JP5691282B2 - Air-fuel ratio control device - Google Patents

Description

  The present invention relates to an air-fuel ratio control apparatus that controls an actual air-fuel ratio to approach a target air-fuel ratio based on the oxygen concentration in exhaust gas.

  Conventionally, this type of air-fuel ratio control apparatus corrects a fuel injection amount command value (air-fuel ratio control command value) from a fuel injection valve based on the oxygen concentration in exhaust gas, thereby achieving a target air-fuel ratio (for example, theoretical air-fuel ratio). Air-fuel ratio feedback control is performed so as to control the combustion at the same time.

  Here, the absolute value of the correction amount when the correction amount repeatedly increases and decreases periodically, that is, when the air-fuel ratio feedback control is stable, increases due to secular change of the fuel injection valve and the like. If the absolute value of the correction amount due to such aging (hereinafter referred to as aging correction amount) increases, the time required for the air-fuel ratio feedback control to become stable becomes longer. The response of the control for achieving the target air-fuel ratio with respect to the change in the valve opening degree deteriorates.

  Therefore, in the control devices described in Patent Documents 1 to 3, etc., the aging correction amount is stored as a learned value during operation of the internal combustion engine, and the air-fuel ratio feedback control is executed while correcting the fuel injection amount based on the learned value. By doing so, the responsiveness of the control for achieving the target air-fuel ratio is improved. Since the aging correction amount varies depending on the operating state quantity such as the engine speed NE and the engine load, the aging correction amount that differs for each operating state quantity is stored as a learning value.

  For example, as shown in FIG. 18, using a learning map Mgx obtained by dividing the intake pressure PM (engine load) and the engine speed NE into a plurality of regions (9 divisions shown by a solid line in the example of FIG. 18), the learning map Mgx is used. Learning values are stored in the respective areas “00” to “08”.

JP 2009-108767 A JP-A-8-284714 Japanese Patent Laid-Open No. 3-145539

  In the conventional learning map Mgx, each of the regions “00” to “08” is divided so as to have the same shape (in the example of FIG. 18, a quadrangle), and the regions “00” to “08” having the same shape are arranged in a plurality of rows. And it is divided equally so that it is regularly arranged in multiple lines. Due to this, the following problems have been encountered in the past.

  That is, simply to finely divide the region of the learning map Mgx in order to control the actual air / fuel ratio to the target air / fuel ratio with high accuracy. However, in the conventional learning map Mgx in which the region is regularly and evenly divided, for example, if the division pitch is reduced and the number of divisions is increased as shown by the dotted line in FIG. The calculation load increases. That is, it is contrary to finely dividing the area of the learning map Mgx to control the actual air-fuel ratio with high accuracy and reducing the memory capacity and reducing the calculation load.

  Further, as described above, increasing the number of divisions increases the number of unlearned areas. Then, when the operating state of the internal combustion engine suddenly changes from the learned region to the unlearned region, the aging correction amount (learned value) changes discontinuously. The change in the fuel injection amount becomes an inappropriate change, which leads to a deterioration in drivability.

  The present invention has been made in order to solve the above-described problems, and the object of the present invention is to store and update the correction amount when the air-fuel ratio control command value is feedback-corrected as a learning value in the map. It is an object of the present invention to provide an air-fuel ratio control apparatus that realizes an improvement in drivability by reducing the region, and achieves high accuracy in air-fuel ratio control while avoiding an increase in storage capacity and an increase in calculation load.

  Hereinafter, means for solving the above-described problems and the operation and effects thereof will be described.

According to the first aspect of the present invention, feedback correction means for correcting the air-fuel ratio control command value output to the actuator for controlling the actual air-fuel ratio of the internal combustion engine based on the oxygen concentration in the exhaust, and the correction amount by the feedback correction means Learning value updating means for storing and updating a learning map in association with the operating state quantity of the internal combustion engine, wherein the learning map is configured by dividing the operating state quantity into a plurality of large regions. And a plurality of large regions are divided so that the target air-fuel ratios in the plurality of large regions are the same, and the shape of at least one large region of the plurality of large regions is defined as the shape of another large region. It is characterized in that it is made into different types of divisions that are differentiated.

  Here, the present inventor has obtained the knowledge that the learning values in the regions where the target air-fuel ratio is the same as a result are almost the same values as a result of earnest research. Specifically, for example, the control map MQ illustrated in FIG. 2 is configured by dividing the engine rotational speed NE and the intake pressure PM (the amount of operation state of the internal combustion engine) into a plurality of regions. 6 is a map in which an air-fuel ratio (or a target fuel injection amount corresponding to a target air-fuel ratio) is stored. All the hatched areas in the map MQ have the same target air-fuel ratio (for example, the theoretical air-fuel ratio). And the part which attached | subjected the code | symbol "04" in the learning map Mg1 illustrated in FIG. 3 is a part corresponding to the shaded part in the control map MQ. That is, the seven areas indicated by “04” in the learning map Mg1 correspond to areas where the target air-fuel ratio is the same on the learning map Mg1, and the learning values in these seven areas “04” Almost the same value. Similarly, in the example of FIG. 3, the learning values of the regions assigned the same numbers “00” to “08” are almost the same value. The outer shapes (the shapes indicated by the thick lines in FIG. 3) of the plurality of regions “00” to “08” having the same learning value in this way cannot all be the same shape, and are regularly ordered. It is impossible to line up evenly.

  In the above invention made based on this knowledge, the learning map for storing and updating the correction amount by the feedback correction means in association with the driving state amount is configured by dividing the driving state amount into a plurality of large regions, The irregular division is performed in which the shapes of the plurality of large regions are differentiated so that the target air-fuel ratio in the corresponding large region is the same. In the example of FIG. 3, nine areas “00” to “08” surrounded by a thick line in the learning map Mg1 correspond to large areas.

  Therefore, in the above-described invention, regions in which the learning values are almost the same in the conventional learning map Mgx (see FIG. 18) obtained by simply dividing the region at an equal pitch regardless of the target air-fuel ratio are grouped as large regions in the above invention. It can be said that. For this reason, it is possible to realize the control so that the actual air-fuel ratio is adjusted to the target air-fuel ratio with the same accuracy as that of the conventional map Mgx while reducing the number of divisions of the large area to the number of divided areas of the conventional map Mgx. Therefore, it is possible to achieve both avoidance of increase in the storage capacity of the memory for storing the learned value, avoidance of increase in calculation load such as processing for storing and updating the learned value, and improvement of control accuracy.

  In addition, as described above, the number of large areas can be optimized by learning value stratification, that is, the number of large areas that have not yet been learned in the learning map can be reduced. To do. Therefore, the chance that the operating state of the internal combustion engine changes from the learned large region to the unlearned large region can be reduced, so that the change of the air-fuel ratio control command value with respect to the change of the operating state can be changed appropriately. As a result, improvement in drivability can be realized.

  Specific examples of the “internal combustion engine operating state quantity” include engine rotational speed NE, engine load, engine temperature, and the like. Specific examples of the “engine load” include an intake pressure detected by an intake pressure sensor, an intake air amount detected by an air flow meter, a throttle opening detected by a throttle valve sensor, and an accelerator operation amount detected by an accelerator sensor. Etc. are raised. Specific examples of the “engine temperature” include engine coolant temperature.

  Specific examples of the “air-fuel ratio control command value” include an injection command value for an electric actuator that opens and closes a valve body of a fuel injection valve, an EGR command value that commands an EGR amount in a diesel engine, and the like. There are various command values that affect the air-fuel ratio.

  According to a second aspect of the present invention, there is provided a small region obtained by further dividing the large region into a plurality of regions, wherein the deformed division is performed so that there are small regions adjacent to two different large regions. .

  Here, the learning map Mg5 shown in FIG. 13 is an example of the above-described invention, and is irregularly divided into six large areas “00” to “05”. Each area divided by a dotted line in the figure corresponds to a “small area” according to the invention. For example, the other small areas adjacent to the small area 03a included in the large area “03” are the surrounding eight areas, three of which are “03”, three are “00”, and two are “04”. Belongs to a large area. Therefore, it can be said that the small area 03a is adjacent to two different large areas “00” and “04”.

  On the other hand, in the conventional learning map Mgx shown in FIG. 18, since the plurality of areas “00” to “08” are regularly divided, for example, the small area 03b included in the large area “03” is different from the other areas. The three large areas “00”, “01”, and “04” are adjacent to each other.

  In short, in the learning map Mg5 exemplifying the above invention, the two large areas “00” and “03” are differentiated from each other so that the small area 03a is adjacent to the two large areas “00” and “04” ( It can be said that it is a variant).

  The invention according to claim 3 is characterized in that the large area is further divided into a plurality of small areas, and the irregular division is performed so that there are four or more different large areas and adjacent small areas. And

  Here, the learning map Mg7 shown in FIG. 15 is an example of the above-described invention, and is irregularly divided into 10 large areas “00” to “09”. Each area divided by a dotted line in the figure corresponds to a “small area” according to the invention. For example, there are eight other small areas adjacent to the small area 08a of the large area “08”, and the small area 08a includes five different large areas “00” “04” “06” “07”. It can be said that it is adjacent to “09”.

  In short, in the learning map Mg7 exemplifying the above invention, the shape of each large region is made different so that the small region 08a is adjacent to the five large regions “00” “04” “06” “07” “09”. It can be said that it is classified (variant classification).

  According to a fourth aspect of the present invention, the deformed segmentation is performed so that a corner located outside the outer edge of the learning map among corners of a predetermined large region contacts only one corner of another large region. It is characterized by being.

  Here, the learning map Mg5 shown in FIG. 13 is an example of the above-described invention. For example, since the shape of the large region “03” is a quadrangle, it has four corners p1, p2, p3, and p4. The corners p1, p2, and p4 are located at the outer edge of the learning map Mg5, and the corner p3 is located at a position other than the outer edge. The corner p3 is in contact with the corner of the large area “04” but is not in contact with the corner of the large area “00”.

  On the other hand, in the conventional learning map Mgx shown in FIG. 18, a plurality of regions “00” to “08” are regularly divided, so that, for example, the corners of the large region “03” other than the outer edge of the learning map Mgx The corner portion p3 located at is in contact with the corner portions of the three large regions “00”, “01”, and “04”.

  In short, in the learning map Mg5 exemplifying the above invention, the two large areas “00” and “03” are set such that the corner p3 located outside the outer edge contacts only one corner of the other large area “04”. It can be said that the shape of each is different (different shape classification).

  The invention according to claim 5 is characterized in that the irregular section is formed so that a large polygonal region having six or more corners exists.

  Here, the learning map Mg5 shown in FIG. 13 is an example of the above-described invention. For example, the shapes of the large regions “01”, “02”, and “04” are hexagons having six corners. On the other hand, in the conventional learning map Mgx shown in FIG. 18, since the plurality of areas “00” to “08” are regularly divided, all the large areas “00” to “08” are square. In short, in the learning map Mg5 exemplifying the above-described invention, when the large areas “01”, “02”, and “04” exist in the hexagonal areas, the shapes of the respective large areas are differentiated (deformation classification). I can say that. Although a large hexagonal region is illustrated in FIG. 13, it may be an n-gon (n ≧ 6), and it is particularly desirable that n is an even number.

  In the invention according to claim 6, the learning map is configured to further divide the large area into a plurality of small areas, and is configured to store the learning value in each of the small areas. When the learning value is stored and updated in a small area, the learning value of the small area is copied to another small area in the same large area.

  According to this, it can be said that the operation state quantity is divided at equal pitches to form a small area, and the small areas are grouped to form a large area. That is, it is possible to configure a large region so that the target air-fuel ratio is the same while the operating state quantities are divided at equal pitches.

  In a seventh aspect of the present invention, for a specific small area set in advance among the plurality of small areas, a copy of a learning value from another small area within the same large area as the specific small area is copied to the specific small area. It is permitted to copy the learning value from the specific small area to the other small area.

  For example, if the learned value of the small area stored and updated is low in reliability, copying the learned value to another small area results in a learning value having low reliability in the entire large area, which is not desirable. Therefore, in the above invention, a small area that is assumed to be low in reliability is set in advance as a “specific small area”, and copying of learning values from the specific small area to other small areas is prohibited. It can avoid that the whole becomes a learning value with low reliability. However, since copying of the learning value from another small area to the specific small area is permitted, the specific small area can be quickly learned.

  As a specific example of the specific small area with low reliability, a small area having a lower learning frequency than the other small areas can be cited. As another specific example, among a plurality of small areas, a small area where the internal combustion engine is assumed to exert a braking force (engine braking), or a small area where combustion in the internal combustion engine is assumed to be unstable Is set as a specific small area.

  In a ninth aspect of the present invention, a learning value is copied from another small area within the same large area as the independent small area to the independent small area for a predetermined independent small area among the plurality of small areas. In addition to prohibiting, copying of the learned value from the independent small area to the other small area is also prohibited.

  Here, even within the same large region, it may be desirable to vary the setting method of the air-fuel ratio control command value for each small region. For example, the fuel injection amount command value (air-fuel ratio control command value) for a small region (WOT small region) that is assumed to be operated by the driver to maximize the output of the internal combustion engine as described in claim 10 There is a concern that the fuel injection amount command value (air-fuel ratio control command value) will not be maximized if the learned value of another small region is copied to such a WOT small region. The Further, if the learning value is copied from the WOT small area to another small area, there is a concern that the fuel injection amount command value (air-fuel ratio control command value) becomes excessive in the other small areas.

  In view of this point, the invention according to claim 9 prohibits copying of the learning value from another small area to the independent small area, and also prohibits copying of the learning value from the independent small area to the other small area. Therefore, the said concern can be eliminated.

  In the invention according to claim 11, the learning map is configured by further dividing the large area into a plurality of small areas, and configured to store the learning value in each of the small areas. When the learning value is stored and updated in a small area, the learning value of the small area is corrected and copied to another small area in the same large area based on a preset weight. To do.

  By the way, when the learning value (reference learning value) is stored and updated in one small area, if the same value as the reference learning value is copied to another small area against the above invention, the learning value of the copy destination There may be excess and deficiency in the change of. Therefore, in the above invention, a weight is set in advance for each small area, the reference learning value is corrected based on the weight, and the corrected value is copied to another small area. According to this, compared with a case where the same value as the reference learning value is copied to another small area, the learning value at the copy destination can be brought close to the optimum value with high accuracy.

  Furthermore, as described in claim 12, as the copy destination small area is farther from the copy source small area, the risk of drastic change in drivability is reduced by setting a smaller weight at the copy destination. it can. Therefore, the learning value can be optimized to improve drivability.

  Further, as described in claim 13, the smaller the copy destination small area is, the more the area where the change in the operation state amount caused by changing the learning value is larger (so-called high load area), the smaller the weight at the copy destination is set. However, it is less likely to affect drastic changes in drivability at the copy destination, and learning can be performed in stages. Therefore, the learning value can be optimized to improve drivability.

  In the invention according to claim 14, the specific small area set in advance among the plurality of small areas is corrected and copied from the other small areas within the same large area as the specific small area. A small area is permitted, and the specific small area is prohibited from the other small area.

  For example, when the reliability of the learned value of the small area that has been stored and updated is low, copying the corrected value of the learned value to another small area (correction copying) causes the entire large area to have a low-reliable learning value. This is undesirable. Therefore, in the above invention, a small area that is assumed to be low in reliability is set in advance as a “specific small area”, and correction copying of learning values from the specific small area to other small areas is prohibited. It is possible to avoid that the entire learning value becomes a learning value with low reliability. However, since the correction copy of the learning value from another small region to the specific small region is permitted, the specific small region can be quickly learned.

  A specific example of the specific small region with low reliability is a small region with a lower learning frequency than the other small regions. As another specific example, among a plurality of small areas, a small area where the internal combustion engine is assumed to exert a braking force (engine braking), or a small area where combustion in the internal combustion engine is assumed to be unstable Is set as a specific small area.

  In the invention according to claim 15, the correction and copying of a preset independent small area among the plurality of small areas is performed from the other small areas within the same large area as the independent small area. Both the small area and the independent small area to the other small area are prohibited.

  Here, even within the same large region, it may be desirable to vary the setting method of the air-fuel ratio control command value for each small region. For example, the fuel injection amount command value (air-fuel ratio control command value) may be maximized for a small region (WOT small region) that is assumed to be operated by the driver to maximize the output of the internal combustion engine. There is a concern that the fuel injection amount command value (air-fuel ratio control command value) will not be maximized if the learned value of another small region is corrected and copied (corrected copy) to such a WOT small region. Is done. Further, if the learning value is corrected and copied from the WOT small area to another small area, there is a concern that the fuel injection amount command value (air-fuel ratio control command value) becomes excessive in the other small areas. In the above invention in view of this point, the correction copy of the learning value from another small region to the independent small region is prohibited, and the correction copy of the learning value from the independent small region to another small region is also prohibited. Can eliminate concerns.

  The invention according to claim 16 is characterized in that when the learning value is stored and updated in one large area, the learning value of another large area is corrected in accordance with the stored and updated learning value. To do.

  Here, the learning values of a plurality of large areas have a certain correlation. For this reason, if the learning value greatly changes in one large region, it is often appropriate to greatly change the learning value in another large region. For example, when the fuel property suddenly changes, such as when the property of the fuel supplied to the fuel tank is different from the property of the remaining fuel, it is appropriate to change the learning value in all large regions. According to the above invention in view of this point, when the learning value is stored and updated in one large region, the learning value in the other large region is corrected according to the stored and updated learning value. The learning values of other large areas that have not been set can be set to optimum values.

  Further, for example, when another area is unlearned, it can be learned by correcting the unlearned large area from the initial value, so that the unlearned area can be eliminated early. Therefore, it is possible to reduce the risk that the drivability changes suddenly when the driving state changes from the learned large area to the unlearned large area. Therefore, the learning value can be optimized to improve drivability.

  The invention according to claim 17 is characterized in that the learning value of another large area is corrected on condition that the amount of change of the learning value due to the memory update is a predetermined amount or more.

  As described above, when the fuel property changes suddenly, it is appropriate to change the learning value in all large regions. However, if the amount of change in the learning value due to the memory update is less than the predetermined amount, there is a high possibility that the fuel property has not changed suddenly. Therefore, in the above invention, the learning value in another large region is corrected on the condition that the amount of change in the learning value is equal to or greater than a predetermined amount, so that the correction is unnecessary even when the fuel property is not suddenly changed. It is possible to avoid correcting the learning value of the large area.

  The invention according to claim 18 is characterized in that the larger the change amount of the learning value due to the memory update, the larger the correction amount for other large regions.

  As described above, the learning values in a plurality of large areas are not a little correlated. Therefore, if the amount of change in the learning value in one large region is large, it is appropriate to greatly change the learning value in the other large region. According to the above invention in view of this point, as the amount of change in the learning value due to the memory update is larger, the correction amount for the other large area is increased, so that the learning value of the other large area is brought closer to the optimum value. it can.

  The invention according to claim 19 is characterized in that, as the other large area is separated from the stored large area, the correction amount for the other large area is reduced.

  In many cases, the correlation is lower as the distance is larger. According to the above-mentioned invention in view of this point, the correction amount for the other large area is reduced as the other large area is separated from the stored and updated large area, so that the learning value of the other large area is excessively corrected. It is possible to promote the approach to the optimum value by suppressing

  The invention according to claim 20 is characterized in that the plurality of large areas are set by dividing the operation state quantity into unequal intervals.

  As shown in FIG. 18, the conventional learning map Mgx divides the region at equal pitches regardless of the target air-fuel ratio. However, when the target air-fuel ratio in the region is divided so as to be the same, and the range of the region is made as large as possible to reduce the number of divisions, there is a limit in the conventional learning map Mgx divided at an equal pitch, Must be divided finely. In the above invention in view of this point, since the operation state quantity is divided at unequal intervals and a large region is set, when dividing so that the target air-fuel ratio in the large region is the same, the range of each large region is set. Enlarging can be easily realized. Therefore, since the number of large areas can be further reduced, it is possible to promote the coexistence of avoiding an increase in memory storage capacity, avoiding an increase in computation load and improving control accuracy, and reducing the number of large areas that have not been learned. To improve drivability.

  The invention according to claim 21 is characterized in that at least one large region among the plurality of large regions is divided into a plurality of regions by other large regions and discretely arranged on the learning map. .

  In short, the above invention can be said to be grouped as one large area when the discretely arranged large areas have substantially the same target air-fuel ratio (within a predetermined range). According to this, the number of large areas can be reduced without lowering the controllability of the air-fuel ratio, compared to the case where large areas that are discretely arranged with substantially the same target air-fuel ratio are divided into different large areas. . In other words, when the target air-fuel ratio in the large region is divided so as to be substantially the same (within a predetermined range), it is possible to further promote increasing the range of each large region. Therefore, it is possible to further promote the coexistence of avoiding an increase in memory storage capacity, avoiding an increase in calculation load and improving control accuracy, and further promoting improving drivability by reducing the number of large areas that have not been learned. .

1 is a diagram showing a schematic configuration of an entire internal combustion engine control system to which an air-fuel ratio control apparatus according to a first embodiment of the present invention is applied. The figure which shows the map in which the optimal value of the target injection quantity with respect to engine speed NE and intake pressure PM was memorize | stored in 1st Embodiment. The figure which shows the learning map concerning 1st Embodiment. The figure which shows the learning map concerning 2nd Embodiment of this invention. 9 is a flowchart illustrating a procedure of processing for copying a learning value in the second embodiment. The figure which shows the learning map concerning 3rd Embodiment of this invention. The figure which shows the learning map concerning 4th Embodiment of this invention. The figure explaining the method of correcting and copying a learning value in 5th Embodiment of this invention. The figure explaining the method of correcting and copying a learning value in 5th Embodiment. The figure which shows the example of a setting of weighting in 5th Embodiment. The figure which shows the modification 1 with respect to FIG. The figure which shows the modification 2 with respect to FIG. The figure which shows the learning map concerning 7th Embodiment of this invention. The figure which shows the modification 1 with respect to FIG. The figure which shows the modification 2 with respect to FIG. The figure which shows the modification 3 with respect to FIG. The figure which shows the learning map concerning 8th Embodiment of this invention. The figure which shows the conventional learning map.

  Hereinafter, embodiments embodying the present invention will be described with reference to the drawings. In the following embodiments, parts that are the same or equivalent to each other are denoted by the same reference numerals in the drawings, and the description of the same reference numerals is used.

(First embodiment)
First, a schematic configuration of the entire control system for an internal combustion engine will be described with reference to FIG. In the present embodiment, a gasoline engine 10 is applied as an internal combustion engine, and the engine 10 functions as a travel drive source mounted on a motorcycle.

  An intake pipe 11 of the engine 10 is provided with an electronic throttle valve 13 whose opening is adjusted by an electric motor 12 and a throttle opening sensor 14 for detecting the throttle opening. Further, an intake pressure sensor 15 for detecting the intake pressure is provided on the downstream side of the throttle valve 13. A fuel injection valve 16 for injecting fuel is attached to a portion of the intake pipe 11 near the intake port of the cylinder head. An ignition plug 17 is attached to the cylinder head of the engine 10, and a high voltage boosted by the ignition coil 18 is sparked by the ignition plug 17 to ignite the air-fuel mixture in the cylinder.

  On the other hand, the exhaust pipe 19 of the engine 10 is provided with a catalyst device 20 having a three-way catalyst or the like for purifying CO, HC, NOx and the like in the exhaust, and the oxygen concentration in the exhaust is upstream of the catalyst device 20. An A / F sensor 21 is provided for detecting. In the present embodiment, the A / F sensor 21 is employed as a sensor for detecting the oxygen concentration in the exhaust gas, but an O2 sensor may be employed. The A / F sensor 21 is a sensor that can linearly detect the air-fuel ratio of the air-fuel mixture by outputting an oxygen concentration detection signal corresponding to the oxygen concentration in the exhaust gas. The O2 sensor detects whether the air-fuel mixture is rich or lean with respect to the predetermined value (theoretical air-fuel ratio) by detecting whether the oxygen concentration in the exhaust gas is higher or lower than the predetermined value. It is a sensor.

  An ECU 30 (electronic control unit) as an air-fuel ratio control device is mainly configured by a microcomputer having a ROM 30a (nonvolatile memory), a RAM 30b (volatile memory), an EEPROM 30c (rewritable nonvolatile memory), and a CPU 30d. The ECU 30 receives detection signals from the throttle opening sensor 14, the intake pressure sensor 15, the water temperature sensor 22, and the crank angle sensor 26 described above. The water temperature sensor 22 is attached to the cylinder block of the engine 10 and detects the cooling water temperature. The crank angle sensor 26 detects the rotational speed NE of the crankshaft of the engine 10.

  The ECU 30 executes various engine control programs stored in the ROM 30a based on detection signals input from various sensors, whereby the fuel injection amount of the fuel injection valve 16, the injection timing, and the opening of the throttle valve 13 (throttle throttle). Opening degree), ignition timing of the spark plug 17, and the like are controlled.

  Next, details of control of the fuel injection amount will be described.

  The ECU 30 controls the operation of an electric actuator (not shown) that opens and closes the valve body of the fuel injection valve 16 to control the fuel injection amount. More specifically, by controlling the valve opening time of the valve body, the fuel injection amount made by one valve opening is controlled. The ECU 30 calculates a target injection amount based on the engine rotational speed NE and the engine load, and corrects the calculated target injection amount based on a feedback correction value and a learning value described later. A command signal (air-fuel ratio control command value) is output to the electric actuator based on the final target injection amount that has been corrected.

  FIG. 2 shows a map MQ in which optimum values of the target injection amount with respect to the engine speed NE and the intake pressure PM (engine load) are stored. This map MQ is stored in the ROM 30a and is a conforming value obtained by a test performed in advance. This adaptive value is basically set so that the air-fuel ratio becomes the ideal air-fuel ratio, but depending on the region in the map MQ, it is set to be richer or leaner than the ideal air-fuel ratio. Therefore, it can be said that this map MQ storing the target injection amount as an appropriate value is a map storing the target air-fuel ratio.

  That is, the air-fuel ratio control apparatus according to the present embodiment has a target air-fuel ratio corresponding to the engine operating state quantity (for example, NE, PM) or a physical quantity corresponding to the target air-fuel ratio (for example, target injection amount, target injection time, etc.). ) Includes storage means (ROM 30a) stored in advance. Then, the ECU 30 calculates the target injection amount using the map MQ based on the engine speed NE and the intake pressure PM based on the detection values of the crank angle sensor 26 and the intake pressure sensor 15.

  Here, there is a concern that the actual air-fuel ratio may deviate from the target air-fuel ratio even if the operation of the fuel injection valve 16 is controlled with the target injection amount based on the map MQ due to the machine difference variation of the fuel injection valve 16 or the like. The Accordingly, the ECU 30 calculates a deviation between the actual air-fuel ratio detected by the A / F sensor 21 and the target air-fuel ratio, and calculates a feedback correction amount based on the deviation. Then, A / F feedback control is performed by correcting the target injection amount using this feedback correction amount. Thereby, the said concern can be eliminated and an actual air fuel ratio can be brought close to a target air fuel ratio accurately.

  The absolute value of the feedback correction amount when the feedback correction amount repeatedly increases and decreases periodically, that is, when the A / F feedback control is stable, increases due to secular change of the fuel injection valve 16 and the like. If the absolute value of the correction amount due to such aging (aging change correction amount) increases, the time required for the A / F feedback control to stabilize becomes longer. Responsiveness of control for achieving the target air-fuel ratio with respect to (change) becomes worse. Therefore, the control for obtaining the target air-fuel ratio by storing the aging correction amount as a learned value during operation of the engine 10 and executing A / F feedback control while correcting the target injection amount based on the learned value. To improve the responsiveness.

  Note that the ECU 30 when controlling to store the aging correction amount as a learned value as described above corresponds to the “learning value updating unit”, and the ECU 30 when performing A / F feedback control as the “feedback correcting unit”. Equivalent to.

  Further, by writing the learning value in the EEPROM 30c, the learning value is retained even after the engine is stopped. Therefore, when the ignition switch is turned on next time, the actual air-fuel ratio can be quickly brought close to the target air-fuel ratio.

  FIG. 3 shows a learning map Mg1 in which the learning value is stored and updated. This learning map Mg1 is stored in the EEPROM 30c, and is configured by dividing the operating state quantity of the engine 10 into a plurality of small regions. That is, the aging correction amount is learned for each operation state amount (each small region). According to this, since the aging correction amount according to the operation state amount is learned, the accuracy of bringing the actual air-fuel ratio closer to the target air-fuel ratio can be improved as compared with the case where the aging change correction amount is uniformly learned.

  In the example shown in FIG. 3, the engine rotational speed NE and the intake pressure PM (engine load) are used as the operating state quantities, and both these parameters NE and PM are divided at equal intervals. Specifically, the engine speed NE is divided into nine sections NE1 to NE9, and the intake pressure PM is divided into nine sections PM1 to PM9. Therefore, the learning map Mg1 is divided into 81 small regions at equal intervals.

  Area numbers “00” to “08” and the like are assigned to each small area. And it can be said that the same learning value is memorize | stored and updated about the small area | region of the same group number, and is grouped. In the example of FIG. 3, the large area surrounded by the thick line in the learning map Mg1 indicates the group range of the small area. In other words, the learning map Mg1 is divided into nine large areas.

  It can be said that the map MQ of FIG. 2 in which the target injection amount is stored represents the target air-fuel ratio for each engine speed NE and intake pressure PM as described above. In the learning map Mg1, small areas in which the target air-fuel ratio set for each NE and PM in the map MQ is the same are grouped as large areas to copy the learning values. In other words, a range in the map MQ where the target air-fuel ratio is the same (same air-fuel ratio range) and a range corresponding to the same large region (same group range) in the learning map Mg1 overlap.

  Note that, when the map MQ and the learning map Mg1 are divided into a plurality of regions, it is desirable that the same air-fuel ratio range and the same group range be completely matched. However, even if the boundary of the same air-fuel ratio range and the boundary of the same group range do not completely coincide, it is sufficient that at least a part of both ranges overlap.

  Further, as a result of dividing the learning map Mg1 into a plurality of areas so that both ranges overlap, in the example of FIG. 3, a plurality of large areas are divided at unequal intervals, and the plurality of large areas are irregularly arranged. Yes. For example, the NE range (NE3 to NE5) of the group to which the region number “01” belongs is wider than the NE range (NE1 to NE2) of the group to which the region number “00” belongs, and the group to which the region number “02” belongs. It is narrower than the NE range (NE6 to NE9). Further, for example, the PM range (PM7 to PM10) of the group to which the area number “06” belongs is wider than the PM range (PM1 to PM3) of the group to which the area number “01” belongs.

  Further, as a result of dividing the learning map Mg1 into a plurality of regions so that the same air-fuel ratio range and the same group range overlap, in the example of FIG. 3, one large region is divided into a plurality of regions by other large regions, and the learning map Discretely arranged on Mg1. For example, small areas corresponding to PM5 and NE3 that are part of the group to which the area number “00” belongs are separated from the small areas corresponding to PM1 to PM10 and NE1 to NE2 by the group to which the area number “03” belongs. Are arranged.

  Incidentally, before shipping the engine 10 to the market, a test is performed in which the engine 10 is operated in advance and the learning value is stored and updated, and small regions where the target air-fuel ratio obtained by the test is the same are grouped. Thus, the above-described large area classification may be set. Alternatively, small areas having the same learning value by the test may be grouped, and the above-described large area classification may be set. In these settings, the learning values of a plurality of small regions that are grouped are not limited to be exactly the same. For example, the existence range of learning values is divided into predetermined ranges, and learning values within the predetermined ranges You may make it group the small area | region which became.

  As described above, the microcomputer of the ECU 30 stores and updates the aging correction amount (learned value) at that time in the corresponding small region in the learning map Mg1, on condition that the A / F feedback control is stable. Then, when learning is performed on any of the small areas as described above, a process of copying the updated learning value to another small area is performed.

  As described above, according to the present embodiment, the learning map Mg1 is configured by dividing the engine rotational speed NE and the intake pressure PM into a plurality of small regions, and the small region is divided into a plurality of groups to be large. An area is set. The same air-fuel ratio range in the map MQ and the same group range in the learning map Mg1 are grouped so as to coincide (or at least partially overlap). In other words, a plurality of small regions are grouped so that the target air-fuel ratio in the corresponding large region is the same. As a result, the learning map Mg1 is divided into different shapes so that the shapes of the plurality of large regions are different from each other. When the learning value is stored and updated in one small area, the learning value in the small area is copied to another small area in the same large area.

  Therefore, the storage capacity of the EEPROM 30c can be reduced as compared with the case where the storage capacity is secured by the number of small regions in the learning map Mg1. Therefore, it is possible to achieve both the avoidance of the increase in the storage capacity of the EEPROM 30c for storing the learning value and the avoidance of the increase in the calculation load of the CPU 30d and the high accuracy of the control for adjusting the actual air-fuel ratio to the target air-fuel ratio.

  Further, since the learning value is copied as described above, the number of small areas that have not been learned in the map can be reduced. Therefore, since the chance that the operation state of the internal combustion engine changes from the learned small region to the unlearned small region can be reduced, the change in the command signal to the fuel injection valve 16 with respect to the change in the operation state is changed appropriately. be able to. That is, since it is possible to avoid discontinuous injection amount control, it is possible to improve drivability.

  Furthermore, in the present embodiment, even if the regions that are discretely arranged on the learning map Mg1 (for example, the region number “01”) are within the same target air-fuel ratio (within a predetermined range), Are grouped as one large area. According to this, enlarging the range of the large region can be further promoted. Therefore, it is possible to further promote the coexistence of avoiding an increase in the storage capacity of the EEPROM 30c, avoiding an increase in the calculation load of the CPU 30d and improving the control accuracy, and improving the drivability by reducing the number of large areas that have not been learned. It can be further promoted.

(Second Embodiment)
In the present embodiment, the learning map Mg1 shown in FIG. 3 is changed to the learning map Mg2 shown in FIG. 4, and a specific small area and an independent small area described below are set in the learning map Mg2.

  In the example shown in FIG. 4, the engine rotational speed NE and the intake pressure PM (engine load) are used as operating state quantities, and both these parameters NE and PM are divided at equal intervals. Specifically, the engine rotation speed NE is divided into nine divisions NE1 to NE9, and the intake pressure PM is divided into ten divisions PM1 to PM10. Therefore, the learning map Mg2 is divided into 90 small areas at equal intervals.

  Area numbers “00” to “08”, “A8”, “B8”, and the like are assigned to each small area. These area numbers are configured by combining group numbers “0” to “8” in the first digit and attribute numbers “0”, “A”, and “B” in the second digit. And it can be said that the same learning value is memorize | stored and updated about the small area | region of the same group number, and is grouped. In the example of FIG. 4, the large area surrounded by the thick line in the learning map Mg2 indicates the group range of the small area. In other words, the learning map Mg2 is divided into nine large areas.

  The attribute number “0” indicates that the corresponding small area is a normal small area, the attribute number “A” indicates that the corresponding small area is a specific small area to be described later, and the attribute number “B” corresponds. This indicates that the small area is an independent small area described later.

  A normal small area is an area that allows both the learning value of another small area to be copied to itself and the copying of its own learning value to another small area. A specific small area is an area where copying of a learning value of another small area to itself is permitted, but copying of its own learning value to another small area is prohibited. An independent small area is an area where copying of learning values of other small areas to itself and copying of learning values of other small areas to other small areas are prohibited.

  For example, when the learning value of the normal small region is newly stored or updated, the learning value of the normal small region is copied to another small region within the same group (within the same large region). . However, when the other small area is an independent small area, the copying is prohibited. If the other small area is a normal small area or a specific small area, the copying is permitted. In addition, when the learning value of a specific small area or independent small area is newly stored or updated, it is prohibited to copy the learning value of the specific small area or independent small area to another small area. To do.

  The specific small region is a region set in advance as a small region that is assumed to have a lower learning frequency than the normal small region. In the example of FIG. 4, the high load PM7, PM8, PM9, and the high rotation NE9 region. Is set in the specific small area A8. Such a high-load high-rotation region is used very infrequently and the time for holding the region is also extremely short. Therefore, even if the operation condition is learned in this high load high rotation region, the learned value is low in reliability.

  As another example of the specific small region, it is assumed that combustion in an internal combustion engine becomes unstable, such as a small region (for example, a region of low load PM1 and high rotation NE9) in which engine braking is assumed to be performed. A small region (for example, a region of high load PM9 and low rotation NE1) is set as the specific small region. Even in these areas, even if learning is performed, the learning value has low reliability.

  The independent small region is a region that is set in advance as a small region that is assumed to be operated by the driver so as to maximize the output of the internal combustion engine by maximizing the throttle valve 13, so-called WOT (wide-open throttle). ) Area. In the example of FIG. 4, the area of the high load PM10 is set to independent small areas B0, B6, B7, and B8.

  FIG. 5 shows a procedure of a process of copying a learned value to another small area when a learned value is stored and updated in a small area in the learning map Mg2 among the processes performed by the microcomputer of the ECU 30. It is a flowchart. As described above, on the condition that the A / F feedback control is stable, the aging correction amount (learning value) at that time is stored and updated in the corresponding small region in the learning map Mg2. As described above, when learning is performed on any of the small regions, the processing of FIG. 5 is executed.

  First, in step S10, it is determined whether or not the learned small region is a specific small region. In subsequent step S20, it is determined whether or not the learned small region is an independent small region. When it is determined that neither the specific small area nor the independent small area (S10: NO and S20: NO), in the following step S30, among other small areas in the same large area as the learned small area, The learned value of the learned small area is copied to all the small areas except the independent small area. On the other hand, when it is determined that the learned small area is a specific small area or an independent small area (S10: YES or S20: YES), the process of FIG. 5 is performed without copying the learning value to another small area. finish.

  As described above, according to the present embodiment, the following effects are exhibited in addition to the effects of the first embodiment. That is, a small area that is assumed to have low reliability is set in advance as a “specific small area”, and copying of a learning value from the specific small area to another small area is prohibited. Therefore, it can be avoided that the entire large area becomes a learning value with low reliability. However, since copying of the learning value from another small area to the specific small area is permitted, the specific small area can be quickly learned.

  Incidentally, in the case of the engine 10 mounted on a motorcycle, the maximum value of the engine rotational speed NE is higher than that of the engine mounted on the motorcycle, and the rise and fall of the rotation are fast (that is, the rotational speed NE). Is abruptly changed). Therefore, in the motorcycle engine 10 according to the present embodiment, a small region (specific small region) that is assumed to be low in reliability is remarkably present. Therefore, according to the present embodiment in which the air-fuel ratio control apparatus of the present invention is applied to the motorcycle engine 10, the effect obtained by installing the specific small region as described above is preferably exhibited.

  In this embodiment, copying of learning values from other small areas to the WOT area (independent small area) is prohibited, and copying of learning values from the WOT area to other small areas is also prohibited. For this reason, there is a concern that the amount of fuel injection will not be maximized by copying the learned value of another small region to the WOT region, or fuel injection in another small region by copying the learned value from the WOT region to another small region. The concern that the amount becomes excessive can be solved.

(Third embodiment)
In the first embodiment, the learning map Mg1 is divided into a plurality of large areas as a result of dividing the learning map Mg1 into small areas and grouping these small areas. On the other hand, in the present embodiment shown in FIG. 6, the learning map Mg3 is divided into a plurality of large regions so that the target air-fuel ratio in the corresponding large region becomes the same by eliminating the division into small regions. Configure. When it is desired to set the specific small area and the independent small area according to the second embodiment, these specific small area and independent small area are set separately from the large area in the learning map Mg3 shown in FIG. That's fine.

  As described above, the present embodiment can provide the same effects as those of the first embodiment. In the example shown in FIG. 6, one large area (for example, large areas indicated by “00”, “01”, and “05”) is divided into a plurality of parts by other large areas and discretely arranged on the learning map Mg3. Has been. In the example shown in FIG. 6, a plurality of large areas are divided at unequal intervals.

(Fourth embodiment)
In the learning map Mg3 according to the third embodiment, one large region is discretely arranged, but in the present embodiment shown in FIG. 7, such discrete arrangement is abolished. However, a plurality of large areas are divided at unequal intervals so that the target air-fuel ratios in the corresponding large areas are the same. When it is desired to set the specific small area and the independent small area according to the second embodiment, the specific small area and the independent small area are set separately from the large area in the learning map Mg4 shown in FIG. That's fine. As described above, the present embodiment can provide the same effects as those of the second embodiment.

(Fifth embodiment)
In each of the above embodiments, when a learning value of a small region is newly stored or updated, the same value as the learning value (reference learning value) of the small region is set within the same group (same large region). (Inside) Copying to other small area. On the other hand, in the present embodiment, the reference learning value is corrected based on a preset weight, and the corrected value is copied to another small area.

  FIG. 8 is a diagram illustrating one aspect of the correction according to the present embodiment. The vertical axis in FIG. 8 indicates the learning value of the region number “01” in the learning map Mg1 in FIG. 3, and the horizontal axis in FIG. Indicates the intake pressure PM, which is the horizontal axis of the learning map Mg1. Further, in the example of FIG. 8, in the state where the learning values g1a, g2a, g3a, and g5a indicated by white circles in the drawing are stored in the learning map Mg1, the learning value g1b (the learning value g1b indicated by the black circle in the PM1 region) is stored. It is assumed that the reference learning value is updated. In this case, the learning values g2a, g3a, and g5a of the other small areas PM2, PM3, and PM5 are similarly changed by ΔQ and set to the respective small areas instead of being changed to g2a ′, g3a ′, and g5a ′. The amount of change ΔQ is corrected according to the weights w2, w3, and w5, and the learning values g2a, g3a, and g5a are changed by the amount of change after the correction.

  In short, in the example of FIG. 8, the learning value of the other small areas PM2, PM3, PM5 is not updated to the same value as the reference learning value g1b, but the value obtained by correcting the reference learning value g1b based on the weights w2, w3, w5. Has been updated. As a result, the small areas in the same group store learning values g1b, g2b, g3b, and g5b having different values. In the example of FIG. 8, the weights w2, w3, and w5 of the small areas PM2, PM3, and PM5 are set to 0.7, 0.4, and 0.0, respectively. Therefore, the small area PM5 has the same value before and after learning.

  If the copy destination small areas PM2, PM3, and PM5 have not been learned, the same value as the reference learning value g1b is copied to the copy destination small area, and the copy destination small areas PM2, PM3, and PM5 have already been learned. In the case of, a value obtained by correcting the reference learning value g1b based on the weights w2, w3, and w5 as described above is copied.

  FIG. 8 shows an embodiment assuming an initial state in which the learning values g1a, g2a, g3a, and g5a are the same, whereas FIG. 9 shows the learning values g1a, g2a, g3a, This is an example of a case where the learning value g1a in the PM1 region is updated to the reference learning value g1b under a situation where g5a has a different value. When the amount of change when updating the learning value of the region of PM1 is ΔQ, the weights w2, w3, and w5 set in advance are corrected by multiplying ΔQ, and each of the small regions PM2, PM3, and PM5 is corrected. Update change amounts ΔQ × w2, ΔQ × w3, ΔQ × w5 are calculated. Then, the update change amount is added to g2b, g3b, and g5b of each of the small regions PM2, PM3, and PM5, and updated so as to become the learning values g2b, g3b, and g5b.

  FIG. 10 is a partially enlarged view of the learning map Mg1 shown in FIG. 3, and the numerical values shown in parentheses indicate weights w1, w2, w3, and w5 set in advance for each small area. In the example of FIG. 10, the weighting is set to a different value according to the intake pressure PM. In other words, the weights are set to the same value for the small regions having the same intake pressure PM in the same group. On the other hand, in the modified example 1 shown in FIG. 11, the weighting is set to the same value for the small areas having the same engine speed NE in the same group. The values of these weights w1, w2, w3, and w5 are set to smaller values as the distance from the copy source small area (PM1, NE3) increases, and the degree of reflection of the reference learning value g1b at the copy destination Is low.

  Further, in the second modification shown in FIG. 12, even when the engine speed NE or the intake pressure PM is the same small region, it is allowed to set the weighting to a different value. However, the farther the distance on the learning map Mg1 from the copy source small area (PM1, NE3) to the copy destination small area, the smaller the weighting value is set, and the degree of reflection of the reference learned value g1b at the copy destination is reduced. doing.

  As described above, according to the present embodiment, the same weight as the reference learning value g1b is not copied to other small areas as shown by the symbols g2b ′, g3b ′, and g5b ′ in FIG. Since the learning value of the other small area is updated to the corresponding value, the learning value at the copy destination can be brought close to the optimum value with high accuracy.

(Sixth embodiment)
In the fifth embodiment, the weighting value is set to be smaller as the distance from the copy source small area (PM1, NE3) to the copy destination small area increases. In the present embodiment, the copy destination small area is set. The smaller the area is, the larger the “impact” described below, the smaller the weight at the copy destination.

  That is, for example, in a small region of low load and low NE, the engine operating state amount changes greatly only by slightly changing the learning value. On the other hand, in the small region of high load and high NE, the engine operating state quantity does not change greatly compared to the small region of low load and low NE. In other words, the change in the engine operating state amount caused by changing the learning value by a predetermined amount is defined as “impact”, and the weighting is reduced in the region where the impact is large to reduce the reference learning value g1b. In this embodiment, the degree of reflection is lowered.

  For example, when the learning value g1a in the areas PM1 and NE3 is updated to the reference learning value g1b in the same manner as in the fifth embodiment, the small areas PM2 and PM3 are areas having a larger impact than the small area PM5. As an example of this embodiment, setting the weights w2 and w3 of the small areas PM2 and PM3 to be smaller than the weight w5 of PM5.

  As described above, according to the present embodiment in which the weighting is reduced and the reflection degree of the reference learning value g1b is lowered as the impact is higher, it is possible to suppress an abrupt change in the operating state quantity of the engine. Can be improved.

(Seventh embodiment)
In the present embodiment, as examples of deformed sections that are divided by different shapes of a plurality of large areas, various examples of deformed sections that are different from the learning map Mg1 illustrated in FIG. 3 are illustrated in FIGS. Will be described.

  The learning map Mg5 shown in FIG. 13 is irregularly divided into six large areas “00” to “05” shown by solid lines. Each area divided by a dotted line in the figure indicates a small area. The three large areas “00”, “03”, and “05” are squares, and the remaining three large areas “01”, “02”, and “04” are hexagons. Further, even if they are the same rectangle, the large area “00” and the large area “03” have different shapes. Further, even in the same hexagon, the large regions “01”, “02”, and “04” have different shapes. That is, except that the large areas “00” and “05” have the same shape, it can be said that all of the other large areas are divided into different shapes (different shape classification).

  That is, when the regions where the target air-fuel ratio is the same in the learning map Mg5 are grouped, the shapes for each group are not regularly arranged in the same shape, but are irregularly arranged in different shapes. . Corresponding to the irregular and irregular arrangement, the large area is divided into irregular shapes. For example, as a result of the irregular division so that the target air-fuel ratio in the large region is the same, the small region 03a included in the large region “03” is adjacent to the two large regions “00” and “04”. Alternatively, the corner p3 of the large region “03” comes into contact with the corner of the large region “04”, but does not come into contact with the corner of the large region “00”.

  A learning map Mg6 shown in FIG. 14 is a modification of Mg5, in which the large area “01” is divided into a plurality of small areas 01a that are divided into a plurality of parts by the other large areas “04”. For example, even if regions where the target air-fuel ratio is the same are distributed in a discrete manner, the same learning value can be grouped so as to be stored by discretely arranging like the small region 01a. In addition, the small area | region 01b shows the example arrange | positioned discretely in the aspect which the small area | region 01c and the corner | angular part p5 contact.

  The small areas 01a and 01b that are discretely arranged in this manner have a higher learning frequency than the other small areas in the same large area 01 (13 small areas located on the left side of the small area 01b). By copying learning values frequently obtained from the areas 01a and 01b to 13 other small areas, the update frequency of learning values in other small areas with relatively low learning frequency increases, and the unlearned areas are reduced. be able to.

  A learning map Mg7 shown in FIG. 15 is a modified example of Mg5, and is an example in which the number of large areas is increased for an area where a highly accurate learning value is required. Specifically, NE = 3500-5500 rpm, Four large regions “06” to “09” are increased in the region of PM = 50 to 80 kPa. This learning map Mg7 assumes a small two-wheeled vehicle with a displacement of 125 cc. The four added large areas “06” to “09” are set smaller than the other large areas. In the example of FIG. 15, one small area constitutes one large area.

  It should be noted that a highly accurate learning value is often required in the low load region at the idle rotation speed, and therefore, a large region corresponding to such an idle region is smaller than other large regions (for example, one It is preferable to set it to a small area. In addition, since a highly accurate learning value is often required for a large region corresponding to a region where the air-fuel ratio changes greatly, the shape of the corresponding large region becomes smaller as the region where the air-fuel ratio change is larger. It is suitable to set.

  A learning map Mg7 shown in FIG. 16 is a modified example of Mg5, and the number of small areas is increased without changing the number of large areas. By increasing the number of small areas in this way, the shape of the large area can be set finely as shown by the dashed line, and the shape setting freedom of the large area can be improved, so the accuracy of the learning value can be improved. The controllability of the air / fuel ratio can be improved. Nevertheless, since it is not necessary to increase the number of sections in the large area, avoiding an increase in the storage capacity of the memory for storing the learning value, avoiding an increase in calculation load such as a process for storing and updating the learning value, and an air-fuel ratio It is possible to achieve both improvement of controllability.

  In the present embodiment, a small two-wheeled vehicle is assumed, but the same effects can be obtained even with a large two-wheeled vehicle and a four-wheeled vehicle.

(Eighth embodiment)
In each of the above embodiments, when the learning value is stored and updated in one small region, other small regions in the same large region as the small region are copied and the learning value is changed. The learning value is not changed for the large area. On the other hand, in the present embodiment, when a learning value is stored and updated in one small region, not only other small regions in the same large region but also learning values in other large regions are stored. Is also changed (corrected).

  Specifically, using the learning map Mg9 shown in FIG. 17, for example, when the learning value is updated for the small area indicated by the reference numeral 01d in the large area “01”, all the small areas in the large area “01” are displayed. The learning value of the small area 01d is copied to the area, and the large areas “07” and “08” designated in advance among the other large areas “00” “02” to “08” or all large areas The learning value is corrected for “00”, “02” to “08”. Alternatively, when the large areas “01” and “07” and “08” are associated in advance and the learning value of the large area “01” is stored and updated, only the associated large areas “07” and “08” are stored. May be corrected.

  Next, the correction amount when performing the above correction will be described. For example, in the case where the value before learning of the small region 01d is A and the value after learning is B, the larger the change amount AB of the learning value of the small region 01d, the other large regions “00” and “02”. Increase the correction amount of “08”.

  For example, the rate of change (A−B) / A in the large region “01” is calculated and corrected by multiplying the existing learning values in the other large regions “07” and “08” by the rate of change. Alternatively, a weight may be set for each of the other large regions “07” and “08”, and the existing learning value may be corrected by multiplying the change rate and the weight. Alternatively, the change amount (A−B) in the large region “01” is calculated, and the change amount is added to the existing learning values in the other large regions “07” and “08” to be corrected. Alternatively, a weight may be set for each of the other large areas “07” and “08”, and a value obtained by multiplying the change amount by the weight may be added to the existing learning value to be corrected. As a result, the larger the change amount AB of the learning value in the small area 01d, the larger the correction amounts of the other large areas “00”, “02” to “08”.

  Further, with respect to the correction amount, as the large areas “07” and “08” are separated from the stored large area “01”, the correction amounts for the other large areas “07” and “08” are reduced. For example, when the learning value of the small area 01d is stored and updated, the distance from the large area “01” corresponding to the small area 01d is farther (separated) from the large area “08” than the “07”. In this case, for example, by setting the weight for the large area “08” smaller than the weight for “07”, the correction amount or the correction ratio of the large area “08” is made smaller than “07”.

  Here, since the learning values of a plurality of large areas have a certain correlation, for example, when the learning value of the large area “01” is increased, the other large areas “00”, “02” to It is often appropriate to increase the learning value of “08”. For example, when the property of the fuel supplied to the fuel tank is different from the property of the remaining fuel, it is appropriate to increase or decrease the learning value in all large regions “00” to “08”. In view of this point, in the present embodiment, when the learning value is stored and updated in one large area “01”, the other large areas “00” and “02” to “02” are changed according to the stored and updated learning value. Since the learning value of “08” is corrected, learning values in other large areas that have not been learned can be set to optimum values that match the characteristics of the vehicle.

  Furthermore, according to the present embodiment, for example, when other areas “00”, “02” to “08” are unlearned, the learning initial values of the unlearned large areas “00”, “02” to “08” are set. Since it can be corrected and learned, unlearned large areas “00”, “02” to “08” can be eliminated early. Therefore, it is possible to reduce the risk that the drivability changes suddenly when the driving state changes from the learned large area to the unlearned large area. Therefore, the learning value can be optimized to improve drivability.

  Further, according to the present embodiment, the larger the learning value change amount A-B due to the memory update is, the larger the correction amount for the other large region is. Can promote. Further, as the other large areas “07” and “08” are separated from the stored large area “01”, the correction amount for the other large areas “07” and “08” is reduced. It is possible to prevent the learning values of “07” and “08” from being excessively corrected and to bring each learning value closer to the optimum value that matches the vehicle characteristics.

(Modification of the eighth embodiment)
In carrying out the eighth embodiment, the learning value of another large region may be corrected on condition that the change amount of the learning value is equal to or greater than a predetermined amount. For example, when the learning value of the small area 01d is stored and updated, if the change amount AB of the learning value of the small area 01d is less than a predetermined amount set in advance, the other large areas “00” and “02”. ”To“ 08 ”is prohibited from being corrected. On the other hand, if the amount of change A-B is equal to or greater than a predetermined amount, the correction is permitted for the other large areas “00”, “02” to “08”.

  Here, when the fuel property changes suddenly, it is appropriate to change the learning value in all large regions “00” to “08”. However, if the amount of change in the learning value due to the memory update is less than the predetermined amount, It is highly possible that the fuel properties have not changed suddenly. In this modified example in view of this point, the learning values of the other large regions “00”, “02” to “08” are corrected on condition that the learning value change amount AB is equal to or greater than a predetermined amount. It is possible to avoid correcting the learned values of the other large areas “00”, “02” to “08” until correction is not necessary, such as when the fuel property has not changed suddenly.

(Other embodiments)
The present invention is not limited to the description of the above embodiment, and may be modified as follows. Moreover, you may make it combine the characteristic structure of each embodiment arbitrarily, respectively.

  When correcting the target injection amount using the learning value (aging correction amount) stored in each of the learning maps Mg1 to Mg9, the current engine operating state is a plurality of small regions (or large regions) in the learning map. Area), the linear interpolation is performed using the learning value of the area corresponding to the current engine operating state and the learning value of the adjacent area, and the value obtained by this linear interpolation is calculated. It may be used to correct the target injection amount.

  In the second embodiment, the independent small area is grouped with the normal small area and the independent small area is included in the large area. However, the learning map Mg1 may be configured so that the independent small area is excluded from the large area. Good.

  In calculating the target fuel injection amount based on the engine speed NE and the engine load, in the first embodiment, the intake pressure PM1 is used as a parameter representing the engine load. In other words, the intake air amount correlated with the engine load is estimated from the intake pressure PM1, and the target injection amount is calculated using the estimated intake air amount. On the other hand, the target injection amount may be calculated using the throttle opening by the throttle opening sensor 14 or using the intake air amount estimated from the throttle opening. The learning map in this case may be configured by dividing the engine speed NE and the throttle opening into a plurality of regions.

  In addition, an air flow meter 27 (see FIG. 1) for detecting the intake air amount (mass) may be provided, and the target injection amount may be calculated using the intake air amount detected by the air flow meter 27. The learning map in this case may be configured by dividing the engine speed NE and the intake air amount into a plurality of regions.

  In each of the above embodiments, the engine rotation speed NE and the engine load (intake pressure PM) are adopted as the operating state quantity of the engine 10, and the learning maps Mg1 to Mg4 are configured by dividing these NE and PM into a plurality. However, the present invention is not limited to this. For example, the engine coolant temperature TW is added to NE and PM, and the NE, PM, and TW are divided into a plurality of parts to learn a three-dimensional map. A map may be constructed. Even in this case, it is required to divide the large region (or group the small regions) so that the target air-fuel ratio in the same large region is the same.

  In the fifth and sixth embodiments, the weighting is limited to a value between 0 and 1, and the change amount of the copy destination learning value is set to be smaller than the change amount ΔQ of the copy source learning value. The weight may be set to one or more values. Also, the copy destination learning value may be increased or decreased in the opposite direction to the increase or decrease of the copy source learning value by setting the weighting to a value less than 0 (negative value).

  In the fifth and sixth embodiments as well, it is desirable to set “normal small region”, “specific small region”, and “independent small region” as in the second embodiment. That is, in the normal small area, the learning value of the other small area is corrected based on the weight w and copied to itself, and the own learning value is corrected based on the weight w and copied to the other small area. Both are acceptable. Further, in the specific small area, it is allowed to correct the learning value of the other small area based on the weight w and copy it to itself, but correct the learning value of the other small area based on the weight w and move to the other small area. Copying is prohibited. In the independent small area, the learning value of the other small area is corrected based on the weight w and copied to itself, and the own learning value is corrected based on the weight w and copied to the other small area. Both of these are prohibited.

  In the various learning maps Mg1 to Mg9 described above, the outer edge of the learning area is set to a quadrangle, but as shown by the two-dot chain line in FIG. In this way, a large region may be set by dividing into different shapes.

  The present invention is not limited to the irregular division so that all of the plurality of large areas in the learning map have different shapes, and it is sufficient that at least one large area has a different shape from the other large areas. However, the target air-fuel ratio in the same large region needs to be substantially the same (within a predetermined range).

  In implementing the eighth embodiment, a large area (for example, “01”) that is predicted to have a high learning frequency is set in advance, and the learning value of the large area “01” is stored and updated. As a condition, correction of other large areas “00”, “02” to “08” is permitted, and learning values of large areas other than the set large area “01” (for example, “08”) are stored and updated. Alternatively, it may be prohibited to correct the learning values of the other large areas “00” to “07”.

  30 ... ECU (feedback correction means, learning value update means), Mg1, Mg3, Mg4, Mg5, Mg5, Mg6, Mg7, Mg7, Mg8, Mg9 ... learning map.

Claims (21)

Feedback correction means for correcting the air-fuel ratio control command value output to the actuator that controls the actual air-fuel ratio of the internal combustion engine based on the oxygen concentration in the exhaust;
Learning value updating means for making a correction amount by the feedback correction means a learning value, and storing and updating the learning map in association with the operating state quantity of the internal combustion engine;
With
The learning map is configured by dividing the operation state quantity into a plurality of large regions, and the plurality of large regions are divided so that the target air-fuel ratios in the plurality of large regions are the same , An air-fuel ratio control apparatus characterized in that at least one large region of the plurality of large regions is shaped to be different from the shape of the other large regions.   2. The air-fuel ratio according to claim 1, wherein the large area is further divided into a plurality of small areas, and the irregular division is performed so that two different large areas are adjacent to each other. Control device.   The small region obtained by further dividing the large region into a plurality of regions, wherein the deformed region is divided so that there are small regions adjacent to four or more different large regions. Air-fuel ratio control device.   Of the corners of a predetermined large area, the corners located outside the outer edge of the learning map are divided so that only one corner of the other large area is in contact with each other. The air-fuel ratio control apparatus according to any one of claims 1 to 3.   The air-fuel ratio control apparatus according to any one of claims 1 to 4, wherein the deformed section is divided so that a large polygonal region having six or more corners exists. The learning map is configured by further dividing the large area into a plurality of small areas, and configured to store the learning value in each of the small areas,
6. When the learning value is stored and updated in one small area, the learning value of the small area is copied to another small area in the same large area. The air-fuel ratio control apparatus according to one.   For a specific small area set in advance among the plurality of small areas, copying of a learning value from another small area within the same large area as the specific small area to the specific small area is permitted, and from the specific small area, The air-fuel ratio control apparatus according to claim 6, wherein copying of the learned value to the other small area is prohibited.   The air-fuel ratio control apparatus according to claim 7, wherein a small region having a lower learning frequency than the other small regions is set as the specific small region.   For the independent small area set in advance among the plurality of small areas, copying of the learning value from another small area within the same large area as the independent small area to the independent small area is prohibited, and the independent small area The air-fuel ratio control apparatus according to any one of claims 6 to 8, wherein copying of a learning value from one to another small area is also prohibited.   The air-fuel ratio control apparatus according to claim 9, wherein a small region assumed to be operated by a driver so as to maximize the output of the internal combustion engine is set as the independent small region. The learning map is configured by further dividing the large area into a plurality of small areas, and configured to store the learning value in each of the small areas,
When the learning value is stored and updated in one small area, the learning value of the small area is corrected and copied to another small area in the same large area based on a preset weight. The air-fuel ratio control apparatus according to any one of claims 1 to 5, wherein   12. The air-fuel ratio control apparatus according to claim 11, wherein the weighting at the copy destination is set to be smaller as the copy destination small area is farther from the copy source small area.   The weighting at the copy destination is set to be smaller as the small area of the copy destination is an area where the change in the operation state amount due to the change of the learning value is larger. Air-fuel ratio control device.   For the specific small area set in advance among the plurality of small areas, the correction and copying are permitted from the other small areas within the same large area as the specific small area to the specific small area, The air-fuel ratio control apparatus according to any one of claims 11 to 13, wherein the specific small area is prohibited from the other small area.   Of the plurality of small areas, a preset independent small area is corrected and copied from another small area within the same large area as the independent small area to the independent small area and the independent small area. The air-fuel ratio control apparatus according to any one of claims 11 to 14, wherein any of the region to the other small region is prohibited.   16. When the learning value is stored and updated in one large area, the learning value of another large area is corrected in accordance with the stored and updated learning value. The air-fuel ratio control apparatus according to any one of the above.   The air-fuel ratio control apparatus according to claim 16, wherein the learning value in another large region is corrected on condition that the amount of change in the learning value due to the memory update is equal to or greater than a predetermined amount.   18. The air-fuel ratio control apparatus according to claim 16, wherein the correction amount for the other large region is increased as the amount of change in the learning value due to the memory update is larger.   The air-fuel ratio control apparatus according to any one of claims 16 to 18, wherein the correction amount for the other large area is decreased as the other large area is separated from the stored large area.   The air-fuel ratio control apparatus according to any one of claims 1 to 19, wherein the plurality of large regions are set by dividing the operation state quantity at unequal intervals.   The at least one large region among the plurality of large regions is divided into a plurality of portions by other large regions and is discretely arranged on the learning map. The air-fuel ratio control apparatus according to one.

Images