JP2013200780A - Control method and control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To effectively suppress hunting of control output by executing feedback control by an appropriate feedback gain suited to an individual present state of a control object.SOLUTION: On the basis of a closed loop control model for which a state feedback control rule is applied to a control object model, an FB term operation unit 63 computes FB gains F(1)-F(9) using a pole αof the control model and model parameters a, band b. Then, on the basis of a control deviation e(i) of a target value and the control output, an FB term u(i) is computed by the state feedback control rule using the FB gains F(1)-F(9). The pole αis set by a pole setting unit 77. When hunting of control output y(i) is detected, the pole setting unit 77 reduces the FB gain and suppresses the hunting by bringing the pole αset at present close to 1.

Description

  The present invention relates to a control device having a function of controlling a control object by feedback control, and a control method used in the control device.

  Conventionally, as a control method for controlling a control target by feedback control, control is performed by modeling the control target, calculating a control input (operation amount) by state feedback, and adding an operation according to the calculated control input to the control target. Model-based feedback control for controlling an object is known (for example, see Patent Document 1). A control method in which feedforward control is combined with this feedback control is also known (see, for example, Patent Document 2).

  In such conventional feedback control, a fixed value designed in advance is generally used as the feedback gain used for the control. However, because of the model-based design, if a modeling error occurs, the controllability becomes unstable, which causes hunting of the control output.

  The modeling error means that the input / output relationship of the control target model fluctuates from the model assumed at the time of design due to various factors such as variations in the control target and aging. For example, in a control system that detects the air-fuel ratio of exhaust gas with a sensor and controls the fuel injection amount so that the air-fuel ratio follows the target air-fuel ratio in a vehicle engine, a sensor or It is conceivable that the actuator (fuel injection device), the engine, etc., change over time (including deterioration), variation, and difference in operating conditions.

  On the other hand, Patent Document 1 describes that each operating condition has a feedback gain. Japanese Patent Application Laid-Open No. 2004-228561 describes that correction in consideration of a variable element is performed on a feedback operation amount calculated using a feedback gain.

Japanese Patent Laid-Open No. 2002-130020 JP 2008-287581 A

  However, since the technique described in Patent Document 1 does not deal with modeling errors, it is difficult to expect an effect of suppressing hunting. The technique described in Patent Document 2 is a limited method for dealing with modeling errors, and cannot sufficiently cope with modeling errors (hunting suppression).

  The present invention has been made in view of the above problems, and is capable of effectively suppressing hunting of a control output by performing feedback control with an appropriate feedback gain suitable for each current state of a control target. Objective.

  The invention according to claim 1, which has been made in order to solve the above problem, uses a control target model that simulates a control target, and inputs an operation to the control target so that the control output of the control target follows the target value. A control method for state feedback control of a quantity, wherein a pole of a closed-loop control model in which a state feedback control law is applied to a control target model is set, and a feedback gain is set using the set pole and the model parameter of the control target model Calculate the control deviation indicating the difference between the target value and the control output, and calculate the feedback manipulated variable as the manipulated variable based on the control deviation based on the state feedback control law using the feedback gain. To do. When hunting of the control output is detected, the pole is reset so that the feedback gain becomes smaller.

  According to the control method of the present invention, when hunting of the control output occurs due to modeling error of the control target, the feedback gain is reduced by resetting the pole, and the control system is further stabilized. Can do. Therefore, feedback control can be performed with an appropriate feedback gain that matches the current status of each control target (considering a modeling error or the like), and control output hunting can be effectively suppressed.

  The invention according to claim 2 is a control device that performs state feedback control on an operation amount input to a control target so that a control output of the control target follows a target value using a control target model that simulates the control target. The pole setting means for setting the pole of the closed-loop control model in which the state feedback control law is applied to the control target model to a value within a range satisfying a predetermined stability condition, and the pole and control set by the pole setting means A feedback gain calculating means for calculating a feedback gain using the model parameters of the target model, a control deviation indicating a difference between the target value and the control output is calculated, and based on the control deviation, the feedback gain calculating means The feedback as a manipulated variable is determined by the state feedback control law using the calculated feedback gain. Comprising a feedback manipulated variable calculation means for calculating a click operation amount, the hunting detection means for detecting a hunting of the control output, a. Then, when the hunting is detected by the hunting detection means, the pole setting means resets the pole in a direction in which the feedback gain is reduced.

  According to the control device configured as described above, the above-described control method of the present invention is realized. That is, pole setting and feedback gain calculation are performed by the above-described control method of the present invention, and pole resetting is performed when hunting is detected. Therefore, the same effect as the control method of the present invention described above can be obtained.

  Various specific methods for resetting the pole when hunting is detected can be considered. However, when the closed-loop control model is represented by a discrete-time control model, the pole is reset as described in claim 3. Good.

  In other words, the pole setting means sets the pole within a range where the absolute value is less than 1. Then, when hunting is detected by the hunting detection means, the pole is reset so that its absolute value is closer to 1 than the currently set value. Thus, by resetting the poles when hunting is detected, hunting can be suppressed appropriately and reliably.

  Invention of Claim 4 is the control apparatus of Claim 2 or Claim 3, Comprising: The hunting non-detection continuation judgment which judges whether the state where hunting detection means does not detect hunting continued for a predetermined time Means. Then, the pole setting means resets the pole in the direction in which the feedback gain increases when it is determined by the hunting non-detection continuation determination means that hunting has not been detected for a predetermined time.

  That is, when hunting is detected, the pole is reset in the direction in which the feedback gain decreases, while when no hunting is detected for a predetermined time, the pole is increased in the direction in which the feedback gain increases. Reset the settings.

  In this way, by increasing or decreasing the feedback gain by resetting the pole according to the detection state of hunting, the controlled object can be controlled with a more appropriate feedback gain that matches the current state of the controlled object. It is possible to achieve both effective suppression of control and improvement of control responsiveness.

  A fifth aspect of the present invention is the control device according to any one of the second to fourth aspects, wherein the target value is subjected to a smoothing process using a preset smoothing coefficient. The smoothing processing means for calculating the smoothed target value by smoothing the target value, and the smoothing coefficient setting for setting the smoothing coefficient that determines the degree of smoothing of the target value in the smoothing process Means, a feedforward manipulated variable computing means for computing a feedforward manipulated variable as an manipulated variable based on the smoothing target value calculated by the anneal processing means, and a feed computed by the feedforward manipulated variable computing means By adding the forward manipulated variable and the feedback manipulated variable calculated by the feedback manipulated variable calculating means, the input manipulated variable calculation for calculating the manipulated variable input to the control target is performed. And means, the. Then, when the hunting is detected by the hunting detection means, the annealing coefficient setting means resets the annealing coefficient to a value larger by a predetermined amount than the currently set value.

  That is, in the calculation of the feedforward manipulated variable, the target value is not used as it is, but the smoothed target value after the annealing process is used. That is, the smoothed target value subjected to the annealing process is set as a substantial target value at the calculation timing (a target value used for calculating the feedforward manipulated variable; hereinafter, also referred to as “actual target value”). The annealing process is a process of limiting the change of the target value, that is, a process of limiting the change amount of the actual target value so that the change amount of the actual target value is smaller than the actual change amount of the target value. is there. In other words, it can be said that this is a process of smoothing a change in the actual target value and setting the smoothed target value as a real target value (an annealed target value).

  In the control device configured as described above, the operation amount is calculated by adding the feedback operation amount and the feedforward operation amount. That is, a control system including feedback control and feedforward control is constructed. The feedforward manipulated variable is calculated based on the target value. However, the target value is not used as it is, but is calculated using the smoothed target value obtained by subjecting the target value to the smoothing process. Therefore, even if the change of the target value is large (stepwise change), it is possible to suppress the feedforward manipulated variable from changing greatly.

  Then, the annealing coefficient that determines the degree of annealing is set to a large value when hunting is detected. In other words, when hunting is detected, the target value is increased to suppress the fluctuation of the target value. Therefore, hunting can be effectively suppressed in a control system having both feedback control and feedforward control.

  The invention according to claim 6 is the control device according to claim 5, further comprising a hunting non-detection continuation judging means for judging whether or not a state in which no hunting is detected by the hunting detection means has continued for a predetermined time, The annealing coefficient setting means sets the annealing coefficient to a value that is smaller by a predetermined amount than the currently set value when the hunting non-detection continuation determination means determines that hunting has not been detected for a predetermined time. Reset it.

  In this way, by adjusting the degree of smoothing of the target value by resetting the smoothing coefficient according to the detection state of hunting, control with a more appropriate feedforward operation amount that matches the current status of the control target The target can be controlled, and effective suppression of hunting and improvement in follow-up of the control output to the target value can both be achieved.

  The invention according to claim 7 is the control device according to claim 5 or 6, wherein the feedback manipulated variable calculation means is controlled from the smoothing target value calculated by the smoothing processing means and the control output. A deviation is calculated, and a feedback operation amount is calculated based on the control deviation.

  In other words, the smoothed target value is also used when calculating the feedback manipulated variable (specifically when calculating the control deviation), so that the control output can be stably set to the target value even if the target value fluctuates greatly. Can be followed.

  The invention according to claim 8 is the control device according to any one of claims 5 to 7, wherein the pole upper limit value indicating the upper limit value of the absolute value is preset for the pole. In addition, an annealing coefficient upper limit value indicating the upper limit value of the annealing coefficient is set in advance. Also, there is provided an extreme upper limit judging means for judging whether or not the pole has reached the extreme upper limit value, and an annealing coefficient upper limit judging means for judging whether or not the annealing coefficient has reached the smoothing coefficient upper limit value. . The input operation amount calculation means does not use at least the feedback operation amount when it is determined that at least one of the maximum upper limit determination means and the smoothing coefficient upper limit determination means has reached the corresponding upper limit value. The operation amount is calculated by a predetermined abnormality time calculation method.

  As a result of the resetting of the poles and the resetting of the smoothing coefficient by hunting detection, the fact that the corresponding upper limit value has been reached in any one of the cases may indicate some abnormality in the control system. Therefore, in such a case, the feedback control is stopped, and for example, the control target is controlled by the operation amount using the feedforward operation amount, so that the necessary minimum control can be performed even after reaching the upper limit. Can be.

It is explanatory drawing showing the whole structure of the vehicle control system of embodiment. It is a functional block diagram showing the structure of the air fuel ratio control system of embodiment. It is a flowchart showing the injector drive control process in the air fuel ratio control system of an embodiment. It is a flowchart showing the detail of the hunting detection process of S150 in the injector drive control process of FIG. It is a flowchart showing the detail of the FB term calculation process of S260 in the injector drive control process of FIG. It is a simulation result showing the behavior when a modeling error occurs, (a) shows the fluctuation of the gain (control target model) of the controlled object, (b) shows the target value r, the FB target value yr, and the controlled variable. (c) represents the variation of the manipulated variable ufinal, and (d) represents the variation of the smoothing coefficient α and the pole α _FB . It is a simulation result showing the behavior when a disturbance is applied, (a) shows the fluctuation of the gain (control target model) of the controlled object, (b) shows the applied state of the disturbance, and (c) shows the FB target value. (d) represents the variation of the manipulated variable ufinal, and (e) represents the variation of the smoothing coefficient α and the pole α _FB .

Preferred embodiments of the present invention will be described below with reference to the drawings.
A vehicle control system 1 according to the present embodiment shown in FIG. 1 controls an engine 10. In the vehicle control system 1, the control of the engine 10 that is an internal combustion engine is realized by an electronic control unit (ECU) 50 included in the system 1 executing various processes. Below, the whole structure of the vehicle control system 1 is demonstrated first, and the content of the process performed by ECU50 is demonstrated concretely after that.

(1) Overall Configuration of Vehicle Control System 1 As shown in FIG. 1, in the vehicle control system 1 of the present embodiment, an air cleaner 15 is provided at the most upstream portion of the intake pipe 13 of the engine 10. An air flow meter 17 for detecting the intake air amount is provided on the downstream side. A throttle valve 20 that is adjusted by a motor 19 and a throttle opening sensor 21 that detects the throttle opening are provided on the downstream side of the air flow meter 17.

  A surge tank 13a is provided on the downstream side of the throttle valve 20, and an intake pipe pressure sensor 23 for detecting the intake pipe pressure is provided in the surge tank 13a. The surge tank 13a is provided with an intake manifold 25 for introducing air into each cylinder of the engine 10, and an injector 27 for injecting fuel is provided near the intake port of the intake manifold 25 of each cylinder.

  In addition, the cylinder head of the engine 10 is provided with a spark plug 29 for each cylinder, and the air-fuel mixture in the cylinder is ignited by spark discharge of the spark plug 29. In addition, each of the intake valve 31 and the exhaust valve 32 included in the engine 10 is configured to be driven by the variable valve timing adjustment mechanisms 33 and 34. In the vehicle control system 1 according to the present embodiment, the intake valve 31 and the exhaust valve 32 correspond to the engine operating state. Thus, the intake / exhaust valve timing (VVT angle) is adjusted.

  Further, the exhaust pipe 35 of the engine 10 is provided with a catalyst 37 such as a three-way catalyst for purifying the exhaust gas, and an air-fuel ratio sensor 39 for detecting the air-fuel ratio λ of the exhaust gas is provided on the upstream side of the catalyst 37. It has been. In addition, the cylinder block of the engine 10 is provided with a water temperature sensor 41 for detecting the cooling water temperature and a rotation speed sensor (crank angle sensor) 43 for detecting the engine rotation speed.

  Outputs of these various sensors are input to the ECU 50, and the ECU 50 executes various processes based on these sensor inputs. Specifically, the ECU 50 is configured mainly with a microcomputer, and executes various engine control programs recorded in a built-in ROM, thereby performing air-fuel ratio control, variable valve timing control, throttle opening control. Various engine controls such as idle speed control are realized. Hereinafter, an example in which the present invention is applied to air-fuel ratio control will be described more specifically.

(2) Outline of Air-Fuel Ratio Control In the vehicle control system 1 of the present embodiment, the reciprocal 1 / λ _T of the target air-fuel ratio λ _T that is the target value of the air-fuel ratio λ is used as the control target (target value r) to enter the operating state The target value r is switched accordingly, and based on the target value r and the actual air-fuel ratio (actual air-fuel ratio) λ detected by the air-fuel ratio sensor 39 (specifically, the reciprocal 1 / λ of the actual air-fuel ratio λ is set to the control amount y). Based on the control amount y), the fuel injection amount of the injector 27 is adjusted to realize air-fuel ratio control. Specifically, in the present embodiment, air-fuel ratio control is realized by a two-degree-of-freedom air-fuel ratio control system in which a feedback control system is combined with a feedforward control system.

In the air-fuel ratio control system of this embodiment, in feedforward control, based on the target value r and the model parameter of the model to be controlled, the feedforward manipulated variable, that is, the feedforward term in the control law (hereinafter referred to as “FF term”). u _ff is calculated. In feedback control, a closed-loop control model in which a state feedback control law is applied to a controlled object model is set, and feedback gain (hereinafter referred to as “FB”) is set by state feedback based on a difference between a target value r and a controlled variable (control output) y. The feedback manipulated variable, that is, a feedback term (hereinafter referred to as “FB term”) u _fb in the control law is calculated so that the controlled variable y follows the target value r using the FB gain. Calculate. Both the feedforward control and the feedback control are so-called model-based controls using a control target model (a plant model from the injector 27 to the air-fuel ratio sensor 39) that simulates the control target. Then, an operation amount to be finally input to the control target is calculated by adding an operation amount correction value to the sum of the FF term u _ff and the FB term u _fb .

In the air-fuel ratio control system of the present embodiment, the target value r is calculated and the control amount y is obtained at every calculation timing of a predetermined cycle, and the FF term u _ff and the FB term u _fb are calculated based on them. These are used to calculate the manipulated variable ufinal that is finally input to the plant. Then, a fuel injection command corresponding to the calculated operation amount ufinal is input to the injector 27, so that an amount of fuel corresponding to the operation amount ufinal is injected from the injector 27.

  Specifically, the air-fuel ratio control system of this embodiment can be represented by a functional block diagram as shown in FIG. The functional block diagram of FIG. 2 is a block diagram showing the functions of the air-fuel ratio control system performed by the ECU 50 (specifically, realized by the microcomputer in the ECU 50 executing the injector driving process (see FIG. 3)). It is a representation.

That is, as shown in FIG. 2, the air-fuel ratio control system of the present embodiment performs the target value r (i) by smoothing the target value r (i). The target value smoothing processing unit 61 for calculating i), the FF term operation unit 62 for calculating the FF term u _ff (i) based on the smoothing target value rfilt (i), and the smoothing target value rfilt (i) ), The FB term operation unit 63 that calculates the FB term u _fb , the other correction operation unit 64 that calculates the operation amount correction value uother (i), the FF term u _ff (i), and the FB term u _fb (i ) And the operation amount correction value uother (i) are added to calculate the operation amount ufinal (i), the model parameters a ₁ , b ₁ , b ₂ and waste of the control target model. A model parameter setting unit 75 for setting the time d; A moderation coefficient setting unit 76 sets moderation coefficient alpha (i) is used in the smoothing process in the target value smoothing processing section 61, the pole is used to calculate the FB gain in FB term calculation unit 63 alpha _FB (i ), And a hunting detection unit 78 that detects hunting of the controlled variable y (i). In the control object (plant) 66, fuel injection corresponding to the input operation amount ufinal is executed by the injector 27, and a control output (control amount) y is output for the operation amount ufinal.

  Note that the control output actually obtained from the plant is the air-fuel ratio λ that is the measured value of the air-fuel ratio sensor 39, but in the air-fuel ratio control system of the present embodiment, the reciprocal of the measured air-fuel ratio λ (excess fuel) Rate) is used as a control amount y for feedback control. Further, “i” indicates the cumulative number (number of times of calculation) of calculation timings that are discretely generated from a certain reference timing (for example, the first calculation timing after the start of control). That is, the air-fuel ratio control system of the present embodiment is configured as a digital (discrete time system) control system that performs various calculations at each calculation timing.

  A discrete-time transfer function H (z) of a control target model that simulates a control target from the injector 27 to the air-fuel ratio sensor 39 can be expressed as the following equation (1).

In the above equation (1), a ₁ , b ₁ , and b ₂ represent model parameters of the control model, and d represents a dead time.
In order to control the control target represented by such a control target model, in the present embodiment, the FF term u _ff (i) is calculated using the target value r and the model parameters a ₁ , b ₁ , b _2. At the same time, the FB term u _fb (i) is calculated according to the state feedback control law.

The model parameters a ₁ , b ₁ , b ₂ are set by the model parameter setting unit 75. In the present embodiment, model parameters a ₁ , b ₁ , and b ₂ are set in advance based on the controlled object model, and these are used for calculation of the FF term u _ff (i) and the FB term u _fb (i). The

  The model parameter setting unit 75 is also set with a dead time d. In the air-fuel ratio control system of the present embodiment, a certain amount of time is required from when the fuel injection command based on the target value r (i) calculated at a certain calculation timing is output until the control amount y reflecting it is observed. Since there is a time difference L, the dead time d is set in consideration of the time difference L. In the present embodiment, the dead time d is obtained by converting the time difference L into an integer based on the calculation cycle (a numerical value representing an integer multiple of the calculation cycle). In the following description, a case where the dead time d = 7 is taken as an example.

Note that the model parameters a ₁ , b ₁ , and b ₂ may be dynamically set at a predetermined cycle (for example, the calculation cycle) dt. Specifically, for example, it may be calculated by the following equations (2) to (4) using the model time constant τ and the calculation cycle dt.

Note that L1 in equation (3) is calculated by L1 = L−d · dt using the time difference L, the dead time d, and the calculation cycle dt. Since a more specific method for dynamically setting the model parameters a ₁ , b ₁ , and b ₂ is described in detail in the aforementioned Patent Documents 1 and 2, detailed description thereof is omitted here.

  The target value r (i) is generated by a target value generation function that the ECU 50 has. The ECU 50 calculates a target value r (i) using a map or the like according to various engine states such as the engine speed and the accelerator opening at each calculation timing.

The target value smoothing processing unit 61 performs a smoothing process on the target value r (i) at each calculation timing, and smoothes the result (that is, the result of the target value r (i) being smoothed). The target value rfilt (i) is output. The annealing process is a process for smoothing a change in the actual target value r (i), and the smoothed value is used in the calculation of the FF term u _ff (i) and the FB term u _fb (i). Target value (annealing target value). In the present embodiment, j-time smoothing processing is performed from the first smoothing processing to the j-th (j is a natural number of 2 or more) smoothing processing at each calculation timing. An n-annealed value r _n (i) (n = 1 to j) is calculated.

  Specifically, the smoothing process is sequentially executed by j operations from the following formulas (5) to (7).

That is, the first first smoothing process according to the above formula (5) is the first smoothing calculated in the first smoothing process at the target value r (i) acquired at the current calculation timing and the previous calculation timing. The first smoothed value r ₁ (i) is calculated by weighting and adding the value r ₁ (i−1) with the smoothing coefficient α.

When the first smoothed value r ₁ (i) is calculated, the second smoothing process according to the above equation (6) is then performed using the first smoothed value r ₁ (i), and the second smoothing value r ₁ (i) is calculated. The preferred value r ₂ (i) is calculated. This second smoothing process includes the first smoothed value r ₁ (i) calculated by the equation (5) and the second smoothed value r ₂ (i) calculated by the second smoothing process at the previous calculation timing. -1) is weighted and added by the smoothing coefficient α. Similarly, the smoothing process is repeated sequentially from the third time to the jth time.

That is, in the n-th annealing process for the n-th time after the second time, the ( _n-1 ) _-th annealing value r _n-1 (i ) And the nth smoothed value r _n (i−1) calculated by the nth smoothing process at the previous calculation timing are weighted and added by the smoothing coefficient α to obtain the nth smoothed value r _n ( i) is calculated. As a result, the j-th j-th annealing process is finally performed in equation (7), and the j-th annealing value r _j (i) is calculated.

The jth smoothing value r _j (i) finally calculated by performing _j smoothing processes in this way is set as the smoothing target value rfilt (i) as shown in the following equation (8). .

  The smoothing coefficient α determines the strength (weight) of the smoothing, and is set by the smoothing coefficient setting unit 76 within a range of 0 ≦ α <1. The greater the annealing coefficient α, the greater the effect of annealing. Note that the number j of the annealing processes can be set as appropriate. The target value smoothing processing unit 61 acquires the smoothing coefficient α set by the smoothing coefficient setting unit 76 at each calculation timing, and uses the smoothing coefficient α as the smoothing coefficient α at the current calculation timing. Used for annealing.

The FF term calculation unit 62 includes the target value rfilt (i) calculated at the current calculation timing in the target value smoothing processing unit 61 and the target value rfilt (i−1) calculated at the previous calculation timing. , And model parameters a ₁ , b ₁ , b ₂ , the FF term u _ff (i) is calculated by the following equation (9).

The FB term calculation unit 63 calculates the FB term u _fb (i) using the smoothing target value rfilt (i) calculated by the target value smoothing processing unit 61. , An FB gain calculation unit 72, a deviation calculation unit 73, and an FB term calculation unit 74.

The FB target value calculation unit 71 performs the smoothing target value rfilt (id) at the calculation timing before the dead time d and the smoothing target value rfilt (id-1) at the calculation timing before the dead time d + 1. ), Model parameters b ₁ and b ₂ , and dead time d, the FB target value yr (i) is calculated by the following equation (10).

  Note that the calculation of the FB target value yr (i) used for calculating the FB term by the above equation (10) is merely an example, and the smoothing target value rfilt (id) that is just before the dead time d is used as the FB. The target value yr (i) may be used.

  The deviation calculating unit 73 calculates a control deviation e (i) that is a difference between the FB target value yr (i) and the controlled variable y (i) as shown in the following equation (11).

The FB term calculation unit 74 calculates the FB term u _fb (i) using the FB gain (F (1) to F (9) in this example) based on the state feedback control law.
In general, when the FB term u _fb (i) is calculated using the FB gains F (1) to F (9) of the state feedback, a polynomial operation using these FB gains F (1) to F (9) is performed. In many cases, the FB term u _fb (i) is directly calculated, and in this embodiment, it may be calculated by such a method. However, in the above method, there is a possibility that the FB term u _fb (i) may be temporarily disturbed by changes in the FB gains F (1) to F (9).

Therefore, in this embodiment, the FB term correction value Δu _fb (i), which is the correction value of the FB term u _fb (i), is calculated at each calculation timing. Then, as shown in the following equation (12), the calculated FB term correction value Δu _fb (i) is added to the FB term u _fb (i−1) calculated at the previous calculation timing, thereby obtaining the current FB. The term u _fb (i) is calculated.

Here, the FB term correction value Δu _fb (i) is calculated by a state feedback control law expressed by the following equation (13).

In the above equation (13), Δe (i) is the difference between the current control deviation e (i) and the previous control deviation e (i−1), as shown in the following equation (14), and Δu _fb (I−n) (where n = 1 to 7), as shown in the following formulas (15) to (21), the FB term u _fb (i−n) calculated at the calculation timing n times before and n + 1 This is the difference from the FB term u _fb (i−n−1) calculated at the previous calculation timing.

In order to calculate the FB term correction value Δu _fb (i) by the state feedback control law using the above equation (13), it is necessary to calculate the FB gains F (1) to F (9). Therefore, in this embodiment, the FB gain calculation unit 72 calculates the FB gains F (1) to F (9).

The FB gain calculation unit 72 calculates the FB gains F (1) to F (9) based on the pole placement method in the state feedback control. Specifically, the pole of the closed-loop control model (root of the closed-loop characteristic equation) is set, and the FB gain is calculated using the set pole and the model parameters a ₁ , b ₁ , and b _{2 of} the controlled object model. To do.

  In the present embodiment, the closed loop characteristic equation is set as in the following equation (22).

That is, the pole of the closed-loop control model of the present embodiment is α _FB , and the pole α _FB is appropriately set to stabilize the closed-loop system. In general, in a discrete-time system control model, it is a stable condition that poles are arranged in a unit circle in the z plane (in other words, the absolute value of the pole is less than 1), or the absolute value of the pole is It is known that the closer to 1, the smaller the FB gain and the more the stability can be improved.

Therefore, also in this embodiment, the pole α _FB is set within a range in which the absolute value is less than 1. The pole α _FB is set by the pole setting unit 77, details of which will be described later.
The left side of the closed-loop characteristic equation (ie, characteristic polynomial) in equation (22) can be expressed in other ways, for example, as shown in equation (23) below.

Patent Documents 1 and 2 describe in detail that the characteristic polynomial is set as in the above equation (23) and the FB gain is calculated based on the characteristic polynomial.
In this embodiment, it is also possible to set the poles using a characteristic polynomial such as the above equation (23). However, in the present embodiment, as shown on the left side of the above equation (22), in order to simplify the calculation. A characteristic polynomial is set and the pole α _FB is set based on this characteristic polynomial.

Therefore, if the pole α _FB is set, the coefficients A1 and A2 of the characteristic polynomial in the above equation (23) can be obtained as the following equations (24) and (25).

The FB gains F (1) to F (9) can be calculated by the following equations (26) to (34) using these coefficients A1 and A2 (that is, using the pole α _FB ).

The FB term calculation unit 74 acquires the pole α _FB set by the pole setting unit 77 and the model parameters a ₁ , b ₁ , and b ₂ set by the model parameter setting unit 75 at each calculation timing. Then, using the obtained pole α _FB and model parameters a ₁ , b ₁ , b ₂ , FB gains F (1) to F (9) are calculated by the above formulas (24) to (34). Then, using the calculated FB gains F (1) to F (9), the FB term correction value Δu _fb (i) is calculated by the above equation (13), and the FB term u _fb is further calculated by using the equation (12). (I) is calculated.

  The hunting detection unit 78 determines the presence / absence of hunting of the control amount y (i) at each calculation timing, and outputs the determination result to the smoothing coefficient setting unit 76 and the pole setting unit 77. A specific method of hunting detection will be described later with reference to FIG.

  The smoothing coefficient setting unit 76 sets the smoothing coefficient α used when the target value smoothing processing unit 61 performs the smoothing process on the target value r (i). Specifically, the initial value of the smoothing coefficient α is set in advance to an appropriate value (for example, 0.5) within the range of 0 ≦ α <1. Thereby, the target value smoothing processing unit 61 performs the smoothing process using the initial value at least when the control is started.

Each time hunting is detected by the hunting detector 78, the smoothing coefficient α is increased from the current value (the value at the previous calculation timing) by a predetermined amount C _P α, so that the degree of annealing is increased. I try to increase it.

In addition, the annealing coefficient setting unit 76 also determines whether or not the state in which hunting is not detected continues for a predetermined time. If the state in which no hunting is detected continues for a predetermined time, the annealing coefficient α is calculated from the current value. The degree of annealing is reduced by reducing the predetermined amount C _M α.

The pole setting unit 77 sets the pole α _FB used when the FB gain calculation unit 72 calculates the FB gains F (1) to F (9). Specifically, first, an initial value of the pole α _FB is set in advance to a predetermined value less than 1 (for example, 0.9), and control is executed using the initial value at the start of control.

Each time hunting is detected by the hunting detection unit 78, the pole α _FB is increased from the current value by a predetermined amount C _p α _FB to be close to 1, thereby reducing the FB gain, thereby hunting. Suppression (stabilization of control output).

The pole setting unit 77 also determines whether or not the state in which hunting is not detected continues for a predetermined time. If the state in which no hunting is detected continues for a predetermined time, the pole α _FB is changed from the current value to a predetermined amount C _By reducing only _M α _FB , the FB gain is increased to improve the control response.

The amount by which the smoothing coefficient α is increased or decreased when detecting hunting (C _P α, C _M α) and the amount by which the pole α _FB is increased / decreased (C _P α _FB , C _M α _FB ) must be set individually. Can do.
The other correction calculation unit 64 performs various corrections on the FF term u _ff (i) and the FB term u _fb (i). For example, the other correction calculation unit 64 performs a predetermined calculation based on the coolant temperature of the engine, the acceleration state of the vehicle, or the like. Then, an operation amount correction value uother (i) is calculated.

The manipulated variable calculator 65 includes the FF term u _ff (i) calculated by the FF term calculator 62, the FB term u _fb (i) calculated by the FB term calculator 63, and the other correction calculator 64. The operation amount correction value uother (i) calculated in the above is added to calculate the operation amount ufinal (i).

(3) Description of Injector Drive Control Process Next, the injector drive control process executed by the ECU 50 to realize the function of the air-fuel ratio control system of FIG. 2 described above will be described with reference to FIG. In the following description, for simplification of description, a case where the smoothing process of the target value r (i) is performed twice (that is, j = 2) will be described as an example.

  When the microcomputer of the ECU 50 starts the injector drive control process at every calculation timing, first, in S110, the microcomputer calculates a target value r (i) from the engine state. Note that the target value r (i) calculated here is the reciprocal of the target air-fuel ratio λT (i).

In S120, the model parameter of the control target model is read out. Specifically, each preset model parameter a ₁ , b ₁ , b ₂ and dead time d are read. These are set by the model parameter setting unit 75 shown in FIG. 2 as described above.

  In S130, the actual air-fuel ratio λ (i) that is a measured value of the air-fuel ratio sensor 39 is read. In S140, the reciprocal of the read actual air-fuel ratio λ (i) is set as a control amount y (i), and a hunting detection process is performed in S150. This hunting detection process is a process for realizing the function as the hunting detection unit 78 shown in FIG.

Details of the hunting detection process of S150 are as shown in FIG. That is, first, in S310, a high-pass filter process of the control amount y (i) is performed. Specifically, the filter processing control amount y _HP (i), which is the control amount from which the low frequency component is removed, is calculated by performing the calculation of the following equation (35).

  Note that it is not essential to perform the high-pass filter processing on the control amount y (i) as described above, and the processing after S320 may be performed using the control amount y (i) as it is. However, in order to prevent erroneous detection of hunting and improve detection accuracy, it is desirable to perform unnecessary high-pass filtering to remove unnecessary components.

After the high-pass filter process, it is determined in S320 whether or not the filter process control amount y _HP (i) is equal to or greater than the positive first threshold value. The first threshold can be appropriately set according to the detection sensitivity required for hunting detection, and is set to 0.005, for example, in this example.

If it is determined in S320 that the filter processing control amount y _HP (i) is greater than or equal to the first threshold, the determination flag s is set to 1 in S330, and the process proceeds to S360.
If the filter processing control amount y _HP (i) is not determined to be greater than or equal to the first threshold value in S320, whether or not the filter processing control amount y _HP (i) is equal to or less than the negative second threshold value in S340. Judging. The second threshold value can also be set as appropriate according to the detection sensitivity required for hunting detection, and is set to, for example, -0.005 in this example.

If it is determined in S340 that the filter processing control amount y _HP (i) is less than or equal to the second threshold, the determination flag s is set to 0 in S350, and the process proceeds to S360. If it is not determined in S340 that the filter processing control amount y _HP (i) is equal to or smaller than the second threshold value, the process proceeds to S360 as it is (that is, the determination flag remains as it is).

In S360, the number of inversions cs of the determination flag s within a predetermined time (for example, 1 second) is updated. The initial value of the number of inversions cs is set to 0, and is cleared to 0 every predetermined time (in this example, 1 second) after the start of control. At each calculation timing, the determination flag s when the process proceeds to S360 at the previous calculation timing is compared with the determination flag s when the process proceeds to S360 at the current calculation timing, and the determination flag is inverted. In this case, the inversion number cs is incremented by one. Therefore, if the filter processing control amount y _HP (i) is repeatedly increased to the first threshold value or more and decreased to the second threshold value or less in one second, the inversion number cs increases accordingly. become.

  In S370, it is determined whether or not the number of inversions cs is equal to or greater than the inversion number threshold. This inversion number threshold value can also be set as appropriate according to the detection sensitivity required for hunting detection, and is set to, for example, twice in this example.

  When it is determined in S370 that the number of inversions cs is equal to or greater than the inversion number threshold, the process proceeds to S380, where it is determined that hunting is present, that is, hunting is occurring. On the other hand, if it is not determined in S370 that the number of inversions cs is equal to or greater than the inversion number threshold, the process proceeds to S390, where it is determined that no hunting has occurred, that is, no hunting has occurred.

After performing the hunting detection process (determination of hunting presence / absence) in this way, the process proceeds to S160 (FIG. 3), and whether or not hunting is detected (that is, it is determined that hunting is present in the hunting detection process of S150). Whether or not). If hunting is detected, the process proceeds to S170, and the smoothing coefficient α and the pole α _FB are increased as shown in the following equations (36) and (37), and the process proceeds to S200. That is, the smoothing effect α is increased to increase the smoothing effect, and the pole α _FB is brought close to 1 to reduce the FB gain, thereby suppressing hunting.

  On the other hand, if it is determined in S160 that hunting has not been detected (that is, if it is determined that hunting is not detected in the hunting detection process in S150), the process proceeds to S180 and the state where hunting is not detected is determined. It is determined whether or not it continues for a predetermined time. This predetermined time can also be set as appropriate.

Then, while the state determined as having no hunting continues for a predetermined time, the process proceeds to S240. However, when the state determined to have no hunting continues for a predetermined time, the process proceeds to S190, and the following equation (38) is satisfied. , (39), the annealing coefficient α and the pole α _FB are respectively decreased, and the process proceeds to S200. In other words, the smoothing coefficient α is made smaller to reduce the effect of smoothing, and the pole α _FB is moved away from 1 (closed to 0) to increase the FB gain, thereby improving the control response. is there.

  More specifically, the determination in S180 is performed every predetermined time. That is, when it is determined that there is no hunting continuously from the calculation timing i−m times before the predetermined time to the current i-th calculation timing after the predetermined time has elapsed, A positive determination is made in S180. Thereafter, it is determined again whether or not it is determined that there is no hunting in the period from the current i-th calculation timing to the i + m-th calculation timing after a predetermined time has elapsed.

  Alternatively, after the determination that there is no hunting continues for a predetermined time, until it is determined that hunting is present, an affirmative determination may be made at S180 for each calculation timing, and the process of S190 may be performed.

In S200, a guard process is performed on the _annealing coefficient α and the pole α _FB that have been increased in S170 or decreased in S190. That is, it is determined whether or not the smoothing coefficient α is within the range of the smoothing coefficient lower limit value Minα to the smoothing coefficient upper limit value Maxα as shown in the following equation (40).

  If the smoothing coefficient lower limit value Minα is smaller than the smoothing coefficient lower limit value Minα, the smoothing coefficient lower limit value Minα is set. Conversely, if the smoothing coefficient lower limit value Maxα is larger than the smoothing coefficient lower limit value Maxα, The smoothing coefficient α is set to the smoothing coefficient upper limit value Maxα.

Similarly, the pole alpha _FB, determines whether within the scope of the pole lower limit Minα _FB ~ pole upper limit Maxarufa _FB as shown in the following equation (41).

If it is smaller than the lower limit value Minα _FB , the pole α _FB is set to the lower limit value Minα _FB , and conversely if it is larger than the upper limit value Maxα _FB , the pole α _{FB is set} to The maximum value Maxα _FB is set.

After the guard process in S200, it is determined in S210 whether any one of the smoothing coefficient α and the pole α _FB has reached the upper limit value. That is, it is determined whether or not the annealing coefficient α is set to the annealing coefficient upper limit value Maxα and whether or not the pole α _FB is set to the pole upper limit value Maxα _FB .

If either of them is set to the upper limit value, it is determined in S220 that the fuel system is abnormal (that is, some abnormality has occurred in the air-fuel ratio control system), for example, the vehicle Predetermined abnormality processing such as urging the driver to evacuate. In subsequent S230, feedback control is prohibited in the air-fuel ratio control system. That is, a predetermined abnormality time calculation method that does not use the FB term u _fb (i) (for example, a sum of the FF term u _ff (i) and the operation amount correction value uother (i) is defined as an operation amount ufinal (i). The operation amount ufinal (i) is calculated by a calculation method or the like. Accordingly, the driver can evacuate even in the event of an abnormality while prompting the driver to recognize the abnormality early and respond appropriately and promptly.

On the other hand, if neither the smoothing coefficient α nor the pole α _FB is set to the upper limit value in S210, the process proceeds to S240. In S240, the target value r (i) is smoothed. Specifically, the annealing process is performed twice by the following equations (42) and (43).

  Then, the result of the two annealing processes is calculated as an annealing target value rfilt (i) as shown in the following equation (44).

After the target value r (i) is annealed, the FF term u _ff (i) is calculated in S250. Specifically, it is calculated by the above-described equation (9).
After calculating the FF term u _ff (i), an FB term calculation process is executed in S260. Details of the FB term calculation processing in S260 are as shown in FIG. That is, first, in S410, the FB target value yr (i) is calculated. Specifically, it is calculated by the above-described equation (10).

In S420, FB gains F (1) to F (9) are calculated. Specifically, the coefficients A1 and A2 of the characteristic polynomial are calculated by the above-described equations (24) and (25) using the currently set pole α _FB . Then, the FB gains F (1) to F (9) are calculated using the equations (26) to (34), respectively.

In S430, each variable e (i), Δe (i), Δu _fb (in) used for calculation of the FB term u _fb (i) (specifically, calculation of the FB term correction value Δu _fb (i)) is set. The calculation is performed using the above-described equations (11) and (14) to (21).

In S440, the FB term correction value Δu _fb (i) is calculated by the above-described equation (13) using the calculated variables, and the FB term correction value Δu _fb (i) is further used. Then, the FB term u _fb (i) is calculated by the above-described equation (12). After calculating the FB term u _fb (i) in this way, the process proceeds to S270 (FIG. 3), and an operation amount correction value uother (i) is calculated.

  In S280, the operation amount ufinal (i) is calculated by the following equation (45), and the injector 27 is driven by outputting a fuel injection command based on the calculated operation amount ufinal (i) to the injector 27. .

(4) Effects of Embodiment, etc. According to the vehicle control system 1 of the present embodiment described above, when the hunting of the controlled variable y (i) occurs due to the modeling error of the controlled object, the pole α _FB is By resetting, the FB gains F (1) to F (9) become small, and the fluctuation of the FB term u _fb becomes gentle. Thereby, the control system can be further stabilized. Therefore, feedback control can be performed with an appropriate FB gain that matches the current status of each control target (considering modeling errors and the like), and hunting of the control amount y (i) is effectively suppressed. Can do.

The pole α _FB is set within the stability condition range (0 ≦ α _FB <1) in advance, and is used. Whenever hunting is detected, a predetermined amount C is added to the current pole α _{FB. By} adding _P α _FB , the pole α _FB is made closer to 1. By re-setting the pole α _FB for hunting in this way, hunting can be appropriately and reliably suppressed with a simple configuration.

In addition, the resetting of the pole α _FB is not only increased by a predetermined amount every time hunting is detected, but is also subtracted by a predetermined amount C _MP α _FB when hunting is not detected for a predetermined time. Yes. That is, when hunting is detected, the pole α _FB is reset in such a direction that the FB gain decreases. On the other hand, if hunting is not detected for a predetermined time, the FB gain increases. In this case, the pole α _FB is reset.

In this way, by increasing / decreasing the FB gain by increasing / decreasing the pole α _FB according to the detection state of hunting, the controlled object can be controlled with a more appropriate feedback gain that matches the current state of the controlled object. It is possible to achieve both effective suppression and improvement in control responsiveness.

In the present embodiment, the target value r (i) is not used as it is for the feedback control, and the feedback control is performed using the smoothed target value rfilt (i) obtained by smoothing the target value r (i). Yes. Therefore, even if the change of the target value r (i) is large (stepwise change), the fluctuation of the target value used for the feedback control can be suppressed, and the fluctuation of the FB term u _fb (i) is thereby also reduced. As a result, fluctuations in the control amount y (i) can be suppressed. That is, by using the smoothed target value rfilt (i) as the target value, the effect of suppressing hunting can be obtained.

In this embodiment, air-fuel ratio control is performed by a control system that combines not only feedback control but also feedforward control. In the feedforward control, the FF term u _ff (i) is calculated using the smoothing target value rfilt (i). Therefore, even for the FF term u _ff (i), even if the change in the target value r (i) is large, the fluctuation can be suppressed, and hunting can be suppressed accordingly.

In addition, the smoothing coefficient α used when the target value r (i) is smoothed is increased or decreased according to the hunting detection state of the control amount y (i), similarly to the pole α _FB . Therefore, hunting can be more effectively suppressed in a control system including both feedback control and feedforward control.

Here, among the effects of the present embodiment, a specific example in which hunting can be suppressed by resetting the smoothing coefficient α and the pole α _FB will be described with reference to the simulation results of FIGS. 6 and 7. In the examples of FIGS. 6 and 7, the smoothing coefficient α and the pole α _FB are set to the same initial value, and the increase amount at the time of detecting hunting is also set to the same value.

  First, FIG. 6 shows the behavior when a modeling error occurs in the controlled object. As shown in FIG. 6A, a modeling error is given from the time 6 seconds to about 15 seconds later in the controlled object. . As shown in FIG. 6B, the target value r is changed stepwise at a predetermined period.

  While no modeling error occurs and when the modeling error is small, as shown in FIG. 6B, the control amount y follows the target value r (FB target value yr). However, as the modeling error increases, the manipulated variable ufinal fluctuates greatly as shown in FIG. 6 (c), and the control variable y fluctuates greatly as shown in FIG. 6 (b). Become. Thereby, hunting is detected around the time 15 seconds.

When hunting is detected, the smoothing coefficient α and the pole α _FB are increased by a predetermined amount at each calculation timing until hunting is not detected, as shown in FIG. As a result, the degree of smoothing of the target value r is increased, and the FB gains F (1) to F (9) are reduced, so that the control amount y is controlled as shown in FIGS. Both the fluctuation and the fluctuation of the manipulated variable ufinal become smaller. By increasing the smoothing coefficient α and the pole α _FB in this way, hunting at the time of the next target value change (after about 21 seconds) is suppressed.

Next, FIG. 7 shows the compensation behavior for the disturbance. Specifically, as shown in FIG. 7A, a modeling error is generated by changing the control target from time 6 seconds to about 15 seconds, and in parallel, as shown in FIG. 7B. 2 is a behavior when a disturbance (for example, purge) of 0.1 is applied in a pulsed manner. Also in this example, due to the disturbance, as shown in FIGS. 7C and 7D, the operation amount ufinal and the control amount y fluctuate and hunting occurs. On the other hand, the behavior (hunting) against disturbance is suppressed by appropriately setting the smoothing coefficient α and the pole α _FB as shown in FIG.

  By applying the present invention to the air-fuel ratio control system, it becomes possible to make the control amount (air-fuel ratio) follow the target value even when a modeling error occurs, improving the operating performance and exhaust gas purification performance. To do.

  In this embodiment, the pole setting unit 77 corresponds to a pole setting unit and a pole upper limit determination unit of the present invention, and the FB gain calculation unit 72 corresponds to a feedback gain calculation unit of the present invention, and the deviation calculation unit 73 and the FB The term calculation unit 74 corresponds to the feedback manipulated variable calculation unit of the present invention, the hunting detection unit 78 corresponds to the hunting detection unit and the hunting non-detection continuation determination unit of the present invention, and the target value smoothing processing unit 61 corresponds to the present invention. The smoothing coefficient setting unit 76 corresponds to the smoothing processing means, the smoothing coefficient setting means and the smoothing coefficient upper limit determination means of the present invention, and the FF term calculation unit 62 corresponds to the feedforward manipulated variable calculation means of the present invention. The operation amount calculator 65 corresponds to the input operation amount calculator of the present invention.

[Modification]
Although the embodiments of the present invention have been described above, the embodiments of the present invention are not limited to the above-described embodiments, and can take various forms as long as they belong to the technical scope of the present invention. Needless to say.

For example, in the above embodiment, the pole alpha _FB, set the first initial value of less than 1, but so as to increase or decrease a predetermined amount in accordance with the hunting detection state, the pole alpha _FB is the unit circle in the z plane (Absolute value is less than 1) can be set as appropriate. Then, every time hunting is detected, the absolute value is corrected so as to approach 1, and conversely, when the non-detected state of hunting continues, the absolute value is corrected so as to approach 0. Good. The detection method of hunting is not limited to the method shown in FIG. 4, and detection can be performed by various methods.

Further, the increase and decrease of the pole α _FB can be performed as follows. For example, a value close to 0 is set as an initial value. Then, gradually increase each time hunting is detected, and if hunting is not detected at a certain point, the value of the current pole α _FB is assumed to be an appropriate value and the value is used thereafter (that is, no subtraction is performed). Like that. Conversely, a value close to 1 is set as the initial value, and it is gradually decreased every time hunting is detected. If hunting is lost at a certain point, the pole α _FB at that point is _assumed to be an appropriate value. Keeps using that value (ie, not adding).

However, as described above, if the pole α _FB is continuously used when hunting is no longer detected, the α _FB may become inappropriate for some reason. For example, the results had very alpha _FB is determined when the chance there is some abnormality, then even though the abnormality is eliminated has been very alpha _FB is no longer a proper value, the pole alpha _FB previously determined The case where it keeps using it can be considered. Therefore, it is preferable that hunting detection is continuously performed as in the above-described embodiment, and the pole α _FB is appropriately increased or decreased according to the detection state. The same applies to the annealing coefficient α.

  Further, in the above embodiment, the target value r is smoothed by the methods shown in the equations (5) to (8). However, such a calculation method is merely an example, and the target value r ( As long as i) can be appropriately performed, various specific methods are conceivable.

  Further, when the annealing process is performed a plurality of times, the annealing coefficient in each annealing process may be set individually. In addition, it is not essential to perform the annealing process on the target value r, and it is not essential to use the annealing target value in both the FB control and the FF control. However, in order to enhance the hunting suppression effect, Thus, it is preferable to smooth the target value and use the smoothed target value in both the FB control and the FF control.

  In the above embodiment, the example in which the present invention is applied to an air-fuel ratio control system has been described. However, the application of the present invention is not limited to an air-fuel ratio control system, and a control system that controls a controlled object by feedback control. In addition, the present invention can be applied to a control system that controls a controlled object by both feedback control and feedforward control.

    DESCRIPTION OF SYMBOLS 1 ... Vehicle control system, 10 ... Engine, 13 ... Intake pipe, 15 ... Air cleaner, 17 ... Air flow meter, 19 ... Motor, 20 ... Throttle valve, 21 ... Throttle opening sensor, 23 ... Intake pipe pressure sensor, 25 ... Intake Manifold 27 ... Injector 29 ... Spark plug 31 ... Intake valve 32 ... Exhaust valve 33,34 ... Variable valve timing adjustment mechanism 35 ... Exhaust pipe 37 ... Catalyst 39 ... Air fuel ratio sensor 41 ... Water temperature sensor , 43 ... Rotational speed sensor, 50 ... ECU, 61 ... Target value smoothing processing unit, 62 ... FF term calculation unit, 63 ... FB term calculation unit, 64 ... Other correction calculation unit, 65 ... Manipulation amount calculation unit, 66 ... Control target (plant), 71... FB target value calculation unit, 72... FB gain calculation unit, 73... Deviation calculation unit, 74. Setting unit, 76 ... averaging coefficient setting unit, 77 ... electrode setting unit, 78 ... hunting detector

Claims (8)

A control method that performs state feedback control of an operation amount input to a control target so that a control output of the control target follows a target value using a control target model that simulates the control target,
Set a pole of a closed-loop control model that applies a state feedback control law to the controlled object model, calculate a feedback gain using the set pole and the model parameter of the controlled object model,
A control deviation indicating a difference between the target value and the control output is calculated, and based on the control deviation, a feedback operation amount as the operation amount is obtained by the state feedback control law using the feedback gain. Operate,
When the hunting of the control output is detected, the pole is reset in a direction in which the feedback gain decreases. A control device that performs state feedback control of an operation amount input to a control target so that a control output of the control target follows a target value using a control target model that simulates the control target,
Pole setting means for setting a pole of a closed loop control model in which a state feedback control law is applied to the controlled object model to a value within a range satisfying a predetermined stability condition;
Feedback gain calculating means for calculating a feedback gain using the pole set by the pole setting means and the model parameters of the controlled object model;
A control deviation indicating a difference between the target value and the control output is calculated, and based on the control deviation, the state feedback control law using the feedback gain calculated by the feedback gain calculation unit is used. A feedback manipulated variable calculating means for calculating a feedback manipulated variable as an operation variable;
Hunting detection means for detecting hunting of the control output;
With
The control device, wherein the pole setting means resets the pole in a direction in which the feedback gain becomes smaller when the hunting detection means detects the hunting. The control device according to claim 2,
The closed-loop control model is a discrete-time control model,
The pole setting means sets the pole within a range where the absolute value is less than 1, and when the hunting detection means detects the hunting, the pole is set to an absolute value higher than the currently set value. A control device, wherein the value is reset so that the value approaches 1. The control device according to claim 2 or 3, wherein
Hunting non-detection continuation determination means for determining whether or not the state where the hunting is not detected by the hunting detection means continues for a predetermined time,
The pole setting means resets the pole in a direction in which the feedback gain increases when the hunting non-detection continuation determination means determines that the hunting is not continuously detected for the predetermined time. Control device characterized. It is a control device given in any 1 paragraph of Claims 2-4,
An annealing processing means for calculating an annealing target value obtained by performing an annealing process on the target value using a preset annealing coefficient;
Smoothing coefficient setting means for setting the smoothing coefficient that determines the degree of smoothing of the target value in the smoothing process;
A feedforward manipulated variable calculation means for calculating a feedforward manipulated variable as the manipulated variable based on the smoothing target value calculated by the smoothing processing means;
The operation amount input to the control target is calculated by adding the feedforward operation amount calculated by the feedforward operation amount calculation means and the feedback operation amount calculated by the feedback operation amount calculation means. Input manipulated variable calculation means;
With
The smoothing coefficient setting means resets the smoothing coefficient to a value larger by a predetermined amount than the currently set value when the hunting is detected by the hunting detection means. apparatus. The control device according to claim 5,
Hunting non-detection continuation determination means for determining whether or not the state where the hunting is not detected by the hunting detection means continues for a predetermined time,
If the hunting non-detection continuation determination unit determines that the hunting is not continuously detected for the predetermined time, the annealing coefficient setting unit sets the annealing coefficient to a value that is currently set. A control device characterized by resetting to a value smaller by a predetermined amount. The control device according to claim 5 or 6, wherein
The feedback manipulated variable calculating means calculates the control deviation from the smoothing target value calculated by the smoothing processing means and the control output, and calculates the feedback manipulated variable based on the control deviation. Control device characterized. It is a control device given in any 1 paragraph of Claims 5-7,
The pole upper limit value indicating the upper limit value of the absolute value is preset for the pole, and the annealing coefficient upper limit value indicating the upper limit value is preset for the annealing coefficient,
Extreme upper limit judging means for judging whether or not the pole has reached the extreme upper limit;
An annealing coefficient upper limit judging means for judging whether or not the annealing coefficient has reached the annealing coefficient upper limit value;
With
When the input operation amount calculation means determines that the corresponding upper limit value has been reached by at least one of the extreme upper limit determination means and the smoothing coefficient upper limit determination means, at least the feedback operation amount The control device is characterized in that the operation amount is calculated by a predetermined abnormality time calculation method without using any of the above.

Images