JP2009052417A - Fuel injection control device of diesel engine and method of learning fuel injection quantity of diesel engine - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fuel injection control device of a diesel engine enabling a reduction in a time requiring for completing a learning. <P>SOLUTION: An ECU instructs an injection for learning to an injector of a specific cylinder when the requirements for learning are satisfied (step S1), detects the rotational speed ω of an engine at that time, calculates a torque proportional amount Tp by calculating a rotational variation amount ▵ω, rotational speed rise amounts δ of cylinders, and their averaged value δx, and calculates the actual injection amount Qest of a fuel from the injector according to the torque proportional amount Tp (steps S2, S3). When an amount obtained by multiplying an adjustment gain k by the difference between the actual injection amount Qest and the injection amount Qtrg instructed to the injector is calculated as a final learned value (step S9), the adjustment gain k is variably set according to the TQ-Q sensitivity of the injector (steps S7, S8). <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a fuel injection control device that performs injection amount learning of a diesel engine and a method for learning the injection amount.

Conventionally, in a diesel engine, a method of performing so-called pilot injection in which a very small amount of fuel is injected prior to main injection is known as means for reducing combustion noise and suppressing NOx (nitrogen oxide). However, when performing pilot injection with a small injection amount command value, it is required to improve the micro-injection accuracy in order to fully exhibit the effects (reduction of combustion noise, suppression of NOx). In order to improve the micro-injection accuracy, it is necessary to perform injection amount learning for detecting and correcting a deviation between the command injection amount for pilot injection and the actually injected fuel amount (hereinafter referred to as an actual injection amount).
However, since it is difficult to directly measure the actual injection amount while the vehicle is running, conventionally, for example, a method for detecting the actual injection amount with an A / F value (air-fuel ratio), an in-cylinder pressure or the like (see Patent Document 1). ) And a method of correcting the injection amount deviation from the correction amount in the rotational speed feedback such as ISC (idle rotational speed control) (see Patent Document 2).

  However, in the method of Patent Document 1, since an A / F sensor or an in-cylinder pressure sensor that is generally not installed in a vehicle is used, additional equipment is required, resulting in a problem of significant cost increase. Further, in the method of Patent Document 2, the engine-to-cylinder variation and engine load factors such as an external load such as an air conditioner fluctuate, so that the unambiguous relationship between the engine speed and the injection amount is lost. There is. Since learning is performed from the relationship between the rotational speed and the command injection amount that are balanced in a state in which such a change in load factor is applied, it is difficult to perform highly accurate injection amount learning.

Therefore, in Patent Document 3, in order to solve the above-described problem, a single injection is performed at the time of non-injection when the command injection amount to the injector is zero or less, and the amount of change in the engine speed (the number of rotations) that is increased by the single injection The product of the increase amount δx) and the engine speed ω0 when single injection is performed is calculated as a torque proportional amount Tp, the actual injection amount is estimated from the generated torque calculated from the torque proportional amount Tp, and the estimated actual injection A method has been proposed in which the injection amount learning is performed with high accuracy without detecting an additional equipment such as a torque sensor by detecting the difference between the amount and the command injection amount as an injection amount deviation.
JP 11-294227 A JP 2002-295291 A JP 2005-36788 A

  However, considering the time required for completing the learning, that is, the convergence of learning in the method of Patent Document 3, the convergence is determined by the sensitivity between the injection period TQ and the injection amount Q of the injector (TQ-Q characteristic). It depends on the size of. As shown in FIG. 7, the magnitude of the slope of the linear TQ-Q characteristic corresponds to the magnitude of the sensitivity of the injector.

FIG. 8 shows a process when learning is performed for an injector having a relatively low sensitivity. A straight line indicated by a two-dot chain line in the figure is a TQ-Q characteristic (characteristic map) given as an injector specification, and learning is started using this characteristic map as a reference. The actual injection characteristic in the unlearned state is indicated by a solid line.
As shown in FIG. 8A, when the target injection amount Qtrg. Is expected and the injection period TQ1 is given based on the characteristic map, it is assumed that the actual injection amount Q1 has a larger value. In order to correct the difference ΔQ between the two, if the injection period difference ΔTQ1 is obtained as a learning value on the characteristic map, (TQ1−ΔTQ1) is set as the injection period in the next learning. However, since the sensitivity of the actual injection characteristic is smaller than that of the characteristic map, the correction is insufficient. As a result, the actual injection amount Q2 is still large, and the injection period difference ΔTQ2 is obtained as the learning value. By repeating the learning pattern as described above, the learning converges so as to approach the “true learning amount”.

  On the other hand, as shown in FIG. 9, when learning is performed for an injector with relatively high sensitivity, if the learning value ΔTQ1 obtained in the first learning shown in (a) is reflected in the next learning, the correction becomes excessive (excessive). Correction), the actual injection amount Q2 falls below the characteristic map [see (b)]. Thus, learning converges including the case where the sign of the learning value is inverted.

  For such learning, in general, an upper limit is set for the number of repetitions of learning from an allowable learning completion time or distance, or learning accuracy, or a restriction for avoiding interference with other controls. This is common (see FIG. 10). As shown in FIG. 7, when the variation in TQ-Q sensitivity based on the characteristic map is within a relatively narrow range A, the sensitivity is relatively small within the range A as shown in FIG. In any of the cases shown in FIG. 11B, which has a relatively high sensitivity within the range A, it can be expected that the learning is completed before the allowable number of repetitions is reached.

On the other hand, as shown in FIG. 7, in the range B where the variation in the TQ-Q sensitivity is relatively wide, in the case of FIG. In the case of FIG. 12B, which is small and has relatively high sensitivity in the range B, the correction amount obtained in each learning greatly changes and vibrates, so that learning is expected to complete before reaching the allowable number of repetitions. become unable. Especially in the latter worst case, there is a possibility that the learning amount may not be completed due to the diverging amount.
As described above, when the time required for completion of learning becomes long, not only the time efficiency is deteriorated, but also the injector is deteriorated, or the detection ability for the change in the injection amount due to the difference in the environment or the difference in the fuel is lowered. As a result, the learning accuracy generally deteriorates.

  The present invention has been made in view of the above circumstances, and its object is to provide a fuel injection control device for a diesel engine capable of reducing the time required for completion of learning even when the sensitivity of the injectors varies, and It is to provide a fuel injection amount learning method for a diesel engine.

According to the fuel injection control device for a diesel engine according to claim 1, when it is determined that the learning condition is satisfied, the injection for learning is instructed to the injector of the specific cylinder, and the rotational speed of the diesel engine in that case is set. Detected as engine speed. Then, the fluctuation amount of the engine speed when the learning injection is performed and when it is not performed is calculated as the engine speed increase amount, and based on the engine speed increase amount, the fuel actually injected from the injector is calculated. The actual injection amount is calculated. When the injection correction amount calculation means calculates the injection correction amount by multiplying the difference between the calculated actual injection amount and the command injection amount commanded to the injector by the adjustment gain, the adjustment gain setting means The gain is variably set according to the sensitivity of the fuel injection amount with respect to the injection period in the injector.
That is, as described above, the learning completion time varies depending on the sensitivity of the injector, and therefore, when calculating the injection correction amount, an adjustment gain that multiplies the difference between the actual injection amount and the command injection amount is determined according to the sensitivity. If it is variably set, the adjustment gain can be optimized, and the time required for completion of learning can be shortened appropriately even when the sensitivity of the injector varies.

  According to the fuel injection control device for a diesel engine according to claim 2, when the injection correction amount calculation means repeats the learning process while the difference between the actual injection amount and the command injection amount is outside the predetermined range, the adjustment gain setting is performed. The means counts the number of repetitions of the learning process, and sets the absolute value of the adjustment gain to increase as the number of repetitions increases. That is, when the number of repetitions of the learning process increases, it is estimated that the sensitivity of the injector is relatively small. Therefore, the convergence of the learning result can be accelerated by increasing the absolute value of the adjustment gain.

  According to the fuel injection control device for a diesel engine according to the third aspect, when the number of repetitions exceeds the upper limit value, the adjustment gain setting means thereafter maintains the adjustment gain constant. That is, if the learning result does not converge even after repeating the learning process to some extent, it is estimated that there is some abnormality in the fuel injection control system. Therefore, in such a case, the adjustment is stopped by keeping the adjustment gain constant without changing it.

  According to the fuel injection control device for a diesel engine as set forth in claim 4, the adjustment gain setting means starts the variable setting of the adjustment gain when the number of repetitions of the learning process exceeds a predetermined threshold. That is, if the sensitivity of the injector is comparatively small and comparatively large, the learning completion time is shorter in the former. Therefore, if the adjustment gain is variably set in the latter case, the learning completion time is reduced. It can be shortened effectively. When the sensitivity of the injector is high to some extent, the learning result tends to vibrate as described above, so that the learning process is repeated more frequently. Therefore, it is possible to appropriately determine the timing for changing the adjustment gain from the initial value based on the number of repetitions.

  According to the fuel injection control device for a diesel engine according to claim 5, the adjustment gain setting means compares the sign of the last injection correction amount of the previous time with that of the current time, and adjusts according to whether the signs of both coincide or differ. Set gain increase / decrease. That is, when the sign of the last injection correction amount of the previous time coincides with the sign of the current injection, the correction amount in learning tends to be insufficient, and conversely, when the sign is different, the correction amount in learning tends to be excessive. It is estimated to be. Therefore, if the increase / decrease of the adjustment gain is set in accordance with the coincidence / non-coincidence of the codes, the correction amount given in the learning process can be optimized and the learning completion time can be further shortened.

  An embodiment of the present invention will be described below with reference to FIGS. FIG. 6 is an overall configuration diagram showing a fuel injection system of a diesel engine. The fuel injection system is applied to, for example, a four-cylinder diesel engine (hereinafter referred to as engine 1). The common rail 2 that stores high-pressure fuel and the fuel pumped from the fuel tank 3 are pressurized and supplied to the common rail 2. A fuel supply pump 4, an injector 5 for injecting high-pressure fuel supplied from the common rail 2 into the cylinder (combustion chamber 1a) of the engine 1, and an electronic control unit (hereinafter referred to as ECU (Electronic Control Unit)) 6 for electronically controlling the system For example).

  The common rail 2 has a target rail pressure set by the ECU 6 and accumulates the high-pressure fuel supplied from the fuel supply pump 4 to the target rail pressure. The common rail 2 includes a pressure sensor 7 that detects accumulated fuel pressure (hereinafter referred to as rail pressure) and outputs it to the ECU 6, and a pressure limiter that limits the rail pressure so as not to exceed a preset upper limit value. 8 is attached.

  The fuel supply pump 4 includes a camshaft 9 that is driven by the engine 1 to rotate, a feed pump 10 that is driven by the camshaft 9 to pump fuel from the fuel tank 3, and a cylinder 11 in synchronization with the rotation of the camshaft 9. It has a plunger 12 that reciprocates inside, and an electromagnetic metering valve 14 that regulates the amount of fuel drawn from the feed pump 10 into the pressurizing chamber 13 in the cylinder 11.

  In the fuel supply pump 4, when the plunger 12 moves in the cylinder 11 from the top dead center toward the bottom dead center, the fuel delivered from the feed pump 10 is metered by the electromagnetic metering valve 14 and sucked. The valve 15 is pushed open and sucked into the pressurizing chamber 13. Thereafter, when the plunger 12 moves in the cylinder 11 from the bottom dead center to the top dead center, the fuel in the pressurizing chamber 13 is pressurized by the plunger 12, and the pressurized fuel passes through the discharge valve 16. Pushed open and pumped to the common rail 2.

The injector 5 is mounted for each cylinder of the engine 1 and is connected to the common rail 2 via a high-pressure pipe 17. The injector 5 includes an electromagnetic valve 5a that operates based on a command from the ECU 6, and a nozzle 5b that injects fuel when the electromagnetic valve 5a is energized.
The solenoid valve 5a opens and closes a low-pressure passage (not shown) that leads from the pressure chamber (not shown) to which the high-pressure fuel of the common rail 2 is applied to the low-pressure side. Shut off the low pressure passage.

  The nozzle 5b incorporates a needle (not shown) that opens and closes the nozzle hole, and the fuel pressure in the pressure chamber urges the needle in the valve closing direction (direction in which the nozzle hole is closed). Accordingly, when the low pressure passage is opened by energization of the electromagnetic valve 5a and the fuel pressure in the pressure chamber decreases, the needle rises in the nozzle 5b and opens (opens the nozzle hole), thereby being supplied from the common rail 2. High pressure fuel is injected from the nozzle hole. On the other hand, when the low pressure passage is blocked by stopping energization of the electromagnetic valve 5a and the fuel pressure in the pressure chamber rises, the needle descends in the nozzle 5b and closes, thereby terminating the injection.

  The ECU 6 includes a rotational speed sensor (rotational speed detection means) 18 that detects an engine rotational speed (rotational speed per minute, engine rotational speed), and an accelerator opening sensor (not shown) that detects an accelerator opening (engine load). And a pressure sensor 7 for detecting the rail pressure, etc., and based on sensor information detected by these sensors, the target rail pressure of the common rail 2 and the injection timing suitable for the operating state of the engine 1 and The injection amount and the like are calculated, and the electromagnetic metering valve 14 of the fuel supply pump 4 and the electromagnetic valve 5a of the injector 5 are electronically controlled according to the calculation result.

  Further, in the injection amount control (control of injection timing and injection amount) by the ECU 6, when performing the minimum amount of pilot injection prior to the main injection, the injection amount learning for the pilot injection is performed. The ECU 6 functions as a learning condition determination unit, a learning injection command unit, a rotation speed increase amount calculation unit, an injection amount calculation unit, an injection correction amount calculation unit, an injection amount correction unit, and an adjustment gain setting unit according to the present invention. have.

  Next, the operation of this embodiment will be described with reference to FIGS. First, a processing procedure in which the ECU 6 performs the injection amount learning will be described based on the flowcharts shown in FIGS. 1 and 2. The basic part of the learning process is the same as that disclosed in Patent Document 3. When a predetermined learning condition is satisfied, the ECU 6 gives a single-quantity injection amount Qtrg. (Command injection amount, actually determined by the injection period TQ) to the injector 5 of the specific cylinder, and single injection for learning. (Step S1), and the state change amount of the engine 1 in that case is detected by each sensor or the like (step S2).

  The detection of the state change amount here is detection of the rotational speed ω of the engine 1, the rotational speed fluctuation amount Δω for each cylinder, as shown in the flowchart of FIG. Calculation of the number increase amount δ, calculation of the average value δx, calculation of the torque proportional amount Tp (Tp = δx · ω0), and the like. FIG. 5 is a view corresponding to FIG. 1 of Patent Document 3 and shows the basic contents of the learning process in the present embodiment. In this example, when the single injection for learning is performed, the rotational speed fluctuation amount Δω3 is calculated for the third cylinder.

In the subsequent step S3, an estimated actual injection amount Qest. Of the injector 5 is calculated. The estimated actual injection amount Qest. Is calculated by the inverse function because the torque proportional amount is represented by a linear function of the estimated actual injection amount Qest. (Tp = a · Qest. + B). Next, when a deviation amount ΔQ between the eye quantity injection amount Qtrg. And the estimated actual injection amount Qest. Is calculated (step S4), whether the deviation amount ΔQ is within a predetermined range A, that is, | ΔQ | ≦ A ( It is determined whether or not it is a learning completion determination value (step S5). If the condition | ΔQ | ≦ A is satisfied (YES), the learning amount is fixed at that time (step S11), and the learning process ends.
On the other hand, if | ΔQ |> A in step S5 (NO), the counter n for counting the number of repetitions of the learning process is counted up (step S6). Then, a data variation pattern and a learning amount are calculated (step S7).

Here, the process of step S7 will be described with reference to FIG. First, when the previous learning amount TQGold is written and stored in the memory (step S7_1), the current learning amount TQG is calculated from the deviation amount ΔQ and the TQ-Q characteristic map (step S7_2). Then, it is determined whether or not the signs of the current learning amount TQG and the previous learning amount TQGold are the same (step S7_3). If they are the same (YES), the learning variation pattern PAT is set to “1” (step S7_4). If the signs are different from each other (NO), the learning variation pattern PAT is set to “2” (step S7_5).
That is, the fact that the sign of the current learning amount and the previous learning amount are the same indicates that the correction is in the process of converging to that value without exceeding the “true learning amount” (insufficient correction). On the other hand, the fact that the signs of the two are different indicates that the previous correction was performed excessively beyond the “true learning amount” (overcorrection).

  Reference is again made to FIG. In step S8 following step S7, the adjustment gain k is determined according to the number of repetitions indicated by the counter n and the value of the learning variation pattern PAT. The adjustment gain k is determined by the table shown in FIG. FIGS. 3A and 3B correspond to the cases of PAT = 1 and 2, respectively. When PAT = 1, the adjustment gain k starts from the initial value “1”, and the number of repetitions n increases to some extent. Increase gradually from the stage. On the other hand, when PAT = 2, the adjustment gain k starts from the initial value “1” and gradually decreases from the stage where the number of repetitions n has increased to some extent.

  That is, in any case, the threshold Nth is set for the number of repetitions n, and the adjustment gain k is set to maintain the initial value “1” in the region where the number of repetitions n is small (n ≦ Nth). Yes. Then, the adjustment gain k is increased or decreased from the initial value “1” from the region (n> Nth). In any case, an upper limit Nlim is set for the number of repetitions n, and after reaching the upper limit Nlim, when PAT = 1, the adjustment gain k is constant at the maximum value “1.7”. In the case of = 2, the adjustment gain k is constant at the minimum value “0.3”. If the number of repetitions n reaches the upper limit Nlim, it means that the correction does not converge to the “true learning amount” even if learning is repeated to some extent, so it is assumed that there is some abnormality in the control system. Therefore, in this case, it is preferable to take measures such as stopping the learning process and notifying abnormality.

  When the adjustment gain k is determined in step S8, the final learning amount is calculated by multiplying the learning amount TQG calculated in step S7 by the adjustment gain k (step S9). Then, the final learning amount is reflected in the next single injection command (step S10), and the process returns to step S1.

  FIG. 4 explains the effect when the adjustment gain k is variably set as described above. As indicated by a broken line in the figure, unlike the present embodiment, when the adjustment gain k is not reflected in the learning process, the sensitivity of the injector 5 is relatively large and the learning pattern vibrates, so that the allowable number of repetitions Nlim is reached. Suppose that there is a case where learning is not completed. When the adjustment gain k is reflected on this learning pattern as in the present embodiment, the adjustment gain k having a large value according to the sensitivity level acts. As a result, the learning converges faster as shown by the solid line. Improvement is made so that learning is completed before the number of repetitions Nlim is reached.

  As described above, according to this embodiment, the ECU 6 instructs the injector 5 for a specific cylinder to perform learning injection when the learning condition is satisfied, detects the rotational speed ω of the engine 1 in that case, and rotates the engine. The torque proportional amount Tp is calculated by calculating the fluctuation amount Δω, the rotational speed increase amount δ of each cylinder, and the average value δx thereof, and the actual fuel injection amount Qest. From the injector 5 is calculated based on the torque proportional amount Tp. To do. When the difference between the actual injection amount Qest. And the injection amount Qtrg. Commanded to the injector 5 is multiplied by the adjustment gain k, the adjustment gain k is calculated as TQ-Q of the injector 5. Variably set according to sensitivity.

That is, the learning completion time varies depending on the sensitivity of the injector 5, so that the adjustment gain k can be optimized if the adjustment gain k to be multiplied when calculating the final learning amount is variably set according to the sensitivity. Thus, even when the sensitivity of the injector 5 varies, the time required for completion of learning can be shortened appropriately.
When the learning process is repeated while the difference ΔQ between the actual injection amount Qest. And the command injection amount Qtrg. Is outside the predetermined range A, the ECU 6 counts the number of repetitions n of the learning process, and the repetition number n is Since the absolute value of the adjustment gain k is set to increase as it increases, the convergence of the learning result can be accelerated according to the sensitivity of the injector 5.

Further, since the ECU 6 keeps the adjustment gain k constant after the repetition number n exceeds the upper limit value Nlim, the ECU 6 changes the adjustment gain k when it is estimated that there is some abnormality in the fuel injection control system. Without adjustment, stop the adjustment. In addition, since the adjustment gain k is changed from the initial value “1” when the number of repetitions n exceeds the threshold value Nth, the adjustment gain k is variably set when the sensitivity of the injector 5 is estimated to be relatively high. The time required for completion of learning can be effectively shortened. When the sensitivity of the injector 5 is high to some extent, the learning result shows a tendency to vibrate, and the learning process is repeated more frequently. Therefore, the timing for changing the adjustment gain k from the initial value can be appropriately determined based on the number of repetitions n. .
Further, the ECU 6 compares the sign of the last injection correction amount of the previous time with that of the current time, and sets the increase / decrease of the adjustment gain k depending on whether the signs are the same or different. The time required for completion of learning can be further reduced by optimization.

The present invention is not limited to the embodiments described above and shown in the drawings, and the following modifications or expansions are possible.
The adjustment gain k may be changed from the initial value in a region where the number of repetitions n of the learning process is relatively small.
The upper and lower limits of the adjustment gain k may be changed and set as appropriate.
The upper limit Nlim for the number of repetitions n and the upper limit and the lower limit for the adjustment gain k may be set as necessary.
A process of determining whether or not the signs of the previous and current learning amounts match and sets the increase / decrease of the adjustment gain k according to the result may be performed as necessary.
The adjustment gain k may be determined by calculation using an appropriate function.

The flowchart which is one Example of this invention, and shows the process sequence which performs injection amount learning The flowchart which shows the detail of step S7 of FIG. The figure which shows the table for determining the adjustment gain k Diagram showing an example of a learning pattern Timing chart showing the process of injection quantity learning Overall configuration diagram showing the fuel injection system of a diesel engine The figure explaining the dispersion | variation in the TQ-Q characteristic sensitivity of an injector The figure explaining the learning process when the sensitivity is comparatively small Equivalent to FIG. 8 when sensitivity is relatively high The figure which shows the relationship between the number of repetitions of learning processing, and learning completion time The figure which shows an example of the learning pattern when the variation range of a sensitivity is comparatively narrow FIG. 11 equivalent diagram when the sensitivity variation range is relatively wide

Explanation of symbols

  In the drawings, 1 is an engine (diesel engine), 2 is a common rail, 5 is an injector, 6 is an ECU (learning condition determination means, learning injection command means, rotation speed increase amount calculation means, injection amount calculation means, injection correction amount calculation) Means, injection amount correction means, adjustment gain setting means), 18 denotes a rotation speed sensor (rotation speed detection means).

Claims (10)

Learning condition determination means for determining whether or not a learning condition for performing fuel learning injection is satisfied for an injector in a specific cylinder of a diesel engine;
A learning injection command means for commanding a learning injection to an injector of the specific cylinder when it is determined that the learning condition is satisfied;
A rotational speed detection means for detecting the rotational speed of the diesel engine as an engine rotational speed;
A rotational speed increase amount calculating means for calculating a fluctuation amount of the engine rotational speed when the learning injection is performed and when it is not performed as a rotational speed increase amount;
An injection amount calculating means for calculating an actual injection amount of fuel actually injected from the injector based on the calculated rotation speed increase amount;
An injection correction amount calculation means for calculating an injection correction amount by multiplying the difference between the calculated actual injection amount and the command injection amount commanded to the injector by an adjustment gain;
A fuel injection control device for a diesel engine, comprising: adjustment gain setting means for variably setting the adjustment gain in accordance with a sensitivity of a fuel injection amount with respect to an injection period in the injector. The injection correction amount calculation means repeats the learning process while the difference between the actual injection amount and the command injection amount is outside a predetermined range,
The diesel engine according to claim 1, wherein the adjustment gain setting means counts the number of repetitions and sets the absolute value of the adjustment gain to increase as the number of repetitions increases. Engine fuel injection control device.   3. The fuel injection control device for a diesel engine according to claim 2, wherein, when the number of repetitions exceeds an upper limit value, the adjustment gain setting means maintains the adjustment gain constant thereafter.   4. The fuel injection control device for a diesel engine according to claim 2, wherein the adjustment gain setting means starts the variable setting of the adjustment gain when the number of repetitions exceeds a predetermined threshold value. 5. .   The adjustment gain setting means compares the sign of the last and current final injection correction amounts, and sets the increase / decrease of the adjustment gain according to whether the signs match or differ. Item 5. The fuel injection control device for a diesel engine according to any one of Items 2 to 4. For the injector in a specific cylinder of the diesel engine, determine whether or not a learning condition for performing fuel learning injection is satisfied,
When it is determined that the learning condition is satisfied, the learning injection is commanded to the injector of the specific cylinder,
Detecting the rotational speed of the diesel engine as the engine speed,
Calculating the amount of fluctuation of the engine speed when the learning injection is performed and the case where it is not performed as the engine speed increase amount;
Based on the calculated rotation speed increase amount, the actual injection amount of the fuel actually injected from the injector is calculated,
A difference between the calculated actual injection amount and the command injection amount commanded to the injector multiplied by an adjustment gain is calculated as an injection correction amount;
A fuel injection amount learning method for a diesel engine, wherein the adjustment gain is variably set according to sensitivity of a fuel injection amount with respect to an injection period in the injector. While the difference between the actual injection amount and the command injection amount is outside the predetermined range, the learning process is repeated,
7. The method for learning the fuel injection amount of a diesel engine according to claim 6, wherein the number of repetitions is counted and the absolute value of the adjustment gain is set to increase as the number of repetitions increases. .   8. The method for learning the fuel injection amount of a diesel engine according to claim 7, wherein when the number of repetitions exceeds an upper limit value, the adjustment gain is maintained constant thereafter.   9. The method for learning the fuel injection amount of a diesel engine according to claim 7, wherein the adjustment gain is variably set when the number of repetitions exceeds a predetermined threshold value.   10. The sign of the final injection correction amount of the previous time and that of the current time are compared, and the increase / decrease of the adjustment gain is set according to whether the signs of the both coincide or differ. The fuel injection amount learning method of the diesel engine as described in 2.

Images