JP2010229968A - Control device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the responsiveness of control by a simple method. <P>SOLUTION: Average values ave_eegr, ave_epim for output data used in a system identification stage are subtracted from actually measured values eegr, epim for control outputs given to a sliding mode controller 51, and average values ave_EGRv, ave_VNT, ave_Dth for input data are added to control inputs u<SB>1</SB>, u<SB>2</SB>, u<SB>3</SB>calculated by the sliding mode controller 51. Thus, deviation between a model and an actual plant is reduced and opportunities of increasing nonlinear input terms and application terms resulting from other elements than disturbance are reduced, thus improving the responsiveness of control. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a control device for controlling an internal combustion engine or a device attached thereto.

  An exhaust gas recirculation system disclosed in Patent Document 1 below controls an EGR rate (or EGR amount) of an internal combustion engine equipped with a supercharger. There is mutual interference between the supercharging pressure and the EGR rate, and it is difficult to simultaneously control both the supercharging pressure and the EGR rate with a one-input one-output controller. In addition, the responsiveness varies depending on the operation region of the internal combustion engine, and the turbocharger has a turbo lag (dead time). Under such circumstances, the system described in Patent Document 1 adopts sliding mode control, which is an effective control method for a non-linear control target, and designs an EGR controller by designing a multi-input multi-output controller in consideration of interaction. I have control.

  When designing a controller, it is necessary to identify the characteristics of the plant to be controlled. As described in Non-Patent Document 1 below, in system identification, input / output data to be controlled is collected by experiment, and after performing a preprocessing to remove a direct current (bias) component of the input / output data, It is customary to try to build a state space model (identify parameters).

  The reason for removing the direct current component from the input / output data and setting the average value to 0 is to shorten the experiment time. For example, when performing identification using the frequency response method, an experiment time is required so that a sine wave having a target frequency can be measured for several cycles or more. The 1 Hz sine wave period is 1 second, but the 0.001 Hz sine wave period is 1000 seconds, reaching about 16 minutes. Thus, the longer the experiment time is consumed, the lower the frequency component. In particular, since the DC component can be considered as a low-frequency signal having an infinite cycle, extremely long input / output data is required to identify the DC component. Even in the case of performing identification using a prediction error method such as the least square method, it is the same in that the measurement takes time for the low frequency component.

  Second, in the numerical calculation using a computer, the removal of the DC component from the data results in smaller numerical digits and higher calculation accuracy.

  A controller designed on the basis of input / output data from which a direct current component has been removed realizes a control output centered on 0 with a control input centered on 0. However, in general, the actual operation unit of the plant is not operated around the operation amount 0, and the control output does not fluctuate around 0. Then, the range of the control input and control output taken by the model and the one taken by the actual plant are offset to compensate for the difference (in the case of adaptive sliding mode control, this is mainly due to the nonlinear input terms and adaptive terms). It takes time to reach the actual control target (due to the increase), resulting in a deterioration in responsiveness.

JP 2007-032462 A

Shuichi Adachi, “Advanced System Identification for Control by MATLAB”, 1st Edition, Tokyo Denki University Press, March 10, 2004, p. 7-9

  The present invention, which has been made by paying attention to the above problem for the first time, is intended to improve control responsiveness by a simple method.

  In the present invention, an internal combustion engine or a device attached thereto is controlled by operating an operation unit, and a servo controller that repeatedly calculates a control input to be given to the operation unit, and an actual control output of the plant, An output offset unit that gives the servo controller a control output of the plant that is obtained by subtracting the average of the output data from the input / output data used when identifying the plant model for the purpose of designing the servo controller, and the servo A control device comprising an input offset unit that gives an operation unit an actual control input obtained by adding an average of the input data of the input / output data to the control input calculated by the controller. .

  That is, the average value of the output data used in the system identification stage is subtracted from the actually measured value of the control output given to the servo controller, and the average value of the input data is added to the control input calculated by the servo controller. Thereby, the divergence between the model and the actual plant can be reduced, and the control responsiveness is improved.

  According to the present invention, it is possible to easily improve control responsiveness.

The hardware resource block diagram of the EGR system in one Embodiment of this invention. Configuration explanatory drawing of the control apparatus of the embodiment. The figure which illustrates the input-output data used for the identification of the model of a plant. The block diagram of the adaptive sliding mode controller of the embodiment.

  An embodiment of the present invention will be described with reference to the drawings. What is shown in FIG. 1 is an EGR system that is one of the objects to which the present invention is applied. This EGR system attached to the internal combustion engine 2 includes measuring devices (or sensors) 11 and 12 for detecting values related to a plurality of fluid pressures or flow rates in the intake and exhaust systems 3 and 4, and target values for these values. An ECU (Electronic Control Unit) 5 serving as a control device that operates the plurality of operation units 45, 42, and 33 to set each value and follow the target value is provided.

  The internal combustion engine 2 is a diesel engine equipped with a supercharger, for example. The intake system 3 of the internal combustion engine 2 is provided with a variable turbo compressor 31 and an intercooler 32 for intake air cooling and a D throttle valve 33 for adjusting the intake air (fresh air) amount downstream thereof. Further, a flow meter 11 for measuring the intake air amount and a pressure meter 12 for measuring the intake pipe pressure are installed.

  The exhaust system 4 of the internal combustion engine 2 is provided with a turbine 41 for driving the compressor 31, and a nozzle vane 42 for increasing or decreasing the A / R ratio of the supercharger is provided at the inlet of the turbine 41. Then, an EGR passage 43 for recirculating a part of the exhaust gas discharged from the combustion chamber of the internal combustion engine 2 to the intake system 3 is formed. The EGR passage 43 is connected downstream of the throttle valve 33 in the intake system 3. The EGR passage 43 is provided with an EGR cooler 44 for cooling the exhaust, and an external EGR valve 45 that adjusts the amount of exhaust gas (EGR gas) that passes therethrough.

  In the present embodiment, a target value is set for each of the EGR rate (or EGR amount) and the intake pipe pressure, and a plurality of operation units, that is, an EGR valve 45, Control is performed to operate the variable turbo nozzle 42 and the throttle valve 33.

  The EGR valve 45, the nozzle vane 42, and the throttle valve 33 are controlled by the ECU 5 to change their opening degrees linearly. Each operation unit 45, 42, 33 is combined with an electric valve that changes the opening by increasing or decreasing the duty ratio of the drive signal, or a vacuum control valve, etc. It uses mechanical valves that change.

  The ECU 5 is a microcomputer including a processor, a RAM (Random Access Memory), a ROM (Read Only Memory) or a flash memory, an A / D conversion circuit, a D / A conversion circuit, and the like. In addition to the measuring instruments 11 and 12 for detecting the EGR rate and the intake pipe pressure, the ECU 5 detects the external atmospheric pressure (atmospheric pressure sensor) 13, engine speed, accelerator pedal depression amount, cooling water temperature Each value can be obtained by electrically connecting to various measuring devices (not shown) for detecting the intake air temperature, the outside air temperature, etc., and receiving signals output from these measuring devices.

  Incidentally, in this embodiment, the EGR rate is not directly measured. The amount of air entering the cylinder of the internal combustion engine 2 can be predicted based on the nozzle opening of the variable turbo. When the predicted value of the air amount is set as gcyl and the intake air amount measured by the flow meter 11 is set as ga, the relationship of eegr = 1−ga / gcyl is established for the estimated EGR rate eegr. The ROM or flash memory of the ECU 5 stores in advance map data that defines the relationship between the variable turbo nozzle opening and the amount of air entering the cylinder. The ECU 5 searches the map using the nozzle opening of the variable turbo as a key, obtains a predicted value of the air amount entering the cylinder, and substitutes this and the intake air amount into the above formula to calculate the EGR rate.

  In addition, the ECU 5 is electrically connected to the EGR valve 45, the variable turbo nozzle 42, the throttle valve 33, an injector for controlling fuel injection, a fuel pump, and the like (not shown), and signals for driving them. Can be entered.

  A program to be executed by the ECU 5 is stored in advance in a ROM or a flash memory, and is read into the RAM at the time of execution and is decoded by the processor. The ECU 5 controls the internal combustion engine 2 according to a program. For example, the required fuel injection amount (in other words, engine load) is determined based on various conditions such as engine speed, accelerator pedal depression amount, and coolant temperature, and a drive signal corresponding to the required injection amount is input to an injector or the like. And control the fuel injection.

  During feedback control, the ECU 5 receives signals output from various measuring instruments (not shown), and knows the engine speed, accelerator depression amount, cooling water temperature, intake air temperature, external temperature and pressure, etc., and the required injection amount To decide. Subsequently, the target EGR rate and the target intake pipe pressure are set based on at least the engine speed and the required injection amount. The ROM or flash memory of the ECU 5 stores in advance map data indicating each target value to be set according to the engine speed and the required injection amount. The ECU 5 searches the map using the engine speed and the required injection amount as keys, and obtains target values for the EGR rate and the intake pipe pressure. Further, the target value obtained by referring to the map is set as a basic value, and this is corrected according to the cooling water temperature, the intake air temperature, the external air temperature, the atmospheric pressure, etc., and final target values ref_egr and ref_epim are set.

  The ECU 5 receives the signals output from the measuring instruments 11 and 12 to obtain the current values eegr and epim of the EGR rate and the intake pipe pressure, and the current values eegr and epim of the control amounts and the target values ref_eegr and ref_epim. Of the EGR valve 45, the opening VNT of the variable turbo nozzle 42, and the opening Dth of the throttle valve 33 are calculated from the deviations, and the drive signals corresponding to the respective operation amounts EGRv, VNT, Dth are operated. Input to parts 45, 42 and 33, and manipulate the opening.

  In addition, the ECU 5 exhibits functions as the servo controller 51, the output offset unit 52, and the input offset unit 53 shown in FIGS. 2 and 4 according to the program.

  The servo controller 51 is a sliding mode controller and is responsible for the sliding mode control of the EGR rate and the intake pipe pressure.

  A supplementary explanation will be given regarding adaptive sliding mode control of the EGR rate. The state equation and the output equation are as shown in the following equation (Equation 1).

  In this embodiment, the state quantity vector X has a structure that can be directly obtained from the output vector Y. In other words, a value that can be detected via the measuring instruments 11 and 12 is directly controlled, so that the state estimation observer. This prevents the deterioration of the control performance due to the estimation error. The output matrix C is known, and is a unit matrix in this embodiment.

  In plant modeling, that is, identification of the coefficient matrix A and the input matrix B in the state equation (Equation 1), as illustrated in FIG. 3, each operation unit 45, 42, 33 has an M-sequence signal having various frequencies. , The opening degree is manipulated, the output values of the EGR rate and the intake pipe pressure are observed, and then the matrices A and B are identified from the input / output data. It is assumed that the M-sequence signals input to the operation units 45, 42, and 33 are uncorrelated with each other. This makes it possible to create a model that takes into account the mutual interference between the values.

  In addition, before performing identification, the process which removes the direct current (bias) component of input-output data is performed beforehand. That is, the time average value ave_EGRv is subtracted from the input data to the EGR valve 45, the time average value ave_VNT is subtracted from the input data to the nozzle vane 42, and the time average value ave_Dth is subtracted from the input data to the D throttle valve 33. In addition, the time average value ave_egr is subtracted from the output data of the EGR rate, and the time average value ave_epim is subtracted from the output data of the intake pipe pressure.

FIG. 4 shows a block diagram of the adaptive sliding mode control system of the present embodiment. The design procedure of the sliding mode controller 51 includes the design of the switching hyperplane and the design of a nonlinear switching input for constraining the state quantity to the switching hyperplane. If a new state quantity vector _Xe is defined by adding an integral value vector Z of the deviation between the target value vector R and the output vector Y to the original state quantity vector X in order to constitute a type 1 servo system, The equation of state of the expanded system shown in (Expression 2) is obtained.

  In consideration of the stability margin, the design method using the zero of the system is used to design the switching hyperplane. That is, the hyperplane is designed so that the equivalent control system is stable when the expansion system of the above equation (Equation 2) is generating the sliding mode. When the switching function σ is defined by Expression (Expression 3), σ = 0 and Expression (Expression 4) holds when the state is constrained to the hyperplane.

  Therefore, the linear input (equivalent control input) when the sliding mode occurs is expressed by the following equation (Equation 5).

  Substituting the linear input of the above equation (Equation 5) into the state equation (Equation 2) of the expanded system results in an equivalent control system of the following equation (Equation 6).

  Since designing a hyperplane so that this equivalent control system is stable is equivalent to designing a system ignoring the target value R, the following equation (Equation 7) holds.

  Taking the stability ε into consideration for the system of the above equation (Equation 7), obtaining the feedback gain using the optimal control theory, and making it a hyperplane, the following equation (Equation 8) is obtained.

The matrix P _s is a positive definite solution of the Riccati equation (Equation 9).

Q _s in the Riccati equation (Equation 9) is a weight matrix for control purposes, and is a non-negative definite symmetric matrix. q ₁ and q ₂ are weights for the integral Z of the deviation, and are determined by the difference in the speed of the frequency response of the control system. q ₃ and q ₄ are weights for the output Y, and are determined by the difference in the magnitude of the gain. Further, R _s in the Riccati equation (Equation 9) is a weight matrix of the control input and is a positive definite symmetric matrix. ε is a stability margin coefficient and is specified so that ε ≧ 0.

  Instead of the above equations (Equation 8) and (Equation 9), the following discrete hyperplane construction equation (Equation 10) and algebraic Riccati equation (Equation 11) may be used.

The final sliding mode method is used to design the input for constraining to the hyperplane. Here, the control input U is expressed by the following expression (Equation 12) as the sum of the linear input U _eq and a new input, that is, a nonlinear input (nonlinear control input) U _nl .

  Since it is desired to stabilize the switching function σ, the Lyapunov function for σ is selected as shown in the following equation (Equation 13) and differentiated to obtain the equation (Equation 14).

  Substituting the equation (Equation 12) into the equation (Equation 14) yields the following equation (Equation 15).

When the nonlinear input U _nl to the following expression (Expression 16), the derivative of the Lyapunov function becomes equation (17).

  Therefore, if the switching gain k is positive, the differential value of the Lyapunov function can be negative, and the sliding mode is guaranteed. The control input U at this time is expressed by the following equation (Equation 18).

  η is a smoothing coefficient introduced to reduce chattering, and η> 0.

  In the sliding mode control, it is necessary to increase the nonlinear gain in order to constrain the state quantity to the hyperplane. However, if the nonlinear gain is increased, chattering occurs in the control input. Therefore, the uncertainty of the model is divided into a definite part whose structure is unknown and whose parameter is unknown, and an uncertain part whose structure is unknown but whose upper bound is known. Uncertainty (f + Δf) is added to the state equation (Equation 1), and is expressed by the following equation (Equation 19).

  The uncertainty determination part f is compensated by identifying the unknown parameter θ. In this case, the switching gain is applied only to the uncertain part Δf of the uncertainty, and the chattering of the control input can be greatly reduced as compared with the case where the switching gain is applied to the entire uncertain component (f + Δf).

The control input U is represented by the following equation (Equation 20) obtained by adding the adaptive term U _ad to the equation (Equation 18).

Γ _{1 at the} control input (Equation 20) is an adaptive gain matrix. The function h is generally a function of the state quantity x and / or the unknown parameter θ. However, in the present embodiment, by making h a simple expression and a constant unrelated to x and θ, x is quickly converged, and θ To increase the adaptation speed. In particular, when h = 1, the estimated parameter can be identified according to the following equation (Equation 21).

In the present embodiment, the EGR ratio y ₁ and the intake pipe pressure y ₂ as a control output variable, the opening degree u ₁ of the EGR valve 45, the opening degree u ₃ of opening u ₂ and the throttle valve 33 of the variable turbo nozzle 42 Performs 3-input 2-output feedback control using control input variables. The number of state variables (system order) is 2 in the initial system (Equation 1), and 4 in the expanded system (Equation 2). By specifying the control output and the state quantity in this way, there is no need to install a measuring instrument such as a flow meter at a location where it directly contacts the exhaust gas.

However, in a three-input two-output system as in this embodiment, det (SB _e ) = 0 holds, and the matrix (SB _e ) is not regular. Therefore, the inverse matrix (SB _e ) ⁻¹ is calculated as a generalized inverse matrix. As the generalized inverse matrix, for example, a Moore-Penrose-type inverse matrix (SB _e ) ^† is used.

Thus, the output offset unit 52 subtracts the average of the output data used when identifying the model of the plant from the control output of the plant and its target value, and gives it to the controller 51. That is, a value y ₁ obtained by subtracting the time average value ave_eegr from the actual measured value eegr of the EGR rate detected via the measuring instrument 11, and the time average ave_epim from the actual measured value epim of the intake pipe pressure detected via the measuring instrument 12. The value y _{2 obtained} by subtracting the value is given to the controller 51 as the control output Y. In addition, a value r ₁ obtained by subtracting the time average value ave_eegr from the target value ref_eegr of the EGR rate set based on the engine speed, the fuel injection amount, and the like, and a time average from the target value ref_epim of the intake pipe pressure that is similarly set The value r ₂ obtained by subtracting the value ave_epim is given to the controller 51 as the target value R.

The input offset unit 53 adds the average of the input data used when identifying the plant model to the control input calculated by the controller 51, and gives this to the operation units 45, 42, and 33. That is, a value EGRv obtained by adding the time average value ave_EGRv to the control input value u ₁ calculated by the controller 51 is given to the EGR valve 45, and a value VNT obtained by adding the time average value ave_VNT to the control input value u ₂ calculated by the controller 51 is given. It is given to the nozzle vane 42. In addition, a value Dth obtained by adding the time average value ave_Dth to the control input value u ₃ calculated by the controller 51 is given to the D throttle valve 33.

According to the present embodiment, the internal combustion engine 2 or a device attached thereto is controlled by operating the plurality of operation units 45, 42, 33, and a control input u to be given to each operation unit 45, 42, 33. Input / output used to identify a plant model for the purpose of designing the servo controller 51 from the servo controller 51 that repeatedly calculates ₁ , u ₂ , u ₃ and the actual control outputs eegr, epim of the plant An output offset unit 52 that gives the average of ave_eegr and ave_epim of the output data of the data to the servo controller 51 as the control outputs y ₁ and y ₂ of the plant, and the control input u ₁ calculated by the servo controller 51, to u _2, u _3, the average of the input data of the input and output data ave_EGRv, ave_VNT, ave Since the control device including the input offset unit 53 that gives the operation units 45, 42, and 33 to the operation units 45, 42, and 33 as the actual control inputs EGR, VNT, and Dth added with Dth is configured, the difference between the model and the actual plant is reduced. be able to. Since there is no need to compensate for the deviation, the opportunity for increasing the nonlinear input term U _nl and the adaptive term U _ad is reduced. Because these terms U _nl and U _ad are less likely to fluctuate due to factors other than disturbances, the time delay required for U _nl and U _ad to converge is shortened, and control response and tracking are improved. To do.

  The present invention is not limited to the embodiment described in detail above. Control input variables in the EGR control are not limited to the EGR valve opening, the variable nozzle turbo opening, and the throttle valve opening. The control output variable is not limited to the EGR rate (or EGR amount) and the intake pipe pressure. It is also possible to construct a tertiary system with four inputs and three outputs by adding new input variables and output variables. For example, when a passage that bypasses the supercharger (compressor) is present in the intake system, a valve provided on the passage may be operated. At this time, the pressure or flow rate in the bypass passage can be included in the control output variable, and the opening of the valve on the bypass passage can be included in the control input variable.

  The method of multi-input feedback control realized by the servo controller is not limited to the sliding mode control, and a method other than the sliding mode control, for example, optimal control, H∞ control, backstepping control, etc. may be adopted.

  Other specific configurations of each part can be variously modified without departing from the spirit of the present invention.

  The present invention can be used, for example, as a control controller for controlling the EGR rate of an EGR device attached to an internal combustion engine including a supercharger.

5 ... ECU (control device)
51 ... Adaptive sliding mode controller (servo controller)
52 ... Output offset part 53 ... Input offset part

Claims (1)

Controlling an internal combustion engine or a device attached thereto by operating an operation unit,
A servo controller that repeatedly calculates a control input to be given to the operation unit;
Subtracting the average of the output data from the input / output data used to identify the plant model for the purpose of designing the servo controller from the actual control output of the plant as the control output of the plant. An output offset section to give,
A control apparatus comprising: an input offset unit that gives an operation unit an actual control input obtained by adding an average of input data of the input / output data to the control input calculated by the servo controller.

Images