JP4368928B2 - Control device for internal combustion engine - Google Patents

Description

  The present invention relates to an internal combustion engine that controls an internal combustion engine by using a model that defines a relationship between a simulated value that simulates the behavior of an internal variable of the internal combustion engine as a plant and a detected value that reflects the behavior of the internal variable. The present invention relates to a control device.

  In recent years, internal combustion engines for vehicles have been required to ensure good exhaust gas characteristics, that is, good catalyst purification rate, due to social demands. On the other hand, in an internal combustion engine having a plurality of cylinders, the air-fuel ratio of the air-fuel mixture supplied to the plurality of cylinders may vary among the cylinders due to problems such as the EGR device, the evaporated fuel processing device, and the injector. In some cases, the catalyst purification rate may be reduced. As an internal combustion engine control device for preventing this, an air-fuel ratio control device for an internal combustion engine that corrects variations in the air-fuel ratio among a plurality of cylinders by applying an observer based on an optimal control theory is known (for example, , See Patent Document 1). This air-fuel ratio control device is provided in a collective portion of an exhaust pipe of an internal combustion engine, and includes a LAF sensor that detects an air-fuel ratio in exhaust gas, and a control unit to which a detection signal (detected air-fuel ratio) of this LAF sensor is input. An intake manifold of an intake pipe of an internal combustion engine is provided for each cylinder, and includes an injector connected to a control unit.

  In this control unit, based on the air-fuel ratio detected by the LAF sensor, the fuel injection amount for each cylinder, which is the fuel injection amount of each injector, is calculated using the observer and PID control as follows, and discharged from a plurality of cylinders. Variation among the cylinders in the air-fuel ratio of the exhaust gas to be discharged, that is, the air-fuel ratio of the air-fuel mixture supplied to the plurality of cylinders is corrected.

  That is, the control unit calculates the basic injection amount according to the operating state of the internal combustion engine, and calculates the output injection amount by multiplying the basic injection amount by various correction coefficients. Next, as will be described later, the estimated air-fuel ratio for each cylinder is estimated by an observer, and the cylinder-by-cylinder feedback correction coefficient is calculated based on the estimated air-fuel ratio for each cylinder by PID control. By multiplying the output injection amount, the fuel injection amount for each cylinder is calculated.

  In the observer, the estimated air-fuel ratio for each cylinder is estimated based on the optimal control theory. Specifically, the estimated air-fuel ratio for each cylinder is calculated by using a discrete time system model that represents the relationship between the fuel-air ratio for each cylinder and the fuel-air ratio of the collecting portion (LAF sensor mounting portion). Further, in the PID control, a value obtained by dividing the collective part air-fuel ratio, that is, the detected air-fuel ratio by the average value of the previous values of the feedback correction coefficients is set as a target value, and this target value and the estimated air-fuel ratio for each cylinder estimated by the observer are The feedback correction coefficient for each cylinder is calculated so that the deviation of.

  Further, as another air-fuel ratio control device, fuel injection is performed based on an estimated intake air amount estimated for each cylinder for an intake air amount in a plurality of cylinders and an estimated air-fuel ratio estimated for each cylinder by an observer similar to the above. One that calculates the amount for each cylinder is known (for example, see Patent Document 2).

  Specifically, in this air-fuel ratio control device, the target intake fuel amount is calculated by searching a map according to the engine speed and the intake pipe internal pressure. Further, by applying a hydrodynamic model to the intake system of the internal combustion engine, the estimated intake air amount is calculated for each cylinder, and the estimated air-fuel ratio is calculated for each cylinder by the above-described observer. Further, by dividing the estimated intake air amount by the estimated air-fuel ratio, the estimated intake fuel amount is calculated for each cylinder, and the adaptive controller determines the final value so that the estimated intake fuel amount matches the target intake fuel amount. A fuel injection amount is calculated.

Japanese Patent No. 3296472 (pages 19 to 23, FIGS. 35 and 36) JP-A-6-74076 (pages 3 to 12, FIGS. 1, 31)

  In recent years, an internal combustion engine has been required to have a high output and a high torque in addition to the above-described requirement for ensuring a good catalyst purification rate. To achieve this, the exhaust system layout has a complicated shape (for example, an exhaust manifold). The exhaust passage has a shape of 4 → 2 → 1 and gathers while gradually decreasing the number of the exhaust passages, and a method of reducing exhaust resistance and exhaust interference is known. However, when the former air-fuel ratio control device is applied to an internal combustion engine with such an exhaust system layout, the conventional optimal control theory does not hold an observer, so the variation in air-fuel ratio between cylinders is corrected appropriately. This may not be possible and may lead to a decrease in the catalyst purification rate. This is because, in the conventional optimal control theory, the assumption model and the optimal control theory itself do not take into account the modeling error and the dynamic characteristics of the model, so the observer's stability margin is small and the robustness is low. This is due to insufficient stability against changes in the contribution of exhaust gas in each cylinder at the detected air-fuel ratio of the LAF sensor, variations in response of the LAF sensor, and aging of the LAF sensor.

  Also, the latter air-fuel ratio control device uses the same observer as the former, so the observer may not be established for the reasons described above. In this case, the fuel injection amount is calculated appropriately for each cylinder. If it becomes impossible, the catalyst purification rate may be reduced. Further, in a multi-cylinder internal combustion engine, the intake air amount generally varies among cylinders, whereas in the latter air-fuel ratio control device, the correction of the variation in intake air amount is taken into consideration. However, the intake air amount is merely estimated for each cylinder by applying the hydrodynamic model. Therefore, variation in intake air amount between cylinders cannot be corrected appropriately, and variation in air-fuel ratio between cylinders may result in further reduction in the catalyst purification rate.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a control device for an internal combustion engine that has a large stability margin and can realize highly robust control.

In order to achieve the above object, a control device 1 for an internal combustion engine 3 according to claim 1 is provided in an exhaust passage (collecting portion 7j) of the internal combustion engine 3 and represents an air-fuel ratio of an exhaust gas in the exhaust passage. An air-fuel ratio detecting means (LAF sensor 14 ) for detecting a detected value (detected air-fuel ratio KACT) and an amount of fuel (basic fuel injection amount TIBS) supplied to the cylinders (1st to 4th cylinders # 1 to 4) of the internal combustion engine 3 ) And a predetermined simulation that presupposes the behavior corresponding to the crank angle of the internal combustion engine 3 of the air-fuel ratio of the exhaust gas discharged from the cylinder. A simulation value generating means (ECU2, adaptive observer 31 ) for generating the value KACT_OS based on the crank angle, and an estimated value KACT_EST of the air-fuel ratio detection value are models that define the relationship between the estimated value and the simulated value [formula (2 )] Based estimating means (ECU 2, the adaptive observer 31) of estimating a, as estimated estimated value matches the air-fuel ratio detected value detected, in response to the detected air-fuel ratio detection value and the generated simulated value , Identification means (ECU 2, adaptive observer 31, step 8 ) for identifying the model parameters ( air-fuel ratio variation coefficient Φ ) of the model on-board, and a first parameter for correcting the fuel amount using the identified model parameters The first correction value calculating means (ECU 2, air-fuel ratio variation correction coefficient calculating unit 32, step 9) for calculating the correction value (air-fuel ratio variation correction coefficient KOBSV) and the calculated first correction value are used to calculate the fuel amount. First fuel amount correcting means for correcting (ECU2, first air-fuel ratio controller 30, step 11) .
According to the control device for an internal combustion engine, an air-fuel ratio detection value representing an air-fuel ratio of the exhaust gas in the exhaust passage is detected, and an estimated value of the air-fuel ratio detection value is calculated from this and the exhaust gas discharged from the cylinder. The air-fuel ratio is estimated by using a model that defines the relationship between the behavior corresponding to the crank angle of the internal combustion engine and a predetermined simulated value that is preliminarily assumed, and the estimated value matches the air-fuel ratio detection value. The model parameters of the model are identified onboard according to the detected fuel ratio and the simulated value. Thus, since the model parameter is identified on-board so that the estimated value matches the air-fuel ratio detection value, the actual behavior of the air-fuel ratio in the exhaust gas exhausted from the cylinder is determined appropriately. And it can identify as a value reflected in real time. Therefore, by calculating the first correction value using the identification value of such a model parameter and correcting the fuel amount using the first correction value, the fuel amount can be calculated with high accuracy, and the empty value can be calculated. The control accuracy of the fuel ratio can be improved.

According to a second aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the first aspect, the cylinder includes a plurality of cylinders (1st to 4th cylinders # 1 to 4), and the exhaust passage includes a plurality of cylinders. And a plurality of exhaust passages (exhaust pipe portions 7c to 7f) and one exhaust passage (collection portion 7j) in which a plurality of exhaust passages are gathered. Provided in the part 7j), the fuel amount determining means determines the fuel amount as the fuel amount supplied to each of the plurality of cylinders, and the simulated value generating means is discharged from the plurality of cylinders as simulated values. A plurality of simulated values KACT_OS pre-assuming the behavior of the air-fuel ratio of the exhaust gas corresponding to the crank angle of the internal combustion engine _i The model is composed of a model that defines the relationship between the estimated value of the air-fuel ratio detection value and a plurality of simulation values, and the identification means is used as a model parameter according to the air-fuel ratio detection value and the plurality of simulation values. , Multiple model parameters (air-fuel ratio variation coefficient Φ _i ) Is identified on-board, and a second amount for correcting the amount of fuel supplied to each cylinder so that the detected air-fuel ratio (detected air-fuel ratio KACT) converges to a predetermined target value (target air-fuel ratio KCMD). Second correction value calculation means (ECU2, second air-fuel ratio controller 40) for calculating a correction value (feedback correction coefficient KSTR), and the amount of fuel supplied to each cylinder is corrected according to the calculated second correction value. And a second fuel amount correcting means (ECU2, second air-fuel ratio controller 40) that further includes a plurality of identified model parameters (air-fuel ratio variation coefficient Φ). _i ) Is calculated for each cylinder so as to converge to the average value (moving average value Φave) of the plurality of identified model parameters (step 9).
According to the control device for the internal combustion engine, the plurality of model parameters are identified on-board so that the estimated value of the air-fuel ratio detection value matches the detected air-fuel ratio detection value. It can be identified as a value reflecting the actual behavior of the air-fuel ratio in the exhaust gas discharged from a plurality of cylinders, that is, the variation in the air-fuel ratio among the cylinders. Therefore, by correcting the fuel amount for each cylinder according to the first correction value calculated according to the identification values of the plurality of model parameters, it is possible to appropriately correct the variation in the air-fuel ratio between the cylinders. it can. Moreover, since a plurality of model parameters are identified on-board, the first correction value can be calculated based on the model parameters identified in real time. As a result, the dynamic characteristics of the controlled object change due to changes in the contribution of each cylinder to the air-fuel ratio detection value due to fuel adhesion in each cylinder, response variations in the air-fuel ratio detection means, and changes over time in the air-fuel ratio detection means, etc. Even when this is done, the fuel amount can be corrected so as to correct (absorb) the variation in the air-fuel ratio among the plurality of cylinders while reflecting the change in the dynamic characteristics of the controlled object in the model, unlike the conventional case. As a result, even when this control device is applied to an internal combustion engine having a complicated exhaust layout, variations in air-fuel ratio among cylinders can be corrected appropriately and quickly, and the air-fuel ratio can be controlled with high accuracy. That is, air-fuel ratio control with a large stability margin and high robustness can be realized. In addition, a second correction value for correcting the amount of fuel supplied to each cylinder is calculated by the second correction value calculation means so that the detected air-fuel ratio detection value converges to a predetermined target value. The fuel amount supplied to each cylinder is corrected by the second fuel amount correcting means in accordance with the second correction value, and the identified plurality of model parameters are identified by the first correction value calculating means. The first correction value is calculated for each cylinder so as to converge to the average value of the plurality of model parameters. In this way, the first correction value is calculated so that the identification values of the plurality of model parameters converge to the average value of these identification values, so that variations in the air-fuel ratio among the plurality of cylinders can be corrected. Thus, it is possible to avoid the control processing for converging the air-fuel ratio detection value to a predetermined target value and the control processing for correcting the variation in the air-fuel ratio between the cylinders from interfering with each other, The stability of air-fuel ratio control can be ensured.

According to a third aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the second aspect, the learning correction value of the first correction value is obtained by using the sequential statistical algorithm [Equations (15) to (21)]. KOBSV_LS _i Is further provided with a learning correction value calculating means (ECU 2, learning correction value calculating unit 33, step 10) for each cylinder, and the first fuel amount correcting means further calculates the fuel amount according to the calculated learning correction value. Correction is performed for each cylinder (step 11).
Here, as an identification calculation algorithm, the least square method is generally used. However, in the identification calculation by the least square method, a predetermined number of various data for calculation are collected, and then a batch calculation is performed based on the collected data. Therefore, at the start of the air-fuel ratio control, the identification of the model parameter is not executed until the data collection is completed, and during this time, the first correction value cannot be calculated according to the identification value of the model parameter. As a result, the controllability of the air-fuel ratio control may be reduced. On the other hand, according to the control device for the internal combustion engine, the learning correction value of the first correction value is calculated by the sequential statistical algorithm, so that it is calculated for each control cycle even at the start of the air-fuel ratio control. The first correction value can be corrected by the learning correction value. Therefore, for example, by setting the initial value of the first correction value in advance, the learning correction value calculated for each control cycle can be obtained at the start of air-fuel ratio control and until the identification of the model parameter is started. Thus, the first correction value can always be corrected, and the controllability at the start of air-fuel ratio control can be improved.

According to a fourth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the third aspect, an operating state parameter detecting means (ECU2) for detecting an operating state parameter (exhaust gas volume ESV) indicating the operating state of the internal combustion engine 3 is provided. , The intake pipe absolute pressure sensor 11 and the crank angle sensor 13), and the learning correction value calculation means includes a learning correction value KOBSV_LS. _i Is calculated by the regression equation [Equation (22)] with the learning correction value as a dependent variable and the detected operating state parameter as an independent variable, and the regression coefficient AOBSV_LS of the regression equation. _i And the constant term BOBSV_LS _i Is calculated by a sequential statistical algorithm [Expressions (15) to (21)].
According to the control apparatus for an internal combustion engine, the learning correction value of the first correction value is calculated by a regression equation using the detected correction value as a dependent variable and the detected operating state parameter as an independent variable, and the regression coefficient of the regression equation And the constant term are calculated by a sequential statistical algorithm, so even if the internal combustion engine is in a rapidly changing operating state such as a transient operating state, the air-fuel ratio suddenly changes due to its influence, and its estimation is difficult. The learning correction value can be calculated as a value that appropriately reflects the actual state of the air-fuel ratio of each cylinder. As a result, the controllability of the air-fuel ratio control can be further improved.

According to a fifth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the second to fourth aspects, the first correction value calculating means includes a first correction value (air-fuel ratio variation correction coefficient KOBSV). _i The correction value component included in the model parameter (the air-fuel ratio variation coefficient Φ _i ) And a predetermined target value (moving average value Φave).
According to the control device for the internal combustion engine, the correction value component included in the first correction value is calculated by the first correction value calculation unit based on the deviation between the model parameter and the predetermined target value. The fuel amount can be corrected for each cylinder so that the air-fuel ratio converges to a predetermined target value, whereby the air-fuel ratio is controlled for each cylinder so as to converge to the predetermined value without causing a steady-state deviation. be able to.

According to a sixth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the fifth aspect, the first correction value calculating means includes a first correction value (air-fuel ratio variation correction coefficient KOBSV). _i ) Is calculated based on the identified model parameter (air-fuel ratio variation coefficient Φ).
According to the control device for an internal combustion engine, in addition to the correction value component determined based on the deviation between the model parameter and the predetermined target value, the other correction determined based on the model parameter Since the value component is further included, when the fuel amount is corrected for each cylinder so that the model parameter converges to the predetermined target value, the air-fuel ratio remains stable without causing overshoot or vibrational behavior. Control can be performed for each cylinder so as to converge to a predetermined value.

According to a seventh aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the second to fifth aspects, the first correction value calculating means includes a response designating control algorithm [Expressions (47) to (49). ], According to the model parameter (air-fuel ratio variation coefficient Φ), the first correction value (air-fuel ratio variation correction coefficient KOBSV _i ) Is calculated.
According to the control device for an internal combustion engine, the first correction value is determined according to the model parameter by the response designation control algorithm. For example, the fuel amount is adjusted so that the model parameter converges to a predetermined target value. The correction can be performed for each cylinder, whereby the air-fuel ratio can be controlled for each cylinder so as to converge to a predetermined value in a stable state without causing overshoot or vibrational behavior.

According to an eighth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the second to seventh aspects, the identifying means uses the model parameter (empty) according to the fixed gain method [Equations (50) to (57)]. It is characterized by identifying a fuel ratio variation coefficient Φ).
According to the control device for the internal combustion engine, the model parameter is identified by the fixed gain method, so that the calculation load of the identification unit can be reduced. As a result, the calculation time of the first correction value can be shortened. Therefore, even when the change speed of the air-fuel ratio of each cylinder such as a transient operation state is fast, the value that appropriately reflects the behavior of the air-fuel ratio is One correction value can be calculated quickly and appropriately for each cylinder. In addition, when using a method for identifying a model parameter by adding a predetermined correction component to the reference value as a fixed gain method, the identification value of the model parameter can be constrained in the vicinity of the reference value. It can be avoided that the actual state of the air-fuel ratio is inappropriately reflected in the identification value of the model parameter due to an increase in the change rate of the air-fuel ratio, and the stability of the air-fuel ratio control can be further improved. .

According to a ninth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the fourth aspect of the present invention, the identification means is an operating state parameter (exhaust gas volume ESV) in which the model parameter reference value (reference value vector φbase) is detected. The model parameter (air-fuel ratio variation coefficient Φ) is identified by adding a predetermined correction component (correction term vector dφ) to the calculated model parameter reference value.
According to the control apparatus for an internal combustion engine, the model parameter is identified by adding a predetermined correction component to the model parameter reference value calculated according to the operating state parameter. Since the parameter identification value can be constrained, the first correction value can be quickly and appropriately set as a value that appropriately reflects the behavior of the air-fuel ratio even when the change speed of the air-fuel ratio is fast due to the influence of fluctuations in the operating state of the internal combustion engine. Therefore, the stability of the control can be further improved.

According to a tenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the second to ninth aspects, the detected air-fuel ratio detected value (detected air-fuel ratio KACT) is set to a predetermined delay time (dead time). Delay means (ECU2, adaptive observer 31) that is delayed by time d), and the identification means is configured to determine the model parameter (air-fuel ratio variation coefficient) according to the delayed air-fuel ratio detection value and the plurality of generated simulation values. Φ) is identified.
In general, in an internal combustion engine, a predetermined dead time exists between the time when the air-fuel mixture supplied to each cylinder reaches the collecting portion of the exhaust passage or the downstream side thereof as exhaust gas after combustion. On the other hand, according to the control device for the internal combustion engine, the model parameter is identified in accordance with the predetermined delay time, the delayed air-fuel ratio detection value, and the plurality of simulation values. While reflecting, it is possible to identify the model parameters with high accuracy, and to further improve the stability of the control.

The control device 1 for the internal combustion engine 3 according to claim 11 is provided in the intake passage of the internal combustion engine, and detects an intake air amount that indicates an intake air amount (intake air amount GAIR, intake pipe absolute pressure PBA). Means (air flow sensor 9, intake pipe absolute pressure sensor 11) and fuel amount determining means (ECU2, basic fuel injection amount calculation unit 20) for determining the amount of fuel (basic fuel injection amount TIBS) supplied to the cylinder of the internal combustion engine And a simulation value generation means (ECU 2, adaptive observer 61) that generates a predetermined simulation value GAIR_OS that presupposes a behavior corresponding to the crank angle of the internal combustion engine 3 of the intake air amount sucked into the cylinder based on the crank angle. ) And estimated means (ECU2) for estimating the estimated value GAIR_EST of the detected intake air amount based on a model [Equation (59)] that defines the relationship between the estimated value and a plurality of simulated values In accordance with the detected intake air amount detected value and the generated simulated value so that the adaptive observer 61) and the estimated estimated value GAIR_EST coincide with the detected intake air amount detected value (intake air amount GAIR) The third correction value for correcting the fuel amount using the identifying means (ECU 2, adaptive observer 61, step 111) for identifying the model parameter (intake air amount variation coefficient Ψ) of the on-board and the identified model parameter Using the third correction value calculation means (ECU 2, intake amount variation correction coefficient calculation unit 62, step 112) for calculating (intake amount variation correction coefficient KICYL) and the calculated third correction value, the fuel amount is corrected. And third fuel amount correction means (ECU2, third air-fuel ratio controller 60, step 114).
According to the control device for the internal combustion engine, an intake air amount detection value representing the intake air amount is detected, and an estimated value of the intake air amount detection value is calculated as the intake air amount sucked into the cylinder. It is estimated by using a model that defines the relationship between the behavior corresponding to the corner and a predetermined simulated value that is assumed in advance, and in accordance with the detected intake air amount and the simulated value so that this estimated value matches the detected intake air amount. Thus, the model parameters of the model are identified on-board. In this way, since the model parameter is identified on-board so that the estimated value matches the intake air amount detection value, the actual behavior of the intake air amount sucked into the cylinder can be determined appropriately and in real time. It can be identified as a reflected value. Accordingly, the third correction value is calculated using the identification value of such a model parameter, and the fuel amount is corrected using the third correction value, so that the fuel amount can be accurately calculated. The control accuracy of the fuel ratio can be improved.

According to a twelfth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the eleventh aspect, the cylinder is composed of a plurality of cylinders (1st to 4th cylinders # 1 to 4), and the intake passage is one intake air. The passage (main pipe portion 4a, collecting portion 4c) and a plurality of intake passages (branch portions 4d) branching from one intake passage and extending to a plurality of cylinders, respectively. The fuel amount determining means determines the fuel amount as a fuel amount supplied to each of the plurality of cylinders, and the simulated value generating means is an intake air sucked into the plurality of cylinders as a simulated value. A plurality of simulated values GAIR_OS that presuppose the behavior corresponding to the crank angle of the internal combustion engine _i The model generates an estimated value GAIR_EST of the detected intake air amount and a plurality of simulated values GAIR_OS. _i The identification means includes a plurality of model parameters (intake air amount variation coefficient Ψ) as model parameters in accordance with the intake air amount detection value and the plurality of simulated values. _i ) Are identified on-board, and a plurality of exhaust passages (exhaust pipe portions 7c to 7f) respectively extending from a plurality of cylinders (1st to 4th cylinders # 1 to 4) are gathered into one exhaust passage (collecting portion 7j). An air-fuel ratio detection means (LAF sensor 14) for detecting an air-fuel ratio detection value (detected air-fuel ratio KACT) representing an air-fuel ratio of the exhaust gas in one exhaust passage, and a detected air-fuel ratio detection value (detection) Fourth correction for calculating a fourth correction value (feedback correction coefficient KSTR) for correcting the amount of fuel supplied to each cylinder so that the air-fuel ratio KACT) converges to a predetermined target value (target air-fuel ratio KCMD). Value calculating means (ECU2, second air-fuel ratio controller 40, step 107) and fourth fuel amount correcting means (ECU2, second fuel ratio correcting means for correcting the amount of fuel supplied to each cylinder in accordance with the calculated fourth correction value. 2 air-fuel ratio control And roller 40), further comprising a third correction value calculation means, a plurality of model parameters identified (intake air amount variation coefficient Ψ _i ) Is calculated for each cylinder so as to converge to the average value (moving average value Ψave) of the plurality of identified model parameters (step 111).
According to the control device for the internal combustion engine, the plurality of model parameters are identified on-board so that the estimated value of the intake air amount detection value coincides with the detected intake air amount detection value. It can be identified as a value reflecting the actual behavior of the intake air amount sucked into a plurality of cylinders, that is, the variation in the intake air amount among the cylinders. Therefore, by correcting the fuel amount for each cylinder in accordance with the third correction value calculated in accordance with the identification values of the plurality of model parameters, it is possible to appropriately correct the variation in the intake air amount between the cylinders. Can do. Moreover, since a plurality of model parameters are identified on-board, the third correction value can be calculated based on the model parameters identified in real time. As a result, even when the dynamic characteristics of the controlled object changes due to response variations and aging in the intake air amount detection means, unlike the conventional case, while reflecting the change in the dynamic characteristics of the controlled object in the model, The fuel amount can be corrected so as to correct (absorb) the variation in the intake air amount. As a result, even when this control device is applied to an internal combustion engine having a complicated exhaust layout, variation in intake air amount between cylinders can be corrected appropriately and quickly, and the air-fuel ratio can be accurately controlled. That is, air-fuel ratio control with a large stability margin and high robustness can be realized. In addition, a fourth correction value for correcting the amount of fuel supplied to each cylinder is calculated by the fourth correction value calculation means so that the air-fuel ratio detection value converges to a predetermined target value. The amount of fuel supplied to each cylinder is corrected by the four fuel amount correcting means in accordance with the fourth correction value, and the plurality of identified model parameters are identified by the third correction value calculating means. The third correction value is calculated for each cylinder so as to converge to the average value of the model parameters. In this way, the third correction value is calculated so that the identification values of the plurality of model parameters converge to the average value of these identification values, so that variations in intake air amounts among the plurality of cylinders can be corrected. Thus, it is possible to prevent the control processing for converging the air-fuel ratio detection value to a predetermined target value and the control processing for correcting the variation in the intake air amount between the cylinders from interfering with each other. And the stability of the air-fuel ratio control can be ensured.

According to a thirteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the twelfth aspect, the learning correction value of the third correction value is obtained by using the sequential statistical algorithm [Expressions (60) to (66)]. KICYL_LS _i Learning correction value calculation means (ECU 2, learning correction value calculation unit 63, step 113) for each cylinder, and the third fuel amount correction means further determines the fuel amount according to the calculated learning correction value. The correction is performed for each cylinder (step 114).
As described above, in the identification calculation based on the general least square method as the identification calculation algorithm, the identification of the model parameter is not executed until the data collection ends at the start of the control. Since the third correction value cannot be calculated according to the value, the controllability of the air-fuel ratio control may be reduced. On the other hand, according to the control device for the internal combustion engine, the learning correction value of the third correction value is calculated by the sequential statistical algorithm, and thus is calculated for each control cycle even at the start of the air-fuel ratio control. The third correction value can be corrected by the learning correction value. Therefore, for example, by setting the initial value of the third correction value in advance, the learning correction value calculated for each control cycle can be obtained at the start of air-fuel ratio control and until the identification of the model parameter is started. Thus, the third correction value can always be corrected, and the controllability at the start of the air-fuel ratio control can be improved.

According to a fourteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the thirteenth aspect, an operation state parameter detecting means (ECU2) for detecting an operation state parameter (exhaust gas volume ESV) representing the operation state of the internal combustion engine 3 , The intake pipe absolute pressure sensor 11 and the crank angle sensor 13), and the learning correction value calculation means includes a learning correction value KICYL_LS. _i Is calculated by the regression equation [Equation (78)] with the learning correction value as the dependent variable and the detected operating state parameter as the independent variable, and the regression coefficient AICYL_LS of the regression equation. _i And the constant term BICYL_LS _i Is calculated by a sequential statistical algorithm [Expressions (71) to (77)].
According to this control device for an internal combustion engine, the learning correction value of the third correction value is calculated by a regression equation having this as a dependent variable and the detected operating state parameter as an independent variable, and the regression coefficient of the regression equation When the internal combustion engine is in an operating state that changes rapidly, such as a transient operating state, the intake air amount changes suddenly and is difficult to estimate. However, the learning correction value can be calculated as a value that appropriately reflects the actual state of the intake air amount of each cylinder. As a result, the controllability of the air-fuel ratio control can be further improved.

According to a fifteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the twelfth to fourteenth aspects, the third correction value calculation means includes a third correction value (intake amount variation correction coefficient KICYL). _i ) Is included in the model parameter (intake air amount variation coefficient Ψ _i ) And a predetermined target value (moving average value Ψave).
According to the control device for the internal combustion engine, the correction value component included in the third correction value is calculated based on the deviation between the model parameter and the predetermined target value by the third correction value calculation means. The amount of fuel can be corrected for each cylinder so that the value converges to a predetermined target value, thereby controlling the amount of intake air for each cylinder so that the amount of intake air converges to a predetermined value without causing a steady deviation. can do.

According to a sixteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the fifteenth aspect, the third correction value calculating means includes a third correction value (intake amount variation correction coefficient KICYL). _i ) Other than the correction value component included in the model parameter (intake air amount variation coefficient Ψ _i ).
According to the control device for an internal combustion engine, the third correction value is determined based on the model parameter in addition to the correction value component determined based on the deviation between the model parameter and the predetermined target value. Since the value component is further included, when the fuel amount is corrected for each cylinder so that the model parameter converges to the predetermined target value, the intake air amount is stable without causing overshoot or vibrational behavior. Can be controlled for each cylinder so as to converge to a predetermined value.

The invention according to claim 17 is the control device 1 for the internal combustion engine 3 according to any one of claims 12 to 15, wherein the third correction value calculating means is a response designating control algorithm [Expressions (89) to (91). ], The model parameter (intake air amount variation coefficient Ψ _i ) In accordance with the third correction value (intake amount variation correction coefficient KICYL _i ) Is calculated.
According to the control device for the internal combustion engine, the third correction value is determined according to the model parameter by the response designation control algorithm. For example, the fuel amount is adjusted so that the model parameter converges to a predetermined target value. Correction can be made for each cylinder, whereby the intake air amount can be controlled for each cylinder so as to converge to a predetermined value in a stable state without causing overshoot or vibrational behavior.

According to an eighteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any one of the twelfth to seventeenth aspects, the identification means uses model parameters (intake air) by a fixed gain method [Equations (92) to (99)]. It is characterized by identifying a quantity variation coefficient (Ψ).
According to the control device for the internal combustion engine, the model parameter is identified by the fixed gain method, so that the calculation load of the identification unit can be reduced. As a result, the calculation time of the third correction value can be shortened, so that the behavior of the intake air amount is appropriately reflected even when the change rate of the intake air amount of each cylinder in a transient operation state is fast. The third correction value can be calculated quickly and appropriately for each cylinder. In addition, when using a method for identifying a model parameter by adding a predetermined correction component to the reference value as a fixed gain method, the identification value of the model parameter can be constrained in the vicinity of the reference value. To prevent the actual state of the intake air amount from being improperly reflected in the model parameter identification value due to an increase in the change rate of the intake air amount, and to further improve the stability of the air-fuel ratio control Can do.

According to a nineteenth aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to the fourteenth aspect, the identification means is an operating state parameter (exhaust gas volume ESV) in which the model parameter reference value (reference value vector ψbase) is detected. The model parameter (intake air amount variation coefficient ψ) is identified by adding a predetermined correction component (correction term vector dψ) to the calculated model parameter reference value.
According to the control apparatus for an internal combustion engine, the model parameter is identified by adding a predetermined correction component to the model parameter reference value calculated according to the operating state parameter. Since the identification value of the parameter can be constrained, the third correction value can be quickly set as a value that appropriately reflects the behavior of the intake air amount even when the change rate of the intake air amount is fast due to the influence of fluctuations in the operating state of the internal combustion engine. And it can calculate appropriately for every cylinder and can further improve the stability of control.

The invention according to claim 20 is the control device 1 for the internal combustion engine 3 according to any one of claims 12 to 19, wherein the plurality of generated simulated values GAIR_OS _i Is further provided with delay means (ECU 2, adaptive observer 61) for delaying by a predetermined delay time (dead time d ′), and the identification means is responsive to the detected intake air amount detection value and the plurality of delayed delay values. Thus, the model parameter (intake air amount variation coefficient Ψ) is identified.
Generally, in an internal combustion engine, a predetermined dead time exists until the air taken into the intake passage reaches each cylinder through the branched intake passage. On the other hand, according to the control apparatus for an internal combustion engine, the model parameter is identified in accordance with the intake air amount detection value and a plurality of simulated values delayed by a predetermined delay time, so that the dead time is reflected. Thus, the model parameters can be identified with high accuracy, and the control stability can be further improved.

According to a twenty-first aspect of the present invention, in the control device 1 for the internal combustion engine 3 according to any of the twelfth to nineteenth aspects, a predetermined filtering process is performed on the detected intake air amount detection value (intake air amount GAIR). And a filter means (filter 61j) for generating a filter value (filter value GAIR_F) of the intake air amount detection value, and the identifying means is configured to generate the filter value GAIR_F of the generated intake air amount detection value and the plurality of generated simulated values. Accordingly, the model parameter (intake air amount variation coefficient Ψ) is identified.

In general, in this type of control device, when the internal combustion engine is in an operating state where the fluctuation range of the absolute value of the intake air amount detection value is large , such as in a transient operating state, the identification process by the identification means cannot follow the fluctuation. As a result, the identification delay of the model parameter occurs, and the identification accuracy may be lowered. On the other hand, according to the control device for the internal combustion engine, the model parameter is identified by the identification unit according to the filter value and the simulated value of the intake amount detection value subjected to the predetermined filtering process. By appropriately setting the filter characteristics of the filtering process, even when the fluctuation range of the absolute value of the intake air amount detection value is large, the information necessary for model parameter identification, that is , the intake behavior (variation, etc.) information of each cylinder The filter value of the intake air amount detection value can be generated as a value in which the fluctuation range of the intake air amount detection value is suppressed while ensuring. Therefore, by identifying model parameters according to such filter values and simulated values, model parameter identification delays can be suppressed, the identification accuracy can be increased, and the stability and responsiveness of air-fuel ratio control can be improved. This can be further improved.

  Hereinafter, a control apparatus for an internal combustion engine according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a schematic configuration of a control device 1 according to the first embodiment and an internal combustion engine 3 to which the control device 1 is applied. As shown in the figure, the control device 1 includes an ECU 2, and the ECU 2 supplies fuel to the internal combustion engine (hereinafter referred to as “engine”) 3 according to the operating state of the internal combustion engine (hereinafter referred to as “engine”) 3 as will be described later. By controlling the amount, the air-fuel ratio of the air-fuel mixture is controlled.

  The engine 3 is an in-line four-cylinder gasoline engine mounted on a vehicle (not shown), and includes first to fourth cylinders # 1 to # 4. The intake pipe 4 of the engine 3 includes a main pipe portion 4a constituting a single intake passage and an intake manifold 4b connected to the main pipe portion 4a. A throttle valve 5 is provided in the middle of the main pipe portion 4a. It has been.

An air flow sensor 9 and an intake pipe absolute pressure sensor 11 are provided upstream and downstream of the throttle valve 5 of the main pipe portion 4a, respectively. The air flow sensor 9 ( intake amount detection means) detects an intake air amount GAIR ( intake amount detection value) taken into the engine 3 via the intake pipe 4 and outputs a detection signal to the ECU 2.

The intake pipe absolute pressure sensor 11 ( operating state parameter detection means, intake air quantity detection means) is constituted by, for example, a semiconductor pressure sensor, and detects the intake pipe absolute pressure PBA ( intake air quantity detection value) in the intake pipe 4. The detection signal is output to the ECU 2.

  Further, a throttle valve opening sensor 10 composed of, for example, a potentiometer is provided in the vicinity of the throttle valve 5 of the main pipe portion 4a. The throttle valve opening sensor 10 detects the opening TH of the throttle valve 5 (hereinafter referred to as “throttle valve opening”) TH and outputs a detection signal to the ECU 2.

  The intake manifold 4b of the intake pipe 4 is composed of a collective part 4c connected to the main pipe part 4a and four branch parts 4d branched from this and connected to the four cylinders # 1 to # 4, respectively. Yes. An injector 6 is attached to each branch portion 4d upstream of an intake port (not shown) of each cylinder. During the operation of the engine 3, each injector 6 is controlled by the drive signal from the ECU 2 for the fuel injection amount and the injection timing that are the valve opening time.

  On the other hand, a water temperature sensor 12 composed of, for example, a thermistor is attached to the main body of the engine 3. The water temperature sensor 12 detects the engine water temperature TW, which is the temperature of the cooling water circulating in the cylinder block of the engine 3, and outputs a detection signal to the ECU 2.

The crankshaft (not shown) of the engine 3 is provided with a crank angle sensor 13 ( operating state parameter detecting means ). The crank angle sensor 13 outputs a CRK signal and a TDC signal, both of which are pulse signals, to the ECU 2 as the crankshaft rotates.

  One pulse of the CRK signal is output every predetermined crank angle (for example, 30 °). The ECU 2 calculates the engine speed (hereinafter referred to as “engine speed”) NE of the engine 3 in accordance with the CRK signal. The TDC signal is a signal indicating that the piston (not shown) of each cylinder is at a predetermined crank angle position slightly before the TDC position of the intake stroke, and one pulse is output for each predetermined crank angle. Is done.

  On the other hand, the exhaust pipe 7 is provided with an exhaust manifold 7b connected to the four cylinders # 1 to # 4 and a main pipe portion 7a connected to the collective portion 7j. The exhaust manifold 7b includes four cylinders. Four exhaust pipe portions 7c to 7f extending from # 1 to # 4 respectively have a shape that gathers in order of 4 → 2 → 1. That is, in the exhaust manifold 7b, a collecting portion 7g in which two exhaust pipe portions 7c and 7f extending from the first and fourth cylinders # 1 and # 4 are gathered together, and the second and third cylinders # 2 and # 3. A collective portion 7h in which exhaust pipe portions 7d and 7e extending from one are gathered together and a collective portion 7j (one exhaust passage) in which these collective portions 7g and 7h are gathered into one are integrally configured. With such a shape, the exhaust resistance of the exhaust manifold 7b is set to a value smaller than that of a normal exhaust manifold in which the four exhaust pipe portions are gathered in the order of 4 → 1. Compared to those having an exhaust manifold, it is configured to generate higher power and torque.

The main pipe portion 7a of the exhaust pipe 7 is provided with first and second catalytic devices 8a and 8b in order from the upstream side with a gap therebetween. Each catalyst device 8 is a combination of a NOx catalyst and a three-way catalyst, and this NOx catalyst is not shown, but an iridium catalyst (a silicon carbide whisker powder carrying iridium and a fired body of silica) is formed on a honeycomb structure. The surface of the material is coated, and a perovskite double oxide (LaCoO ₃ powder and silica fired body) is further coated thereon. The catalyst device 8 purifies NOx in the exhaust gas during the lean burn operation by the oxidation / reduction action of the NOx catalyst, and CO 2 in the exhaust gas during the operation other than the lean burn operation by the oxidation / reduction action of the three-way catalyst. , Purifies HC and NOx.

  An oxygen concentration sensor (hereinafter referred to as “O2 sensor”) 16 is attached to the main pipe portion 7a between the first and second catalytic devices 8a and 8b. The O2 sensor 15 is composed of zirconia and a platinum electrode, and sends an output Vout based on the oxygen concentration in the exhaust gas on the downstream side of the first catalyst device 8a to the ECU 2. The output Vout of the O2 sensor 15 becomes a high level voltage value (for example, 0.8 V) when the air-fuel mixture richer than the stoichiometric air-fuel ratio burns, and when the air-fuel mixture is lean, the output level Vout (for example, low level) 0.2V), and when the air-fuel mixture is in the vicinity of the stoichiometric air-fuel ratio, it becomes a predetermined target value Vop (eg, 0.6V) between the high level and the low level.

A LAF sensor 14 is attached in the vicinity of the collecting portion 7j of the exhaust manifold 7a. The LAF sensor 14 ( air-fuel ratio detection means) is configured by combining a sensor similar to the O2 sensor 15 and a detection circuit such as a linearizer, and exhausts in a wide range of air-fuel ratios from the rich region to the lean region. The oxygen concentration in the gas is detected linearly, and a detection signal proportional to the oxygen concentration is output to the ECU 2. Based on the detection signal of the LAF sensor 14, the ECU 2 calculates a detected air-fuel ratio KACT ( air-fuel ratio detection value) representing the air-fuel ratio in the exhaust gas near the collecting portion 7j. The detected air-fuel ratio KACT is specifically calculated as an equivalence ratio.

  Further, an accelerator opening sensor 16, an atmospheric pressure sensor 17, an intake air temperature sensor 18, a vehicle speed sensor 19 and the like are connected to the ECU 2. The accelerator opening sensor 16 detects a depression amount (hereinafter referred to as “accelerator opening”) AP of an accelerator pedal (not shown) of the vehicle, and outputs a detection signal to the ECU 2. The atmospheric pressure sensor 17, the intake air temperature sensor 18, and the vehicle speed sensor 19 detect the atmospheric pressure PA, the intake air temperature TA, and the vehicle speed VP, respectively, and output detection signals to the ECU 2.

Next, the ECU 2 will be described. The ECU 2 includes a microcomputer including an I / O interface, a CPU, a RAM, and a ROM. The ECU 2 determines the operating state of the engine 3 according to the outputs of the various sensors 9 to 19 described above, and the ROM. The target air-fuel ratio KCMD, the feedback correction coefficient KSTR, the air-fuel ratio variation correction coefficient KOBSV _i, and the learning correction thereof are executed by executing an air-fuel ratio control process, which will be described later, in accordance with a control program stored in advance in FIG. A value KOBSV_LS _i or the like is calculated. Furthermore, as will be described later, these KCMD, KSTR, based on such calculated value of KOBSV _i and KOBSV_LS _i, to calculate the final fuel injection amount TOUT _i of the injector 6 for each cylinder, final fuel injection amount TOUT that this calculated By driving the injector 6 with a drive signal based on _i , the air-fuel ratio of the air-fuel mixture, that is, the air-fuel ratio of the exhaust gas is controlled for each cylinder. The subscript “i” in the final fuel injection amount TOUT _i is a cylinder number value representing a cylinder number (i = 1 to 4), and this point is the air-fuel ratio variation correction coefficient KOBSV _i , the learning correction value. The same applies to KOBSV_LS _i and parameters to be described later.

In the present embodiment, the ECU 2 causes the simulation value generation means, estimation means, identification means , learning correction value calculation means, delay means , fuel amount determination means, first correction value calculation means, first fuel amount correction means, 2 correction value calculating means, second fuel amount correcting means, operating state parameter detecting means, third correction value calculating means, third fuel amount correcting means, fourth correction value calculating means, and fourth fuel amount correcting means are configured. Yes.

  As shown in FIG. 2, the control device 1 includes a basic fuel injection amount calculation unit 20, a first air-fuel ratio controller 30, a second air-fuel ratio controller 40, an adhesion correction unit 50, and the like. Specifically, the ECU 2 is configured. In this control apparatus 1, the basic fuel injection amount TIBS is calculated by searching a table (not shown) according to the intake air amount GAIR by the basic fuel injection amount calculation unit 20 as the fuel amount determining means.

Further, as will be described later, in order to correct the variation in the air-fuel ratio between the cylinders, the first air-fuel ratio controller 30 calculates the air-fuel ratio variation correction coefficient KOBSV _i and its learning correction value KOBSV_LS _i , respectively. A feedback correction coefficient KSTR is calculated by the fuel ratio controller 40 in order to converge the detected air fuel ratio KACT to the target air fuel ratio KCMD. Then, the required fuel injection is obtained by multiplying the basic fuel injection amount TIBS by the corrected target air-fuel ratio KCMDM, the total correction coefficient KTOTAL, the feedback correction coefficient KSTR, the air-fuel ratio variation correction coefficient KOBSV _i , and the learning correction value KOBSV_LS _i. The amount TCYL _i is calculated for each cylinder. Next, the final fuel injection amount TOUT _i is calculated for each cylinder by the adhesion correction unit 50 based on the required fuel injection amount TCYL _i .

  Next, the first air-fuel ratio controller 30 will be described. The first air-fuel ratio controller 30 is for correcting variations in the air-fuel ratio between cylinders, and includes an adaptive observer 31, an air-fuel ratio variation correction coefficient calculation unit 32, a learning correction value calculation unit 33, and a multiplication unit 34. Has been.

In the first air-fuel ratio controller 30 (first fuel amount correction means) , the adaptive observer 31 (simulated value generation means, estimation means, identification means, delay means) determines the air-fuel ratio variation coefficient Φ _{i according} to the algorithm described below. An air-fuel ratio variation correction coefficient calculation unit 32 (first correction value calculation unit) is calculated for each cylinder, and an air-fuel ratio variation correction coefficient KOBSV _i is calculated for each cylinder, and a learning correction value calculation unit 33 (learning correction value calculation unit). ), The learning correction value KOBSV_LS _i of the air-fuel ratio variation correction coefficient is calculated for each cylinder. Even more to the multiplier 3 4, the air-fuel ratio variation correction coefficient KOBSV ₁ ~KOBSV _4, the learning correction value KOBSV_LS ₁ ~KOBSV_LS ₄ are respectively multiplied. That is, the air-fuel ratio variation correction coefficient KOBSV _i is corrected by the learning correction value KOBSV_LS _i .

Next, the algorithm of the adaptive observer 31 will be described. First, as shown in FIG. 3, the exhaust system of the engine 3 is regarded as a system represented by _four simulated values KACT_OS _{1 to} KACT_OS ₄ and four air-fuel ratio variation coefficients Φ _{1 to} Φ ₄ . These simulated values KACT_OS _i are values obtained by simulating the exhaust gas discharge timing and the exhaust behavior for each cylinder, and the air-fuel ratio variation coefficient Φ _i is the variation in the air-fuel ratio of the exhaust gas between the cylinders and the variation in the exhaust behavior. A value representing minutes. When this system is modeled as a discrete-time system model, Equation (1) shown in FIG. 4 is obtained. In the equation (1), symbol k represents a discretized time, and each discrete data (time series data) with symbol (k) is data sampled every time a TDC signal is generated. Show. This is the same for the other discrete data in the following specification (note that the discrete data may be data sampled every time the CRK signal is generated). Further, d represents a dead time (predetermined delay time) until the exhaust gas discharged from each cylinder reaches the LAF sensor 14, and is preset to a predetermined constant value in this embodiment. The dead time d may be set according to the operating state of the engine 3 (engine speed NE or the like).

In the adaptive observer 31 of the present embodiment, an equation in which the left side of the equation (1) is replaced with the detected air-fuel ratio estimated value KACT_EST (k), that is, the equation (2) in FIG. 4 is used as a model, and the simulated value KACT_OS _i Is generated by the signal generator 31a as will be described later, and the vector φ (k) of the air-fuel ratio variation coefficient Φ _i as the model parameter of the equation (2) is the estimated value KACT_EST (k) is the detected air-fuel ratio KACT. It is identified by a variable gain type sequential least squares algorithm shown in equations (3) to (9) of FIG. 4 so as to coincide with (k).

In Equation (3), KP (k) represents a gain coefficient vector, and ide (k) represents an identification error. In addition, φ (k) ^T in Equation (4) represents a transposed matrix of φ (k). In the following description, the expression “vector” is omitted as appropriate. The identification error ide (k) in equation (3) is calculated by equations (5) to (7) in FIG. 4, and ζ (k) in equation (6) is defined as in equation (7). It is a vector of simulated values. Further, the gain coefficient vector KP (k) is calculated by the equation (8) in FIG. 4, and P (k) in the equation (8) is a fourth-order defined by the equation (9) in FIG. It is a square matrix.

In this adaptive observer 31, the vector φ (k) of the air-fuel ratio variation coefficient Φ _i is identified by the sequential least square algorithm shown in the above equations (2) to (9). Thereby, the noise fluctuation component of the exhaust behavior associated with such that the operation state of the engine 3 is suddenly changed, the air-fuel ratio can be variations removed from the coefficient [Phi _i (filtering), the air-fuel ratio variation coefficient [Phi _i, cylinder It can be calculated as a value that substantially shows the variation in the air-fuel ratio between the two.

The configuration of the above adaptive observer 31 is shown in a block diagram in FIG. That is, as shown in the figure, in this adaptive observer 31, a vector ζ (k) of the simulated value KACT_OS _i is generated by the signal generator 31a. Specifically, in the signal generator 31a, as shown in FIG. 6, the simulated value KACT_OS _i has a waveform that is obtained by alternately combining triangular waves, trapezoidal waves, or the like so that the sum of the values is always 1. Are generated as signal values. Further, in the multiplier 31b, an estimated value KACT_EST (k) of the detected air-fuel ratio is generated as a value obtained by multiplying the vector ζ (k) of the simulated value by the vector φ (k−1) of the air-fuel ratio variation coefficient. . Then, the difference 31d generates an identification error ide (k) as a deviation between the detected air-fuel ratio KACT (k) and the estimated value KACT_EST (k).

  Further, the logic unit 31e generates a gain coefficient vector KP (k) based on the simulated value vector ζ (k), and the multiplier 31f generates an identification error ide (k) and a gain coefficient vector KP ( k) product [ide (k) · KP (k)] is generated. Next, the adder 31g uses the product [ide (k) · KP (k)] and the delayed air-fuel ratio variation coefficient vector φ (k−1) as the sum of the air-fuel ratio variation coefficient vector φ ( k) is generated.

Next, an algorithm for calculating the air-fuel ratio variation correction coefficient KOBSV _i ( first correction value ) in the air-fuel ratio variation correction coefficient calculation unit 32 will be described. First, the air-fuel ratio variation correction coefficient calculation unit 32 moves the air-fuel ratio variation coefficient based on the air-fuel ratio variation coefficient Φ _i (k) calculated for each cylinder by the adaptive observer 31 according to the equation (10) in FIG. An average value Φave (k) is calculated. Next, the air-fuel ratio variation correction coefficient KOBSV _i is cylinderized by an I-PD control (proportional / differential preceding PID control) algorithm so that the air-fuel ratio variation coefficient Φ _i (k) converges to the moving average value Φave (k). Calculate every time. This I-PD control algorithm is shown in equations (11) and (12) in FIG. In this equation (12), e (k) represents a tracking error.

As described above, the air-fuel ratio variation correction coefficient calculation unit 32 corrects the air-fuel ratio variation so that the air-fuel ratio variation coefficient Φ _i (k) converges to the moving average value Φave (k) by the I-PD control algorithm. A coefficient KOBSV _i is calculated. This is because the air-fuel ratio variation coefficient Φ _i (k) is controlled so that overshoot does not occur in the convergence behavior of the air-fuel ratio variation coefficient Φ _i (k) to the moving average value Φave (k). This is to prevent the air-fuel ratio control for correcting the variation in the fuel ratio from interfering with the air-fuel ratio control by the second air-fuel ratio controller 40 described later.

Next, an algorithm for calculating the learning correction value KOBSV_LS _i of the air-fuel ratio variation correction coefficient KOBSV _{i by} the learning correction value calculation unit 33 will be described. The air-fuel ratio variation correction coefficient KOBSV _i is easily affected by the operating state of the engine 3, and therefore changes accordingly when the operating state of the engine 3 changes. FIG. 8A shows the relationship between the exhaust gas volume ESV (k) as an operating state parameter representing the operating state of the engine 3 and the air-fuel ratio variation correction coefficient KOBSV _i (k). The exhaust gas volume ESV (k ) is an estimated value of the space velocity and is calculated by the equation (13) in FIG. In the equation (13), SVPRA is a predetermined coefficient determined in advance by the displacement of the engine 3.

Referring to FIG. 8A, in the air-fuel ratio variation correction coefficient KOBSV _i (k), this is a dependent variable, and the air-fuel ratio variation correction coefficient is expressed by a linear expression with the exhaust gas volume ESV (k) as an independent variable. It can be seen that an approximate value, that is, an estimated value of KOBSV _i (k) can be calculated (see FIG. 8B). Therefore, the learning correction value calculation unit 33 defines the learning correction value KOBSV_LS _i (k) of the air-fuel ratio variation correction coefficient as an estimated value calculated by the regression equation shown in the equation (14) of FIG. A vector (hereinafter referred to as “regression coefficient vector”) θOBSV_LS _i (k) of the coefficient AOBSV_LS _i and the constant term BOBSV_LS _i is calculated by the sequential least square method shown in the equations (15) to (21) of FIG.

In this equation (15), KQ _i (k) represents a vector of gain coefficients, and Eov _i (k) represents an error. Further, the error Eov _i (k) is calculated by the equations (17) to (19) in FIG. Further, the vector KQ _i (k) of the gain coefficient is calculated by the equation (20) in FIG. 9, and Q _i (k) in the equation (20) is 2 defined by the equation (21) in FIG. The following square matrix.

Further, the learning correction value KOBSV_LS _i (k) is specifically calculated by the equation (22) in FIG. As will be described later, when the engine 3 is in an extreme operating state or operating environment, the calculation of the regression coefficient AOBSV_LS _i and the constant term BOBSV_LS _i by the above-described sequential least square method is avoided, and the previous regression coefficient vector is calculated. The value θOBSV_LS _i (k−1) is used as the current value θOBSV_LS _i (k) in the calculation of the learning correction value KOBSV_LS _i (k).

According to the algorithms shown in the above equations (13), (15) to (22), the learning correction value calculation unit 33 uses the learning correction value KOBSV_LS _i (k) as the air-fuel ratio variation correction coefficient KOBSV _i (k). To converge to the product of

Next, the second air-fuel ratio controller 40 ( second correction value calculating means, second fuel amount correcting means, fourth correction value calculating means, fourth fuel amount correcting means ) will be described. Specifically, the second air-fuel ratio controller 40 is configured as an STR (Self Tuning Regulator) including an on-board identifier 41 and an STR controller 42. In the second air-fuel ratio controller 40, the feedback correction coefficient KSTR is calculated so that the detected air-fuel ratio KACT converges to the target air-fuel ratio KCMD (predetermined target value). More specifically, the on-board identifier 41 identifies the model parameter vector θ of the first cylinder # 1 by the algorithm described below, and the STR controller 42 feeds back the feedback correction coefficient KSTR ( second correction value, fourth correction value). ) Is calculated.

  First, when the first cylinder # 1 is regarded as a control target having the feedback correction coefficient KSTR as an input and the detected air-fuel ratio KACT as an output, and modeling this control target as a discrete time system model, the equation (23) in FIG. It will be shown in In the equation (23), the symbol n represents the discretized time, and each discrete data with the symbol (n) is sampled every one combustion cycle, that is, every time the TDC signal is generated four times in succession. Data. This also applies to the following discrete data.

  Here, since the dead time of the detected air-fuel ratio KACT with respect to the target air-fuel ratio KCMD is estimated to be about three combustion cycles, the relationship of KCMD (n) = KACT (n + 3) is established, and this is expressed by the equation (23). When applied, equation (24) in FIG. 10 is derived.

Further, the vector θ (n) of the model parameters b0 (n), r1 (n), r2 (n), r3 (n), and s0 (n) in the equation (23) is expressed by the equations (25) to ( It is identified by the identification algorithm of 31). In the equation (25), KΓ (n) represents a vector of gain coefficients, and ide_st (n) represents an identification error. Further, θ (n) ^T in the equation (26) represents a transposed matrix of θ (n).

  The identification error ide_st (n) in the above equation (25) is calculated by the equations (27) to (29) in FIG. 10, and KACTHAT (n) in the equation (28) represents the identification value of the detected air-fuel ratio KACT. Yes. Further, the gain coefficient vector KΓ (n) is calculated by the equation (30) of FIG. 10, and Γ (n) of the equation (30) is a fifth order defined by the equation (31) of FIG. It is a square matrix.

  In the control system as in the present embodiment, when the air-fuel ratio control is executed by the algorithm of the above formulas (24) to (31), when the low-pass characteristic of the LAF sensor 14 is strong, the update cycle of the model parameter vector θ The resonance of the control system may occur at an integral multiple of the cycle. In order to solve this problem, the second air-fuel ratio controller 40 of the present embodiment calculates the feedback correction coefficient KSTR as follows.

  That is, in the second air-fuel ratio controller 40 of the present embodiment, the model parameter vector θ of the first cylinder # 1 identified by the on-board identifier 41 is oversampled in synchronization with the generation timing of the TDC signal, and the A moving average value θ_ave is calculated. Specifically, the moving average value θ_ave (k) of the model parameter vector θ is calculated by the equation (32) in FIG. 11, and using this, the feedback correction coefficient KSTR ( k) is calculated. In the equation (32), θbuf indicates the oversampling value of the model parameter vector θ of the first cylinder # 1, and the moving average value θ_ave (k) is defined as the equation (33) in FIG. The Further, m in the expression (32) is a predetermined integer, and is set to m = 11 in the present embodiment.

  As described above, each piece of discrete data with the symbol (k) in the equations (32) to (34) is data sampled in synchronization with the generation timing of the TDC signal, so that n−f = k−. When the relationship of 4 · f (f: integer) is established and this is applied to the equation (24) in FIG. 10, the above equation (34) is derived. Further, an identification algorithm for identifying the model parameter vector θ (k) is as shown in equations (35) to (41) in FIG.

  As described above, in the onboard identifier 41 of the second air-fuel ratio controller 40 of the present embodiment, the model parameter vector θ is identified by the identification algorithm shown in the equations (35) to (41) in FIG. At 42, the feedback correction coefficient KSTR (k) is calculated by the equations (32) to (34) in FIG.

Hereinafter, the air-fuel ratio control executed by the ECU 2 will be described with reference to FIGS. In the following description, symbols (k) and (n) indicating the current value are omitted as appropriate. FIG. 12 shows the main routine of this control process, and this process is interrupted in synchronization with the input of the TDC signal. In this process, as described below, the final fuel injection amount TOUT _i is calculated for each cylinder.

  First, in step 1 (abbreviated as “S1” in the figure, the same applies hereinafter), the outputs of the various sensors 9 to 19 described above are read, and the read data is stored in the RAM.

  Next, the process proceeds to step 2 to calculate a basic fuel injection amount TIBS. In this process, the basic fuel injection amount TIBS is calculated by searching a table (not shown) according to the intake air amount GAIR.

  Next, the process proceeds to step 3 where a total correction coefficient KTOTAL is calculated. This total correction coefficient KTOTAL searches various tables and maps according to various operating parameters (for example, intake air temperature TA, atmospheric pressure PA, engine water temperature TW, accelerator pedal opening AP, throttle valve opening TH, etc.). Thus, various correction coefficients are calculated, and these various correction coefficients are multiplied by each other.

  Next, the routine proceeds to step 4 where the target air-fuel ratio KCMD is calculated. The content of the calculation process of the target air-fuel ratio KCMD is executed in the same manner as the control method described in Japanese Patent Laid-Open No. 2000-179385, although not shown here. That is, the target air-fuel ratio KCMD is calculated according to the operating state of the engine 3 so that the output Vout of the O2 sensor 15 converges to the predetermined target value Vop by the sliding mode control process or the map search process.

  Next, the routine proceeds to step 5 where a corrected target air-fuel ratio KCMDM is calculated. The corrected target air-fuel ratio KCMDM is for compensating for the change in charging efficiency due to the change in the air-fuel ratio A / F, and a table (not shown) is searched according to the target air-fuel ratio KCMD calculated in step 4 above. Is calculated by

  Next, in steps 6 and 7, the model parameter vector θ and the feedback correction coefficient KSTR of the first cylinder # 1 are calculated, respectively. These calculation processes will be described later.

Next, in steps 8 to 10, an air-fuel ratio variation coefficient vector φ, an air-fuel ratio variation correction coefficient KOBSV _i and a learning correction value KOBSV_LS _i are calculated. These calculation processes will be described later.

Next, the routine proceeds to step 11 where the basic fuel injection amount TIBS, total correction coefficient KTOTAL, correction target air-fuel ratio KCMDM, feedback correction coefficient KSTR, air-fuel ratio variation correction coefficient KOBSV _i and learning correction value KOBSV_LS _i calculated as described above are obtained. Using the following equation (42), the required fuel injection amount TCYL _i is calculated.

TCYL _i = TIBS, KTOTAL, KCMDM, KSTR, KOBSV _i , KOBSV_LS _i (42)

Next, the routine proceeds to step 12, where the final fuel injection amount TOUT _i is calculated by adhering the required fuel injection amount TCYL _i . Specifically, the final fuel injection amount TOUT _i is calculated according to the operating state of the engine 3 such as the ratio of the fuel injected from the injector 6 in the current combustion cycle adhering to the inner wall surface of the combustion chamber. It is calculated by correcting the required fuel injection amount TCYL _i based on the calculated ratio.

Next, the process proceeds to step 13 where a drive signal based on the final fuel injection amount TOUT _i calculated as described above is output to the injector 6 of the corresponding cylinder, and then this process is terminated.

  Next, the calculation process of the model parameter vector θ in step 6 will be described. In this process, first, in step 20, a process for setting the cylinder number value i corresponding to the subscript “i” of each parameter is executed.

  In this process, although not shown, the cylinder number value i is set as follows based on the previous value PRVi of the cylinder number value i set in the previous loop stored in the RAM. Specifically, i = 3 when PRVi = 1, i = 1 when PRVi = 2, i = 4 when PRVi = 3, and i = 2 when PRVi = 4. As described above, the cylinder number value i is repeatedly set in the order of “1 → 3 → 4 → 2 → 1 → 3 → 4 → 2 → 1...”, For example.

  Next, the routine proceeds to step 21, where it is determined whether or not the cylinder number value i set at step 20 is a value of 1. When the determination result is YES and the model parameter vector θ of the first cylinder # 1 is to be calculated, the process proceeds to step 22 where the calculated value of the model parameter vector θ in the previous loop stored in the RAM is It is set as the previous value PRVθ [= θ (n−1)].

  Next, the process proceeds to step 23, and after calculating the vector ξ by the above-described equation (39) in FIG. 11, at step 24, the KACT identification value KACT_HAT is calculated by the above-described equation (38) in FIG. 11.

  Next, the process proceeds to step 25, and after the identification error ide_st is calculated by the above-described equation (37) in FIG. 11, the next value NEXΓ [= of the square matrix in the previous loop stored in the RAM is stored in step 26. The calculated value of Γ (n + 1)] is set as the current value Γ.

  Next, the process proceeds to step 27, the gain coefficient vector KΓ is calculated from the above-described equation (40) in FIG. 11, and then the process proceeds to step 28, where the model parameter vector θ is determined from the above-described equation (35) in FIG. calculate.

  Next, the process proceeds to step 29, the next value NEXΓ of the square matrix is calculated by the above-described equation (41) in FIG. 11, and then the process proceeds to step 30, where a predetermined number (previous embodiment) stored in the RAM is stored. 12) of the detected air-fuel ratio KACT is updated. Specifically, each value of KACT in the RAM is set as an old value for one control cycle (for example, the current value KACT (k) is set as the previous value KACT (k−1), and the previous value KACT (k−1) is set. -1) is set as the previous value KACT (k-2)).

  Next, the process proceeds to step 31, and the oversampling value θbuf of the model parameter vector θ of the predetermined number (12 in this embodiment) of the first cylinder # 1 stored in the RAM is updated. Specifically, as in step 30 above, each value of θbuf in the RAM is set as an old value for one control cycle (for example, the current value θbuf (k) is set to the previous value θbuf (k−1). The previous value θbuf (k−1) is set as the previous value θbuf (k−2)). Then, this process is complete | finished.

  On the other hand, if the determination result in step 21 is NO and it is not necessary to calculate the model parameter vector θ, steps 22 to 29 are skipped, steps 30 and 31 are executed as described above, and the process is terminated. .

  Next, the calculation process of the feedback correction coefficient KSTR in step 7 will be described with reference to FIG. In this process, first, in step 40, the moving average value θ_ave of the model parameter vector is calculated based on the oversampling value θbuf updated in step 31 by the above-described equation (32) of FIG.

  Next, in step 41, the feedback correction coefficient KSTR is calculated based on the moving average value θ_ave calculated in step 41 by the above-described equation (34) in FIG.

  Next, the process proceeds to step 42, and the predetermined number (12 in the present embodiment) of feedback correction coefficients KSTR stored in the RAM are updated. Specifically, each value of KSTR in the RAM is set as an old value for one control cycle (for example, the current value KSTR (k) is set as the previous value KSTR (k−1), and the previous value KSTR (k−1). -1) is set as the previous value KSTR (k-2)). Then, this process is complete | finished.

  Next, the calculation processing of the vector φ of the air-fuel ratio variation coefficient in step 8 will be described with reference to FIG. In this process, first, in step 50, the calculated value of the vector φ of the air-fuel ratio variation coefficient in the previous loop stored in the RAM is set as the previous value PRVφ [= φ (k−1)]. .

  Next, the process proceeds to step 51, where the simulated value vector ζ is calculated by the equation (7) in FIG. 4, and then the process proceeds to step 52, where the estimated value KACT_EST of the detected air-fuel ratio is calculated by the equation (6) in FIG. calculate.

  Next, the process proceeds to step 53, where the identification error ide is calculated by the above-described equation (5) in FIG. 4, and then the process proceeds to step 54 where the next value NEXP of the square matrix in the previous loop stored in the RAM. The calculated value of [= P (k + 1)] is set as the current value P.

  Next, the process proceeds to step 55, the gain coefficient vector KP is calculated from the above-described equation (8) in FIG. 4, and then the process proceeds to step 56, where the air-fuel ratio variation coefficient vector is calculated from the above-described equation (3) in FIG. Calculate φ.

Next, the process proceeds to step 57, the next value NEXP [= P (k + 1)] of the square matrix is calculated by the above-described equation (9) in FIG. 4, and then the process proceeds to step 58, where a predetermined number stored in the RAM is stored. The time series data of the simulation value KACT_OS _i (12 in this embodiment) is updated. Specifically, each value of the simulated value KACT_OS _i in the RAM is set as an old value for one control cycle (for example, the current value KACT_OS _i (k) is set as the previous value KACT_OS _i (k−1)). The previous value KACT_OS _i (k−1) is set as the previous value KACT_OS _i (k−2)).

Next, the process proceeds to step 59, where the present value KACT_OS _i of the simulated value is calculated, and then this process ends.

Next, the calculation process of the air-fuel ratio variation correction coefficient KOBSV _i in step 9 will be described with reference to FIG. In this process, first, in step 70, the moving average value Φave of the air-fuel ratio variation coefficient is calculated by the above-described equation (10) in FIG.

Next, the process proceeds to step 71, where the tracking error e is calculated by the above-described equation (12) in FIG. Next, the process proceeds to step 73, and the air-fuel ratio variation correction is performed by using the moving average value Φave of the air-fuel ratio variation coefficient calculated in steps 70 and 72 and the integral value Σe of the tracking error according to the above-described equation (11) in FIG. After the coefficient KOBSV _i is calculated, this process is terminated.

Next, the calculation process of the learning correction value KOBSV_LS _i for the air-fuel ratio variation correction coefficient in step 10 will be described with reference to FIG. In this process, first, in step 80, the exhaust gas volume ESV is calculated by the aforementioned equation (13) of FIG.

Next, the process proceeds to step 81, where the calculated value of the regression coefficient vector θOBSV_LS _{i in} the previous loop stored in the RAM is set as the previous value PRVθOBSV_LS _i [= θOBSV_LS _i (k−1)].

Next, the routine proceeds to step 82, where the learning correction value KOBSV_LS _i is calculated by the equation (22) shown in FIG. Thereafter, the process proceeds to step 83 to determine whether or not any of the following five conditions (a1) to (a5) is satisfied.
(A1) The engine water temperature TW is higher than a predetermined lower limit value TWOBSL and lower than a predetermined upper limit value TWOBSH.
(A2) The intake air temperature TA is higher than a predetermined lower limit value TAOBSL and lower than a predetermined upper limit value TAOBSH.
(A3) The engine speed NE is higher than a predetermined lower limit value NEOBSL and lower than a predetermined upper limit value NEOBSH.
(A4) The intake pipe absolute pressure PBA is higher than a predetermined lower limit value PBOBSL and lower than a predetermined upper limit value PBOBSH.
(A5) The vehicle speed VP is higher than a predetermined lower limit value VPOBSL and lower than a predetermined upper limit value VPOBSH.

When all of the above five conditions (a1) to (a5) are satisfied, it is determined that the regression coefficient vector θOBSV_LS _i is to be calculated by the recursive least square method, and the process proceeds to step 84 and described above. The exhaust gas volume vector Z is calculated by the equation (19) in FIG.

Next, the process proceeds to step 85, and after calculating the error Eov _i by the above-described equation (17) in FIG. 9, the process proceeds to step 86, and the next value NEXQ of the square matrix in the previous loop stored in the RAM. The calculated value of _i [= Q _i (k + 1)] is set as the current value Q _i .

Next, the process proceeds to step 87, the gain coefficient vector KQ _i is calculated from the above-described equation (20) in FIG. 9, and then the process proceeds to step 88, where the regression coefficient vector θOBSV_LS _i is calculated from the above-described equation (15) in FIG. Is calculated. Next, the process proceeds to step 89, and the next value NEXQ _i [= Q _i (k + 1)] of the square matrix is calculated by the above-described equation (21) in FIG.

On the other hand, when the determination result in step 83 is NO and at least one of the five conditions (a1) to (a5) is not satisfied, the process proceeds to step 90 and the previous value of the regression coefficient vector set in step 81 is obtained. After setting PRVθOBSV_LS _i to the current value θOBSV_LS _i , the present process is terminated. Thereby, for example, in the process of step 81 in the next loop, the value calculated by the sequential least square method of steps 84 to 89 in the current loop is used as the previous value PRVθOBSV_LS _i of the regression coefficient vector.

Next, the operation when the air-fuel ratio is controlled by the above control device 1 will be described with reference to FIGS. 18 and 19. FIG. 18 shows a case where the air-fuel ratio is controlled by the control device 1 of the present embodiment. More specifically, the detected air-fuel ratio KACT is set to a value 1 (corresponding to the theoretical air-fuel ratio) by the second air-fuel ratio controller 40. When the first air-fuel ratio controller 30 is operated from a stopped state, that is, by the first air-fuel ratio controller 30, the air-fuel ratio variation coefficient Φ _i and the air-fuel ratio variation are controlled. The operation example when the calculation of the correction coefficient KOBSV _i and the learning correction value KOBSV_LS _i is started is shown.

For comparison, FIG. 19 shows that the learning correction value KOBSV_LS _i is replaced with the normal PID control algorithm (formula (43 in FIG. 20) instead of the I-PD control algorithm of formulas (11) and (12). ) And (44) (algorithm shown in FIG. 44) are comparative examples of operations. In both figures, the values of KACT ₁ to 4 are the air-fuel ratios (equivalent ratio converted values) in the exhaust gases that are exhausted from the first to fourth cylinders # 1 to # 4 and are not mixed with each other. Specifically, four LAF sensors for measurement (not shown) are additionally provided immediately after the exhaust ports of cylinders # 1 to # 4 of the exhaust manifold 7a. Based on the output of the LAF sensor, the values of KACT ₁ to ₄ are calculated.

As shown in FIG. 18, in the operation example of the present embodiment, when the first air-fuel ratio controller 30 is stopped, KACT ₁ to ₄ representing the air-fuel ratio of the exhaust gas discharged from each cylinder are unstable. As a result, the detected air-fuel ratio KACT also becomes slightly unstable. However, when the first air-fuel ratio controller 30 is activated (time t1), after some time has elapsed, all of KACT ₁ to ₄ converge to the value 1 (equivalent ratio corresponding to the theoretical air-fuel ratio), and accordingly. It can be seen that the detected air-fuel ratio KACT also converges to the value 1. That is, it can be seen that the variation in the air-fuel ratio between the cylinders is appropriately corrected. It can also be seen that the value of the product KOBSV _i · KOBSV_LS _i (i = 1 to 4) of the air-fuel ratio variation correction coefficient and the learning correction value is also stable.

On the other hand, in the comparative example of FIG. 19, the settling time from when the first air-fuel ratio controller 30 is activated (time t2) until KACT ₁ to ₄ all converge to the value 1 is the operation of this embodiment. It can be seen that the detected air-fuel ratio KACT does not easily converge to the value 1, which is longer than the example. In addition, it can be seen that the value of the product KOBSV _i · KOBSV_LS _i of the air-fuel ratio variation correction coefficient and the learning correction value is not stable. That is, it can be seen that the variation of the air-fuel ratio between the cylinders can be corrected quickly and appropriately by using the I-PD control algorithm as in the present embodiment, compared to the case of using the normal PID control algorithm. This is because the I-PD control algorithm calculates the learning correction value KOBSV_LS _i so that overshoot does not occur in the convergence behavior of the air-fuel ratio variation coefficient Φ _{i to} the moving average value Φave than the PID control algorithm. It depends on what you can do.

As described above, according to the control device 1 of the present embodiment, the air-fuel ratio variation coefficient Φ _i is calculated by the first air-fuel ratio controller 30, and the air-fuel ratio variation correction coefficient is converged to the moving average value Φave. KOBSV _i is calculated, and its learning correction value KOBSV_LS _i is calculated. Further, the second air-fuel ratio controller 40 calculates the feedback correction coefficient KSTR so that the detected air-fuel ratio KACT converges to the target air-fuel ratio KCMD. Then, the basic fuel injection amount TIBS is corrected by the calculated feedback correction coefficient KSTR, air-fuel ratio variation correction coefficient KOBSV _i, and learning correction value KOBSV_LS _i , whereby the final fuel injection amount TOUT _i is calculated for each cylinder. .

In the adaptive observer 31 of the first air-fuel ratio controller 30, an estimated value KACT_EST of the detected air-fuel ratio KACT is expressed by a model [Expression (2)] defined by the simulated value KACT_OS _i and the air-fuel ratio variation coefficient Φ _i . Further, the air-fuel ratio variation coefficient Φ _i as a model parameter is identified by the sequential least square method so that the estimated value KACT_EST matches the detected air-fuel ratio KACT. Thus, the noise fluctuation component of the exhaust behavior associated with such that the operation state of the engine 3 is suddenly changed, the air-fuel ratio can be variations removed from the coefficient [Phi _i (filtering), the air-fuel ratio variation coefficient [Phi _i, cylinder It can be calculated as a value that substantially shows the variation in the air-fuel ratio between the two. Therefore, since the basic fuel injection amount TIBS is corrected for each cylinder by the air-fuel ratio variation correction coefficient KOBSV _i calculated based on such an air-fuel ratio variation coefficient Φ _i , the detected air-fuel ratio due to fuel adhesion in each cylinder is corrected. Even when the dynamic characteristics of the controlled object changes due to changes in the contribution of each cylinder to KACT, the response variation of the LAF sensor 14 and aging, etc., the change in the dynamic characteristics of the controlled object is reflected in the model unlike the conventional case. However, the final fuel injection amount TOUT _i can be calculated for each cylinder so as to correct the variation in the air-fuel ratio between the cylinders. As a result, even when the air-fuel ratio of the engine 3 having a complicated exhaust layout is controlled as in the first embodiment, the air-fuel ratio control with a large stability margin and high robustness can be realized, and a good catalyst A purification rate can be secured.

In the first air-fuel ratio controller 30, since the air-fuel ratio variation correction coefficient KOBSV _i is calculated by the I-PD control algorithm, overshoot occurs in the convergence behavior of the air-fuel ratio variation coefficient Φ _{i to} the moving average value Φave. Thus, the air-fuel ratio variation correction coefficient KOBSV _i can be calculated. As a result, it is possible to correct the variation in the air-fuel ratio among the cylinders while avoiding the behavior of the air-fuel ratio of each cylinder from oscillating. Further, since the air-fuel ratio variation correction coefficient KOBSV _i is calculated so that the air-fuel ratio variation coefficient Φ _i converges to the moving average value Φave, the air-fuel ratio control by the first air-fuel ratio controller 30 is performed by the second air-fuel ratio controller 40. It is possible to correct the variation in the air-fuel ratio among the cylinders while avoiding the interference with the air-fuel ratio control.

Further, in the first air-fuel ratio controller 30, the learning correction value KOBSV_LS _{i of} the air-fuel ratio variation correction coefficient KOBSV _i is calculated by a regression equation [Equation (22)] with the exhaust gas volume ESV as an independent variable. A regression coefficient vector θOBSV_LS _i that is a vector of the regression coefficient AOBSV_LS _i and the constant term BOBSV_LS _{i in} the equation is calculated by a sequential least square method. Therefore, even when the variation state of the air-fuel ratio between the cylinders changes abruptly due to the influence of the engine 3 in an operating state that changes rapidly, such as a transient operation state, the learning correction value KOBSV_LS _i is set to It can be calculated as a value appropriately reflecting the variation state of the fuel ratio. As a result, even when the engine 3 is in a transient operation state, the air-fuel ratio can be appropriately controlled while compensating for the variation in the air-fuel ratio between the cylinders.

In addition, since the air-fuel ratio variation coefficient Φ _i and the regression coefficient vector θOBSV_LS _i are calculated by the sequential least square method, compared to the case of using a general least square method as a statistical algorithm, even at the start of air-fuel ratio control. The air-fuel ratio variation correction coefficient KOBSV _i and the learning correction value KOBSV_LS _i can be calculated for each control cycle. Therefore, for example, by setting the initial values of KOBSV _i and KOBSV _i in advance, the learning correction value KOBSV_LS _i calculated for each control cycle and the air-fuel ratio variation correction coefficient KOBSV _i at the start of the air-fuel ratio control The final fuel injection amount TOUT _i can be calculated as a value always corrected by the product of the above, and the controllability at the start of air-fuel ratio control can be improved. Thereby, the catalyst purification rate at the start of air-fuel ratio control can be further improved.

The first embodiment is an example in which a linear expression is used as a regression expression used for calculating the learning correction value KOBSV_LS _i , but the regression expression is not limited to this, and an n-order expression (n is an integer of 2 or more). But you can. Even in such a case, the same effect as that of the first embodiment can be obtained by calculating the regression coefficient and the constant term of the nth order equation by the sequential least square method. Further, the learning correction value KOBSV_LS _i may be calculated by using a predetermined value set in advance for each of a plurality of operation regions as the regression coefficient and constant term of the regression equation. In this way, the calculation time of the learning correction value KOBSV_LS _i can be shortened, and the calculation load on the ECU 2 can be reduced.

Further, the control algorithm for calculating the air-fuel ratio variation correction coefficient KOBSV _i in order to converge the air-fuel ratio variation coefficient Φ _i to the moving average value Φave is not limited to the I-PD control algorithm of the first embodiment. It goes without saying that a control algorithm may be used. For example, instead of the I-PD control algorithm, the air-fuel ratio variation correction coefficient KOBSV _i is calculated by using the IP-D control algorithm (differential preceding PID control algorithm) shown in equations (45) and (46) of FIG. The air-fuel ratio variation correction coefficient KOBSV _i may be calculated by using a response designating control algorithm (sliding mode control algorithm or backstepping control algorithm) shown in equations (47) to (49) in FIG. It may be calculated. Even when these control algorithms are used, overshoot does not occur in the convergence behavior of the air-fuel ratio variation coefficient Φ _{i to} the moving average value Φave, as in the case of using the I-PD control algorithm of the embodiment. Thus, the air-fuel ratio variation correction coefficient KOBSV _i can be calculated, and as a result, variations in the air-fuel ratio among the cylinders can be corrected quickly and appropriately.

Further, as described above, when the I-PD control algorithm, the IP-D control algorithm, and the response designation control algorithm are used in the calculation of the air-fuel ratio variation correction coefficient KOBSV _i , the feedback gain is set to the optimum regulator theory or It may be determined based on _H∞ control theory. In this way, in the convergence behavior of the air-fuel ratio variation coefficient Φ _{i to} the moving average value Φave, it is possible to more effectively suppress the occurrence of overshoot, and as a result, the accuracy of correcting the variation in air-fuel ratio among cylinders can be further increased. Can be improved.

Further, when the settling time until the air-fuel ratio variation coefficient Φ _i converges to the moving average value Φave may be long, the PID control algorithm shown in the equations (43) and (44) of FIG. It goes without saying that the fuel ratio variation correction coefficient KOBSV _i may be calculated. Furthermore, the average value of the air-fuel ratio variation coefficient as a target value for converging the air-fuel ratio variation coefficient Φ _i is not limited to the moving average value Φave of the embodiment, and may be a weighted average value.

  Further, the adaptive observer 31 of the first air-fuel ratio controller 30 according to the first embodiment uses the variable gain type sequential least square method shown in equations (3) to (9) of FIG. In the example in which (k) is identified, it goes without saying that the identification algorithm of the vector φ (k) of the air-fuel ratio variation coefficient in the adaptive observer 31 is not limited to this. For example, the vector φ (k) of the air-fuel ratio variation coefficient may be identified by a fixed gain method to which the δ correction method shown in equations (50) to (57) of FIG. 21 is applied.

  Φbase in equation (50) in the figure is a reference value vector (model parameter reference value) defined as in equation (51), and reference values Φbase1 to Φbase4 which are four elements of this vector are exhaust gas. It is calculated by searching the table shown in FIG. 22 according to the volume ESV. As shown in the figure, the four reference values Φbase1 to Φbase4 are all set as values in the vicinity of the value 1. Further, dφ (k) in Expression (50) is a correction term (correction component) defined as in Expression (52), and is calculated by Expressions (53) to (57).

  When the vector φ (k) of the air-fuel ratio variation coefficient is identified by the fixed gain method to which the above δ correction method is applied, the calculation time can be shortened compared with the case of the sequential least square method, and the calculation in the ECU 2 The load can be reduced. As a result, the ECU 2 can be reduced in size and cost. In addition to this, the identification value of the vector φ (k) can be constrained to the vicinity of the value 1. Therefore, even when the engine 3 is in an operating state in which the change of the air-fuel ratio is severe, such as a transient operating state, the air-fuel ratio between the cylinders. It is possible to calculate the air-fuel ratio variation coefficient vector φ (k) representing the variation of the air-fuel ratio promptly and appropriately as a value that appropriately reflects the behavior of the air-fuel ratio, and to improve the stability of the air-fuel ratio control. it can.

  In the above-described fixed gain method, when the table shown in FIG. 22 cannot be prepared in advance, any of the four elements Φbase1 to Φbase4 may be set to a value 1.

  Further, the calculation method of the basic fuel injection amount TIBS in step 2 of FIG. 12 is not limited to the example of the first embodiment in which the basic fuel injection amount TIBS is calculated by searching a table according to the intake air amount GAIR. For example, the basic fuel injection amount TIBS may be calculated by searching the map according to the intake pipe absolute pressure PBA and the engine speed NE.

  Next, the control apparatus 1 which concerns on 2nd Embodiment of this invention is demonstrated. This control device 1 differs from the control device 1 according to the first embodiment described above only in that it includes a third air-fuel ratio controller 60 as shown in FIGS. Has been. Therefore, hereinafter, the third air-fuel ratio controller 60 will be mainly described, and the same components as those in the first embodiment will be denoted by the same reference numerals, and the description thereof will be omitted as appropriate.

In this control device 1, as will be described later, the third air-fuel ratio controller 60 calculates the intake air amount variation correction coefficient KICYL _i and its learning correction value KICYL_LS _i in order to correct the variation in the intake air amount between the cylinders. Is done. Then, the basic target fuel injection amount TIBS includes a correction target air-fuel ratio KCMDM, a total correction coefficient KTOTAL, a feedback correction coefficient KSTR, an air-fuel ratio variation correction coefficient KOBSV _i , an air-fuel ratio variation correction coefficient learning correction value KOBSV_LS _i , and an intake amount variation correction. The required fuel injection amount TCYL _i is calculated for each cylinder by multiplying the coefficient KICYL _i and the learning correction value KICYL_LS _i of the intake air amount variation correction coefficient. Next, the final fuel injection amount TOUT _i is calculated for each cylinder by the adhesion correction unit 50 based on the required fuel injection amount TCYL _i .

  Next, the third air-fuel ratio controller 60 will be described. As shown in FIG. 25, when the intake air amount GAIR to the engine 3 is detected by the air flow sensor 9, intake pulsation is also detected due to the intake behavior of each cylinder. The intake pulsation becomes irregular as shown in the figure when variations in intake air amount occur between cylinders. That is, this figure shows an example in which the intake air amount in the fourth cylinder # 4 is smaller than in the other cylinders.

The third air-fuel ratio controller 60 (third fuel amount correction means) estimates the variation in intake air amount between the cylinders as described above, and corrects the fuel injection amount accordingly, and is an adaptive observer. 61, an intake air amount variation correction coefficient calculation unit 62, a learning correction value calculation unit 63, and a multiplication unit 64. In the third air-fuel ratio controller 60, the intake air amount variation coefficient Ψ _i is calculated for each cylinder in the adaptive observer 61 (simulated value generating means, estimating means, identifying means, delay means) according to the algorithm described below. The variation correction coefficient calculation unit 62 (third correction value calculation unit) calculates the intake amount variation correction coefficient KICYL _i for each cylinder, and the learning correction value calculation unit 63 (learning correction value calculation unit) calculates the intake amount variation correction coefficient. learning correction value KICYL_LS _i is calculated for each cylinder of the. Even more to the multiplier 6 4, the intake air amount variation correction coefficient KICYL ₁ ~KICYL _4, the learning correction value KICYL_LS ₁ ~KICYL_LS ₄ are respectively multiplied. That is, the intake air amount variation correction coefficient KICYL _i is corrected by the learning correction value KICYL_LS _i .

Next, the algorithm of the adaptive observer 61 will be described. First, as shown in FIG. 26, the intake system of the engine 3 is regarded as a system represented by _four simulated values GAIR_OS _{1 to} GAIR_OS ₄ and four intake air amount variation coefficients Ψ _{1 to} Ψ ₄ . These simulated values GAIR_OS _i are values obtained by simulating the intake air intake start timing and intake behavior for each cylinder, and the intake air amount variation coefficient Ψ _i is a variation in intake air amount between cylinders and a variation in intake behavior. Is a value representing When this system is modeled as a discrete time system model, an equation (58) shown in FIG. 27 is obtained. In the equation (58), d ′ represents a dead time (predetermined delay time) until the air flowing in the intake pipe 4 reaches each cylinder from the airflow sensor 9. It is preset to a constant value. The dead time d ′ may be set according to the operation state of the engine 3 (engine speed NE or the like).

In the adaptive observer 61 of the present embodiment, an equation in which the left side of the equation (58) is replaced with the estimated value GAIR_EST (k) of the intake air amount, that is, the equation (59) in FIG. 27 is used as a model, and the simulated value GAIR_OS _i Is generated by the signal generator 61a as described later, and the vector ψ (k) of the intake air amount variation coefficient ψ _i as the model parameter of the equation (59) is the estimated value GAIR_EST (k) is the intake air amount GAIR. It is identified by a variable gain type recursive least square algorithm shown in equations (60) to (66) of FIG. 27 so as to coincide with (kd ′).

In the equation (60), KR (k) represents a vector of gain coefficients, and ide ′ (k) represents an identification error. Further, ψ (k) ^T in the equation (61) represents a transposed matrix of ψ (k). The identification error ide ′ (k) in the equation (60) is calculated by the equations (62) to (64) in FIG. 27, and ζ ′ (k) in the equation (63) is defined as the equation (64). This is a vector of simulated values. Further, the gain coefficient vector KR (k) is calculated by the equation (65) in FIG. 27, and R (k) in the equation (65) is a fourth-order defined by the equation (66) in FIG. It is a square matrix.

As described above, in this adaptive observer 61, the vector ψ (k) of the intake air amount variation coefficient ψ _i is identified by the sequential least square algorithm shown in the above equations (60) to (66). Thereby, the noise fluctuation component of the intake behavior caused by such that the operating conditions of the engine 3 suddenly changes, can be removed from the intake air amount variation coefficient [psi _i (filtering), the intake air amount variation coefficient [psi _i, cylinder It can be calculated as a value that substantially shows the variation in the intake air amount during the period.

The configuration of the adaptive observer 61 described above is as shown in the block diagram of FIG. 28 in the same manner as the adaptive observer 31 of the first air-fuel ratio controller 30 described above. That is, as shown in the figure, in this adaptive observer 61, the signal generator 61a generates a vector ζ ′ (k) of the simulated value GAIR_OS _i . More specifically, in the signal generator 61a, as shown in FIG. 29, the simulated value GAIR_OS _i is obtained by alternately combining triangular waves, trapezoidal waves, or the like so that the sum of the values is always 1. Generated as a signal value. Further, in the multiplier 61b, an estimated value GAIR_EST (k) of the intake air amount is generated as a value obtained by multiplying the vector ζ ′ (k) of the simulated value by the vector ψ (k−1) of the intake air amount variation coefficient. The Then, the difference 61d generates an identification error ide '(k) as a deviation between the intake air amount GAIR (kd') and the estimated value GAIR_EST (k).

  Further, the logic unit 61e generates a gain coefficient vector KR (k) based on the simulated value vector ζ ′ (k). In the multiplier 61f, the identification error ide ′ (k) and the gain coefficient vector are generated. A product [ide ′ (k) · KR (k)] of KR (k) is generated. Next, the adder 61g uses the product [ide '(k) · KR (k)] and the delayed intake air amount variation coefficient vector ψ (k-1) as the sum of the intake air amount variation coefficient vector ψ. (K) is generated.

Next, an algorithm for calculating the intake air amount variation correction coefficient KICYL _i ( third correction value ) in the intake air amount variation correction coefficient calculating unit 62 will be described. In the intake air amount variation correction coefficient calculating section 62, first, an intake air amount variation coefficient vector ψ (k) calculated by the adaptive observer 61, that is, four intake air amount variation coefficients ψ ₁ ( k) to Ψ ₄ (k), the moving average value Ψave (k) of the intake air amount variation coefficient is calculated. Next, in order to converge the intake air amount variation coefficient ψ _i (k) to the moving average value ψave (k), the intake air amount variation correction coefficient KICYL _i is determined by the I-PD control (proportional / differential precedence type PID control) cylinder. Calculate every time. This I-PD control algorithm is shown in equations (68) and (69) in FIG.

Here, the air-fuel ratio control for correcting the variation in the intake air amount between the cylinders by the third air-fuel ratio controller 60 is the air-fuel ratio control for correcting the variation in the air-fuel ratio between the cylinders by the first air-fuel ratio controller 30. In order to avoid this, in the two controllers 30, 60, the convergence speed of the air-fuel ratio variation coefficient Φ _i (k) to the moving average value Φave and the intake air amount variation coefficient are avoided. It is necessary to make the convergence speed of Ψ _i (k) to the moving average value Ψave (k) different from each other.

In the present embodiment, the feedback gains FI ′, GI ′, and HI ′ in the above formula (68) are larger in absolute value than the absolute values of the feedback gains FI, GI, and HI in the above-described formula (11). Is set to be That is, each feedback gain is set so that the relationship of 0 <| FI | <| FI ′ |, 0 <| GI | <| GI ′ |, 0 <| HI | <| HI ′ | Thereby, the convergence speed of the intake air amount variation coefficient Ψ _i (k) to the moving average value Ψave (k) is faster than the convergence speed of the air-fuel ratio variation coefficient Φ _i (k) to the moving average value Φave. The air-fuel ratio is controlled. This is because the air flow sensor 9 has a better S / N ratio than the LAF sensor 14, so that the air-fuel ratio control by the two controllers 30 and 60 can be mutually performed by setting the feedback gains as described above. This is because the stability of the air-fuel ratio control can be ensured as a whole while avoiding interference.

In addition, since the intake air amount variation correction coefficient KICYL _i is calculated by the I-PD control algorithm so that the intake air amount variation coefficient Ψ _i (k) converges to the moving average value Ψave (k), In the convergence behavior of the intake air amount variation coefficient ψ _i (k) to the moving average value ψave (k), control can be performed so that no overshoot occurs. Thus, when the third air-fuel ratio controller 60 executes air-fuel ratio control for correcting variation in the air-fuel ratio between the cylinders, the second air-fuel ratio controller is changed by the change in the collective portion air-fuel ratio by the air-fuel ratio control. It can be avoided that the controllability of the air-fuel ratio control by 40 is lowered.

Next, an algorithm for calculating the learning correction value KICYL_LS _i of the intake air amount variation correction coefficient KICYL _{i by} the learning correction value calculation unit 63 will be described. The intake air amount variation correction coefficient KICYL _i is easily affected by the operating state of the engine 3, and changes accordingly when the operating state of the engine 3 changes. FIG. 31 shows the relationship between the exhaust gas volume ESV (k) as an operation state parameter representing the operation state of the engine 3 and the intake air amount variation correction coefficient KICYL _i (k).

Referring to FIG. 31, in this intake air amount variation correction coefficient KICYL _i (k), similarly to the air-fuel ratio variation correction coefficient KOBSV _i (k) described above, the intake air amount variation correction coefficient KICYL _i (k) is used as a dependent variable. It can be seen that an approximate value, that is, an estimated value of the intake air amount variation correction coefficient KICYL _i (k) can be calculated by a linear expression using the exhaust gas volume ESV (k) as an independent variable. Therefore, the learning correction value calculation unit 63 defines the learning correction value KICYL_LS _i (k) of the intake air amount variation correction coefficient as an estimated value calculated by the regression equation shown in the equation (70) of FIG. A vector of coefficient AICYL_LS _i and constant term BICYL_LS _i (hereinafter referred to as “regression coefficient vector”) θICYL_LS _i (k) is calculated by the sequential least square method shown in equations (71) to (77) of FIG.

In this equation (71), KU _i (k) represents a vector of gain coefficients, and Eic _i (k) represents an error. The error Eic _i (k) is calculated by the equation (73) in FIG. Further, the gain coefficient vector KU _i (k) is calculated by the equation (76) in FIG. 32, and U _i (k) in the equation (76) is defined by the equation (77) in FIG. The following square matrix.

Further, the learning correction value KICYL_LS _i (k) is specifically calculated by the equation (78) in FIG. As will be described later, when the engine 3 is in an extreme operating state or operating environment, the calculation of the regression coefficient AICYL_LS _i and the constant term BICYL_LS _i by the above-described sequential least square method is avoided, and the previous regression coefficient vector is calculated. The value θICYL_LS _i (k−1) is used as the current value θICYL_LS _i (k) in the calculation of the learning correction value KICYL_LS _i (k).

In the learning correction value calculation unit 63, the learning correction value KICYL_LS _i (k) converges to the product of this and the intake air amount variation correction coefficient KICYL _i (k) by the algorithm shown in the above equations (71) to (78). To be calculated.

  As shown in FIG. 25, even when the intake pipe absolute pressure PBA is detected by the intake pipe absolute pressure sensor 11, the pulsation of the intake air can be detected. Therefore, in the above equations (58) to (78), By using the algorithm in which the parameter represented by “GAIR” is replaced with the parameter represented by “PBA” and the intake pipe absolute pressure PBA detected by the intake pipe absolute pressure sensor 11, the intake air amount between the cylinders is used. It is possible to configure an air-fuel ratio controller for correcting variations in the above.

  Hereinafter, the air-fuel ratio control process in the second embodiment will be described with reference to FIGS. FIG. 33 shows the main routine of this control process, and this process is interrupted in synchronization with the input of the TDC signal. As shown in the figure, since this process is the same as steps 1 to 13 in FIG. 12 of the first embodiment, steps other than steps 111 to 113 are described here, focusing on steps 111 to 113. To do.

That is, after the learning correction value KOBSV_LS _i of the air-fuel ratio variation correction coefficient is calculated in step 110, the vector ψ of the intake air amount variation coefficient is calculated in step 111, as will be described later.

Next, the routine proceeds to step 112, and after calculating the intake air amount variation correction coefficient KICYL _i , the routine proceeds to step 113, where the learning correction value KICYL_LS _i of the intake amount variation correction coefficient is calculated. Next, similarly to steps 11 to 13 described above, after executing steps 114 to 116, the present process is terminated.

  Next, the processing for calculating the vector ψ of the intake air amount variation coefficient in step 111 will be described with reference to FIG. In this process, the vector ψ of the intake air amount variation coefficient is calculated by the same method as the calculation process of the vector φ of the air-fuel ratio variation coefficient in FIG. That is, in step 120, the calculated value of the vector ψ of the intake air amount variation coefficient in the previous loop stored in the RAM is set as the previous value RRVψ [= ψ (k−1)].

Next, the process proceeds to step 121, where the simulation current value GAIR_OS _i is calculated. Then, the process proceeds to step 122, where the simulation value vector ζ ′ is calculated by the equation (64) in FIG.

  Next, the process proceeds to step 123, and after calculating the intake air amount estimated value GAIR_EST from the equation (63) in FIG. 27, the process proceeds to step 124, and the identification error ide ′ is calculated from the equation (62) in FIG. Is calculated.

  Next, the process proceeds to step 125, where the calculated value of the next value NEXR [= R (k + 1)] of the square matrix in the previous loop stored in the RAM is set as the current value R, and then the process proceeds to step 126. The gain coefficient vector KR is calculated by the above-described equation (65) in FIG.

  Next, the process proceeds to step 127, and after calculating the vector ψ of the intake air amount variation coefficient by the above-described equation (60) in FIG. 27, the process proceeds to step 128, and the square matrix is calculated by the above-described equation (66) in FIG. The next value NEXR [= R (k + 1)] is calculated.

  Next, the process proceeds to step 129, and the time series data of a predetermined number (in this embodiment, 12) of intake air amounts GAIR stored in the RAM are updated. Specifically, each value of the intake air amount GAIR in the RAM is set as an old value for one control cycle (for example, the current value GAIR (k) is set as the previous value GAIR (k−1), and the previous value is set. GAIR (k-1) is set as the previous value GAIR (k-2), respectively). Then, this process is complete | finished.

Next, the calculation process of the intake air amount variation correction coefficient KICYL _i in step 112 will be described with reference to FIG. In this process, the intake air amount variation correction coefficient KICYL _i is calculated by the same method as the calculation process of the air-fuel ratio variation correction coefficient KOBSV _i in FIG. That is, first, in step 140, the moving average value Ψave of the intake air amount variation coefficient is calculated by the aforementioned equation (67) in FIG.

Next, the process proceeds to step 141, where the tracking error e ′ is calculated by the above-described equation (69) in FIG. 30, and then, in step 142, the integrated value Σe ′ of the tracking error is calculated. Next, the routine proceeds to step 143, where the moving average value Ψave of the intake air amount variation coefficient calculated in steps 140 and 142 and the integral value Σe ′ of the tracking error are used, and the air-fuel ratio variation is expressed by the above-described equation (68) in FIG. After calculating the correction coefficient KICYL _i , this process is terminated.

Next, the calculation processing of the learning correction value KICYL_LS _i of the intake air amount variation correction coefficient in the step 113 will be described with reference to FIG. In this process, by the same method as calculating process of the learning correction value KOBSV_LS _i of the air-fuel ratio variation correction coefficient of FIG. 17 described above, the learning correction value KICYL_LS _i of the intake air amount variation correction coefficient is calculated. That is, first, in step 150, the exhaust gas volume ESV is calculated by the above-described equation (13) in FIG.

Next, the process proceeds to step 151, where the calculated value of the regression coefficient vector θICYL_LS _{i in} the previous loop stored in the RAM is set as the previous value PRVθICYL_LS _i [= θICYL_LS _i (k−1)].

Next, the routine proceeds to step 152, where the learning correction value KICYL_LS _i is calculated by the equation (78) shown in FIG. Thereafter, the process proceeds to step 153, where it is determined whether or not any of the following five conditions (a6) to (a10) is satisfied.
(A6) The engine water temperature TW is higher than a predetermined lower limit value TWICYL and lower than a predetermined upper limit value TWICYH.
(A7) The intake air temperature TA is higher than a predetermined lower limit value TAICYL and lower than a predetermined upper limit value TAICYH.
(A8) The engine speed NE is higher than a predetermined lower limit value NEICYL and lower than a predetermined upper limit value NEICYH.
(A9) The intake pipe absolute pressure PBA is higher than a predetermined lower limit value PBICYL and lower than a predetermined upper limit value PBICYH.
(A10) The vehicle speed VP is higher than a predetermined lower limit value VPICYL and lower than a predetermined upper limit value VPICYH.

When all of the above five conditions (a6) to (a10) are satisfied, it is determined that the regression coefficient vector θICYL_LS _i is to be calculated by the successive least squares method, and the process proceeds to step 154 and described above. The vector Z ′ of the exhaust gas volume is calculated from the equation (75) in FIG.

Next, the process proceeds to step 155, and after calculating the error Eic _i by the above-described equation (73) in FIG. 32, the process proceeds to step 156 and the next value NEXU of the square matrix in the previous loop stored in the RAM. The calculated value of _i [= U _i (k + 1)] is set as the current value U _i .

Next, the process proceeds to step 157, and after calculating the gain coefficient vector KU _i by the above-described equation (76) in FIG. 32, the process proceeds to step 158, and the regression coefficient vector θICYL_LS _i by the above-described equation (71) in FIG. Is calculated. Next, the process proceeds to step 159, and the next value NEXU _i [= U _i (k + 1)] of the square matrix is calculated by the above-described equation (77) of FIG.

On the other hand, if the determination result in step 153 is NO and at least one of the five conditions (a6) to (a10) is not satisfied, the process proceeds to step 160 and the previous value of the regression coefficient vector set in step 151 is obtained. After setting PRVθICYL_LS _i to the current value θICYL_LS _i , the present process is terminated. Thus, for example, in the processing of step 151 in the next loop, the value calculated by the sequential least square method of steps 154 to 159 in the current loop is used as the previous value PRVθICYL_LS _i of the regression coefficient vector.

As described above, according to the control device 1 of the second embodiment, the first air-fuel ratio controller 30 calculates the air-fuel ratio variation correction coefficient KOBSV _i and the learning correction value KOBSV_LS _i , and the second air-fuel ratio controller 40 A feedback correction coefficient KSTR is calculated. Further, the third air-fuel ratio controller 60 calculates the intake air amount variation coefficient ψ _i , calculates the intake air amount variation correction coefficient KICYL _i so that it converges to the moving average value Ψ ave, and calculates its learning correction value KICYL_LS _i. Is done. Then, the basic fuel injection amount TIBS is corrected by the calculated feedback correction coefficient KSTR, air-fuel ratio variation correction coefficient KOBSV _i , learning correction value KOBSV_LS _i , intake air amount variation correction coefficient KICYL _i and learning correction value KICYL_LS _i . The final fuel injection amount TOUT _i is calculated for each cylinder.

In the adaptive observer 61 of the third air-fuel ratio controller 60, the estimated value GAIR_EST of the intake air amount GAIR is calculated from a model [Equation (59)] defined by the simulated value GAIR_OS _i and the intake air amount variation coefficient ψ _i . Further, the intake air amount variation coefficient Ψ _i as a model parameter is identified by the sequential least square method so that the estimated value GAIR_EST matches the intake air amount GAIR. Thus, the noise fluctuation component of the intake behavior caused by such that the operating conditions of the engine 3 suddenly changes, can be removed from the intake air amount variation coefficient [psi _i (filtering), the intake air amount variation coefficient [psi _i, cylinder It can be calculated as a value that substantially shows the variation in the air-fuel ratio between the two. Therefore, since the basic fuel injection amount TIBS is corrected for each cylinder by the intake air amount variation correction coefficient KICYL _i calculated based on the intake air amount variation coefficient Ψ _i , the response variation of the air flow sensor 9 and the secular change, etc. Thus, even when the dynamic characteristic of the controlled object changes, the final fuel injection amount TOUT _i is set for each cylinder so as to correct the variation in the intake air amount between the cylinders while reflecting the change in the dynamic characteristic of the controlled object in the model. Can be calculated. As a result, even when the air-fuel ratio of the engine 3 having a complicated exhaust layout is controlled as in the present embodiment, the air-fuel ratio has a larger stability margin and higher robustness than the control device 1 of the first embodiment. Control can be realized, and a better catalyst purification rate can be secured.

Further, in the third air-fuel ratio controller 60, since the intake air amount variation correction coefficient KICYL _i is calculated by the I-PD control algorithm, overshoot occurs in the convergence behavior of the intake air amount variation coefficient ψ _{i to} the moving average value ψave. In this way, the intake air amount variation correction coefficient KICYL _i can be calculated. As a result, it is possible to correct the variation in the intake air amount between the cylinders while avoiding the behavior of the intake air amount in each cylinder from being oscillating. In addition, in the I-PD control algorithm, the convergence speed of the intake air amount variation coefficient ψ _{i to} the moving average value ψ ave is faster than the convergence speed of the air-fuel ratio variation coefficient Φ _{i to} the moving average value Φ ave. Since the values of the feedback gains FI ′, GI ′, and HI ′ are set, the air-fuel ratio control by the third air-fuel ratio controller 60 and the air-fuel ratio control by the first air-fuel ratio controller 30 interfere with each other. Can be avoided. Further, since the intake air amount variation correction coefficient KICYL _i is calculated so that the intake air amount variation coefficient ψ _i converges to the moving average value ψ ave, the air-fuel ratio control by the third air-fuel ratio controller 60 and the second air-fuel ratio controller 40 are calculated. It is possible to avoid the air-fuel ratio control by the interference with each other. As described above, the air-fuel ratio control by the third air-fuel ratio controller 60, the air-fuel ratio control by the first air-fuel ratio controller, and the air-fuel ratio control by the second air-fuel ratio controller 40 are avoided while interfering with each other. Variations in air volume can be corrected.

Further, in the third air-fuel ratio controller 60, the learning correction value KICYL_LS _{i of} the intake air amount variation correction coefficient KICYL _i is calculated by a regression equation [Equation (78)] with the exhaust gas volume ESV as an independent variable. Since the regression coefficient vector θICYL_LS _i that is a vector of the regression coefficient AICYL_LS _i and the constant term BICYL_LS _{i in} the equation is calculated by the sequential least square method, the engine 3 is in an operating state that changes rapidly, such as a transient operating state. Thus, even when the variation state of the intake air amount between the cylinders suddenly changes due to the influence, the learning correction value KICYL_LS _i can be calculated as a value appropriately reflecting the variation state of the intake air amount between the cylinders. . As a result, even when the engine 3 is in a transient operation state, it is possible to appropriately control the air-fuel ratio while compensating for variations in the intake air amount between the cylinders.

Further, since the intake air amount variation coefficient ψ _i and the regression coefficient vector θICYL_LS _i are calculated by the sequential least square method, unlike the case of using a general least square method as a statistical algorithm, even at the start of air-fuel ratio control, The intake air amount variation correction coefficient KICYL _i and the learning correction value KICYL_LS _i can be calculated for each control cycle. Therefore, for example, by setting the initial values of KICYL _i and KICYL_LS _i in advance, at the start of air-fuel ratio control, the intake air amount variation correction coefficient KICYL _i calculated for each control cycle and the learning correction value KICYL_LS _i The final fuel injection amount TOUT _i can be calculated as a value always corrected by the product, and controllability at the start of air-fuel ratio control can be improved. Thereby, the catalyst purification rate at the start of air-fuel ratio control can be further improved.

  Note that in the air-fuel ratio control by the third air-fuel ratio controller, the amount of fluctuation of the absolute value may be very large compared to the detected air-fuel ratio KACT as a characteristic of the intake air amount GAIR. In the adaptive observer 61, the fluctuation amount of the identification value of the intake air amount variation coefficient vector ψ (k) identified by the identification algorithms of the equations (60) to (66) becomes very large, so that the control system becomes unstable. There is a risk of becoming. To avoid this, an adaptive observer 61 may be configured as shown in FIG. That is, in the adaptive observer 61, a filter 61j (filter means) comprising any one of the high-pass filters [1] and [2] shown in FIG. 38 and the band-pass filters [1] to [3] shown in FIG. The air-fuel ratio control may be executed using the filter value GAIR_F (k) obtained by filtering the intake air amount GAIR (k) by the filter 61j.

This filter 61j is represented by the equation (79) in FIG. In the formula (79), m ^* and n ^* represent a predetermined integer. Further, the identification algorithm of the vector ψ (k) of the intake air amount variation coefficient in the adaptive observer 61 is as shown in equations (80) to (86) in FIG. With the above configuration, the intake air amount is secured while ensuring the information necessary for identifying the intake air amount variation coefficient vector ψ (k) even in an operating state in which the intake air amount GAIR (k) varies greatly. The filter value GAIR_F (k) can be generated as a value in which the fluctuation range of the air amount GAIR (k) is suppressed. Therefore, by identifying the vector ψ (k) of the intake air amount variation coefficient according to such a filter value GAIR_F (k), the identification delay can be suppressed, the identification accuracy can be increased, and the air-fuel ratio control Stability and quick response can be further improved.

The second embodiment is an example in which a linear expression is used as a regression expression used for calculating the learning correction value KICYL_LS _i , but the regression expression is not limited to this, and an n-order expression (n is an integer of 2 or more). But you can. Even in such a case, the same effect as that of the second embodiment can be obtained by calculating the regression coefficient and the constant term of the nth order equation by the sequential least square method. Further, the learning correction value KICYL_LS _i may be calculated by using a predetermined value set in advance for each of a plurality of operation regions as the regression coefficient and constant term of the regression equation. In this way, the calculation time of the learning correction value KICYL_LS _i can be shortened, and the calculation load on the ECU 2 can be reduced.

Furthermore, it goes without saying that the control algorithm for converging the intake air amount variation coefficient ψ _i to the moving average value ψ ave is not limited to the I-PD control algorithm of the second embodiment, and other control algorithms may be used. . For example, instead of the I-PD control algorithm, an IP-D control algorithm represented by equations (87) and (88) in FIG. 41 or a response designating control algorithm represented by equations (89) to (91) in FIG. By using this, the intake air amount variation correction coefficient KICYL _i may be calculated. Even when these control algorithms are used, as in the case of using the I-PD control algorithm of the embodiment, overshoot does not occur in the convergence behavior of the intake air amount variation coefficient Ψ _{i to} the moving average value Ψ ave. Thus, the intake air amount variation correction coefficient KICYL _i can be calculated, and as a result, the variation in the intake air amount between the cylinders can be corrected quickly and appropriately.

As described above, the moving average value Ψave of the intake air amount variation coefficient ψ _i (k) is also used in the calculation of the intake air amount variation correction coefficient KICYL _i when the IP-D control algorithm or the response designating control algorithm is used. As described above, by setting the values of the feedback gains and the switching function setting parameters so that the convergence speed of the air-fuel ratio becomes faster than the convergence speed of the air-fuel ratio variation coefficient Φ _i (k) to the moving average value Φave. In addition, the air-fuel ratio control by the third air-fuel ratio controller 60 and the air-fuel ratio control by the first air-fuel ratio controller 30 can be prevented from interfering with each other. In addition, each feedback gain may be determined based on optimal regulator theory or H _∞ control theory. In this way, in the convergence behavior of the intake air amount variation coefficient Ψ _{i to} the moving average value Ψ ave, the occurrence of overshoot can be more effectively suppressed, and as a result, the accuracy of correcting the variation in intake air amount between cylinders can be improved. Further improvement can be achieved.

Furthermore, when the settling time until the intake air amount variation coefficient ψ _i converges to the moving average value ψ ave may be long, the intake air amount variation correction coefficient KICYL _i may be calculated by the PID control algorithm described above. Further, the average value of the intake air amount variation coefficient as a target value for converging the intake air amount variation coefficient ψ _i is not limited to the moving average value ψ ave of the second embodiment, but may be a weighted average value or the like.

  Further, the adaptive observer 61 of the third air-fuel ratio controller 60 of the second embodiment uses the variable gain type sequential least square method shown in equations (60) to (66) of FIG. In the example in which (k) is identified, for example, the δ correction method shown in the equations (92) to (99) in FIG. 42 is applied as the identification algorithm of the vector ψ (k) of the intake air amount variation coefficient in the adaptive observer 61. The vector ψ (k) of the intake air amount variation coefficient may be identified by the fixed gain method.

  In the equation (92) in the figure, ψbase is a reference value vector (reference value) defined as in the equation (93), and the reference values ψbase1 to ψbase4 which are four elements of this vector are the exhaust gas volume. It is calculated by searching the table shown in FIG. 43 according to the ESV. Further, dψ (k) in the equation (92) is a correction term (correction component) defined as in the equation (52), and is calculated by the equations (94) to (99).

  When the vector ψ (k) of the intake air amount variation coefficient is identified by the fixed gain method to which the above δ correction method is applied, the calculation time can be shortened compared to the case of the sequential least square method, and the calculation in the ECU 2 The load can be reduced. As a result, the ECU 2 can be reduced in size and cost. In addition, since the identification value of the vector ψ (k) can be constrained in the vicinity of the value 1, intake air between the cylinders can be obtained even when the engine 3 is in an operating state in which the air-fuel ratio changes drastically, such as a transient operating state. The vector ψ (k) of the intake air amount variation coefficient representing the variation in the amount can be quickly and appropriately calculated as a value appropriately reflecting the behavior of the air-fuel ratio, and the stability of the air-fuel ratio control is improved. Can do.

  Moreover, although each above embodiment is an example which applied the control apparatus of this invention to the control apparatus which controls the air fuel ratio of the engine 3 for vehicles, the control apparatus of this invention is not restricted to this, For ships Needless to say, the present invention is also applicable to other internal combustion engines and other industrial equipment.

1 is a diagram illustrating a schematic configuration of a control device according to a first embodiment of the present invention and an internal combustion engine to which the control device is applied. FIG. It is a block diagram of the control apparatus of 1st Embodiment. It is a schematic diagram for demonstrating the calculation algorithm of the air fuel ratio variation coefficient (PHI) in the adaptive observer of a 1st air fuel ratio controller. It is a figure which shows the numerical formula of the calculation algorithm of the air fuel ratio variation coefficient (PHI) in an adaptive observer. It is a block diagram which shows the structure of an adaptive observer. It is a figure which shows the simulation value KACT_OS output from the signal generator of an adaptive observer. It is a figure which shows the numerical formula of the I-PD control algorithm used for calculation of the air-fuel ratio variation correction coefficient KOBSV. (A) A diagram showing the relationship between the exhaust gas volume ESV and the air-fuel ratio variation correction coefficient KOBSV, and (b) a diagram showing the relationship between the exhaust gas volume ESV, the air-fuel ratio variation correction coefficient KOBSV, and the learning correction value KOBSV_LS. It is a diagram showing a formula of the air-fuel ratio variation correction coefficient calculation algorithm of the learning correction value KOBSV_LS _i of. It is a figure which shows the numerical formula for demonstrating the calculation algorithm of the feedback correction coefficient KSTR in a 2nd air fuel ratio controller. It is a figure which shows the numerical formula of the calculation algorithm of the feedback correction coefficient KSTR in a 2nd air fuel ratio controller. It is a flowchart which shows an air fuel ratio control process. It is a flowchart which shows the calculation process of the model parameter vector (theta) in step 6 of FIG. It is a flowchart which shows the KSTR calculation process in step 7 of FIG. 13 is a flowchart showing a calculation process of an air-fuel ratio variation coefficient vector φ in step 8 of FIG. 12. 13 is a flowchart showing a calculation process of an air-fuel ratio variation correction coefficient KOBSV _i in step 9 of FIG. 13 is a flowchart showing a calculation process of a learning correction value KOBSV_LS _i of an air-fuel ratio variation correction coefficient in step 10 of FIG. It is a timing chart which shows the operation example of the air fuel ratio control by the control apparatus of 1st Embodiment. It is a timing chart which shows the comparative example of the operation | movement of air fuel ratio control. PID control algorithm, a diagram showing a calculation formula for air-fuel ratio variation correction coefficient KOBSV _i by IP-D control algorithm and the response-specifying control algorithm. It is a figure which shows the identification algorithm of the vector (phi) of the air-fuel ratio variation coefficient by the fixed gain method which applied (delta) correction method. Is a diagram showing an example of a table used for calculation of Φbase _i. It is a block diagram which shows the structure centering on the 3rd air fuel ratio controller in the control apparatus which concerns on 2nd Embodiment of this invention. It is a block diagram which shows the structure centering on the 1st and 2nd air fuel ratio controller in the control apparatus of 2nd Embodiment. It is a figure which shows the pulsation of the intake air detected by an airflow sensor. It is a schematic diagram for demonstrating the calculation algorithm of the intake air quantity variation coefficient (PSI) in the adaptive observer of a 3rd air fuel ratio controller. It is a figure which shows the numerical formula of the calculation algorithm of the intake air amount variation coefficient (PSI) in an adaptive observer. It is a block diagram which shows the structure of an adaptive observer. It is a figure which shows the simulation value GAIR_OS output from the signal generator of an adaptive observer. It is a figure which shows the numerical formula of the I-PD control algorithm used for calculation of the intake air amount variation correction coefficient KICYL. It is a figure which shows the relationship between the exhaust gas volume ESV, the intake air amount variation correction coefficient KICYL, and the learning correction value KICYL_LS. It is a figure which shows the numerical formula of the calculation algorithm of the learning correction value KICYL_LS of an intake air amount variation correction coefficient. It is a flowchart which shows the air fuel ratio control process performed by the control apparatus which concerns on 2nd Embodiment. FIG. 35 is a flowchart showing a calculation process of an intake air amount variation coefficient vector ψ in step 111 of FIG. 34. FIG. FIG. 35 is a flowchart showing a calculation process of an intake air amount variation correction coefficient KICYL _i in step 112 of FIG. 34. FIG. FIG. 35 is a flowchart showing a calculation process of a learning correction value KICYL_LS _i of an intake air amount variation correction coefficient in step 113 of FIG. 34. FIG. It is a block diagram which shows the structure of the modification of the adaptive observer of a 3rd air fuel ratio controller. It is a figure which shows an example of the filter in the modification of an adaptive observer. It is a figure which shows the other example of the filter in the modification of an adaptive observer. It is a figure which shows the numerical formula showing the filter in the modification of an adaptive observer, and the numerical formula of the calculation algorithm of intake air amount variation coefficient (PSI). In the calculation of the intake air amount variation correction coefficient KICYL _i, is a diagram showing a formula in the case of using the IP-D control algorithm or response-specifying control algorithm. It is a figure which shows the identification algorithm of the vector (psi) of the intake air amount variation coefficient by the fixed gain method to which (delta) correction method is applied. Is a diagram showing an example of a table used for calculation of Ψbase _i.

Explanation of symbols

1 control device 2 ECU (simulated value generating means, estimating means, identifying means, the learning correction value calculating means, delaying means, fuel amount determining means, the first correction value calculation means, first fuel amount correcting means, second compensation values calculating means, second fuel amount correcting means, operating condition parameter-detecting means, the third compensation value calculating means, third fuel amount correcting means, the fourth correction value calculating means, the fourth fuel amount correcting means)
3 Internal combustion engine
4a Main pipe part of intake pipe (one intake passage)
4c Intake pipe assembly (one intake passage)
4d intake pipe branch (multiple intake passages)
7c-7f Exhaust pipe part of exhaust pipe (multiple exhaust passages)
7j Exhaust pipe assembly (one exhaust passage)
9 Air flow sensor ( intake air amount detection means)
11 Intake pipe absolute pressure sensor ( operating state parameter detecting means, intake air amount detecting means)
13 crank angle sensor (operating state parameter detecting means)
14 LAF sensor ( air-fuel ratio detection means)
20 basic fuel injection amount calculation unit (fuel amount determination means)
30 1st air fuel ratio controller (1st fuel quantity correction means)
31 Adaptive observer (simulated value generation means, estimation means, identification means, delay means)
32 Air-fuel ratio variation correction coefficient calculation unit ( first correction value means)
33 learning correction value calculation unit (learning correction value calculation means)
40 Second air-fuel ratio controller ( second correction value calculating means, second fuel amount correcting means, fourth correction value calculating means, fourth fuel amount correcting means )
60 third air-fuel ratio controller (third fuel amount correcting means)
61 Adaptive observer (simulated value generation means, estimation means, identification means, delay means)
61j Filter (filter means)
62 Intake air amount variation correction coefficient calculation unit ( third correction value calculation means )
63 Learning correction value calculation unit (learning correction value calculation means)
# 1 to # 4 1st to 4th cylinders (multiple cylinders)
KACT detection air-fuel ratio ( air-fuel ratio detection value)
KACT_OS _i 4 simulated values (predetermined multiple simulated values)
Estimate of KACT_EST detected air-fuel ratio (the estimated value of the air-fuel ratio detected value)
Φ _i Four air-fuel ratio variation coefficients (multiple model parameters)
Φave Moving average value (average value)
e Deviation φbase reference value vector (model parameter reference value)
dφ correction term vector (correction component)
KOBSV _i air-fuel ratio variation correction coefficient ( first correction value )
KOBSV_LS _i learning correction value of air-fuel ratio variation correction coefficient (learning value of first correction value)
AOBSV_LS _i regression coefficient BOBSV_LS _i constant term KCMD target air-fuel ratio (predetermined target value)
KSTR feedback correction coefficient ( second correction value, fourth correction value )
ESV Exhaust gas volume ( operating condition parameter )
TIBS basic fuel injection amount (determined fuel amount)
d Dead time (predetermined delay time)
GAIR intake air volume ( intake volume detection value)
GAIR_OS _i 4 simulated values (predetermined multiple simulated values)
GAIR_EST intake air amount estimated value (the estimated value of the intake air quantity sensing value)
GAIR_F Intake air amount filter value ( Intake amount detection value filter value)
Ψ _i Four intake air quantity variation coefficients (multiple model parameters)
Ψave moving average value (average value)
e 'deviation ψbase reference value vector (model parameter reference value)
dψ correction term vector (correction component)
KICYL _i intake air amount variation correction coefficient ( third correction value )
Learning correction value of KICYL_LS _i intake air amount variation correction coefficient (learned value of the third correction value)
AICYL_LS _i regression coefficient BICYL_LS _i constant term
d 'dead time (predetermined delay time)
PBA Intake pipe absolute pressure ( Intake amount detection value)

Claims (21)

An air-fuel ratio detecting means provided in an exhaust passage of the internal combustion engine for detecting an air-fuel ratio detection value representing an air-fuel ratio of the exhaust gas in the exhaust passage ;
Fuel amount determining means for determining the amount of fuel supplied to the cylinder of the internal combustion engine;
A simulation value generating means for generating a predetermined simulation value based on the crank angle, which presupposes a behavior corresponding to the crank angle of the internal combustion engine of the air-fuel ratio of the exhaust gas discharged from the cylinder ;
Estimating means for estimating an estimated value of the air-fuel ratio detected value based on a model defining a relationship between the estimated value and the simulated value;
The so estimated estimated value matches to the detected air-fuel ratio detection value, in response to the detected air-fuel ratio detection value and the generated simulated value, identifying the model parameters of the model on board Identification means to
First correction value calculating means for calculating a first correction value for correcting the fuel amount using the identified model parameter ;
First fuel amount correcting means for correcting the fuel amount using the calculated first correction value;
A control device for an internal combustion engine, comprising: The cylinder is composed of a plurality of cylinders,
The exhaust passage includes a plurality of exhaust passages extending from the plurality of cylinders and a single exhaust passage in which the plurality of exhaust passages are gathered.
The air-fuel ratio detection means is provided in the one exhaust passage,
The fuel amount determining means determines the fuel amount as a fuel amount supplied to each of the plurality of cylinders,
The simulated value generating means generates a plurality of simulated values that presuppose a behavior corresponding to a crank angle of the internal combustion engine of the air-fuel ratio of exhaust gas discharged from the plurality of cylinders as the simulated value,
The model includes a model that defines a relationship between the estimated value of the air-fuel ratio detection value and the plurality of simulated values,
The identification means identifies a plurality of model parameters on-board as the model parameters in accordance with the air-fuel ratio detection value and the plurality of simulated values,
Second correction value calculation means for calculating a second correction value for correcting the amount of fuel supplied to each cylinder so that the air-fuel ratio detection value converges to a predetermined target value;
Second fuel amount correction means for correcting the amount of fuel supplied to each cylinder according to the calculated second correction value;
Further comprising
The first correction value calculating means calculates the first correction value for each cylinder so that the plurality of identified model parameters converges to an average value of the plurality of identified model parameters. The control device for an internal combustion engine according to claim 1. A learning correction value calculating means for calculating a learning correction value of the first correction value for each cylinder by using a sequential statistical algorithm;
3. The control device for an internal combustion engine according to claim 2 , wherein the first fuel amount correction means corrects the fuel amount for each cylinder further according to the calculated learning correction value . An operation state parameter detecting means for detecting an operation state parameter representing the operation state of the internal combustion engine;
The learning correction value calculating means, the pre-Symbol learning correction value, to calculate the regression equation for the operating state parameters of the learning correction value is dependent variable Toshikatsu the detection and independent variables, the regression coefficient of the regression equation and 4. The control apparatus for an internal combustion engine according to claim 3, wherein a constant term is calculated by the sequential statistical algorithm. The first correction value calculating means, either the correction value component first included in the correction value, of claims 2 to 4, characterized in that calculated on the basis of the deviation between the model parameters and a predetermined target value A control device for an internal combustion engine according to claim 1. The first correction value calculating means, an internal combustion according to claim 5, characterized in that the other correction value components other than the correction value component included in the first correction value is calculated based on the model parameter Engine control device. The first correction value calculating means, the response-specifying control algorithm, the control of the internal combustion engine according to any one of claims 2 to 5, and calculates the first correction value in response to said model parameters apparatus. The control device for an internal combustion engine according to claim 2 , wherein the identification unit identifies the model parameter by a fixed gain method. The identification means calculates a model parameter reference value according to the detected operating state parameter , and identifies the model parameter by adding a predetermined correction component to the calculated model parameter reference value. The control apparatus for an internal combustion engine according to claim 4. Delay means for delaying the air-fuel ratio detection value by a predetermined delay time;
10. The internal combustion engine according to claim 2 , wherein the identification unit identifies the model parameter in accordance with the delayed air-fuel ratio detection value and the generated plurality of simulated values. Control device. An intake air amount detecting means provided in an intake passage of the internal combustion engine for detecting an intake air amount detection value representing an intake air amount;
Fuel amount determining means for determining the amount of fuel supplied to the cylinder of the internal combustion engine;
Simulated value generation means for generating a predetermined simulated value of the intake air amount sucked into the cylinder based on the crank angle assuming a behavior corresponding to the crank angle of the internal combustion engine in advance;
Estimating means for estimating an estimated value of the intake air amount detection value based on a model defining a relationship between the estimated value and the simulated value;
On-board identification of model parameters of the model according to the detected intake air amount detection value and the generated simulated value so that the estimated estimation value matches the detected intake air amount detection value Identification means to
Third correction value calculating means for calculating a third correction value for correcting the fuel amount using the identified model parameter;
Third fuel amount correcting means for correcting the fuel amount using the calculated third correction value;
A control device for an internal combustion engine, comprising: The cylinder is composed of a plurality of cylinders,
The intake passage is composed of one intake passage and a plurality of intake passages branched from the one intake passage and extending to the plurality of cylinders, respectively.
The intake air amount detecting means is provided in the one intake passage,
The fuel amount determining means determines the fuel amount as a fuel amount supplied to each of the plurality of cylinders,
The simulated value generating means generates a plurality of simulated values that presuppose behavior corresponding to the crank angle of the internal combustion engine of the intake air amount sucked into the plurality of cylinders as the simulated value,
The model includes a model that defines a relationship between the estimated value of the intake air amount detection value and the plurality of simulated values,
The identifying means identifies, as the model parameter, a plurality of model parameters on-board according to the intake air amount detection value and the plurality of simulated values,
A plurality of exhaust passages extending from the plurality of cylinders are gathered together in one exhaust passage;
Air-fuel ratio detection means for detecting an air-fuel ratio detection value representing an air-fuel ratio of the exhaust gas in the one exhaust passage;
Fourth correction value calculating means for calculating a fourth correction value for correcting the amount of fuel supplied to each cylinder so that the detected air-fuel ratio detection value converges to a predetermined target value;
Fourth fuel amount correction means for correcting the amount of fuel supplied to each cylinder according to the calculated fourth correction value;
Further comprising
The third correction value calculating means calculates the third correction value for each cylinder so that the plurality of identified model parameters converge to an average value of the plurality of identified model parameters. The control device for an internal combustion engine according to claim 11. A learning correction value calculating means for calculating a learning correction value of the third correction value for each cylinder by using a sequential statistical algorithm;
13. The control device for an internal combustion engine according to claim 12, wherein the third fuel amount correcting means corrects the fuel amount for each cylinder further according to the calculated learning correction value. An operating state parameter detecting means for detecting an operating state parameter representing the operating state of the internal combustion engine;
The learning correction value calculating means calculates the learning correction value by a regression equation having the learning correction value as a dependent variable and the detected operating state parameter as an independent variable, and a regression coefficient and a constant of the regression equation. 14. The control apparatus for an internal combustion engine according to claim 13, wherein a term is calculated by the sequential statistical algorithm. 15. The third correction value calculating means calculates a correction value component included in the third correction value based on a deviation between the model parameter and a predetermined target value. A control device for an internal combustion engine according to claim 1. The internal combustion engine according to claim 15, wherein the third correction value calculation unit calculates a correction value component other than the correction value component included in the third correction value based on the model parameter. Engine control device. The internal combustion engine control according to any one of claims 12 to 15, wherein the third correction value calculating means calculates the third correction value according to the model parameter by a response designation control algorithm. apparatus. 18. The control apparatus for an internal combustion engine according to claim 12, wherein the identification unit identifies the model parameter by a fixed gain method. The identification means calculates a model parameter reference value according to the detected operating state parameter, and identifies the model parameter by adding a predetermined correction component to the calculated model parameter reference value. The control device for an internal combustion engine according to claim 14. A delay means for delaying the generated simulated values by a predetermined delay time;
20. The internal combustion engine according to claim 12, wherein the identification unit identifies the model parameter in accordance with the detected intake air amount detection value and the plurality of delayed simulated values. Control device. Filter means for generating a filter value of the intake air amount detection value by performing a predetermined filtering process on the detected intake air amount detection value,
20. The internal combustion engine according to claim 12, wherein the identification unit identifies the model parameter in accordance with a filter value of the generated intake air amount detection value and the plurality of simulated values. Control device.

Images