JP6213351B2 - Injection amount learning device for internal combustion engine - Google Patents

Description

  The present invention relates to an injection amount learning device for an internal combustion engine, and more particularly to an injection amount learning device for learning an injection amount of fuel injected from an injector in a fuel supply system for a diesel engine.

  Conventionally, in a diesel engine, a technique for performing so-called pilot injection in which a minute amount of fuel is injected prior to main injection is known in order to reduce combustion noise and NOx emission. Further, in order to sufficiently obtain the effect of pilot injection, it is important to precisely adjust the injection amount in pilot injection. Therefore, various techniques related to injection amount learning for micro injection have been proposed (see, for example, Patent Document 1). In Patent Document 1, when a fuel injection request is not made, such as when the vehicle is decelerated, the deviation of the relationship between the energization period of the injector and the actual injection amount is learned, and the fuel is determined according to the deviation. It is disclosed that the injection amount is corrected. Further, it is disclosed that the injection amount learning is performed in accordance with the pressure in the common rail in consideration that the fuel injection amount in the pilot injection changes in accordance with the pressure in the common rail.

Japanese Patent No. 4715821

  In the injection amount learning, if the learning accuracy is to be ensured, there is a problem that learning can be performed only in a limited driving state such as during stable deceleration, and the learning frequency is reduced. On the other hand, in the pilot injection, the same injection amount is not used in all operating states, but the injection amount corresponding to the engine operating state is injected. From the viewpoint of ensuring learning accuracy, it is conceivable that learning is performed for each of a plurality of injection amounts within the injection amount range of pilot injection, but in such a case, there is a trade-off that the travel distance until completion of learning becomes long. .

  The present invention has been made to solve the above-described problem, and provides an injection amount learning device for an internal combustion engine that can ensure learning accuracy while shortening a period required for completion of learning in injection amount learning as much as possible. One purpose.

  The present invention employs the following means in order to solve the above problems.

  The present invention is applied to a fuel supply system (50) of an internal combustion engine including a fuel injection valve (32) for injecting high-pressure fuel in a pressure accumulating pipe (13), and learns the fuel injection amount from the fuel injection valve. The present invention relates to an injection amount learning device for an engine (30). According to the first aspect of the present invention, when the learning execution condition is satisfied, the actual torque generated with the fuel injection from the fuel injection valve is a predetermined minute injection amount of fuel. An injection control means for injecting fuel from the fuel injection valve a plurality of times as a learning injection so as to coincide with a prescribed torque that is a generated torque when the fuel is actually injected from the injection valve; and the injection control means First learning for learning the minute injection amount by using the data of the injection times in which the actual torque belongs to a predetermined torque determination range including the prescribed torque among the data relating to the injections of the fuel injections of the plurality of times. Among the data relating to the injection at each injection in the plurality of fuel injections by the injection control means and the data on the injections in which the actual torque belongs to the torque determination range, and the torque By using the injection times of the data out of the determination range, characterized in that it comprises a second learning means for learning a correlation parameter indicating a correlation between the actual torque and the command injection quantity of the fuel injection valve.

  In short, in the above configuration, the fuel injection is performed a plurality of times as the learning injection so that the actual torque of the internal combustion engine coincides with the prescribed torque generated by the fuel of a predetermined minute injection amount. Of the data relating to the injection in the learning injection, the minute injection amount is learned using the data of the injection times in which the actual torque belongs to the predetermined torque determination range. Further, among the data relating to the injection in the learning injection, the fuel injection is performed using the data of the injection times where the actual torque belongs to the predetermined torque determination range and the data of the injection times where the actual torque does not belong to the predetermined torque determination range. A correlation parameter indicating the correlation between the command injection amount of the valve and the actual torque is learned.

  In the process of matching the actual torque with the specified torque, data that is not used for learning the minute injection amount, that is, data that does not belong to the predetermined torque determination range is also acquired. By utilizing this, the relationship between the command injection amount and the actual torque can be grasped using actual data. This makes it possible to calculate the injection amount within the injection amount range by simply calculating the learning value of the injection amount at one point within the injection amount range to be learned and the learning value of the correlation parameter indicating the correlation between the command injection amount and the actual torque. Can be accurately performed. In this case, it is not necessary to calculate a learning value for each of a plurality of injection amounts within the injection amount range to be learned, so the time required to complete the learning can be shortened. In addition, the correlation parameter indicating the correlation between the command injection amount and the actual torque is highly accurate because it is learned using actual data instead of using a value that is adapted in advance. Therefore, according to the above configuration, it is possible to ensure learning accuracy while shortening the period required for completion of learning in injection amount learning as much as possible.

  Here, the data relating to the injection includes the actual torque generated by performing the single injection, the actual injection amount calculated based on the actual torque, and the command injection amount when the single injection is performed.

1 is an overall schematic configuration diagram of an engine fuel supply system. FIG. The figure which shows transition of the actual injection quantity of each injection in injection for learning. The figure which showed on the xy plane the relationship between the command injection quantity in single injection and an actual torque. The flowchart which shows the process sequence of injection amount learning.

  Hereinafter, embodiments will be described with reference to the drawings. In the present embodiment, a fuel supply system that supplies fuel to an in-vehicle diesel engine as an internal combustion engine is constructed. The fuel supply system performs various controls related to fuel supply to the engine with an electronic control unit (hereinafter referred to as ECU) as a center. An overall schematic configuration diagram of this control system is shown in FIG.

  In the fuel supply system 50 of FIG. 1, the fuel tank 11 is connected to the fuel pump 20 via the fuel pipe 12. The fuel pump 20 includes a low pressure pump 21 that pumps fuel from the fuel tank 11 and a high pressure pump 22 that increases the pressure of the fuel pumped by the low pressure pump 21. In the present embodiment, the low pressure pump 21 and the high pressure pump 22 are integrated, but may be provided separately.

  The high-pressure pump 22 is a mechanical pump that sucks and discharges fuel as the engine 30 rotates. Specifically, the high-pressure pump 22 sucks low-pressure fuel as the plunger is reciprocated in synchronization with the rotation of the output shaft 31 (crankshaft) of the engine 30, increases the pressure thereof, and discharges the high-pressure fuel. To do. The fuel suction portion of the high-pressure pump 22 is provided with an electromagnetically driven suction metering valve (SCV) 23, and the fuel discharge amount (pump discharge amount) from the fuel pump 20 by energization control of the suction metering valve 23. Has been adjusted. Instead of the intake metering valve 23, a discharge metering valve (PCV) that adjusts the fuel discharge amount from the fuel pump 20 at the discharge start timing may be used.

  A common rail 13 as a pressure accumulation pipe is connected to the fuel pump 20 via a fuel pipe. High pressure fuel discharged from the fuel pump 20 is sequentially supplied to the common rail 13. Thereby, the fuel in the common rail 13 is held in a high pressure state. The common rail 13 is provided with a fuel pressure sensor 14 as pressure detecting means, and the fuel pressure sensor 14 detects the fuel pressure (rail pressure) in the common rail 13.

  The engine 30 is a multi-cylinder diesel engine, and is configured as a four-cylinder engine in this embodiment. The engine 30 is provided with an electromagnetically driven injector 32 for each cylinder 33, and high pressure fuel from the common rail 13 is supplied to each injector 32 through the high pressure fuel pipe 15. Further, fuel is injected and supplied to each cylinder 33 of the engine 30 by driving the injector 32. A recirculation pipe 16 is connected to the fuel pump 20 and the injector 32, and surplus fuel in the fuel pump 20 and the injector 32 is returned to the fuel tank 11 via the recirculation pipe 16.

  The ECU 40 is an electronic control device including a known microcomputer including a CPU, a ROM, a RAM, and the like. In addition to the detection signal of the fuel pressure sensor 14 described above, detection signals are sequentially input to the ECU 40 from various sensors such as a crank angle sensor 41 that detects the rotational speed of the engine 30 and an accelerator sensor that detects an accelerator operation amount. The ECU 40 determines an optimal fuel injection amount and injection timing based on engine operation information such as engine rotation speed and accelerator operation amount, and outputs an injection control signal corresponding to the fuel injection amount to the injector 32. Thereby, the fuel injection from the injector 32 is controlled in each cylinder 33. Further, the ECU 40 sets a target rail pressure that is a target value of the common rail pressure (injection pressure) based on the engine rotational speed and the fuel injection amount each time, and the actual rail pressure detected by the fuel pressure sensor 14 is set to the target rail pressure. The pump discharge amount is controlled so as to be a pressure.

  In the fuel injection control of this system, pilot injection that injects a minute amount of fuel as compared with main injection is performed prior to main injection of fuel (main injection). Here, in the pilot injection, it is required to perform the fine injection with high accuracy in order to sufficiently exhibit the effects of reducing the combustion noise and suppressing the generation of NOx. On the other hand, the injector 32 has a command injection amount for the injector 32, that is, an energization pulse period, and a fuel injection amount (injected from the injector 32 during the energization pulse period) due to individual differences and deterioration over time. Deviation appears between the actual injection amount). Therefore, in this system, a deviation between the command value corresponding to the minute injection amount used in the pilot injection and the actual injection amount is detected, the deviation amount is stored as a learned value, and the injector 32 is injected using the stored learned value. The amount of fuel is corrected.

  Specifically, in the injection amount learning (small injection amount learning) of the present system, for example, when the injection request of the injector 32 is not generated as in the case of vehicle deceleration, the injection amount learning occurs along with the fuel injection from the injector 32. Fuel injection from the injector 32 as a learning injection so that the actual torque matches the generated torque (hereinafter referred to as the specified torque) when fuel of a predetermined specified injection amount Qo is actually injected from the injector 32. Repeat several times. This prescribed injection amount Qo corresponds to a value within a small injection amount range used in pilot injection, and is set to 1 mm 3 / st, for example. The learning injection is executed as a single injection. The actual torque is calculated based on, for example, the engine rotation speed NEb immediately before the single injection and the fluctuation amount ΔNE of the engine rotation speed due to the single injection, and specifically, the engine rotation speed NEb, the fluctuation amount ΔNE, and a predetermined proportional constant. Is calculated by adding up.

  Then, the injection quantity Qo is injected from the injector 32 by using the injection time data belonging to the predetermined torque determination range RT in which the actual torque includes the specified torque, among the data relating to the injection of each injection time in a plurality of single injections. The difference between the command injection amount and the actual injection amount is calculated, and this difference is stored as a learning value. The “data relating to injection” includes the actual torque generated by performing the single injection, the actual injection amount calculated based on this actual torque, the command injection amount when performing the single injection, and their correlation values. .

  Further, in the present embodiment, among the data relating to the injection acquired by performing the single injection of a plurality of times, not only the data of the injection times that the actual torque belongs to the torque determination range RT but also the actual torque until the torque enters the torque determination range RT. The micro injection amount learning is carried out by effectively utilizing the injection time data, that is, the data not used for learning the specified injection amount Qo. Specifically, the correlation between the command injection amount of the injector 32 and the actual torque is shown by using the data of the injection times where the actual torque belongs to the torque determination range RT and the data of the injection times outside the torque determination range RT. The injection amount correction coefficient Kq is learned as a correlation parameter.

  FIG. 2 is a diagram showing the transition of the actual injection amount at each injection in the learning injection. In the figure, the horizontal axis indicates the number of times of single injection, and the vertical axis indicates the actual injection amount (DetectionQ). Note that the actual injection amount shown on the vertical axis is a value calculated based on the data of the actual torque generated by performing the single injection. In the engine 30, since the fuel injection amount and the generated torque are in a proportional relationship, the actual injection amount can be calculated if the actual torque is known. In FIG. 2, Q1 is an injection amount corresponding to the upper limit value of the torque determination range RT, and Q2 is an injection amount corresponding to the lower limit value of the torque determination range RT. That is, when the actual injection amount in the single injection is a numerical value within the injection amount range RQ between the upper limit value Q1 and the lower limit value Q2, the actual torque is within the torque determination range RT and is outside the injection amount range RQ. When the numerical value is, it means that the actual torque is outside the torque determination range RT. In FIG. 2, white circles indicate data outside the torque determination range RT, and black circles indicate data within the torque determination range RT.

  In the present embodiment, injection amount feedback control based on the deviation between the actual torque and the specified torque is performed as the injection amount control when the actual torque is made to coincide with the specified torque by the learning injection. Here, if the fuel injection characteristic of the injector 32 is deviated due to individual differences or aging deterioration, the actual injection amount is set to the specified injection amount in spite of commanding the injector 32 to inject the fuel of the specified injection amount Qo. It will not match Qo. In such a state, when the injection amount feedback control based on the deviation between the actual torque and the specified torque is performed, even if the deviation amount of the actual injection amount from the specified injection amount Qo is large in the period immediately after the start of the injection amount learning, FIG. As shown in FIG. 2, the actual injection amount eventually converges to the specified injection amount Qo.

  In the present embodiment, among the data relating to a plurality of injections acquired in the process until the actual torque generated by the learning injection converges to the specified torque, the injection times data (in which the actual torque substantially matches the specified torque) ( The specified injection amount Qo is learned using the black circles in FIG. In the present embodiment, learning injection is continued until a plurality of actual torque data belonging to the torque determination range RT is acquired, and there are a plurality of data belonging to the torque determination range RT. Here, the learning value is calculated from the average value of the plurality of data.

  Note that the learning value of the prescribed injection amount Qo is a correction amount for eliminating an error between the command injection amount and the actual injection amount. This learning value is a correction amount for correcting the command injection amount directly or indirectly. That is, the learning value of the prescribed injection amount Qo may be a correction amount for correcting the command injection amount itself, or may be a correction amount for correcting the injection period (energization pulse time) corresponding to the command injection amount. Good.

  Next, learning of the injection amount correction coefficient Kq will be described with reference to FIGS. FIG. 3 is a diagram showing the relationship between the command injection amount for each injection and the actual torque in the learning injection on the xy plane. The white circles and black circles in FIG. 3 correspond to the data in FIG. In the present embodiment, the command injection amount and the actual injection amount are also used by using the injection time data out of the torque determination range RT among the data relating to the injection acquired in the process until the actual torque in the learning injection converges to the specified torque. The correspondence relationship with the torque is expressed by a linear function (linear function L1), and the slope m of the linear function L1 is stored as a learning value of the injection amount correction coefficient Kq.

  The learned value thus obtained is used for correcting the injection amount of pilot injection. That is, the fuel injection amount of the pilot injection is set for each engine operation region, and there is an operation region where the specified injection amount Qo is set as the pilot injection amount, or there is an injection amount or a specified injection smaller than the specified injection amount Qo. There is also an operation region in which an injection amount larger than the amount Qo is set. In the operation region where the prescribed injection amount Qo is set as the pilot injection amount, the basic injection amount calculated based on the engine operating state (engine speed, engine load, etc.) is corrected using the learning value of the prescribed injection amount Qo. Thus, the command injection amount in the pilot injection is calculated. On the other hand, in the operation region where the value near the specified injection amount Qo is set as the pilot injection amount, the basic injection amount is corrected by using the integrated value of the specified injection amount Qo and the injection amount correction coefficient Kq. A command injection amount is calculated.

  Next, the injection amount learning processing procedure of the present embodiment will be described with reference to the flowchart of FIG. This process is executed at predetermined intervals by the microcomputer of the ECU 40.

  In FIG. 4, in step S101, it is determined whether or not a learning execution condition for injection amount learning is satisfied. In this embodiment, the learning execution condition includes (a) that the injection request of the injector 32 is not generated, (b) that the rail pressure is in a stable state, and the like. As a case corresponding to (a), for example, when the vehicle decelerates due to the accelerator being off. In step S101, an affirmative determination is made when both (a) and (b) are established.

  When the learning execution condition is satisfied, the process proceeds to step S102, and it is determined whether or not there is an unlearned engine operation region. If there is no unlearned engine operation region, this routine is terminated as it is. If there is an unlearned engine operation region, the current engine operation region to be learned is selected from the routine, and the process proceeds to step S103.

  In step S103, the pump discharge amount is adjusted by the intake metering valve 23 so that the actual rail pressure detected by the fuel pressure sensor 14 becomes the learning rail pressure. In the present embodiment, the injection amount learning is performed according to the rail pressure. Here, the rail pressure to be learned this time is selected from the unlearned rail pressure, and this is used as the learned rail pressure.

  In the following step S104, single injection as learning injection is performed. In the single injection, in the first injection after the learning execution condition is satisfied, the prescribed injection amount Qo is set as the command injection amount, and thereafter, the injection amount corresponding to the deviation between the actual torque and the prescribed torque is set. Further, in step S105, it is determined whether or not the actual torque generated by performing the single injection substantially matches the specified torque. Here, when it is determined a plurality of times that the actual torque calculated based on the engine rotation speed immediately before the single injection, the fluctuation amount of the engine rotation speed by the single injection, and a predetermined proportional constant is a value within the torque determination range RT In addition, it is determined that the actual torque substantially matches the specified torque.

  When a negative determination is made in step S105, the process proceeds to step S106, and the command injection amount and the actual torque in the current single injection are stored in association with the injection time. In step S107, the command injection amount is calculated according to the deviation between the actual torque and the specified torque. Thereafter, the processes in steps S104 to S107 are repeated until an affirmative determination is made in step S105.

  Now, when the actual torque by single injection substantially coincides with the prescribed torque, an affirmative determination is made in step S105 and the routine proceeds to step S108. In step S108, the learning value of the specified injection amount Qo is calculated using the data of the injection times in which the actual torque belongs to the torque determination range RT among the data relating to the injection acquired until the actual torque converges to the specified torque. This is memorized / updated. Specifically, the difference between the command injection amount when the actual torque and the specified torque substantially coincide with each other and the command injection amount before the start of feedback is stored and updated as a learned value of the specified injection amount Qo.

  In step S109, among the actual torque data acquired in the process until the actual torque converges to the specified torque, both the injection speed data that the actual torque belongs to the torque determination range RT and the injection speed data that does not belong to the torque determination range RT. Is used to calculate the injection amount correction coefficient Kq, and this is stored and updated as a learning value. In the present embodiment, the injection amount correction coefficient Kq is calculated using the data of all injection times acquired in the process until the actual torque converges to the specified torque.

  The learning value may be updated by comparing the currently stored learning value with the newly acquired learning value and performing the updating process when the difference is equal to or greater than a predetermined value. Although omitted in FIG. 4, the learning values of the prescribed injection amount Qo and the injection amount correction coefficient Kq are acquired for each cylinder 33 of the engine 30. Therefore, the series of processing in steps S101 to S109 is executed the number of times corresponding to the number of cylinders (four times in the present embodiment).

  According to the embodiment described in detail above, the following excellent effects can be obtained.

  When performing multiple fuel injections as learning injections, in the process of matching the actual torque to the specified torque, data that is not used for learning the specified injection amount Qo, that is, data that does not belong to the torque determination range RT is also acquired. However, in the above configuration, such data is also effectively used to calculate the injection amount correction coefficient Kq, thereby grasping the relationship between the command injection amount and the actual torque using actual data. In this case, the injection amount correction coefficient Kq, which is a correlation parameter indicating the correlation between the learning value of one point of injection amount (specified injection amount Qo in this embodiment) within the injection amount range to be learned and the command injection amount and the actual torque, By simply calculating the learning value, the injection amount within the injection amount range can be corrected with high accuracy. In addition, since it is not necessary to calculate the learning value for each of a plurality of injection amounts within the injection amount range to be learned, the time required to complete the learning can be shortened. Furthermore, the injection amount correction coefficient Kq is not a value adapted in advance but is learned using actual data, so that the accuracy is high. Therefore, according to the above configuration, it is possible to ensure learning accuracy while shortening the period required for completion of learning in injection amount learning as much as possible.

  In the minute injection amount learning, the fuel injection by the injector 32 is performed a plurality of times by injection amount feedback control based on the deviation between the actual torque and the specified torque so that the actual torque generated by the single injection coincides with the specified torque. did. In this case, the actual torque can be accurately converged to the specified torque, and as a result, the minute injection amount learning can be performed with high accuracy.

  Regarding the correlation parameter indicating the correlation between the command injection amount of the injector 32 and the actual torque, specifically, the inclination m when the relationship between the command injection amount and the actual torque at each injection time is expressed by a linear function is corrected for the injection amount. The coefficient Kq was calculated, and the calculated gradient m was used as the learning value for the injection amount correction coefficient Kq. By performing linear regression on the relationship between the command injection amount and the actual torque, the ratio (inclination m) between the change amount of the command injection amount and the change amount of the actual torque can be grasped. If the slope m is known, the actual injection amount with respect to the command injection amount in the injection amount range in the vicinity of the prescribed injection amount is accurately calculated by correcting the prescribed injection amount Qo with the slope m (= injection amount correction coefficient Kq). can do. Therefore, it is not necessary to learn the deviation amount between the command injection amount and the actual injection amount for each of the plurality of injection amounts. In particular, in the present system, since the injection amount correction coefficient Kq is actually obtained using each engine 30, when calculating the injection amount in the vicinity of the specified injection amount Qo using the learning value of the specified injection amount Qo, The accuracy can be improved as compared with the case where correction is performed using a value adapted in advance.

  For example, in a driving state where the injection request of the injector 32 is not generated, such as when the fuel is cut during deceleration of the vehicle, the engine speed is monotonously reduced unless the learning injection is performed. Therefore, by including that the injection request of the injector 32 is not generated in the learning execution condition, it is possible to accurately estimate the actual torque when the minute injection is performed as the learning injection. Thereby, the learning accuracy can be improved.

  In the pilot injection of the diesel engine 30, it is necessary to perform the injection amount control of the minute injection with high accuracy in order to sufficiently exhibit the effects of reducing combustion noise and suppressing NOx. On the other hand, in the injector 32, there is a deviation in the relationship between the energization period (command injection amount) of the injector 32 and the actual injection amount due to individual differences and deterioration over time. In this respect, in the present embodiment, since the present invention is applied to learning of a small injection amount of pilot injection, the accuracy of pilot injection can be improved.

(Other embodiments)
The present invention is not limited to the description of the above embodiment, and may be implemented as follows, for example.

  In the above embodiment, the injection amount feedback control based on the deviation between the actual torque and the specified torque is performed as the injection amount control when the actual torque matches the specified torque by the learning injection. The actual torque may be made equal to the specified torque by increasing or decreasing by a predetermined amount. For example, when the actual torque is smaller than the specified torque in the first injection after the learning execution condition is satisfied, the command injection amount is increased by a predetermined amount in the subsequent single injection.

  -If the actual torque belongs to the torque determination range RT from the first injection after the learning execution condition is satisfied, or if the number of data deviating from the torque determination range RT is small, the learning accuracy of the injection amount correction coefficient Kq is guaranteed. It may not be possible. In consideration of this point, in the data relating to the injection at each injection time obtained by the learning injection, if there is no data for which the actual torque deviates from the torque determination range RT or the predetermined number or less, the injection amount correction coefficient Kq It is good also as a structure which prohibits learning. With such a configuration, it is possible to avoid a decrease in learning accuracy of the injection amount correction coefficient Kq. In addition, when learning of the injection amount correction coefficient Kq is prohibited, a value adapted in advance may be used, or the previous learned value may be used as it is.

  In the above embodiment, the command injection amount of the initial injection after the learning execution condition is satisfied is the prescribed injection amount Qo (for example, 1 mm 3 / st), but may be a value other than the prescribed injection amount Qo. For example, the command injection amount of the initial injection is set to a value (for example, 2 mm 3 / st) larger than the specified injection amount Qo, and feedback based on the deviation between the actual torque and the specified torque is performed in the subsequent single injection. In this case, the number of data from which the actual torque deviates from the torque determination range RT can be intentionally increased, and the learning accuracy of the injection amount correction coefficient Kq can be increased.

  In the above embodiment, the actual torque is calculated based on the engine rotation speed immediately before the single injection, the fluctuation amount of the engine rotation speed by the single injection, and a predetermined proportional constant. However, the torque of the engine 30 can be directly detected. A simple sensor may be provided, and the injection amount learning may be performed using the actual torque detected by the sensor.

  In the above embodiment, the specified injection amount Qo is learned using the data of a plurality of injections whose actual torque belongs to the torque determination range RT, but the specified injection amount Qo is learned using the data of one injection. May be.

  The present invention can be applied to a vehicle other than a common rail fuel injection system for a diesel engine. For example, the present invention can be applied to a pressure accumulation type fuel injection system of a direct injection type gasoline engine. It can also be applied to engines other than those for vehicles.

  DESCRIPTION OF SYMBOLS 10 ... Engine, 13 ... Common rail (pressure accumulation piping), 30 ... Engine (internal combustion engine), 32 ... Injector (fuel injection valve), 40 ... ECU (injection control means, 1st learning means, 2nd learning means), 50 ... Fuel supply system, RT ... Torque judgment range.

Claims (5)

An internal combustion engine (30) that is applied to a fuel supply system (50) of an internal combustion engine including a fuel injection valve (32) that injects high-pressure fuel in a pressure accumulating pipe (13) and learns the fuel injection amount from the fuel injection valve. The injection amount learning device of
When the learning execution condition is satisfied, the actual torque generated by the fuel injection from the fuel injection valve is actually injected from the fuel injection valve by a predetermined small injection amount of fuel. Injection control means for performing fuel injection from the fuel injection valve a plurality of times as learning injection so as to coincide with a prescribed torque that is a generated torque of
Of the data relating to the injection of each injection in the multiple fuel injections by the injection control means, the minute injection amount is determined using the data of the injections that belong to the predetermined torque determination range including the specified torque. A first learning means for learning;
Of the data relating to the injection at each injection in a plurality of times of fuel injection by the injection control means, the data of the injection that the actual torque belongs to the torque determination range and the data of the injection that is out of the torque determination range are used. A second learning means for learning a correlation parameter indicating a correlation between the command injection amount of the fuel injection valve and the actual torque;
An injection amount learning device for an internal combustion engine, comprising:   The injection amount learning device for an internal combustion engine according to claim 1, wherein the injection control means performs the learning injection by injection amount feedback control based on a deviation between the actual torque and the specified torque.   The second learning means calculates, as the correlation parameter, an inclination when the relationship between the command injection amount and the actual torque at each injection time is expressed as a linear function, and stores the calculated inclination as a learning value. Item 3. An injection amount learning device for an internal combustion engine according to Item 1 or 2. The execution condition includes that an injection request for the fuel injection valve is not generated,
The injection amount learning device for an internal combustion engine according to any one of claims 1 to 3, wherein the injection control unit performs the learning injection when the injection request is not generated. The minute injection amount is a value within an injection amount range of pilot injection performed prior to main injection,
The first learning hand stage, the internal combustion engine of the injection amount learning device according to claim 1 to carry out the injection amount learning of the pilot injection as the learning.

Images