.
  • Statistical processing can improve S / N with white noise of 5 to 10 digits or more, as well as variations in digit size and hysteresis. Only when the lysis can be smoothed out. If the nature of digitization is poor, the relationship between the average and true values of the measured values will be non-linear even with white noise over a large number of digits, and the S / N will not improve. Since the response function is a coefficient that connects the cause and the result, the same can be said not only for the measurement of the result R but also for the measurement and setting of the causes C and D. Variations, hysteresis, instability, etc. of one digit width of C have a great effect when the fluctuation width of C becomes about several digits.
  • a ⁇ •, 0, 0, ai, a 0, 0, ⁇ ⁇ ⁇ ⁇ ( 1, ⁇ a)
  • f a / (l-q)
  • g b / (l-q) (6D)
  • a (l-q) f
  • b (l-q) g
  • a f — qf
  • b g — qg (6F)
  • f a + qf
  • g b + qg (6G)
  • 6G recurrence formula
  • (4F) and (5F) are equivalent expressions, and it is unnecessary to discuss which is the basic equation.
  • (4F) is interpreted as follows, where r, c, and d are control values, operation values, and intellectual disturbances. In the Z-transform theory, r, c, and d were impulse responses. However, since the Z operator is a left-regular sequence, there is no reason to treat it differently from other sequences. (4F) It suffices if the control system can be represented by (5F) without excess or deficiency, and r, c, and d are simply considered as a sequence representing a time series. It does not matter whether the output value of the double-integral A / D converter is statistically processed or r, c, or d has different measurement and setting timings.
  • q, a, and b have a first rank of 1 or more and are qualified as response functions.
  • a, b, and q are obtained using a finite number of known data r, c, and d.
  • the calculation method can be either the least squares method or the sequential identification method. If a, b, and q are obtained, calculate f and g by (6G).
  • the response function is approximated by the response function of (41A) with ⁇ as the first term, and the deviation is assumed to be non-zero from the ⁇ term (42 ⁇ ).
  • (41A) is a continuous system used by Ziegler and Nichols for the marginal sensitivity method and is used for tuning PID control.
  • Fn ⁇ Z ⁇ -0, af — Z (41A) ez «-ei + e 2 + e 3 + h ez Eze n ⁇ z ⁇ ⁇ 0, E n ⁇ z — 0 (42A) c ' .
  • Is obtained, (4343) is obtained from (6 ⁇ ).
  • (5R ') is (42D), which is (5S), but (27G) is (271).
  • the operation value is a solution of the form (4.3H) (43H ').
  • C 'n Hn, ZEZ + Hn, z + , E z ⁇ ! + H n,,-E z + 2 +''' + ⁇ , ⁇ (43H)
  • C 'n hn, zEz + hn, z + ⁇ z + + hn, + ⁇ ⁇ + ⁇ ⁇ (43H ')
  • Least squares are preferred for early identification because they are faster than sequential identification. To eliminate the unknown mutual interference, it is desirable to use the least squares method as a set of q, a, and b, rather than dividing them into q, a, and b.
  • intellectual disturbances sometimes do not occur at all for long periods of time. If the updated least squares method is used, the response function of the intellectual disturbance may suddenly go wrong. This is because very few parts related to components that have non-zero values cause digit loss.
  • the least squares method calculates (32B).
  • C MAX and C M ' N ; D MAX and D M ' N are the maximum and minimum values of the cause, R MAX and R MIN are the maximum and minimum values of the control value, both of which are safety margins as devices, Take the maximum and minimum values of the minimum width of the output range, input range, etc., and set c, r, d to (45C) c-c / k a *, r-r / k Q *, d-d / k b * (45C) Then, the maximum amplitude will be ⁇ 1.
  • X 1 (identity matrix) can be set by (31D).
  • Wl is weighted if normalized to 1 with the representative value of X i 2 .
  • C MAX and C M ' N ; D MAX and D M ' i are the maximum and minimum values of the cause, R MAX and R M1 N are the maximum and minimum values of the control values, both of which are safety margins as devices.
  • D MAX and D M ' i are the maximum and minimum values of the cause
  • R MAX and R M1 N are the maximum and minimum values of the control values, both of which are safety margins as devices.
  • X 1 (identity matrix) can be set by (31D).
  • V 2 ⁇ 2 + dgt z > (46A) dgt is one digit large. You can simply calculate the average,
  • W i ⁇ S gi / Sg The square of the amplitude is called power and is grasped by the concept of energy. Examples include electromagnetic waves, sound waves, pendulums, and voltages.
  • Sg the total signal power including noise
  • S gi X i 2 is the power of the signal of interest, and is the ratio of Sgi (signal ratio).
  • w of (47B) is the overall signal rate.
  • the signal ratio for a, b, and q can be defined by (47C).
  • Ru amplitude Xi signal rate can be defined.
  • the noise is Noise includes the unbiasedness of the estimator by the method of least squares, the validity of the terms a, b, and q, and the nonlinearity of the control system. These noises are referred to as model inadequacy, and are separated from ".
  • the response function is determined by the least squares method and the accuracy is improved to some extent, and then switched to the sequential identification method, the power and statistical theory will be reduced. This can be corrected without using the method of making the least-squares estimator closer to the unbiased estimator, etc. As the noise increases, the degree of inconsistency in the least-squares method increases. Even in the sequential identification method, the value becomes unstable if the noise is large, and in this sense, the inadequacy of the model can be called the distribution ratio to noise.
  • Modern control is a method that relies on a response function. If the key response function turns out to be inaccurate, it may be prudent to reduce output fluctuations. Normally, a smooth state with a slight delay is preferred to repeating the vibration state due to excessive output.
  • the original form of the updated least squares method (3 1 F) modifies the original data by 1 / n in the nth valid data set. This is interpreted that the response function obtained from n sets of effective data has an error that needs to be corrected by 1 / n.
  • n is the effective number of sets, it is the sum of the weights. However, in (49C), ⁇ eventually overflows the numerical value.
  • the control system changes environmental temperature
  • Corrected form (Rule 91) These include the unbiasedness of the least squares estimator, the validity of the terms a, b, and d, and the nonlinearity of the control system. And to the child say these noise p n model unsuitable with the, to separate the p n from p r. If the response function is obtained by the least squares method and the accuracy is improved to some extent and then switched to the sequential identification method, the estimator by the least squares method should be approached to the unbiased estimator by making full use of statistic theory. You can correct it with Pn. The higher the noise, the greater the least squares mismatch. Even in the sequential identification method, the value becomes unstable if the noise is large. In this sense, the inadequacy of the model can be described as the distribution to noise.
  • Modern control is a method that relies on a response function. If the key response function turns out to be inaccurate, it may be prudent to reduce output fluctuations. Normally, a smooth state with a slight delay is preferred to repeating the vibration state due to excessive output.
  • the original form of the updated least squares method (31F) modifies the original data by 1 / n in the nth valid data set. This is interpreted that the response function obtained from the n sets of effective data has an error to be corrected by 1 / n.
  • n is the effective number of sets, it is the sum of weights w ,. However, in (49C), n eventually causes a numeric overflow.
  • the control system changes environmental temperature O 99/44105
  • the response function changes slightly over time due to various factors such as degradation and deterioration. Being able to follow changes over time is the real thrill of controlling while identifying the response function.
  • p r if (p r M 'N rather p r; Pr / (Wl ⁇ ⁇ + 1)) (49D)
  • Z if (X; Y) " ⁇ If it is not the ZYX If X is a constant" represents the in .
  • Fig.2 shows a graph with the coordinates on the ordinate and the coordinates on the 44. Regardless of (47D), I experienced that even if ⁇ was set to 0.01 to 0.3, there was a noise suppression effect.
  • the identification data should be saved and identified by time-division processing. Identify at the next control start.
  • V l 2 is a long-term average of the square of the noise with the lower limit of the digit. That is, it is necessary to distinguish between the signal rate used for operation and the signal rate used for identification.
  • w ° also have problems.
  • Wl not only is rather small, but rapidly decreases the frequency of occurrence of large data, the proportion of the additional data that causes the loss of significant digits will be dominant. Chile also leads to the destruction of the response function, as the saying goes.
  • k d is set to a value smaller than the specification at the time of equipment design such as the change range of the target value and the maximum number of digits that occur during control during operation observation, for example, 0.7 times the maximum number of digits. . Wl '-' M is a significant digit that cannot be secured for one or two digits and is lost. Normally, when a safe operation range is set and the control value or operation value enters the danger area, the limit circuit or limit switch Inhibit further changes with switches, mechanical means, safety devices, etc. Such a condition is not the relationship between normal operation value and control value.
  • d I are used.
  • K the reference value for K is I ⁇ ' M
  • the period is K PRD .
  • the data during K P RD after C M ⁇ K or K L ' M K that is not abnormal is used for identification.
  • K L 1 M is, surely the CM rather than ⁇ K occurs that K is enough to choose as S / N is Naru rather than the size in the control. You can automatically set the C M in the next technique (Select update method).
  • K M K L, M ⁇ K or k M K L ' M ⁇ K When done
  • K PRD of Sgf / v Sgd / v 2 is 1,
  • the magnitude K of the identification signal is calculated using the past values of the difference between the operation value, the control value, and the intellectual disturbance or the noise and the magnitude of the noise, so that abnormal or small ⁇ data is not used for identification. This method is called the identification maintenance method. If the response function is broken, the response function f; q, a of the control value and control value (memory effect) that directly affects the calculation of the operation value will be fatal if the response function is destroyed.
  • the identification maintenance method prevents destruction of the response function during stable operation.
  • Noise amplification method can suppress noise amplification.
  • the intelligent disturbance capture method reduces noise disturbance by the amount of the intelligent disturbance.
  • the biggest drawback of modern control methods for identifying response functions is the sudden breakdown. It is important to avoid this failure with the identification maintenance method and the error distribution method. By combining this with the noise compression method, it is possible to avoid a state that may lead to a failure and realize more stable control.
  • the intellectual disturbance is not always available, but when it is available, the intellectual disturbance capture method can reduce the disturbance that leads to bankruptcy by removing the adverse effects of the disturbance.
  • the control unit is equipped with an input device for intellectual disturbances, target values, control values, and abnormal signals (signals to detect abnormal conditions), and operation values are calculated based on the predicted values while the response function is identified by the arithmetic unit.
  • the identification maintenance method and the error distribution method are indispensable means, and the noise compression method and the intelligent disturbance acquisition method are implemented as necessary. As a result, a control device that does not break down and is not in a state where noise is amplified is realized.
  • the operation includes an output device for the operation value C, an arithmetic device U, a storage device M, and an input device for the control value R, the target value S, and the intelligent disturbance D.
  • the identified response function or the data for identifying the response function is stored in M, and the predicted value of R estimated using R and C and D in the past, present, and future usable range is stored.
  • FI G. 1 shows the configuration of the control device that constitutes the present invention.
  • V Calculate the correction amount c 'of the operation value based on the deviation E between the target value S and the predicted value R °
  • FIG. 2 shows the functions used for the noise compression method.
  • the abscissa is the signal power and the ordinate is the function value.
  • FIG. 3 shows the identified signal rate of the data set identifying the response function as a solid line.
  • the dashed line shows the concept of the function when the identification signal rate is binarized.
  • the horizontal axis is If a, b, and q can be completely determined by analysis alone, there is no need to identify them during control.
  • the control unit can set a timer interrupt according to the control cycle or check the timer at any time to process.
  • d—, i Prepare n and K.
  • the current time is the n-th time point
  • the control start time is the ⁇ -th, —1 time point.
  • ⁇ — ⁇ is X and ⁇ 1 u is y and k, so it is expressed without using X and y.
  • K 1 (qi, q 2 , ⁇ ⁇ ⁇ , q, q, a, a, a ",", bl, b:.. ⁇ , B, b)
  • c co (A. I. '"—'" ") (51A) c is usually set to 1, but in order to avoid excessive operation, a value that will cause r to be sufficiently larger than noise during the response test. In some cases.
  • the control device can set a timer interrupt according to the control cycle, or check and process the timer at any time.
  • the current time is the n-th time and the start of the control is the first a1 time.
  • M—M is X and M-u is y and k, so it is expressed without using X and y.
  • k 1 (ql, q, ⁇ , q ⁇ q , a ⁇ , a, a., a , b ,, b: ⁇ , b. b)
  • c c. ( ⁇ .- ⁇ . '" + Wq ) ri (51A) c. Is normally set to 1, but it may be set to a value that causes r to be sufficiently larger than noise during the response test to avoid excessive operation.
  • the initial value is set to (51C), and only the observation is performed until the 0th point.
  • pn is adjusted appropriately with 0.05 as the initial value.
  • Icount Icount-1 (52B)
  • p r if (p r M, N ⁇ p r ; p r / (p r + 1)) (49F)
  • R n a n + l C- l + an + 2 C- 2 + * * '+ a Struktur, n C n- ⁇ a
  • the present invention is a high-speed and accurate control method according to the present invention which avoids control failure, has good stability, and is capable of feedforward, and a device for realizing the control method.

Landscapes

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

Abstract

Pour produire la variable manipulée calculée par une unité arithmétique (U) comportant un dispositif de mémoire rémanente (M), par l'introduction d'une valeur voulue (S), d'une variable commandée (R), et d'une valeur de perturbation (D), une fonction de réponse est identifiée au moyen de données correspondant à des variations importantes, quoique non anormales, de (R), (C) et (D); (R) est prévu au moyen des valeurs de (R), (C) et (D) du (plan) passé, présent et futur, et la valeur de correction de (C) servant à égaliser (R) par rapport à (S) est calculée comme valeur présente de la différentielle de (C), sur la base de la valeur prévue de (R). Le produit de la valeur de correction, le coefficient qui constitue le reste obtenu en soustrayant l'incertitude de la fonction de réponse et l'inadéquation du modèle de 1, et la compressibilité du bruit servent de différentielle de la valeur de sortie présente; et la différentielle est ajoutée à la valeur de sortie de la période de commande précédente, ce qui permet de produire la somme. Les données, telles que la fonction de réponse, obtenues au cours d'une commande sont stockées dans la mémoire (M). Le procédé permet d'éviter toute défaillance soudaine se produisant au cours d'une commande numérique classique dans laquelle une fonction de réponse est identifiée et commandée, et tout état d'amplification de bruit. L'invention permet de mettre en oeuvre une régulation par anticipation et d'empêcher des dérangements de commande dus à des perturbations prévisibles.
PCT/JP1999/000837 1998-02-25 1999-02-24 Procede et dispositif de commande Ceased WO1999044105A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP54029799A JP3309142B2 (ja) 1998-02-25 1999-02-24 制御方法とその装置
AU26394/99A AU2639499A (en) 1998-02-25 1999-02-24 Control method and its device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP8788798 1998-02-25
JP10/87887 1998-02-25

Publications (2)

Publication Number Publication Date
WO1999044105A1 WO1999044105A1 (fr) 1999-09-02
WO1999044105A9 true WO1999044105A9 (fr) 2000-01-20

Family

ID=13927397

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP1999/000837 Ceased WO1999044105A1 (fr) 1998-02-25 1999-02-24 Procede et dispositif de commande

Country Status (3)

Country Link
JP (1) JP3309142B2 (fr)
AU (1) AU2639499A (fr)
WO (1) WO1999044105A1 (fr)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102929141B (zh) * 2012-10-10 2015-03-18 西北工业大学 飞行器时间滞后时变模型逼近及控制器设计方法
JP7151546B2 (ja) * 2019-02-25 2022-10-12 オムロン株式会社 制御装置、制御方法、及び制御プログラム

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08123507A (ja) * 1994-10-27 1996-05-17 Fujitsu Ltd ロバスト制御装置

Also Published As

Publication number Publication date
JP3309142B2 (ja) 2002-07-29
AU2639499A (en) 1999-09-15
WO1999044105A1 (fr) 1999-09-02

Similar Documents

Publication Publication Date Title
Rahideh et al. Constrained output feedback model predictive control for nonlinear systems
Chen et al. Machine learning based iterative learning control for non‐repetitive time‐varying systems
Vesely et al. Robust MPC controller design for switched systems using multi-parameter dependent Lyapunov function
Guan et al. Spline adaptive filtering algorithm based on different iterative gradients: Performance analysis and comparison
US9199860B2 (en) Apparatus and method for nonlinear process identification using orthonormal bases and ordinal splines in a process control system
Ahmad et al. Observer design for one‐sided Lipschitz nonlinear systems subject to measurement delays
Chi et al. A model predictive control approach with relevant identification in dynamic PLS framework
Patel Computation of matrix fraction descriptions of linear time-invariant systems
Sakthivel et al. Fault-tolerant sampled-data control of singular networked cascade control systems
Wen et al. Multivariable adaptive output rejection of unmatched input disturbances
Bonnet et al. Stability of neutral systems with commensurate delays and poles asymptotic to the imaginary axis
Jiang et al. Critic-only adaptive dynamic programming algorithms’ applications to the secure control of cyber–physical systems
San Yip et al. Multiple data sets for model updating in real-time operations optimization
Golshan et al. Latent variable model predictive control for trajectory tracking in batch processes: Alternative modeling approaches
Luo et al. Adaptive Predictive Control: A Data‐Driven Closed‐Loop Subspace Identification Approach
Ye et al. Generalized predictive control of a class of MIMO models via a projection neural network
Han et al. Fault-tolerant optimised tracking control for unknown discrete-time linear systems using a combined reinforcement learning and residual compensation methodology
Nguyen et al. Finite‐Time Output Feedback Controller Based on Observer for the Time‐Varying Delayed Systems: A Moore‐Penrose Inverse Approach
WO1999044105A9 (fr) Procede et dispositif de commande
Li et al. Interval estimation for discrete-time linear time-delay systems based on state augmentation
Elloumi et al. Design of self‐tuning regulator for large‐scale interconnected Hammerstein systems
Shi et al. Fuzzy Predictive Fault‐Tolerant Control for Time‐Delay Nonlinear Systems with Partial Actuator Failures
Ding et al. On the sample complexity of observers for unknown linear systems with biased dynamics estimations
Seeber Computing and estimating the reaching time of the super-twisting algorithm
Hammar et al. Fractional Hammerstein CAR system identification

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AL AU BA BB BG BR CA CN CU CZ EE GE GM HU ID IL IS JP KR LC LK LR LT LV MG MK MN MX NO NZ PL RO SG SI SK SL TR TT UA US UZ VN YU

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW SD SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
121 Ep: the epo has been informed by wipo that ep was designated in this application
AK Designated states

Kind code of ref document: C2

Designated state(s): AL AU BA BB BG BR CA CN CU CZ EE GE GM HU ID IL IS JP KR LC LK LR LT LV MG MK MN MX NO NZ PL RO SG SI SK SL TR TT UA US UZ VN YU

AL Designated countries for regional patents

Kind code of ref document: C2

Designated state(s): GH GM KE LS MW SD SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

COP Corrected version of pamphlet

Free format text: PAGES 6, 7, 9-11, 14-17, 19, 22, 25, 27, 29, 35-37, 41 AND 51, DESCRIPTION, REPLACED BY NEW PAGES 6, 7, 9-11, 14-17, 19, 22, 25, 27, 29, 35-37, 41 AND 51; AFTER RECTIFICATION OF OBVIOUS ERRORS AS AUTHORIZED BY THE INTERNATIONAL SEARCHING AUTHORITY

NENP Non-entry into the national phase in:

Ref country code: KR

122 Ep: pct application non-entry in european phase