JP2017193968A - Control device of internal combustion engine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform learning processing in a form that high learning accuracy can be obtained while shortening a time necessary for learning when performing the feedback control of a fuel injection amount so that an actual ignition delay period approximates a target ignition delay period corresponding to a target air-fuel ratio.SOLUTION: A feedback control of a fuel injection amount is performed so that an actual ignition delay index value approximates a target ignition delay index value. A learning map is learnt in which a learning value (α-β)of a correction rate (α-β) being a difference between a correction rate α of the fuel injection amount by the feedback control and a correction rate β of the fuel injection amount based on a difference between a target air-fuel ratio and an actual air-fuel ratio is associated with a fuel injection amount correlation value and an intake air amount correlation value. A learning value is learnt in which a learning value βof the correction rate β is associated with an engine torque correlation value and an engine rotation number correlation value. In the feedback control, not only the correction rate α but also the sum of the learning value] (α-β)and the learning value β is reflected to the correction of the fuel injection value.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a control device for an internal combustion engine.

  For example, Patent Document 1 discloses a control device for an internal combustion engine including an in-cylinder pressure sensor. This control device uses the output value of the in-cylinder pressure sensor to perform feedback control of the fuel injection amount so that the actual crank angle period from the ignition timing to the crank angle at which a predetermined combustion mass ratio is obtained approaches the target crank angle period. Is executed.

Japanese Patent Laying-Open No. 2015-094339 JP 2005-163633 A

  The crank angle period from the ignition timing described in Patent Document 1 to the crank angle at which a predetermined combustion mass ratio is obtained is a parameter representing an ignition delay period in combustion in the cylinder of the internal combustion engine. When the fuel injection amount feedback control is performed so that the actual ignition delay period approaches the target ignition delay period corresponding to the target air-fuel ratio, the actual ignition delay period is constantly deviated from the target ignition delay period due to the following two factors. obtain. The first factor is that the actual air-fuel ratio steadily deviates from the target air-fuel ratio, and the second factor is that the combustion characteristic changes from the target characteristic (initial characteristic). In this feedback control, regardless of which of these two factors the actual ignition delay period is deviated, the correction of the fuel injection amount to bring the actual ignition delay period closer to the target ignition delay period Comprehensive execution regardless of

  Here, it is conceivable to perform a learning process in which the correction amount of the fuel injection amount in the feedback control is taken into a predetermined learning map as a learning value, and to reflect the learning value in the correction of the fuel injection amount. When the learning process is executed without special consideration for the above two factors that cause the actual ignition delay period to deviate, the deviation caused by these two factors is collectively incorporated into the learning value. As a result, each learning value held by the learning map is a collection of values based on correction amounts for different factors. For this reason, learning values that are adjacent to each other on a predetermined map axis are likely to be derived from different factors, and accurate values are calculated when interpolation or extrapolation is performed based on these learning values. Becomes difficult. In order to suppress the occurrence of such a situation, it is conceivable to increase the number of map axes to improve learning accuracy. However, since the map area to be learned increases as a result, it takes a long time to obtain a sufficient learning result, which is not practical.

  The present invention has been made to solve the above-described problems. In the case where feedback control of the fuel injection amount is performed so that the actual ignition delay period approaches the target ignition delay period corresponding to the target air-fuel ratio, learning is performed. It is an object of the present invention to provide a control device for an internal combustion engine that can be accompanied by a learning process in a manner in which a high learning accuracy is obtained while shortening the time required for the engine.

  An internal combustion engine control apparatus according to the present invention includes an ignition device that ignites an air-fuel mixture in a cylinder, a fuel injection valve that supplies fuel into the cylinder, an in-cylinder pressure sensor that detects in-cylinder pressure, and an actual air-fuel ratio. An internal combustion engine including an air-fuel ratio sensor to be detected is controlled. The control device includes index value calculation means, feedback control means, calculation means, first learning means, and second learning means. The index value calculation means calculates an actual ignition delay index value based on the output value of the in-cylinder pressure sensor. The feedback control means calculates a first correction rate of the fuel injection amount based on a difference between a target ignition delay index value corresponding to a target air-fuel ratio and the actual ignition delay index value, and the actual ignition delay index value is The feedback control of the fuel injection amount is executed so as to approach the target ignition delay index value. The calculating means calculates a second correction rate of the fuel injection amount based on a difference between the target air-fuel ratio and the actual air-fuel ratio. The first learning means performs first learning in which a learned value of a difference obtained by subtracting the second correction rate from the first correction rate is associated with a fuel injection amount correlation value and an intake air amount correlation value. Learn the map. The second learning means learns a second learning map in which the learning value of the second correction factor is associated with the engine torque correlation value and the engine rotation speed correlation value. The feedback control means reflects not only the first correction rate but also the sum of the learned value of the difference and the learned value of the second correction rate in the correction of the fuel injection amount.

  According to the present invention, the feedback control of the fuel injection amount is executed so that the actual ignition delay index value approaches the target ignition delay index value. Further, according to the present invention, learning of the first learning map and the second learning map is executed. The first learning map includes a first correction rate of the fuel injection amount based on a difference between a target ignition delay index value and an actual ignition delay index value corresponding to the target air-fuel ratio, and a difference between the target air-fuel ratio and the actual air-fuel ratio. The learning value of the difference between the fuel injection amount based on the second correction factor and the fuel injection amount correlation value and the intake air amount correlation value are associated with each other. The above difference corresponds to a correction factor for reducing a steady deviation of the actual air-fuel ratio from the target air-fuel ratio among the first correction factors of the feedback control. This deviation is based on an error of at least one of the intake air amount and the fuel injection amount. According to the first learning map, by using a learning map having the intake air amount index value and the fuel injection amount index value as the map axis, the change in the learning value with respect to the change in the map axis value is continuous (consistent). A tendency to be obtained. For this reason, it is possible to easily ensure the calculation accuracy of the learning value when performing interpolation or extrapolation. On the other hand, the second correction rate corresponds to a correction for the change in combustion characteristics in the first correction rate. It can be said that the tendency of the correction amount (second correction factor) regarding the change in the combustion characteristics is empirically easily obtained according to the engine operating condition specified by the engine torque and the engine speed. According to the second learning map, by using the engine torque index value and the engine rotation speed index value as map axes, learning when performing interpolation or extrapolation while minimizing the number of map axes is necessary. It is easy to ensure the accuracy of value calculation. As described above, according to the present invention, learning is performed by allocating the first correction factor by the factor of steady deviation of the actual ignition delay index value with respect to the target ignition delay index value, and as described above. By appropriately selecting the map axis, the learning process can be accompanied by the feedback control in such a manner that high learning accuracy can be obtained while shortening the time required for learning.

It is a figure for demonstrating the system configuration | structure of Embodiment 1 of this invention. It is a figure for demonstrating the structure of the learning map used by the learning process in Embodiment 1 of this invention. It is a flowchart showing the main routine of control performed at the time of lean burn operation in Embodiment 1 of the present invention. It is a flowchart showing the subroutine showing the process regarding SA-CA10 feedback control. It is a time chart for demonstrating the effect acquired by learning accuracy becoming high.

Embodiment 1 FIG.
A first embodiment of the present invention will be described with reference to FIGS.

[System Configuration of Embodiment 1]
FIG. 1 is a diagram for explaining a system configuration according to the first embodiment of the present invention. The system shown in FIG. 1 includes a spark ignition type internal combustion engine (a gasoline engine as an example) 10. A piston 12 is provided in the cylinder of the internal combustion engine 10. A combustion chamber 14 is formed on the top side of the piston 12 in the cylinder. An intake passage 16 and an exhaust passage 18 communicate with the combustion chamber 14.

  The intake port of the intake passage 16 is provided with an intake valve 20 that opens and closes the intake port, and the exhaust port of the exhaust passage 18 is provided with an exhaust valve 22 that opens and closes the exhaust port. The intake passage 16 is provided with an electronically controlled throttle valve 24. Each cylinder of the internal combustion engine 10 has a fuel injection valve (for example, a direct injection type fuel injection valve) 26 for supplying fuel into the combustion chamber 14 (in-cylinder), and an ignition for igniting an air-fuel mixture. A device (only the spark plug is shown) 28 is provided.

  Each cylinder incorporates a cylinder pressure sensor 30 for detecting the cylinder pressure. An air-fuel ratio sensor 32 that detects the air-fuel ratio of the gas discharged from the cylinder is incorporated in the exhaust passage 18. Furthermore, the system of the present embodiment includes a drive circuit (not shown) for driving the following various actuators as well as an electronic control unit (ECU) 40 as a control device for controlling the internal combustion engine 10.

  In addition to the in-cylinder pressure sensor 30 and the air-fuel ratio sensor 32 described above, the ECU 40 receives signals from the crank angle sensor 42 disposed near the crankshaft (not shown), and disposed near the inlet of the intake passage 16. Various sensors for acquiring the engine operating state such as the airflow sensor 44 are included. The ECU 40 is electrically connected to an accelerator position sensor 46 that detects the amount of depression of the accelerator pedal (accelerator opening) of the vehicle on which the internal combustion engine 10 is mounted.

  The actuator from which the ECU 40 outputs an operation signal includes various actuators for controlling the engine operation such as the throttle valve 24, the fuel injection valve 26, and the ignition device 28 described above. Further, the ECU 40 has a function of acquiring an output signal of the in-cylinder pressure sensor 30 by performing AD conversion in synchronization with the crank angle. Thereby, the in-cylinder pressure at an arbitrary crank angle timing can be detected within a range allowed by the AD conversion resolution. Further, the ECU 40 stores a map that defines the relationship between the crank angle and the in-cylinder volume in the memory, and can calculate the in-cylinder volume corresponding to the crank angle with reference to such a map.

[Control of Embodiment 1]
(Engine torque control)
The operating region of the internal combustion engine 10 can be specified on a two-dimensional plane with the engine torque and the engine rotation speed as axes. In the present embodiment, the operation mode is switched between stoichiometric burn operation and lean burn operation according to the operation region. The stoichiometric burn operation is an operation performed while controlling the air-fuel ratio of the air-fuel mixture to the stoichiometric air-fuel ratio, while the lean burn operation is an air-fuel mixture air-fuel ratio so that the lean air-fuel ratio becomes larger than the stoichiometric air-fuel ratio. This is an operation performed while controlling the fuel ratio.

  In the present embodiment, the engine torque is controlled so that the target torque is obtained. In this torque control, the target air-fuel ratio is determined as a value according to engine operating conditions (for example, engine load factor and engine speed). Then, an intake air amount, a fuel injection amount, and a feed forward value of the ignition timing for realizing the target torque while satisfying the target air-fuel ratio are set.

  More specifically, the target torque is basically set based on the accelerator opening obtained by the accelerator position sensor 46. The feedforward value of the intake air amount is determined as the intake air amount (target intake air amount) necessary for realizing the target torque under the target air-fuel ratio. The feedforward value of the fuel injection amount corresponds to the basic fuel injection amount, and is determined as a value necessary to satisfy the target air-fuel ratio under the target intake air amount. The feedforward value of the ignition timing corresponds to the basic ignition timing, and is basically an MBT (Minimum advance for the Best Torque) ignition timing.

ただし、上記（１）式において、θ_ｍｉｎは燃焼開始点であり、θ_ｍａｘは燃焼終了点である。 (Calculation of MFB and CAX based on this)
According to the system of the present embodiment including the in-cylinder pressure sensor 30 and the crank angle sensor 42, it is possible to acquire measured data (in-cylinder pressure waveform) of the in-cylinder pressure in synchronization with the crank angle in each cycle of the internal combustion engine 10. . The in-cylinder heat generation amount Q at an arbitrary crank angle θ can be calculated using the obtained in-cylinder pressure waveform and the first law of thermodynamics. Then, using the actually measured data (heat generation amount waveform) of the heat generation amount Q in the cylinder, the combustion mass ratio (hereinafter referred to as “MFB”) at an arbitrary crank angle θ is expressed by the following equation (1): It can be calculated according to In addition, by executing the MFB calculation process for each predetermined crank angle, it is possible to calculate MFB actual measurement data (MFB waveform) in synchronization with the crank angle. Further, when the MFB actual measurement data is obtained, the crank angle CAX when the MFB becomes the specific ratio X can be calculated.

However, in the above equation (1), θ _min is the combustion start point, and θ _max is the combustion end point.

(Feedback control of fuel injection amount using SA-CA10)
In the present embodiment, the following feedback control is executed to control the fuel injection amount during the lean burn operation. Here, the crank angle period from the ignition timing (SA) to the crank angle (10% combustion point) CA10 when the MFB becomes 10% (more specifically, obtained by subtracting the ignition timing (SA) from CA10). Difference) is referred to as “SA-CA10”. SA-CA10 is one of the ignition index values representative of the ignition delay.

  There is a certain correlation between SA-CA10 and the air-fuel ratio. More specifically, in a lean air-fuel ratio region where the air-fuel ratio is larger than the stoichiometric air-fuel ratio, there is a relationship that SA-CA10 increases as the air-fuel ratio becomes leaner. In the present embodiment, the fuel injection amount is controlled so that the actual SA-CA10, which is the calculated value of SA-CA10 based on the output value of the in-cylinder pressure sensor 30, approaches the target SA-CA10 corresponding to the target air-fuel ratio. This control is referred to as “SA-CA10 feedback control” for convenience. Specifically, according to the SA-CA10 feedback control, in the cylinder in which the actual SA-CA10 smaller than the target SA-CA10 is obtained, in order to increase the actual SA-CA10 by making the air-fuel ratio lean, Correction for reducing the fuel injection amount used in the combustion cycle is executed. On the other hand, in the cylinder in which the actual SA-CA10 larger than the target SA-CA10 is obtained, the fuel injection amount used in the next combustion cycle is reduced in order to reduce the actual SA-CA10 by enriching the air-fuel ratio. Increasing correction is performed.

(Learning process with SA-CA10 feedback control)
In the present embodiment, a learning process is executed in which the correction amount of the fuel injection amount for reducing the steady deviation of the actual SA-CA10 with respect to the target SA-CA10 is taken as a learning value in a learning map described later. By correcting the fuel injection amount based on the learning value obtained by such learning processing, the convergence of the actual SA-CA10 when the SA-CA10 feedback control is started is accelerated (that is, the response delay of this feedback control). (See FIG. 5 to be described later as an example).

  Here, there are the following two factors that cause the actual SA-CA10 to steadily deviate from the target SA-CA10 in this feedback control. The first factor is that the actual air-fuel ratio steadily deviates from the target air-fuel ratio. This phenomenon is caused by an error in at least one of the intake air amount and the fuel injection amount. This error may be caused by, for example, an error in the air flow sensor 44, a clogging of the injection hole of the fuel injection valve 26 due to deposit, or a decrease in fuel pressure from a specified value. The second factor is that the combustion characteristic changes from the target characteristic (initial characteristic). This phenomenon may be caused by, for example, individual differences (for example, variation in tumble flow or variation in the mounting angle of the spark plug), aging (for example, change in tumble flow due to deposit accumulation on the intake port), or environment. It can be caused by a change (eg, a change in the humidity of the outside air).

  In this feedback control, the fuel injection amount is corrected to reduce the divergence without distinguishing the factors causing the divergence of the actual SA-CA10. Therefore, when the learning process is executed without special consideration for the above two factors, the divergence due to these two factors is collectively taken into the learning value. As a result, each learning value held by the learning map is a collection of values based on correction amounts for different factors. For this reason, learning values that are adjacent to each other on a predetermined map axis are likely to be derived from different factors, and accurate values are calculated when interpolation or extrapolation is performed based on these learning values. Becomes difficult. In order to suppress the occurrence of such a situation, it is conceivable to increase the number of map axes to improve learning accuracy. However, since the map area to be learned increases as a result, it takes a long time to obtain a sufficient learning result, which is not practical.

  Therefore, in the present embodiment, the learning process is executed in the following manner in order to obtain high learning accuracy while shortening the time required for learning. That is, under the learning execution conditions described later, the correction rate α (%) of the fuel injection amount by SA-CA10 feedback control is learned for each divergence factor of actual SA-CA10.

  More specifically, the fuel injection amount correction factor based on the difference between the target air-fuel ratio and the actual air-fuel ratio is β (%). And this correction factor (beta) is handled as a correction | amendment part for making small deviation of real SA-CA10 resulting from the change of a combustion characteristic among the said correction factors (alpha). The reason why it is appropriate to handle in this way is as follows. That is, when the combustion characteristic changes, the correlation between the air-fuel ratio and the target SA-CA10 changes, and as a result, the actual air-fuel ratio when the actual SA-CA10 converges to the target SA-CA10 by this feedback control. The value is thought to change due to changes in combustion characteristics. For this reason, it is appropriate to learn that the correction amount (correction rate β) that appears as the change between the target air-fuel ratio and the actual air-fuel ratio in the correction rate α is derived from the change in combustion characteristics.

The correction rate β can be calculated according to the following equation (2). A / F in the equation (2) is an actual air-fuel ratio, and can be obtained using the air-fuel ratio sensor 32. A / F _tgt in the equation (2) is a target air-fuel ratio, and can be obtained from a map (not shown) as a value according to engine operating conditions (for example, engine load factor and engine speed) as described above. it can.

  On the other hand, the correction rate (α−β) obtained by subtracting the correction rate β corresponding to the change in combustion characteristics from the overall correction rate α in this feedback control is a correction rate necessary for realizing the target air-fuel ratio, more specifically. Corresponds to a correction factor for reducing the steady deviation of the actual air-fuel ratio from the target air-fuel ratio that occurs when combustion is performed with the above-described feedforward values (basic intake air amount and basic fuel injection amount). . That is, this correction factor (α−β) corresponds to the correction factor for the above-mentioned other factor (the actual air-fuel ratio constantly deviates from the target air-fuel ratio), and an error between the intake air amount and the fuel injection amount. The value is based on the sum of.

  FIGS. 2A and 2B are diagrams for explaining the configuration of a learning map used in the learning process according to Embodiment 1 of the present invention. FIG. 2A shows a learning map of the correction rate (α−β). In this learning map, the intake air amount correlation value and the fuel injection amount index value are used as map axes, and the correction rate (α−β) is associated with these values. In the present embodiment, the intake air amount acquired using the air flow sensor 44 is used as an example of the intake air amount index value, and the target (final) fuel injection amount (fuel injection valve) is used as an example of the fuel injection amount index value. The fuel injection amount indicated for H.26) is used.

FIG. 2B shows a learning map of the correction rate β. In this learning map, the engine torque index value and the engine rotational speed correlation value are used as map axes, and the correction factor β is associated with these values. In the present embodiment, an engine load factor (in-cylinder air filling factor) is used as an example of an engine torque index value, and an engine rotational speed acquired using the crank angle sensor 42 as an example of an engine rotational speed correlation value. Is used. Each learning value of these two learning maps is stored in association with the value of the map axis at each lattice point on the learning map. Then, when it is necessary to calculate the correction rate (α−β) and the learning value (α−β) _L and β _L of the position outside the lattice point, the values are the values of the adjacent lattice points. Based on the interpolation or extrapolation.

(Specific processing in Embodiment 1)
FIG. 3 is a flowchart showing a main routine of control executed during lean burn operation in the first embodiment of the present invention. This routine is started at the timing when the opening timing of the exhaust valve 22 has elapsed in each cylinder (that is, when the acquisition of the data of the in-cylinder pressure P, which is the basis of calculation of the MFB actual measurement data), and Repeated every combustion cycle.

  In the main routine shown in FIG. 2, the ECU 40 first determines whether or not the lean burn operation is being performed (step 100). This determination can be made based on, for example, whether or not the current operation region corresponds to an operation region in which lean burn operation is performed.

When it is determined in step 100 that the lean burn operation is being performed, the ECU 40 learns the fuel injection amount correction factor (α-β) used in the SA-CA10 feedback control (α-β) _L (%). Is acquired (step 102). Specifically, the ECU 40 refers to the learning map shown in FIG. 2A and uses the current intake air amount and the current combustion cycle (using interpolation or extrapolation as necessary). A learning value (α−β) _L is calculated as a value corresponding to the fuel injection amount.

Next, the ECU 40 acquires another learned value β _L (%) of the correction factor β used in the SA-CA10 feedback control (step 104). Specifically, the ECU 40 refers to the learning map shown in FIG. 2B, and uses values corresponding to the current engine load factor and engine speed (using interpolation or extrapolation as necessary). As a result, a learning value β _L is calculated.

Next, the ECU 40 corrects the fuel injection amount used in the next combustion cycle (step 106). Specifically, the sum ((α−β) _L + β _L ) of the learning value (α−β) _L and the learning value β _L acquired in steps 102 and 104 is calculated as a correction factor by the learning process. In step 106, the fuel injection amount correction value is calculated by multiplying the basic fuel injection amount by the obtained correction factor ((α−β) _L + β _L ) (%). The fuel injection amount is corrected by adding to the injection amount.

Next, the ECU 40 executes processing related to SA-CA10 feedback control described below with reference to FIG. 4 (step 108). As a result, the target (final) fuel injection amount of the next combustion cycle is a correction term (PI term) α (% by processing in step 214 described later with respect to the basic fuel injection amount based on the engine load factor and the engine speed. ) And the correction factor ((α−β) _L + β _L ) (%) by the processing of step 106.

  FIG. 4 is a flowchart showing a subroutine representing processing related to SA-CA10 feedback control. In the subroutine shown in FIG. 4, the ECU 40 first acquires the target ignition efficiency together with the engine load factor and the engine speed (step 200). The ignition efficiency is an index related to the engine torque generation efficiency by adjusting the ignition timing, and becomes 1 (peak value) at CA50 when the ignition timing is controlled to the MBT ignition timing. The CA50 is higher than the CA50 by adjusting the ignition timing. It decreases when it is advanced or retarded.

  Next, the ECU 40 calculates a target SA-CA10 (step 202). The ECU 40 stores a map (not shown) that defines the target SA-CA10 in relation to the engine load factor and the engine speed. In step 202, the target SA-CA10 corresponding to the engine load factor and the engine speed is calculated with reference to such a map. More specifically, the target SA-CA10 depends on the engine load factor and the engine speed so that the target air-fuel ratio is set to a value corresponding to the engine load factor and the engine speed. Value is set.

  Next, the ECU 40 uses the output value of the in-cylinder pressure sensor 30 to acquire in-cylinder pressure data on a crank angle basis (step 204). Next, the ECU 40 acquires a target ignition timing (step 206). The target ignition timing is a value obtained by adding the latest correction amount of the ignition timing by CA50 feedback control described later to the basic ignition timing according to the engine load factor and the engine speed.

  Next, the ECU 40 calculates actual SA-CA10 (step 208). The actual SA-CA10 is calculated as a crank angle period from the target ignition timing acquired in step 208 to the actual CA10 obtained as an analysis result of the in-cylinder pressure data acquired in step 204. Next, the ECU 40 calculates a difference ΔSA-CA10 between the target SA-CA10 calculated in steps 202 and 208 and the actual SA-CA10 (step 210).

  Next, the ECU 40 uses the calculated difference ΔSA−CA10 and a predetermined PI gain (proportional term gain and integral term gain) to correct the fuel injection amount in accordance with the difference ΔSA−CA10 and the magnitude of the accumulated value. The rate α is calculated (step 212). Then, the ECU 40 calculates a correction value of the fuel injection amount by multiplying the basic fuel injection amount by the correction rate α (%) calculated in step 212, and adds the correction value to the basic fuel injection amount. The injection amount is corrected (step 214).

  In the main routine shown in FIG. 2, the ECU 40 executes a process related to the CA50 feedback control following the process of step 108 (step 110). The process related to the CA50 feedback control is described in, for example, Japanese Patent Application Laid-Open No. 2015-094339, and a detailed description thereof is omitted here. It is not essential to combine this CA50 feedback control with SA-CA10 feedback control. However, as described in the above publication, CA50 feedback control is preferably performed in combination with SA-CA10 feedback control.

  Next, the ECU 40 determines whether or not a predetermined learning execution condition for the learning process described above is satisfied (step 112). The correction amount (correction rate α) to be acquired as learning data is an amount required to reduce the steady deviation of the actual SA-CA10 from the target SA-CA10. Therefore, in order to ensure a high learning accuracy, it is preferable that the learning process be executed under a situation where the change of the actual SA-CA10 with respect to the target SA-CA10 has subsided. For this reason, for example, the learning execution condition is satisfied when the time change rate of the engine operating condition such as the engine load factor and the engine speed is equal to or less than a predetermined value (when the internal combustion engine 10 is in a steady operating state). Can be determined. In addition, the learning execution condition is, for example, when a predetermined time has elapsed after the target SA-CA10 is changed, or when the change rate of the correction rate α calculated by the process of step 212 is reduced to a predetermined value or less. It can also be determined that it is established. Furthermore, any of these determination methods may be used in combination.

  When the learning execution condition is satisfied, the ECU 40 executes a learning process in steps 114 to 120. In step 114, the correction factor α calculated in step 212 is acquired for the learning process. In step 116, a correction rate β (combustion characteristic change) according to the current target air-fuel ratio and the actual air-fuel ratio detected by the air-fuel ratio sensor 32 is acquired using the above-described equation (2).

  In step 118, the correction rate (α−β), which is the difference between the correction rates α and β acquired in steps 114 and 116, is calculated, and the calculated correction rate (α−β) is used as learning data. As shown in FIG. 2A, learning of the learning map (that is, learning of the learning map relating to the change in the air-fuel ratio) is performed. The specific method of the process of reflecting the correction rate (α−β) that is the learning data in the learning map is not particularly limited. Therefore, for example, the learning values may be immediately updated using the obtained learning data, or the learning data may be stored after storing a predetermined number of learning data under the condition that the map axis values are the same. The learning value may be updated using the average value of.

  In step 120, learning of the learning map shown in FIG. 2B (that is, learning of the learning map related to the change in combustion characteristics) is executed using the correction rate β acquired in step 116 as learning data. The processing in step 120 can also be performed using the same method as in step 118.

  According to the routines shown in FIGS. 2 and 3 described above, the learning process is not performed using the correction rate α of the SA-CA10 feedback control as it is as the learning data, but the deviation of the actual SA-CA10 from the target SA-CA10. The correction rate α is separated into a correction rate (α−β) and a correction rate β in consideration of the factors. Then, the learning process is executed using the learning map with the minimum number of parameters preferably associated with the correction factors (α−β) and β as map axes.

  More specifically, since the correction rate (α−β) deals with errors in the intake air amount and the fuel injection amount that are the basis of the air-fuel ratio change, the intake air amount index value and the fuel injection amount index value are By using the learning map as the map axis, a continuous (consistent) tendency can be obtained with respect to the change of the learning value with respect to the change of the value of the map axis. For this reason, when calculating the map value at a position other than the grid point using interpolation or extrapolation, calculation with high accuracy can be performed. Further, during the stoichiometric burn operation, generally, air-fuel ratio feedback control is performed to correct the fuel injection amount so that the actual air-fuel ratio detected by the air-fuel ratio sensor 32 approaches the target air-fuel ratio (theoretical air-fuel ratio). The During the execution of the air-fuel ratio feedback control, a process for taking in a correction amount as a learning value for reducing the steady deviation of the actual air-fuel ratio from the target air-fuel ratio is executed. In the learning of the learning map shown in FIG. 2A, since the intake air amount index value and the fuel injection amount index value are used as the map axes, not only during lean burn operation in which SA-CA10 feedback control is performed, but also stoichiometric burn. The learning value obtained during the air-fuel ratio feedback control during operation can be diverted to the main learning about the correction rate (α−β). For this reason, it becomes easy to ensure a learning opportunity.

Further, it can be said that the tendency of the correction amount (correction rate β) with respect to the change in the combustion characteristics is empirically easily obtained according to the engine operating condition specified by the engine torque and the engine speed. For this reason, the learning of the correction rate β related to the change in the combustion characteristics is minimized by using the engine torque and the engine rotation speed as the map axes as shown in FIG. 2B. While suppressing, it is easy to ensure the calculation accuracy of the learning value β _L when performing interpolation or extrapolation.

  As described above, according to the learning process of the present embodiment, learning is performed by allocating the correction factor α according to the factor of the steady divergence of the actual SA-CA10 with respect to the target SA-CA10, and as described above. By appropriately selecting the map axis, the learning process can be accompanied by the SA-CA10 feedback control in such a manner that high learning accuracy can be obtained while shortening the time required for learning.

FIG. 5 is a time chart for explaining an effect obtained by increasing the learning accuracy. The “FB control correction amount” on the vertical axis in FIG. 5 is the sum of the correction amount corresponding to the correction rate α by the SA-CA10 feedback control and the correction amount corresponding to the correction rate ((α−β) _L + β _L ). It is. The FB correction amount Y shown in FIG. 5 indicates a value required for the actual SA-CA10 to converge to the target SA-CA10 after the start of control. As shown in FIG. 5, the higher the learning accuracy, the smaller the difference between the FB correction amount (that is, the value based on the correction rate by the learning process) and the FB correction amount Y at the control start time t0. The correction rate α calculated after the start of is reduced. As a result, the higher the learning accuracy, the shorter the time required for convergence (that is, the response delay in feedback control). Thereby, the drivability and exhaust emission of the internal combustion engine 10 can be improved. More specifically, when the actual SA-CA10 is larger than the target SA-CA10 at the control start time t0 (when the air-fuel ratio is shifted to the lean side), the drivability can be improved by improving the response delay. Conversely, when the actual SA-CA10 is smaller than the target SA-CA10 at the time point t0 (when shifted to the rich side), the exhaust emission (NOx emission) can be improved by improving the response delay.

  Regarding the change in the combustion characteristics, as described above, the learning map having two map axes, the engine torque index value and the engine rotation speed index value, is used to organize the relationship, thereby suppressing the complexity of the learning map. It becomes like this. However, even under the same lean burn operation, the degree of influence of the air-fuel ratio on the actual SA-CA 10 when the combustion characteristics change varies depending on the lean degree of the air-fuel ratio. Therefore, in consideration of this point, a learning map for learning a change in combustion characteristics may be divided into a lean air-fuel ratio region near the lean combustion limit and a lean air-fuel ratio region richer than that. In consideration of the above points, the following minimum addition may be made to the map axis of the learning map related to the change in combustion characteristics. That is, in addition to the engine torque index value and the engine speed index value, the target air-fuel ratio or G / F (an index obtained by dividing the sum of the amount of air charged into the cylinder and the amount of EGR gas by the fuel injection amount) Value) may be added as a map axis.

By the way, in Embodiment 1 mentioned above, SA-CA10 was illustrated as an ignition delay index value representing an ignition delay. However, the “ignition delay index value” in the present invention only needs to include an ignition delay period (a crank angle period from the ignition timing to the time when heat generation starts (combustion start point CA0)), and instead of SA-CA10, For example, a crank angle period from the ignition timing (SA) to an arbitrary specific combustion point CAX1 other than CA10 can be used. Further, the “combustion time index value” may be an index value representative of the combustion time, and may be, for example, any specific ratio combustion point CAX2 other than CA50 instead of the exemplified CA50, or MBT. It may be an ignition timing change amount from the ignition timing. Furthermore, the cylinder pressure maximum crank angle θ _Pmax may be used.

DESCRIPTION OF SYMBOLS 10 Internal combustion engine 12 Piston 14 Combustion chamber 16 Intake passage 18 Exhaust passage 20 Intake valve 22 Exhaust valve 24 Throttle valve 26 Fuel injection valve 28 Ignition device 30 In-cylinder pressure sensor 32 Air-fuel ratio sensor 40 Electronic control unit (ECU)
42 Crank angle sensor 44 Air flow sensor 46 Accelerator position sensor

Claims (1)

An internal combustion engine comprising an ignition device that ignites an air-fuel mixture in a cylinder, a fuel injection valve that supplies fuel into the cylinder, an in-cylinder pressure sensor that detects an in-cylinder pressure, and an air-fuel ratio sensor that detects an actual air-fuel ratio A control device for controlling,
Index value calculating means for calculating an actual ignition delay index value based on the output value of the in-cylinder pressure sensor;
A first correction rate of the fuel injection amount is calculated based on a difference between a target ignition delay index value corresponding to a target air-fuel ratio and the actual ignition delay index value, and the actual ignition delay index value becomes the target ignition delay index value. Feedback control means for executing feedback control of the fuel injection amount so as to approach,
Calculating means for calculating a second correction rate of the fuel injection amount based on a difference between the target air-fuel ratio and the actual air-fuel ratio;
First learning for learning a first learning map in which a learning value of a difference obtained by subtracting the second correction factor from the first correction factor is associated with a fuel injection amount correlation value and an intake air amount correlation value Means,
Second learning means for learning a second learning map in which a learning value of the second correction factor is associated with an engine torque correlation value and an engine speed correlation value;
With
The feedback control means reflects not only the first correction rate but also the sum of the learned value of the difference and the learned value of the second correction rate in the correction of the fuel injection amount. Engine control device.

Images