JP2009002301A - Fuel injection control device of diesel engine - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fuel injection control device of a diesel engine which calculates the deviation between an actual fuel injection volume and a command injection volume from a fuel injection device as a learning value, which corrects the fuel injection volume, and which can prevent a deterioration in operation performance of the diesel engine by the learning. <P>SOLUTION: In the learning processing which sets a correction value of the fuel injection volume by detecting a deviation between the actual fuel injection volume and the command injection volume from the injector as a learning value, speed-reduction type learning control (S110) and multi-stage type learning control (S130) are performed in turns. The speed-reduction type learning control (S110) calculates a learning value ▵Q1i for every cylinder at the time of speed reduction of the engine, and the multi-stage type learning control (S130) calculates a learning value ▵Q2i for every cylinder at the time of idle operation of the engine. When the learning values ▵Q1i, ▵Q2i obtained from each of the learning control are nearly in agreement with each other (S160-NO), the average value of each of the learning values is set as a correction value ▵Qi. When that is not the case (S160-YES), a learning value of which the deviation between the average value of all the cylinders is smaller is set as a correction value ▵Qi (S180 to S210). <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a fuel injection control device for a diesel engine that learns a difference between an actual fuel injection amount to a diesel engine and a command injection amount and corrects the fuel injection amount.

  Conventionally, in a diesel engine for a vehicle, pilot injection is performed in which a very small amount of fuel is injected prior to main injection in order to reduce combustion noise and to suppress NOx. In this pilot injection, fuel injection is performed. If there is a difference between the actual fuel injection amount from the device and the command injection amount, the effect cannot be fully exhibited. Also in main injection, if there is a difference between the actual fuel injection amount from the fuel injection device and the command injection amount, combustion noise, large vibrations, and emission deterioration of the diesel engine occur.

  Therefore, in a fuel injection control device for a diesel engine, the difference between the actual fuel injection amount from the fuel injection device and the command injection amount is usually calculated as a fuel injection learning value, and the diesel engine is supplied to the diesel engine based on the learning value. A so-called learning control for correcting the fuel injection amount is performed.

As this learning control, various methods are known as in the following (1) to (6), and one of them is selectively used in a diesel engine.
(1) During deceleration operation in which fuel injection to the diesel engine is stopped, a minute amount of fuel is injected from each fuel injection device for each cylinder, and the rotational speed and rotation fluctuation of the diesel engine caused by the single injection are detected. Detecting, from the detection result, the torque generated in the diesel engine, that is, the actual fuel injection amount is estimated, the deviation between the actual fuel injection amount and the command injection amount is obtained, and set as a learning value, a deceleration type Learning control (see, for example, Patent Document 1).
(2) During the idling operation of the diesel engine, the fuel injection from the injector is performed for each cylinder in a plurality of times, and the rotational fluctuation between the cylinders is determined based on the rotational speed and rotational fluctuation of the diesel engine generated at that time. A first fuel injection correction amount for smoothing and a second fuel injection correction amount for controlling the average rotation speed of all the cylinders to the target rotation speed (idle rotation speed) are obtained, and these two fuel injection correction amounts are obtained. Multi-stage learning control in which a learning value is set as a parameter representing a deviation between the actual fuel injection amount and the command injection amount (see, for example, Patent Document 2).
(3) When the operation state of the diesel engine is stable, the in-cylinder pressure after fuel injection is detected for each cylinder by using the in-cylinder pressure sensor, and obtained from the detected in-cylinder pressure and the operation state of the diesel engine. In-cylinder pressure detection learning control (for example, patent), which calculates a pressure difference from the in-cylinder pressure as a parameter representing a deviation between the actual fuel injection amount and the command injection amount and updates the learning value in a direction to eliminate the pressure difference. Reference 3 etc.).
(4) When the operation state of the diesel engine is stable and the fuel supply to the common rail that stores high-pressure fuel is stopped, the injectors of each cylinder are driven within a range that does not cause fuel injection into the cylinders ( The amount of fuel that leaks from the injector of each cylinder to the fuel discharge passage during fuel injection is determined by measuring the pressure drop per unit time of the common rail pressure. Learning control of a common rail pressure detection method in which a learning value is set as a parameter representing a deviation from the amount (see, for example, Patent Document 4).
(5) Under a predetermined operating condition of the diesel engine, for each cylinder, by detecting an ionic current flowing in the glow plug due to ions generated in the cylinder at the time of fuel fuel, an ignition period of the fuel is obtained. The difference between the average ignition period of all cylinders and the average ignition period of all cylinders, and the difference between the average ignition period of all cylinders and the target ignition period are obtained as parameters representing the deviation between the actual fuel injection amount and the command injection amount. Learning control of an ion current detection method in which a learning value is updated so as to eliminate a shift in the ignition period (see, for example, Patent Document 5).
(6) A predetermined exhaust gas composition such as a hydrocarbon concentration is measured under a predetermined operating condition of the diesel engine, and a deviation of the measurement result from a target value is determined as a deviation between the actual fuel injection amount and the command injection amount. Learning control of an exhaust gas composition detection method that is obtained as a parameter to be expressed and updates the learning value so that the deviation is eliminated (see, for example, Patent Document 6).
JP 2005-36788 A JP 2003-254139 A JP 2004-251208 A Japanese Patent Application Laid-Open No. 2005-307885 Japanese Patent Laid-Open No. 10-252542 JP 11-294227 A

  However, in the conventional fuel injection control device, the learning value obtained by one of the plurality of learning controls is used to correct the fuel injection amount injected and supplied into the cylinder of the diesel engine. If erroneously learned due to a disturbance or the like, the fuel injection amount cannot be corrected appropriately.

  In this case, the intended purpose of the learning control such as reduction of combustion noise and NOx suppression cannot be achieved, and in some cases, combustion noise and vibration are increased, causing driver discomfort. It can also be considered.

  The present invention has been made in view of these problems, and calculates the deviation between the actual fuel injection amount from the fuel injection device and the command injection amount as a learning value and corrects the fuel injection amount, thereby correcting the fuel injection amount. The purpose is to prevent the operation performance of the diesel engine from deteriorating due to mislearning.

  The fuel injection control device for a diesel engine according to claim 1, which has been made to achieve the above object, operates under different learning conditions, and the difference between the command injection amount for the fuel injection device and the actual fuel injection amount. A learning value that represents a learning value is provided, and the comparing means compares the learning values calculated by the learning means, and the difference between the learning values is within an allowable range. In addition, a correction value for the fuel injection amount is set using at least one of the plurality of learning values.

  Thus, in the fuel injection control device of the present invention, the learning values obtained by the plurality of learning means are mutually monitored, and when the learning values substantially match, the fuel injection is performed using the learning values. Since the correction value of the quantity is set, the reliability of the correction value can be improved, and the operation performance (noise, vibration, exhaust, etc.) of the diesel engine can be prevented from deteriorating due to erroneous learning of the learning means.

  Here, when the difference between the learning values obtained by the learning means is not within the allowable range, the comparison means may discard all the learning values and prohibit the setting (updating) of the correction value. However, if it does in this way, the probability that a correction value will be updated based on the learning value obtained by each learning means will become low, and it will also be considered that the effect by learning control cannot fully be exhibited.

  Therefore, as described in claim 2, when the difference between the learning values obtained by the respective learning means is not within the allowable range, the comparison means discards the learning value having a large fluctuation range and uses the remaining learning value. The correction value may be set.

  That is, the learning value obtained by each learning means usually does not fluctuate greatly and changes gradually due to changes in the characteristics of the fuel injection device. Therefore, if the comparison means is configured as described in claim 2, when the difference between the learning values obtained by the learning means is not within the allowable range, the fluctuation range is small and reliable from the plurality of learning values. It is possible to set a correction value by selecting a learning value having high characteristics.

  Therefore, according to the fuel injection control device of the second aspect, not only can the reliability of the correction value for the fuel injection amount be improved, but also the correction is made each time the learning value to be compared is calculated by each learning means. Since the value can be set, it is possible to prevent the update frequency of the correction value from decreasing.

By the way, the plurality of learning means may calculate the learning value by any learning method among various learning methods conventionally known.
Then, when each learning means cannot calculate the learning value for each cylinder of the diesel engine as in the above-described learning control of the exhaust gas composition detection method (6), the invention according to claim 2 is applied. Thus, in order to discard the learning value having a large fluctuation range by the comparison means, the deviation between the latest value of the learning value obtained by the learning means and the previous value is obtained as the fluctuation width of each learning value. do it.

  In addition, when each learning means is configured to calculate a learning value for each cylinder of the diesel engine, the comparison means compares the learning value calculated by each learning means for each cylinder of the diesel engine. Then, the correction value may be set, but in this case, the invention according to claim 2 is applied to discard the learning value having a large fluctuation range for each cylinder by the comparison means. The comparison means may be configured as described in claim 3 or claim 4.

  That is, in the fuel injection control device according to claim 3, the comparison means compares the learning values calculated by the learning means for each cylinder of the diesel engine, and the difference between the learning values obtained by the learning means is calculated. When there is a cylinder that is not within the allowable range, a learning value having a large difference from the average value of the learning values of all the cylinders is selected as a learning value having a large fluctuation range from among a plurality of learning values for the cylinder. Is discarded.

  Further, in the fuel injection control device according to claim 4, the comparison means compares the learning values calculated by the learning means for each cylinder of the diesel engine, as in claim 3. When there is a cylinder whose difference between the learned values is not within the allowable range, a learning value whose change tendency is greatly different from the change tendency of the learning value of the other cylinders among the learning values for the cylinder is learned with a large fluctuation range. Select as a value and discard the learned value.

  Therefore, according to the fuel injection control device of the third or fourth aspect, not only the correction value for correcting the deviation between the actual fuel injection amount and the command injection amount can be set for each cylinder of the diesel engine. When any of the learning means has an abnormality in the learning value of the specific cylinder, the learning value in which the abnormality has occurred in the specific cylinder can be more accurately specified using the learning value of the other cylinder. It becomes like this.

  In order to calculate the learning value for each cylinder of the diesel engine by each learning means as in the fuel injection control device according to claim 3 or 4, each learning means is described above. What is necessary is just to comprise so that any one of learning control (1)-(5) may be performed.

  Of these learning controls (1) to (5), in particular, the deceleration learning control (1) and the multistage learning control (2) are rotational speed sensors normally provided in a diesel engine. Accordingly, the difference between the actual fuel injection amount and the command injection amount can be obtained. Therefore, in the fuel injection control device according to claim 3 or 4, a plurality of learnings are performed as described in claim 5. As means, it is preferable to provide first learning means for executing deceleration type learning control and second learning means for executing multistage learning control.

  That is, among the above-described learning controls (1) to (5), in-cylinder pressure detection learning control (3) and ionic current detection learning control (5), learning control such as an in-cylinder pressure sensor and an ion current sensor is used. It is necessary to provide a dedicated special sensor. Further, the learning control (4) of the common rail pressure detection method calculates the amount of fuel leaking from the injector to the fuel discharge passage as a learning value based on a change in the common rail pressure caused by the idling of the injector. When a difference occurs between the actual fuel injection amount and the command injection amount due to factors other than the leaking fuel, the fuel injection amount cannot be corrected.

  On the other hand, in the case of the deceleration type and multistage learning control (1) and (2), the actual fuel injection amount and the command injection generated in the entire fuel injection system using the rotational speed sensor normally provided in the diesel engine. Since the deviation from the quantity can be obtained as a learning value, if it is adopted as a plurality of learning means of the present invention, the control accuracy of the fuel injection can be improved without increasing the cost of the fuel injection control device.

  On the other hand, in the fuel injection control apparatus according to the present invention (claims 1 to 5), each learning means operates under different learning conditions, so that, for example, the learning condition is established by one of a plurality of learning means. It takes time to do this, and it is conceivable that the time required for setting (updating) the correction value by comparing each learning value with the comparison means becomes longer.

Therefore, the fuel injection control device of the present invention (Claims 1 to 5) may further be provided with a comparison operation limiting means as described in Claim 6.
That is, in the fuel injection device according to claim 6, the comparison operation restriction unit acquires the learning value from at least one of the plurality of learning units, and if the variation range of the learning value is within the set range, the learning operation is performed. A correction value is set based on the value, and if the fluctuation range of the learning value is not within the setting range, the comparison unit is operated.

  Therefore, according to the fuel injection control device of the sixth aspect, when the fluctuation range of the learning value obtained by one of the plurality of learning means is small and the reliability of the learning value is high, the learning is performed. The correction value can be set (updated) only by the value, and the correction value update frequency can be increased as compared with the case where the correction value is set (updated) only by the comparison means.

Embodiments of the present invention will be described below with reference to the drawings.
(First embodiment)
FIG. 1 is a schematic configuration diagram illustrating the overall configuration of a pressure accumulation fuel injection system 10 according to the first embodiment.

  The fuel injection system 10 of this embodiment is for supplying fuel to, for example, a four-cylinder diesel engine 2 for automobiles. The common rail 20 that stores high-pressure fuel and the high-pressure fuel supplied from the common rail 20 are diesel-powered. An injector 30 that injects into the combustion chamber of each cylinder of the engine 2 and an electronic control unit (ECU) 50 that controls the system are provided.

  Further, in order to supply fuel to the common rail 20, the fuel injection system 10 includes a feed pump 14 that pumps fuel from the fuel tank 12, and a high pressure that pressurizes the fuel supplied from the feed pump 14 and supplies the fuel to the common rail 20. A pump 16 is provided.

  Here, the high-pressure pump 16 is a known pump that pressurizes the fuel sucked into the pressurizing chamber when the plunger reciprocates as the cam of the camshaft rotates. The high-pressure pump 16 is provided with a metering valve 18 for metering the amount of fuel sucked from the feed pump 14 in the suction stroke.

  The common rail 20 is provided with a pressure sensor 22 for detecting internal fuel pressure (common rail pressure) and a pressure reducing valve 24 for reducing the internal fuel pressure by overflowing the internal fuel to the fuel tank 12 side. It has been.

  Further, the diesel engine 2 includes a rotation speed sensor 32 that detects a rotation speed NE, an accelerator sensor 34 that detects an accelerator operation amount (accelerator opening ACC) by a driver, and cooling water as sensors that detect the operation state. A water temperature sensor 36 for detecting the temperature (cooling water temperature THW), an intake air temperature sensor 38 for detecting the temperature of intake air (intake air temperature TA), and the like are provided.

On the other hand, the ECU 50 is configured by a microcomputer centered on a CPU, ROM, RAM, and the like.
The ECU 50 takes in detection signals from the pressure sensor 22 provided on the common rail 20 and various sensors 32, 34, 36, 38,... Provided on the diesel engine 2, and the common rail pressure and the fuel injection amount from the injector 30 and Control the fuel injection timing.

  That is, the ECU 50 calculates the target pressure of the common rail 20 based on the operating state of the diesel engine 2, and controls the energization of the metering valve 18 and the pressure reducing valve 24 so that the common rail pressure detected by the pressure sensor 22 becomes the target pressure. By calculating the fuel injection amount and the fuel injection timing based on the common rail pressure control and the operating state of the diesel engine 2 and opening the injector 30 of each cylinder for a predetermined time at a predetermined timing according to the calculation result, The fuel injection control is performed to inject and supply the fuel.

  In this fuel injection control, the ECU 50 executes pilot injection prior to main injection. The difference between the fuel injection command value (command injection amount) for the injector 30 and the fuel amount actually injected from the injector 30 (actual fuel injection amount) causes combustion noise deterioration, large vibration, emission deterioration, and the like. Therefore, in order to prevent these, the ECU 50 executes a learning process in which a deviation between the actual fuel injection amount from the injector 30 and the command injection amount is calculated as a correction value (learning value) for the fuel injection amount.

Hereinafter, this learning process will be described with reference to the flowchart shown in FIG.
This learning process is executed in parallel with the fuel injection control in the ECU 50 during operation of the diesel engine 2. As shown in FIG. 2, when the process is started, first, S110 (S is a step). The above-described deceleration type learning control is executed.

  That is, in S110, when the accelerator opening ACC is zero, the diesel engine 2 enters a deceleration operation, and the fuel injection amount to each cylinder is zero, the cylinder i (i is a cylinder to be learned) The fuel injection is performed by the learning fuel amount (a constant amount) from the injector 30 (representing the number), and the generated torque of the diesel engine 2 is obtained from the rotational fluctuation amount and the rotational speed generated in the diesel engine 2 thereafter. The learning value Δ representing the difference between the actual fuel injection amount and the command injection amount for each cylinder i in the procedure of estimating the actual fuel injection amount from the generated torque and calculating the difference from the command injection amount. Q1i is calculated.

  When the learning value ΔQ1i of all the cylinders i (i: 1 to 4) is calculated by the deceleration type learning control in S110, the process proceeds to S120, and each cylinder i calculated by the deceleration type learning control is transferred. An average value ΔQ1AVE of the learning values ΔQ1i is calculated, and the process proceeds to S130.

  In S130, the multistage learning control described above is executed. That is, in S130, when the accelerator opening ACC is zero and the diesel engine 2 is idling, the fuel injection from the injectors 30 of each cylinder i is performed in a plurality of times (multistage). The first fuel injection correction amount for detecting the rotational speed and the rotational fluctuation of the diesel engine 2 caused by the fuel injection for each cylinder i and smoothing the rotational fluctuation between the cylinders i, and the average rotational speed of all the cylinders The second fuel injection correction amount for controlling the engine speed to the target rotational speed (idle rotational speed) is calculated, and by adding these correction amounts and taking a moving average, the actual fuel injection amount is calculated for each cylinder i. A learning value ΔQ2i representing a deviation from the command injection amount is calculated.

  Then, when the learning value ΔQ2i of all the cylinders i (i: 1 to 4) is calculated by the multistage learning control of S130, the process proceeds to S140, and each cylinder i calculated by the multistage learning control is transferred. An average value ΔQ2AVE of the learning values ΔQ2i is calculated, and the process proceeds to S150.

  Note that the details of the deceleration type and multistage learning control executed in S110 and S130 are described in the above-mentioned Patent Documents 1 and 2 and the like, and since they are conventionally known, more detailed description is given here. Description is omitted.

  Next, in S150, in the subsequent processing, the fuel injection amount correction value is set for each cylinder i of the diesel engine 2 based on the learning value obtained by the two types of learning control. An initial value “1” is set to a counter i for specifying the cylinder i.

  In the subsequent S160, learning values ΔQ1i and ΔQ2i for the cylinder i are compared among the learning values for each cylinder obtained by the learning control in S110 and S130, and the absolute value of the deviation (ΔQ1i−ΔQ2i) is compared. Is determined to exceed a preset threshold value (in other words, an allowable range) ΔQTH.

  If it is determined in S160 that the absolute value of the deviation between the learned values ΔQ1i and ΔQ2i does not exceed the threshold value ΔQTH, the process proceeds to S170, and the correction value ΔQi for the fuel injection amount of the cylinder i is set as the cylinder The average value “(ΔQ1i + ΔQ2i) / 2” of the two learning values ΔQ1i and ΔQ2i of i is set, and the process proceeds to S220.

  On the other hand, if it is determined in S160 that the absolute value of the deviation between the learning values ΔQ1i and ΔQ2i exceeds the threshold value ΔQTH, the process proceeds to S180, where the cylinder i obtained by the deceleration type learning control is obtained. The difference ΔQ1Ai between the learning value ΔQ1i and the average value of all cylinders ΔQ1AVE and the difference ΔQ2Ai between the learning value ΔQ2i of the cylinder i obtained by the multistage learning control and the average value of all cylinders ΔQ2AVE are calculated. To do.

  In the subsequent S190, the differences ΔQ1Ai and ΔQ2Ai for each learning control calculated in S180 are compared, and whether or not the difference ΔQ1Ai in the deceleration learning control is larger than the difference ΔQ2Ai in the multistage learning control. Determine whether.

  In S190, when it is determined that the difference ΔQ1Ai in the deceleration learning control is larger than the difference ΔQ2Ai in the multistage learning control, it is larger than the learning value ΔQ1i obtained in the deceleration learning control. The learning value ΔQ2i obtained by the multistage learning control is determined to have higher reliability, and the process proceeds to S200, where the learning value ΔQ2i determined to have high reliability is set to the fuel injection amount of the cylinder i. The correction value ΔQi is set, and the process proceeds to S220.

  In S190, when it is determined that the difference ΔQ1Ai in the deceleration learning control is equal to or less than the difference ΔQ2Ai in the multistage learning control, the learning value ΔQ2i obtained by the multistage learning control is larger than the learning value ΔQ2i obtained in the multistage learning control. It is determined that the learning value ΔQ1i obtained by the deceleration learning control is higher in reliability, the process proceeds to S210, and the learning value ΔQ1i determined as higher in reliability is determined with respect to the fuel injection amount of the cylinder i. The correction value ΔQi is set, and the process proceeds to S220.

  Next, in S220, whether or not the value of the counter i is equal to or greater than the number n of cylinders of the diesel engine 2 (n = 4 in the present embodiment), that is, the processing of S160 to S210 is performed for all the cylinders i of the diesel engine 2. It is judged whether it went to.

  If the value of the counter i is equal to or greater than the number of cylinders n of the diesel engine 2 and the processes of S160 to S210 are performed for all the cylinders of the diesel engine 2, the learning process is temporarily terminated. If the value of i is smaller than the number n of cylinders of the diesel engine 2 and the cylinders that have not been subjected to the processing of S160 to S210 remain, the counter i is incremented (+1) in S230, and the process proceeds to S160. The fuel injection amount correction value ΔQi is set for the next cylinder i in the same procedure as described above.

  As described above, in the fuel injection system 10 of the present embodiment, the ECU 50 as the fuel injection control device detects a deviation between the actual fuel injection amount from the injector 30 of each cylinder of the diesel engine 2 and the command injection amount. Then, a learning process for correcting the fuel injection amount is executed.

  In this learning process, a deceleration type learning control that calculates a learning value ΔQ1i for each cylinder when the diesel engine 2 decelerates, and a multistage type that calculates a learning value ΔQ2i for each cylinder when the diesel engine 2 is idling. When the learning values ΔQ1i and ΔQ2i obtained by executing the two types of learning control in order with the learning control are substantially the same (| ΔQ1i−ΔQ2i | ≦ ΔQTH), An average value of the two kinds of learning values ΔQ1i and ΔQ2i is set as a correction value ΔQi for the fuel injection amount of the cylinder i.

  Further, when the deviation between the learning values ΔQ1i and ΔQ2i obtained by each learning control is large (| ΔQ1i−ΔQ2i |> ΔQTH), all of the two learning values ΔQ1i and ΔQ2i Discard the larger deviations ΔQ1Ai and ΔQ2Ai from the cylinder average values ΔQ1AVE and ΔQ2AVE, and use the remaining learned values ΔQ1i or ΔQ2i (in other words, the smaller deviation ΔQ1Ai or ΔQ2Ai) It is set as a correction value ΔQi for the fuel injection amount of i.

  For this reason, according to the present embodiment, for example, as shown in FIG. 3A, the learning value ΔQ1i (A shown in the figure) of the first cylinder # 1 obtained by the deceleration learning control is changed to other values. When there is a large change from the previous value compared to the learning value ΔQ1i of the cylinders # 2 to # 4, the learning of the first cylinder # 1 obtained by the multistage learning control as shown by the dotted line in FIG. The value ΔQ2i (B shown in the figure) is also greatly changed from the previous value in the same manner as the learning value B obtained by the deceleration learning control, and when these values A and B substantially match, It is determined that no learning has occurred in each learning control, and the average value of these learning values A and B is set as a correction value for the fuel injection amount of the first cylinder # 1.

  In contrast, as indicated by a solid line in FIG. 3B, the learning value ΔQ2i (C shown in the figure) of the first cylinder # 1 obtained by the multistage learning control is set to the other cylinders # 2 to # 4. If the learning values A and C of the first cylinder # 1 are not significantly different from the previous value as in the case of the learning value ΔQ2i, it is determined that erroneous learning has occurred in one learning control. The learning value C by the multistage learning control close to the all cylinder average value ΔQ2AVE is set as the correction value for the fuel injection amount of the first cylinder # 1.

  Therefore, according to the present embodiment, by comparing the learning values ΔQ1i and ΔQ2i obtained by the learning control of each method for each cylinder, it is accurately determined that an erroneous learning of one learning control has occurred. In addition, when the occurrence of erroneous learning is determined, it is possible to select a normal learning value from two types of learning values ΔQ1i and ΔQ2i and set a correction value for the fuel injection amount. it can. Therefore, according to the present embodiment, it is possible to prevent the operation performance (noise, vibration, exhaust, etc.) of the diesel engine 2 from deteriorating due to erroneous learning of the learned value.

  In the present embodiment, two types of learning control, a deceleration type and a multistage type, are adopted as the learning control to be executed in the learning process. Therefore, in each learning control, a detection signal from the rotation speed sensor 32 is used. In order to calculate the learning value for each cylinder of the diesel engine 2, a special sensor for learning control such as an in-cylinder pressure sensor and an ion current sensor is separately provided in order to calculate the learning value for each cylinder of the diesel engine. There is no need. Therefore, according to the present embodiment, a fuel injection control device that can accurately set the correction value for the fuel injection amount for each cylinder of the diesel engine 2 can be realized at low cost.

In the present embodiment, among the learning processes executed by the ECU 50 as the fuel injection control device, the process of S110 corresponds to the first learning means of the present invention, and the process of S130 is the first of the present invention. 2 corresponds to learning means, and a series of processing from S150 to S220 corresponds to comparison means of the present invention.
(Second Embodiment)
Next, a second embodiment of the present invention will be described.

  In this embodiment, the difference from the first embodiment is a part of the learning process executed by the ECU 50 as the fuel injection control device, and the entire configuration of the fuel injection system 10 and the like are all in the first embodiment. Since the configuration is the same as that of the embodiment, only the differences from the first embodiment in the learning process will be described below, and other descriptions such as the configuration will be omitted.

FIG. 4 is a flowchart showing a learning process executed by the ECU 50 of the present embodiment.
As shown in FIG. 4, in the learning process of the present embodiment, when the learning value ΔQ1i of each cylinder is calculated in the deceleration type learning control in S110, the process proceeds to S125, and this time in S110 for each cylinder i. The amount of change ΔQ1Di from the previous learning value of the calculated learning value ΔQ1i is calculated, and the process proceeds to S130.

  Similarly, when the learning value ΔQ2i of each cylinder is calculated in the multistage learning control of S130, the process proceeds to S145, and for each cylinder i, from the previous learning value of the learning value ΔQ2i calculated this time in S130. Change amount ΔQ2Di is calculated, and the process proceeds to S150.

  On the other hand, if it is determined in S160 that the absolute value of the deviation between the learned values ΔQ1i and ΔQ2i exceeds the threshold value ΔQTH, the process proceeds to S185, and the cylinder i calculated in S125 and S145 is calculated. Differences ΔQ1Ei, ΔQ2Ei (ΔQ1Ei = | ΔQ1ADi−ΔQ1Di) between the learning value variation ΔQ1Di, ΔQ2Di and the average value ΔQ1AD, ΔQ2AD of the learning value variation of the other cylinders excluding the cylinder i |, ΔQ2Ei = | ΔQ2ADi−ΔQ2Di |) are calculated, and the process proceeds to S195.

  In S195, the differences ΔQ1Ei and ΔQ2Ei between the learning value variation calculated in S185 and the average value of the other cylinders are compared, and the difference ΔQ1Ei in the deceleration learning control is the difference Δ in the multistage learning control. When it is determined whether or not the difference ΔQ1Ei in the deceleration learning control is larger than the difference ΔQ2Ei in the multistage learning control, the process proceeds to S200, and the deceleration learning control is performed. If it is determined that the difference ΔQ1Ei at is not larger than the difference ΔQ2Ei in the multistage learning control, the process proceeds to S210.

  As described above, in this embodiment, when the deviation between the learning values ΔQ1i and ΔQ2i obtained by the deceleration type and multistage learning control is large (| ΔQ1i−ΔQ2i |> ΔQTH), For each learning value ΔQ1i, ΔQ2i, the difference ΔQ1Ei, ΔQ1Di, ΔQ2Di between the learning values ΔQ1Di, ΔQ2Di from the previous learning value and the average values ΔQ1AD, ΔQ2AD of the other cylinders ΔQ2Ei is calculated as a parameter representing the difference in change tendency of each learning value ΔQ1i, ΔQ2i from the other cylinders, and the learning value ΔQ1i or ΔQ2i of which the difference ΔQ1Ei, ΔQ2Ei is smaller is It is set as a correction value ΔQi for the fuel injection amount of i.

For this reason, in this embodiment as well, as in the first embodiment, the learning values ΔQ1i and ΔQ2i obtained by the learning control of each method are compared for each cylinder, thereby causing erroneous learning of one learning control. When the occurrence of mislearning is determined, a normal learning value is selected from two types of learning values ΔQ1i and ΔQ2i, and the fuel injection amount is determined. A correction value can be set, and it is possible to prevent the operation performance (noise, vibration, exhaust, etc.) of the diesel engine 2 from deteriorating due to erroneous learning of the learning value.
(Third embodiment)
Next, a third embodiment of the present invention will be described.

  In this embodiment, the difference from each of the above embodiments is a learning process executed by the ECU 50 as the fuel injection control device, and the overall configuration of the fuel injection system 10 is the same as that of each of the above embodiments. Therefore, in the following description, only the learning process will be described.

FIG. 5 is a flowchart showing a learning process executed by the ECU 50 of the present embodiment.
As shown in FIG. 5, when the learning process of the present embodiment is started, first, in S310, the flag XQDA common to all cylinders used in the subsequent processes and the flag XCYLi for each cylinder i are reset (OFF). In the subsequent S320, deceleration type learning control is executed in the same procedure as in S110 described above.

  In the subsequent S330, a difference ΔQ1Ai between the learning value ΔQ1i of each cylinder i obtained by the learning control in S320 and the average value ΔQ1AVE of all the cylinders of the learning value ΔQ1i is calculated for each cylinder i. In S340, the initial value “1” is set in the counter i, and then the process proceeds to S350.

  In S350, the difference ΔQ1Ai calculated for the cylinder i in S330 is compared with a preset first threshold value ΔQTH1, and it is determined whether or not the difference ΔQ1Ai exceeds the first threshold value ΔQTH1. If the difference ΔQ1Ai does not exceed the first threshold value ΔQTH1, the learning value ΔQ1i of the cylinder i calculated in S320 is set as a correction value ΔQi for the fuel injection amount of the cylinder i in S370. Thereafter, the process proceeds to S380. Conversely, if the difference ΔQ1Ai exceeds the first threshold value ΔQTH1, the flag XQDA and the flag XCYLi for the cylinder i are set (ON) in S360, and then the process proceeds to S380.

  Next, in S380, whether or not the value of the counter i is equal to or greater than the number of cylinders n of the diesel engine 2 (n = 4 in the present embodiment), that is, the processing of S350 to S370 is performed for all the cylinders i of the diesel engine 2. It is judged whether it went to.

  If the value of the counter i is equal to or greater than the number of cylinders n of the diesel engine 2 and the processes of S350 to S370 are performed for all the cylinders of the diesel engine 2, the process proceeds to S400. Is smaller than the number of cylinders n of the diesel engine 2 and there are remaining cylinders that have not been subjected to the processing of S350 to S370, the counter i is incremented (+1) in S390, and then the process proceeds to S350. The same processing as described above is executed for the cylinder i.

  Next, in S400, whether or not the flag XQDA is set (ON), that is, in S350, the difference ΔQ1Ai between the learning value ΔQ1i of any cylinder i and the average value ΔQ1AVE of all cylinders is the first. It is determined whether or not it is determined that the threshold value ΔQTH1 has been exceeded.

  If the flag XQDA is in a reset (OFF) state and correction values have been set for all the cylinders i by the processing of S350 to S380, the processing is temporarily terminated, and conversely, the flag XQDA Is in the reset (OFF) state, that is, for at least one cylinder i, it is determined in S350 that the difference ΔQ1Ai between the learning value ΔQ1i and the average value of all cylinders ΔQ1AVE exceeds the first threshold value ΔQTH1. If the correction value for the fuel injection amount is not set, the process proceeds to S410.

  In S410, multistage learning control is executed in the same procedure as in S130 described above, and in S420, the learning value ΔQ2i of each cylinder i obtained by this multistage learning control and its learning value ΔQ2i. A difference ΔQ2Ai from the average value of all cylinders ΔQ2AVE is calculated for each cylinder i. In the subsequent S430, the initial value “1” is set in the counter i, and the process proceeds to S440.

  In S440, it is determined whether or not the flag XCYLi for the cylinder i corresponding to the counter i is set (ON), that is, whether or not the correction value of the fuel injection amount for the cylinder i is not set. If the flag XCYLi is not set (ON), the process proceeds to S500. If the flag XCYLi is set (ON), the process proceeds to S450.

  In S450, the learning values ΔQ1i and ΔQ2i for the cylinder i among the learning values for each cylinder obtained by the learning control in S320 and S420 are compared as in S160 described above, and the deviation (ΔQ1i−ΔQ2i) is compared. ) Exceeds the preset second threshold value ΔQTH2.

  If it is determined in S450 that the absolute value of the deviation between the learned values ΔQ1i and ΔQ2i does not exceed the second threshold value ΔQTH2, the process proceeds to S460, where the correction value Δ for the fuel injection amount of the cylinder i is Δ. As Qi, an average value “(ΔQ1i + ΔQ2i) / 2” of the two types of learning values ΔQ1i and ΔQ2i of the cylinder i is set, and the process proceeds to S500.

  On the other hand, if it is determined in S450 that the absolute value of the deviation between the learned values ΔQ1i and ΔQ2i exceeds the second threshold value ΔQTH2, the process proceeds to S470, and the cylinder calculated in S330 is the same as in S190. The difference ΔQ1Ai between the learning value ΔQ1i of i and the average value of all cylinders ΔQ1AVE and the difference ΔQ2Ai between the learning value ΔQ2i of the cylinder i calculated in S420 and the average value of all cylinders ΔQ2AVE are compared.

  In S470, if it is determined that the difference ΔQ1Ai is larger than the difference ΔQ2Ai, the learning value ΔQ2i obtained by the multistage learning control rather than the learning value ΔQ1i obtained by the deceleration learning control. Therefore, the process proceeds to S480, and after the learning value ΔQ2i is set as the correction value ΔQi for the cylinder i as in S200 described above, the process proceeds to S500.

  In S470, when it is determined that the difference ΔQ1Ai is equal to or less than the difference ΔQ2Ai, the learning value ΔQ1i obtained by the deceleration learning control rather than the learning value ΔQ2i obtained by the multistage learning control. Therefore, the process proceeds to S490, and similarly to S210 described above, the learning value ΔQ1i is set as the correction value ΔQi for the fuel injection amount of the cylinder i, and the process proceeds to S500.

  In S500, whether or not the value of the counter i is equal to or greater than the number of cylinders n of the diesel engine 2 (n = 4 in the present embodiment), that is, the processing of S440 to S490 is performed on all the cylinders i of the diesel engine 2. If the value of the counter i is equal to or greater than the number of cylinders n of the diesel engine 2 and the processes of S440 to S490 are performed for all the cylinders of the diesel engine 2, the learning process is performed. On the contrary, if the counter i is smaller than the number n of cylinders of the diesel engine 2 and there are cylinders not subjected to the processing of S440 to S490, the counter i is incremented (+1) in S510. ), The process proceeds to S440, and the fuel injection amount correction value ΔQi is set for the next cylinder i in the same procedure as described above.

  Thus, in the learning process of the present embodiment, the learning value ΔQ1i obtained by the deceleration type learning control after executing the deceleration type learning control is greatly different from the all cylinder average value ΔQ1AVE ( It is determined whether or not there is a learning value (ΔQ1A1> ΔQTH1), and if there is no learning value greatly different from the all-cylinder average value ΔQ1AVE, deceleration is performed without executing multistage learning control. The learning value ΔQ1i obtained by the learning control of the equation is set as it is as the correction value ΔQi of each cylinder i.

  If there is a learning value greatly different from the average value ΔQ1AVE for all cylinders in the learning value ΔQ1i obtained by the deceleration type learning control, the multistage learning control is executed, and the cylinder i As in the first embodiment, the correction value ΔQi is set by comparing the learning values ΔQ1i and ΔQ2i obtained by each learning control.

  Therefore, according to the fuel injection system 10 of the present embodiment, the same effect as that of the first embodiment can be obtained, and the average value of all cylinders ΔQ1AVE among the learning values ΔQ1i obtained by the deceleration type learning control. If there is no learning value that is significantly different from the learning value, the correction value ΔQi can be set without executing multi-stage learning control, so that the correction value ΔQi can be set more quickly. The update frequency of ΔQi can be increased.

  In the present embodiment, among the learning processes shown in FIG. 5, the process of S320 corresponds to the first learning means of the present invention, and the processes of S340 to S400 correspond to the comparison operation limiting means of the present invention. The process of S410 corresponds to the second learning unit of the present invention, and the series of processes of S430 to S500 corresponds to the comparison unit of the present invention.

As mentioned above, although embodiment of this invention was described, this invention is not limited to the said embodiment, A various aspect can be taken in the range which does not deviate from the summary of this invention.
For example, in the third embodiment, the deceleration type learning control is preferentially executed, and the deviation ΔQ1Ai from the all-cylinder average value ΔQ1AVE is included in the learning value ΔQ1i obtained by the deceleration type learning control. Although it has been described that multistage learning control is executed when there is a learning value ΔQ1i exceeding the first threshold ΔQTH1, deceleration type learning control and multistage learning control are executed when each learning condition is satisfied. After each learning control is executed, the learning value obtained by the learning control is compared with the average value of all cylinders. If the deviation is small, the learning value obtained by the learning control is obtained. The correction value using the learning value is set only when there is a learning value with a large deviation from the average value of all cylinders among the learning values obtained by the learning control. Ban renewal and then other It may be updated correction value by comparing the learning value obtained by the learning control.

  In each of the above embodiments, the fuel injection amount learning control has been described as performing two types of learning control, the deceleration type and the multistage type. However, the above-described in-cylinder pressure detection method, common rail pressure detection method, ion current You may make it perform combining learning control other than the deceleration type or multistage type learning control, such as a detection system, an exhaust gas composition detection system,.

  In each of the above embodiments, the case where the present invention is applied to an accumulator fuel injection system including the common rail 20 has been described. However, the present invention is a fuel injection system including a distributed fuel injection pump. Can also be applied.

It is a schematic structure figure showing the composition of the whole fuel injection system of an embodiment. It is a flowchart showing the learning process of 1st Embodiment. It is explanatory drawing explaining operation | movement of the learning process of 1st Embodiment. It is a flowchart showing the learning process of 2nd Embodiment. It is a flowchart showing the learning process of 3rd Embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 2 ... Diesel engine, 10 ... Fuel injection system, 12 ... Fuel tank, 14 ... Feed pump, 16 ... High pressure pump, 18 ... Metering valve, 20 ... Common rail, 22 ... Pressure sensor, 24 ... Pressure reducing valve, 30 ... Injector, 32 ... Rotational speed sensor, 34 ... Accelerator sensor, 36 ... Water temperature sensor, 38 ... Intake air temperature sensor, 50 ... ECU (electronic control unit).

Claims (6)

In a fuel injection control device that calculates a fuel injection amount based on an operation state of a diesel engine, and injects fuel into a cylinder of the diesel engine by driving the fuel injection device according to the calculation result.
A plurality of learning means that operate under different learning conditions and calculate a learning value representing a deviation between the command injection amount and the actual fuel injection amount for the fuel injection device based on a predetermined operation state of the diesel engine after fuel injection When,
The learning values calculated by the plurality of learning means are compared, and when the difference between the learning values is within an allowable range, a correction value for the fuel injection amount is obtained using at least one of the plurality of learning values. A comparison means to set;
A fuel injection control device for a diesel engine, comprising:   2. The comparison unit according to claim 1, wherein when the difference between the learning values is not within an allowable range, the learning value having a large fluctuation range is discarded and the correction value is set using the remaining learning value. A fuel injection control device for a diesel engine according to claim 1. Each of the learning means calculates the learning value for each of a plurality of cylinders of the diesel engine,
The comparison means compares the learning value calculated by each learning means for each cylinder of the diesel engine, and when there is a cylinder in which the difference between the learning values obtained by each learning means is not within an allowable range, Among the plurality of learning values, the learning value having a large difference from the average value of the learning values of all the cylinders is discarded, and a correction value for the fuel injection amount of the cylinder is set using the remaining learning values. The diesel engine fuel injection control device according to claim 2. Each of the learning means calculates the learning value for each of a plurality of cylinders of the diesel engine,
The comparison means compares the learning value calculated by each learning means for each cylinder of the diesel engine, and when there is a cylinder in which the difference between the learning values obtained by each learning means is not within an allowable range, Of the plurality of learning values, the learning value whose change tendency is significantly different from the learning tendency of the other cylinders is discarded, and the correction value for the fuel injection amount of the cylinder is set using the remaining learning values. The fuel injection control device for a diesel engine according to claim 2. As the plurality of learning means,
During deceleration operation in which the fuel injection amount to the diesel engine becomes zero, for each cylinder of the diesel engine, fuel injection is executed from the fuel injection device in a single shot to detect rotational fluctuations of the diesel engine that occur after fuel injection, A first learning means for calculating an actual fuel injection amount from a detection result to obtain a deviation from a command injection amount, and setting the learning value;
During idle operation of the diesel engine, fuel injection from the fuel injection device is performed in multiple times for each cylinder of the diesel engine, and the rotation fluctuation between the cylinders based on the rotation speed and rotation fluctuation of the diesel engine that occurred at that time The first fuel injection correction amount for smoothing the engine and the second fuel injection correction amount for controlling the average rotation speed of the cylinder to the target rotation speed are obtained, and the learning value is calculated based on these two fuel injection correction amounts. A second learning means for setting
The fuel injection control device for a diesel engine according to claim 3 or 4, characterized by comprising: A learning value is acquired from at least one of the plurality of learning means, and if the fluctuation range of the learning value is within a setting range, the correction value is set based on the learning value, and the fluctuation range of the learning value is set. If not within the range, the comparison operation limiting means for operating the comparison means,
The fuel injection control device for a diesel engine according to any one of claims 1 to 5, wherein

Images