JP2008095615A - Fuel injection control device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fuel injection control device capable of learning a deviation in an injection characteristic of a fuel injection valve in a wider region. <P>SOLUTION: An exhaust gas oxygen concentration estimating part B6 calculates an estimation concentration base value Db on the oxygen concentration in exhaust gas based on an intake air amount or a command injection amount base value. A final oxygen concentration in the exhaust gas (final estimation concentration Df) calculated based on the estimation concentration base value Db is calculated. The deviation of the injection characteristic of the fuel injection valve is learned based on a difference ΔD between the final estimation oxygen concentration Df and a detected value of oxygen concentration detected by an oxygen sensor. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a fuel injection control device that learns a deviation in injection characteristics of a fuel injection valve of an internal combustion engine.

  It is well known to perform exhaust gas recirculation control (EGR control) for recirculating a part of the exhaust gas to the intake system for the purpose of reducing the amount of harmful components in the exhaust gas. In this case, it is desirable to determine the operation amounts of various actuators of the internal combustion engine in accordance with the exhaust amount to be recirculated (EGR amount) and the oxygen concentration immediately after combustion in the combustion chamber. However, during transient operation of the internal combustion engine, the output of a sensor mounted on the internal combustion engine such as an air flow meter cannot be directly used as a parameter having a correlation with the EGR amount or the oxygen concentration.

  Therefore, conventionally, as seen in Patent Documents 1 and 2 below, for example, it has been proposed to estimate the EGR amount and the oxygen concentration on a model basis based on the operating state of the internal combustion engine. As a result, even when the internal combustion engine is in transition, it is possible to acquire highly accurate information such as the EGR amount and the oxygen concentration. Therefore, the operation amounts of the various actuators of the internal combustion engine can be set appropriately, and as a result, the output characteristics of the internal combustion engine can be favorably controlled.

  By the way, in the fuel injection valve of the internal combustion engine, a deviation in injection characteristics may occur due to a solid difference or a secular change. And when the deviation of an injection characteristic arises, even if EGR amount and oxygen concentration are estimated based on the said model, an estimated value will not become an appropriate value. For this reason, it is desirable to learn the deviation of the injection characteristic of the fuel injection valve and perform learning control to compensate for this.

  However, since the learning of the deviation of the injection characteristic is normally performed at the time of idle rotation speed control as seen in Patent Document 3 below, the effective range of the learning of the deviation is limited to the region where the injection amount is small. For this reason, it has been difficult to perform learning control that compensates for deviations in injection characteristics in a region where the injection amount is large.

In addition, not only the above-mentioned EGR amount and the like are estimated by a model, but also in the fuel injection control device, in general, such an actual situation is limited in that the region where the deviation of the injection characteristic of the fuel injection valve can be learned is limited. It has become.
JP-A-64-63647 JP 2002-327634 A JP 2003-343328 A

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a fuel injection control device capable of learning a deviation in injection characteristics of a fuel injection valve in a wider range. is there.

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

  According to a first aspect of the present invention, there is provided prediction means for calculating a predicted value of oxygen concentration in exhaust gas based on a physical quantity of intake air sucked into an intake system of an internal combustion engine and a command injection amount for a fuel injection valve; And a deviation learning means for learning a deviation of the injection characteristic of the fuel injection valve based on the difference between the detected value of the oxygen concentration and the predicted value.

  When a deviation occurs between the predicted value and the detected value of the oxygen concentration in the exhaust gas, this is considered to be caused by a deviation in the injection characteristics of the fuel injection valve. Therefore, it is possible to learn the deviation of the injection characteristics based on the difference between the predicted value and the detected value. In addition, since it is possible to calculate the predicted value of the oxygen concentration in the exhaust gas not during the idle rotation speed control, it is possible to learn the deviation of the injection characteristics in a wider range.

  According to a second aspect of the present invention, in the first aspect of the invention, the deviation learning means uses the information about the mass of the exhaust gas at the time of predicting the oxygen concentration in the exhaust gas, and uses the difference between the detected value and the predicted value. The learning based on the above is performed.

  The difference between the predicted value and the detected value of the oxygen concentration in the exhaust gas is considered to be a value obtained by dividing the deviation amount of the oxygen amount by the mass of the exhaust gas. On the other hand, since the oxygen consumption amount per unit injection amount can be theoretically determined, the deviation amount of the oxygen amount can be converted into the deviation amount of the fuel injection amount. Therefore, in the above configuration, it is possible to suitably calculate the deviation in the injection characteristics by using the information regarding the mass of the exhaust gas at the time of predicting the oxygen concentration in the exhaust gas and the difference between the detected value and the predicted value as inputs.

  The invention according to claim 3 is the invention according to claim 1 or 2, further comprising error learning means for learning an error of the prediction means based on the detected value of the oxygen concentration in the exhaust gas and the predicted value. The deviation learning means performs learning of the deviation based on the detected value and the predicted value through the error learned by the error learning means when the error learning by the error learning means is a predetermined number of times or more. Features.

  In the above configuration, in order to learn the deviation of the injection characteristic through the error learned when the error learning becomes a predetermined number of times or more, the deviation learning can be performed after the error learning converges, and thus the deviation learning accuracy. Can be improved.

  The invention according to claim 4 is the invention according to claim 3, wherein the error learning means learns the error for each region divided by the rotational speed and load of the output shaft of the internal combustion engine, and When learning the error, information on the mass of the exhaust gas is also stored.

  The exhaust mass flow rate has a correlation with the intake mass flow rate. The intake mass flow rate changes in accordance with the intake air temperature and the like. For this reason, the mass of the exhaust gas cannot be accurately calculated only by the rotation speed and the load. For this reason, when learning an error for each of a plurality of regions divided by the rotational speed and load, it is desirable to store some information for calculating the mass of the exhaust gas. In this regard, in the above configuration, in order to store information related to the mass of the exhaust when learning the error, it is preferable to learn the injection characteristic deviation based on the difference between the predicted value and the detected value of the oxygen concentration in the exhaust. Can do.

  According to a fifth aspect of the present invention, in the invention according to the fourth aspect, the deviation learning means converts the area where the error learning by the error learning means is made into an area divided by fuel pressure, and is learned. The deviation of the injection characteristic in the converted region is learned based on the error and the information on the exhaust mass.

  The injection characteristic of the fuel injection valve is highly dependent on the fuel pressure. For this reason, the deviation of the injection characteristics is also highly dependent on the fuel pressure. Therefore, it is desirable to learn the injection characteristic deviation for each region divided by the fuel pressure. In this regard, in the above-described configuration, the deviation of the injection characteristic can be learned for each region divided by the fuel pressure by converting the region where the error is learned into a region divided by the fuel pressure. Accordingly, it is possible to learn the deviation of the injection characteristic in a manner that suitably reflects the fuel pressure dependency of the injection characteristic.

  According to a sixth aspect of the present invention, in the invention according to any one of the third to fifth aspects, when the learning by the deviation learning means is performed, the error learning means learns according to the learned deviation in the injection characteristic. The apparatus further comprises means for correcting the result.

  When learning by the deviation learning means is performed, the deviation of the injection characteristic is compensated for in the fuel injection control, so that the error of the predicted value with respect to the detected value of the oxygen concentration in the exhaust gas is reduced. In this regard, in the above configuration, when the deviation in the injection characteristic is learned, the learning result of the error learning means is corrected according to this learning, so that the learning result of the error learning means is converted into an error of the predicted value with respect to the detected value. It can be updated to an appropriate value depending on the situation.

  According to a seventh aspect of the invention, in the sixth aspect of the invention, the deviation learning unit filters the error learned by the error learning unit, and then learns the deviation using the value after the filtering process. It is characterized by doing.

  In the above configuration, since the deviation of the injection characteristic is learned using the learned error after filtering, it is possible to suitably suppress the influence of noise or the like when learning the deviation, thereby improving the learning accuracy of deviation. Can be improved. However, in this case, the learning result of the deviation is not a direct conversion of the learning result of the error into the deviation of the injection characteristic. For this reason, the deviation of the injection characteristic is compensated based on the learning of the deviation, but the error learned by the error learning means is not canceled by this compensation. In this regard, in the above configuration, by having the configuration of the sixth aspect, the learning result of the error learning means is updated to an appropriate value even though the deviation is learned based on the value after the filter processing. be able to.

  The invention according to claim 8 is the invention according to any one of claims 1 to 7, wherein the internal combustion engine includes an exhaust gas recirculation passage for recirculating exhaust gas exhausted to the exhaust system to an intake system, and the exhaust gas recirculation passage. Area variable means for making the flow path area variable, and further comprising area adjusting means for operating the area variable means based on the prediction value of the prediction means.

  In the above configuration, in order to make the flow area of the exhaust gas recirculation passage variable based on the predicted value of the oxygen concentration in the exhaust, the oxygen concentration in the exhaust regardless of whether the internal combustion engine is in a transient state or in a steady state The flow path area can be made variable based on accurate information about. In addition, since the process for calculating the predicted value of the oxygen concentration in the exhaust gas can be shared by this process and the learning of the deviation in the injection characteristics, the configuration of the control program, the control logic circuit, etc. can be shared. As a result, the configuration can be simplified.

  The invention according to claim 9 is the invention according to any one of claims 1 to 8, wherein target value setting means for setting a target value of the oxygen concentration in the exhaust gas, and feedback control of the predicted value to the target value. Feedback control means, and means for correcting the predicted value used for the learning based on the aspect of the feedback control.

  In the above configuration, the oxygen concentration in the exhaust gas can be made to suitably follow the target value by feedback control. In addition, the process for calculating the predicted value of the oxygen concentration in the exhaust gas can be shared between this process and the learning of the deviation in the injection characteristics, so that the configuration of the control program, the control logic circuit, etc. can be shared. Can be simplified. However, in this case, if the influence of feedback control is not reflected on the predicted value of the oxygen concentration in the exhaust gas, the difference between the predicted value and the detected value does not become an error of the prediction means or an injection characteristic. In this regard, in the above configuration, such a problem can be avoided by correcting the predicted value used for learning to a value predicted as the oxygen concentration in the exhaust gas, reflecting the influence of feedback control.

(First embodiment)
Hereinafter, a first embodiment in which a fuel injection control device for an internal combustion engine according to the present invention is applied to a fuel injection control device for an in-vehicle diesel engine will be described with reference to the drawings.

  FIG. 1 shows the overall configuration of the engine system according to the present embodiment.

  As illustrated, an air cleaner 14, an air flow meter 16, a cooler 18, a throttle valve 20, and the like are provided upstream of the intake passage 12 of the diesel engine 10. Further, an intake air temperature sensor 22 and an intake pressure sensor 24 are provided downstream of the throttle valve 20 in the intake passage 12. The intake passage 12 can communicate with the combustion chamber 26 of each cylinder (here, four cylinders of the first cylinder # 1 to the fourth cylinder # 4 are illustrated) via the manifold 25. High-pressure fuel stored in the common rail 28 is injected into the combustion chambers 26 through the fuel injection valve 30. Thereby, the fuel-air mixture in the combustion chamber 26 is used for combustion, and the rotational force of the diesel engine 10 is generated.

  On the other hand, the exhaust gas, which is the air used for combustion, is discharged to the exhaust passage 32. The exhaust passage 32 is provided with an oxygen concentration sensor 34 that detects the oxygen concentration in the exhaust.

  A variable nozzle turbocharger 36 is provided in the intake passage 12 and the exhaust passage 32. Further, the intake passage 12 and the exhaust passage 32 are provided with an exhaust gas recirculation passage (EGR passage 38) that allows these to communicate with each other, and the flow passage area between the intake passage 12 and the EGR passage 38 is an EGR valve. 40 is adjustable.

  The engine system includes a crank angle sensor 42 that detects the rotation angle of the crankshaft of the diesel engine 10 as a sensor that detects the operating state of the diesel engine 10 in addition to the sensor described above. The engine system also includes various sensors that detect user requests, such as an accelerator sensor 44 that detects an operation amount of an accelerator pedal.

  The electronic control unit (ECU 50) includes a central processing unit, a constant memory holding memory 52, and the like. Here, the always-on memory holding memory 52 is a backup RAM that always holds the power supply state regardless of the state of the ignition switch (power switch of the ECU 50), a non-volatile memory that always holds data regardless of whether power is supplied, etc. A storage device that holds data regardless of the state of the power switch.

  The ECU 50 operates various actuators such as the fuel injection valve 30 on the basis of detection values of various sensors that detect the operation state of the diesel engine 10 and user requests, and thereby outputs characteristics (output torque, exhaust characteristics) of the diesel engine 10. ) To control. FIG. 2 shows a process performed by the ECU 50.

  The injection amount setting unit B2 is a part for setting a basic value (command injection amount base value) of a command value of the injection amount for the fuel injection valve 30. The setting of the command injection amount base value can be performed based on information on the rotation speed and the load. In the present embodiment, a case where the setting is made based on the operation amount of the accelerator pedal detected by the accelerator sensor 44 and the rotation speed of the crankshaft corresponding to the detected value of the crank angle sensor 42 is illustrated.

  The exhaust oxygen concentration target setting unit B4 is a part for setting a target value Da of the oxygen concentration in the exhaust. It is desirable that the target value of the oxygen concentration is determined by adapting the relationship with the operation state of the diesel engine 10. Here, the operation state of the diesel engine 10 is desirably performed based on at least one of the rotational speed and the load of the diesel engine 10. In this embodiment, the case where the target value Da is set based on the command injection amount base value as the load and the rotation speed is illustrated.

  The exhaust oxygen concentration prediction unit B6 calculates a predicted concentration base value Db that is a predicted value of the oxygen concentration in the exhaust gas based on the operating state of the diesel engine 10 using a model of the gas flow mode in the diesel engine 10. It is. Here, as parameters relating to the operating state of the diesel engine 10 used for prediction, there are a physical quantity of intake air sucked into the intake passage 12 and a command injection quantity for the fuel injection valve 30. In the present embodiment, the prediction is performed based on the intake air amount detected by the air flow meter 16, the intake air temperature detected by the intake air temperature sensor 22, the intake pressure detected by the intake air pressure sensor 24, and the command injection amount base value. Is illustrated.

  The injection amount correction amount setting unit B8 is a part that performs feedback control of the predicted value of the exhaust oxygen concentration to the target value Da. Specifically, the feedback correction amount is calculated and output to the adder B10. As a result, the adder B10 calculates the command injection amount base value Qb corrected by the feedback correction amount as the final command injection amount.

  The injection period setting unit B12 is based on the command value of the injection period for the fuel injection valve 30 for injecting fuel of the command injection quantity (command injection period base) based on the fuel pressure NPC in the common rail 28 and the command injection quantity. A value TQb) is calculated.

  The final predicted value setting unit B14 is a part that reflects the influence of feedback control on the target value Da in the predicted value of the oxygen concentration in the exhaust gas. That is, the predicted concentration base value is calculated based on the command injection amount base value Qb, and is not calculated based on the command injection amount QFIN. In other words, the influence of feedback control is not considered in the predicted concentration base value. For this reason, the influence amount of the feedback control appears as an error in the predicted density base value. Therefore, the final predicted value setting unit B14 corrects the predicted density base value to reflect the influence of feedback control, and calculates the final predicted density Df.

  The error calculation unit B16 is a part that calculates a difference ΔD between the oxygen concentration detected by the oxygen concentration sensor 34 and the final predicted concentration Df.

  The model error learning unit B18 is a part that learns the model error learning value LM based on the difference ΔD and stores it in the constant memory holding memory 52 and corrects the predicted concentration base value Db to compensate for the model error. . That is, the sum of the model error learning value LM output from the model error learning unit B18 and the predicted density base value Db is calculated by the predicted value correction unit B20. By outputting the output of the predicted value correcting unit to the deviation calculating unit B22, a difference between the target value Da used for the feedback control and the predicted value is calculated.

  The EGR base value setting unit B24 is a part for setting the opening degree base value θb when operating the EGR valve 40. Here, it is desirable that the relationship between the operating state of the diesel engine 10 and the opening degree base value θb is determined in advance in a map or the like by adaptation. In the present embodiment, the opening degree base value θb is map-calculated by creating in advance a map that defines the relationship between the command injection amount base value and the rotational speed and the opening degree base value θb.

  The EGR correction amount setting unit B26 is an EGR valve for performing feedback control of the predicted value to the target value Da based on the difference between the predicted value of the oxygen concentration in the exhaust gas and the target value Da, in other words, based on the output of the deviation calculating unit B22. This is a part for setting a correction amount of 40 openings.

  The EGR manipulated variable setting unit B28 is a part that calculates a final EGR valve opening command value (EGR manipulated variable θ) by correcting the opening base value θb with the correction amount.

  The target fuel pressure setting unit B30 is a part that sets a target value (target fuel pressure) of the fuel pressure in the common rail 28. Note that the target fuel pressure is preferably calculated based on information on the rotational speed and information on the load. In this embodiment, the case where the target fuel pressure is set based on the command injection amount as the load and the rotation speed is illustrated.

  According to the above configuration, the fuel injection valve is based on accurate information on the oxygen concentration in the exhaust discharged from the combustion chamber 26 regardless of whether the diesel engine 10 is in a steady operation state or a transient operation state. 30 and the EGR valve 40 can be operated.

  By the way, the fuel injection valve 30 may have variations in injection characteristics due to individual differences or aging. When the injection characteristic deviates from the reference characteristic, the control accuracy of the fuel injection control decreases. Therefore, in this embodiment, the deviation of the injection characteristics is learned based on the difference between the predicted value and the detected value of the oxygen concentration in the exhaust gas. Here, the factors that cause the predicted value of the oxygen concentration in the exhaust gas to deviate from the detected value include the output error of the sensors used for calculating the predicted values of the air flow meter 16 and the oxygen concentration sensor 34, and the injection characteristics of the fuel injection valve 30. Deviation is considered. However, among these factors, the inventors have found that the dominant factor causing the deviation between the predicted value and the detected value is the deviation in the injection characteristics.

  For this reason, in this embodiment, the injection amount learning unit B32 is provided, and the deviation of the injection characteristic is learned based on the model error learning value LM learned by the model error learning unit B18, and is stored in the constant memory holding memory 52. . Specifically, learning of the deviation of the injection characteristic is performed as a storage of a correction amount that compensates for the deviation of the injection characteristic. This correction amount is a quantification of the deviation in injection characteristics. In the present embodiment, this correction amount is not the command injection amount QFIN but an amount for correcting the command injection period base value TQb. This is because the command injection amount QFIN is used by controlling the output characteristics of the diesel engine 10 (not shown). The injection amount learning unit B32 compensates for a deviation in the injection characteristic of the fuel injection valve 30, and calculates a correction amount for accurately injecting the fuel of the command injection amount QFIN. For this reason, when the command injection amount QFIN is corrected, it is required to set such that the corrected value is not used as information regarding the fuel amount actually injected from the fuel injection valve 30. On the other hand, such a problem can be avoided by correcting the command injection period base value as in the present embodiment.

  Note that the command injection period base value is corrected in the final injection period setting unit B34 by the correction amount ΔTQ learned by the injection amount learning unit B32, and the final injection period is set.

  Here, the learning of the deviation of the injection characteristic will be described in detail. In order to describe the process related to the learning process of the injection characteristic deviation in detail, the process in the exhaust oxygen concentration prediction unit B6, the process in the model error learning unit B18, and the process in the injection amount learning unit B32 will be described below. .

  FIG. 3 shows processing performed by the exhaust oxygen concentration prediction unit B6. This process is repeatedly executed by the ECU 50, for example, at a predetermined cycle.

  In this series of processes, first, in step S10, the mass flow rate of gas flowing into the combustion chamber 26 of the diesel engine 10 (cylinder inflow gas amount Mcld) is calculated. Here, first, the volumetric efficiency is calculated from the intake pressure Pm and the rotational speed NE. Then, the cylinder inflow gas amount Mcld is calculated from the gas state equation using the intake air temperature Tm, the intake pressure Pm, and the volumetric efficiency.

  In the subsequent step S12, the amount of fresh air flowing into the manifold 25 is calculated based on, for example, the intake air amount MAF, the intake pressure Pm, and the intake air temperature Tm. As this calculation method, for example, any method described in Patent Document 2 may be used. In the subsequent step S14, for example, the EGR amount flowing into the manifold 25 is calculated based on the intake pressure Pm, the intake air temperature Tm, the cylinder inflow gas amount Mcld, and the new air amount. As this calculation method, for example, the method described in Patent Document 2 may be used. In the subsequent step S16, for example, the oxygen concentration in the gas flowing into the combustion chamber 26 is calculated based on the EGR amount, the fresh air amount, the final predicted concentration Df output from the final predicted value setting unit B14 of FIG. . This calculation method may be described in, for example, Patent Document 2 described above.

  In the subsequent step S18, the predicted concentration base value Db is calculated. Here, for example, the predicted concentration base value Db is calculated based on the cylinder inflow gas amount Mcld, the oxygen amount flowing into the combustion chamber 26, and the command injection amount base value Qb. Here, first, the amount of oxygen consumed by the command injection amount base value Qb is calculated by multiplying the oxygen consumption amount k per unit fuel amount by the command injection amount base value Qb. Next, the residual oxygen amount is calculated by subtracting the consumed oxygen amount from the oxygen amount flowing into the combustion chamber 26. Then, the predicted concentration base value Db is calculated by dividing the residual oxygen amount by the sum of the value obtained by multiplying the command injection amount by the fuel specific gravity kp and the cylinder inflow gas amount Mcld.

  In addition, when the process of step S18 is completed, this series of processes is once complete | finished.

  FIG. 4 shows a procedure of processing performed by the model error learning unit B18 (and error calculation unit B16). This process is repeatedly executed by the ECU 50, for example, at a predetermined cycle.

  In this series of processes, first, in step S20, it is determined whether or not the diesel engine 10 is in a steady state. Since the model error learning is performed based on the difference between the final predicted concentration Df and the detected value, it is desirable that the model error learning be performed under a situation where they are assumed to be equal if there is no individual difference in the fuel injection valve 30 or the like. . In the present embodiment, the steady operation state of the diesel engine 10 is determined to be in such a situation. In the subsequent step S22, a difference ΔD between the final predicted concentration Df and the detected value Dr with respect to the oxygen concentration in the exhaust gas is calculated. In the subsequent step S24, a model error learning value LM is calculated. Specifically, the model error learning value LM is calculated by integral control based on the difference ΔD. That is, the current model error learning value is calculated by adding the integral constant K multiplied by the difference ΔD calculated this time to the previous model error learning value LM. In the subsequent step S26, the model error learning value is stored in the constantly storing and holding memory 52 as the value of the corresponding region among a plurality of regions for storing different learning values. Here, the plurality of regions are obtained by dividing the entire operation region of the diesel engine for each region in which the model error learning value LM has a similar value. In the present embodiment, as shown in FIG. 5A, the rotation speed NE and the injection amount Q are divided into a plurality of regions.

  In the subsequent step S28, the model error learning number C indicating the number of model error learning is incremented. As shown in FIG. 5B, the model error learning frequency C is used to count the learning frequency separately for each area divided in the same manner as each learning area of the model error learning value LM. In subsequent step S30, the stored value McldL is calculated by filtering the cylinder inflow gas amount Mcld when the difference ΔD is calculated this time. In this embodiment, as the filter process, a so-called weighted average process that calculates the sum of a value obtained by multiplying the previous stored value McldL by the coefficient α and a value obtained by multiplying the current cylinder inflow gas amount Mcld by the coefficient β (α + β = 1). Is used. In step S32, the stored value McldL stored in the always-on storage holding memory 52 is updated. That is, as shown in FIG. 5C, the stored value McldL is stored in a corresponding area among the areas divided in the same manner as each learning area of the model error learning value LM. When the process of step S32 is completed or when a negative determination is made in step S20, this series of processes is temporarily terminated.

  FIG. 6 shows a process performed by the injection amount learning unit B32. The process shown in FIG. 6 is repeatedly executed by the ECU 50, for example, at a predetermined cycle.

  In this series of processing, first, in step S40, it is determined whether or not the number of learning times C of the model error learning value LM is equal to or greater than the predetermined number A. This process is to determine whether or not the model error learning value LM is a highly reliable value. That is, if the number of learning times C of the model error learning value LM is equal to or greater than the predetermined number A, the model error learning is sufficiently performed, so that the value is considered to have converged, and is a highly reliable value. Judge. Here, for each region divided by the rotational speed NE and the injection amount Q, it is determined whether or not the number of learning of the model error learning value LM is a predetermined number A or more. The predetermined number A is set to the same value among a plurality of regions divided by the rotation speed and the injection amount.

  In the subsequent step S42, the fuel pressure in the common rail 28 is calculated based on the rotation speed and the injection amount for the region determined to be equal to or greater than the predetermined number A in step S40. Here, the fuel pressure may be estimated by a logic that calculates the target fuel pressure from the rotation speed and the injection amount.

  In the subsequent step S44, the injection amount learning value LQ is calculated for each region divided by the fuel pressure and the injection amount, based on the memory value McldL of the cylinder inflow gas amount and the model error learning value LM. Here, the region divided by the fuel pressure and the injection amount is converted from the region divided by the rotation speed and the injection amount, which is the region where the model error learning value LM is learned, into a region in the two-dimensional space of the fuel pressure and the injection amount. It is a thing. This conversion is performed because the command injection period is obtained from the fuel pressure and the injection amount, and the fuel pressure dependency of the injection characteristics (or the deviation amount) is large. By learning for each region divided by the fuel pressure and the injection amount, the learning result can be easily used for fuel injection control. In addition, by performing learning for each region divided by the fuel pressure, the fuel pressure dependency of the injection characteristics is managed by learning.

  On the other hand, the injection amount learning value LQ is expressed by the following equation using the exhaust mass flow rate EM, the oxygen consumption amount k per unit fuel amount, and the fuel specific gravity kp.

LQ = (EM × LM) / (k × kp)
The mass of the exhaust flow rate can be calculated as the sum of the memory value McldL of the cylinder inflow gas amount and the injection amount Q per unit time.

  In a succeeding step S46, it is determined whether or not the injection amount learning value LQ is equal to or higher than a lower limit guard. If it is determined that it is less than the lower limit guard, the injection amount learning value LQ is set as the lower limit guard value in step S48. In step S50, it is determined whether or not the injection amount learning value LQ is equal to or lower than the upper limit guard. When the injection amount learning value LQ exceeds the upper limit guard, the injection amount learning value LQ is set as the upper limit guard value in step S52.

  In the following step S54, the injection amount learning value LQ is filtered. As this filtering process, in this embodiment, the current injection amount learning value LQ (i) is multiplied by a coefficient γ (0 <γ <1)) and the previous injection amount learning value LQ (i−1) is a coefficient. A so-called weighted average process is used to calculate the sum of those multiplied by “1-γ”.

  In the following step S56, the injection amount learning value LQ (i) is converted into the above-described injection period correction amount ΔTQ, which is a parameter obtained by quantifying the deviation in the injection characteristics. Here, based on the information for obtaining the injection period from the fuel pressure and the injection amount by the injection period setting unit B12 shown in FIG. 2, the correction amount ΔTQ is calculated from the injection amount learning value LQ and the fuel pressure. When the correction amount ΔTQ is calculated in this way, the correction amount ΔTQ is stored in the constant storage holding memory 52 in step S58. That is, as shown in FIG. 7, the correction amount ΔTQ is stored in the corresponding region among the regions divided by the fuel pressure P and the injection amount Q.

  In subsequent step S60, the model error learning value LM is updated in accordance with the learning (updating) of the correction amount ΔTQ. That is, when the correction amount ΔTQ is updated, the command injection period is set by correcting the command injection period base value TQb with the correction amount ΔTQ. Therefore, after the command injection period base value is corrected by the correction amount ΔTQ, the difference ΔD between the predicted value of the oxygen concentration in the exhaust gas and the detected value is considered to be different from the value before the correction amount ΔTQ is updated. . For this reason, it is considered that the error learned by the model error learning unit B18 is also different from that before the correction amount ΔTQ is updated. In the present embodiment, focusing on this point, the model error learning value LM is set according to the degree to which the difference between the predicted value and the detected value of the oxygen concentration in the exhaust gas is compensated by updating the correction amount ΔTQ. to correct. The correction of the model error learning value LM according to the assumed degree does not necessarily mean initialization of the model error learning value LM. This is because the injection amount learning value LQ converted into the correction amount ΔTQ for the guard processing in steps S48 and S52 and the filtering processing in step S54 converts the model error learning value LM into the injection amount (step S44). This is because the value calculated by (1) does not necessarily match.

  When the model error learning value LM is updated in this way, the learning frequency C in the updated area is initialized. At this time, the corresponding stored value McldL may also be initialized. When the process of step S60 is completed in this way, or when a negative determination is made in step S40, this series of processes is temporarily terminated.

  FIG. 8 shows a learning mode of the correction amount ΔTQ by the above processing. Specifically, FIG. 8A shows the transition of the model error learning value LM (NE, Q) in a specific region, and FIG. 8B shows the model error learning frequency C (NE, Q) in the region. Shows the transition. FIG. 8C shows the transition of the injection amount learning value LQ (P, Q) in the corresponding region, and FIG. 8D shows the transition of the correction amount ΔTQ (P, Q) in the region. .

  As shown in the figure, when the learning number C reaches the predetermined number A, the injection amount learning value LQ and the correction amount ΔTQ are updated. Along with this, the model error learning value LM is corrected, and the learning count C is initialized.

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

  (1) A predicted value of oxygen concentration in the exhaust gas is calculated based on the physical quantity of the intake air sucked into the intake passage 12 of the diesel engine 10 and the command injection amount for the fuel injection valve 30, and the detected value of the oxygen concentration sensor 34 Based on the difference between the predicted values, the deviation (correction amount ΔTQ) of the injection characteristic of the fuel injection valve 30 was learned. Thereby, it is possible to learn the deviation of the injection characteristics in a wide area.

  (2) Learning of the injection amount learning value LQ based on the difference between the detected value and the predicted value is performed using information related to the mass of the exhaust gas at the time of predicting the oxygen concentration in the exhaust gas (stored value McldL of the cylinder inflow gas amount Mcld). It was. Thereby, the injection amount learning value LQ can be suitably calculated.

  (3) When the learning number C of the model error learning value LM is equal to or greater than the predetermined number A, the injection amount learning value LQ is learned. Thereby, learning accuracy can be improved.

  (4) When learning the model error learning value LM for each region divided by the rotational speed and load (injection amount Q) of the output shaft of the diesel engine 10, information related to the exhaust mass (memory value McldL) is also stored. . Thereby, the injection amount learning value LQ can be suitably calculated from the model error learning value LM.

  (5) A region where the model error learning value LM is learned is converted into a region divided by the fuel pressure, and the injection amount learning value LQ is learned in the converted region. As a result, the injection amount learning value LQ can be learned in a manner that suitably reflects the fuel pressure dependency of the injection characteristics.

  (6) When the deviation of the injection characteristic (correction amount ΔTQ, injection amount learning value LQ) is learned, the model error learning value LM is corrected according to the learning. Thereby, the model error learning value LM can be updated to an appropriate value according to the error of the predicted value with respect to the detected value.

  (7) After filtering the model error learning value LM, the injection amount learning value LQ is learned using the value after the filter processing. Thereby, the influence of noise or the like can be suitably suppressed when learning the correction amount ΔTQ and the injection amount learning value LQ, and as a result, the learning accuracy of the correction amount ΔTQ and the injection amount learning value LQ can be improved. However, in this case, the deviation of the injection characteristic is compensated based on the learning of the correction amount ΔTQ, but the model error learning value LM is not directly canceled by this compensation. In this regard, according to the present embodiment, the correction is performed according to the learning of the correction amount ΔTQ. Therefore, the model error learning value is obtained even though the correction amount ΔTQ is learned based on the value after the filter processing. The learning result of LM can be updated to an appropriate value.

  (8) The EGR valve 40 was operated based on the predicted value of the oxygen concentration in the exhaust. Thus, the EGR valve 40 can be operated based on accurate information about the oxygen concentration in the exhaust gas regardless of whether the diesel engine 10 is in a transient state or a steady state.

  (9) Based on feedback control of the predicted value of oxygen concentration in the exhaust gas (predicted concentration base value Db) to the target value, the predicted value was corrected and the final predicted concentration Df was calculated. In this manner, learning can be performed with higher accuracy by correcting the predicted value (final predicted concentration Df) used for learning based on the influence of feedback control.

(Other embodiments)
Each of the above embodiments may be modified as follows.

  In the above embodiment, in order to simplify the description, the injection amount (required injection amount) required for realizing the required torque according to the accelerator pedal operation amount and the rotational speed is divided into a plurality of times and 1 Although the multistage injection control for injecting within the combustion cycle has not been described, in the diesel engine 10, the multistage injection is usually performed. At this time, the injection amount setting unit B2 may calculate each command injection amount base value Qb by dividing the required injection amount into a plurality of times. Further, the exhaust oxygen concentration prediction unit B6 may calculate the predicted concentration base value Db using the sum of these command injection amount base values Qb.

  At this time, it becomes a problem in the injection amount learning unit B32 whether the model error learning value LM is considered as a deviation of the injection characteristic at the time of fuel injection in the multistage injection control. Here, all the fuel injections may contribute to the model error learning value LM, but it is assumed that the model error learning value LM is generated only by fuel injection or the like that is assumed to have the largest contribution to the model error learning value LM. May be considered. In this case, it is convenient to consider that the model error learning value LM is generated by the main injection having the maximum injection amount during the multi-stage injection. In particular, at the time of idle speed control of the diesel engine 10, the injection amount required when the actual rotation speed is feedback-controlled to the target rotation speed is divided into equal amounts, and each injection amount is a minute injection that is about the pilot injection amount. And has a function of learning the deviation of the injection characteristic of the fuel injection valve 30 at the minute injection amount based on the difference between the required injection amount at this time and the reference injection amount (for example, JP In Japanese Patent Publication No. 2003-343328, it is desirable to use main injection. That is, the learning of the injection characteristic deviation at the injection amount equal to or less than the injection amount required at the idle rotation speed control can be performed at the idle rotation speed control, while the learning of the injection amount larger than this is performed. Because it is difficult. On the other hand, according to the method exemplified in the above embodiment, it is possible to perform learning even when the required injection amount is larger than that during idle rotation speed control. Then, if learning about the minute injection amount such as pilot injection is completed and fuel injection is performed in the form of correction by learning for pilot injection or the like, if learning illustrated in the above embodiment is performed, main injection or the like This makes it possible to learn the deviation of the injection characteristic with high accuracy.

  In the above embodiment, when learning the model error learning value LM, the cylinder inflow gas flow rate Mcld at that time is stored, but the information regarding the exhaust mass is not limited to this. For example, the intake air temperature Tm may be stored.

  In the above embodiment, the correction amount TQ of the injection period is used as the deviation of the injection characteristics. However, the present invention is not limited to this, and the injection amount learning value LQ may be used. In this case, it should be noted that the value before being corrected by the injection amount learning value LQ (command injection amount base value Qb) is the actual injection amount.

  In the above embodiment, each filter process is a weighted average process, but the present invention is not limited to this. For example, a moving average process may be used.

  In view of the fact that the model error learning value LM is calculated by the filtering process of the difference ΔD, the injection amount learning value LQ is not calculated by the filtering process (without providing the process of step S54 in FIG. 6). A value obtained by performing the guard process on the value calculated in step S44 of FIG. 6 may be used as the final injection amount learning value LQ. Moreover, it is not necessary to perform a guard process.

  Considering that the model error learning value LM is calculated by the filtering process of the difference ΔD, it is considered that the value calculated in step S44 will not change suddenly. The injection amount learning value LQ may be learned each time the operation is performed.

  In the above embodiment, the injection amount is used as the load. However, when performing the torque base control, the required torque may be used as the load.

  In the above embodiment, the final predicted concentration Df is calculated based on the control mode (operation mode of the fuel injection valve 30) when the predicted value of the oxygen concentration in the exhaust gas is feedback-controlled to the detected value. The operation mode of the EGR valve 40 may be taken into consideration.

  In the above embodiment, the feedback control based on the difference between the predicted value of the oxygen concentration in the exhaust gas and the target value is performed, but such a function may not be provided. That is, the predicted value may be calculated only for learning the injection characteristic deviation based on the difference between the predicted value and the detected value.

  In the above embodiment, the difference between the predicted value and the detected value of the oxygen concentration in the exhaust gas is caused by the deviation of the injection characteristic, but actually it is also caused by the output error of the sensor. For this reason, if it has a function to detect the output error of the sensor, it is desirable that a value obtained by subtracting this error in advance from the difference between the predicted value and the detected value is an amount resulting from the deviation of the injection characteristics.

  In the above embodiment, a diesel engine is used as the internal combustion engine, but a gasoline engine or the like may be used.

The figure which shows the whole structure of the engine system concerning one Embodiment. The block diagram which shows the learning process concerning the embodiment. 6 is a flowchart showing a procedure of exhaust oxygen concentration calculation processing according to the embodiment. 9 is a flowchart showing a procedure for calculating a model error learning value according to the embodiment. The figure which shows the storage method of the parameter at the time of calculation of the model error learning value concerning the embodiment. The flowchart which shows the procedure of the injection quantity learning process concerning the embodiment. The figure which shows the memory | storage method of the correction amount concerning the embodiment. The time chart which shows the aspect of the learning process concerning the embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Diesel engine, 34 ... Oxygen concentration, 50 ... ECU (one Embodiment of a fuel-injection control apparatus).

Claims (9)

Prediction means for calculating a predicted value of the oxygen concentration in the exhaust based on a physical quantity of intake air sucked into the intake system of the internal combustion engine and a command injection amount for the fuel injection valve;
A fuel injection control device comprising: a deviation learning means for learning a deviation in injection characteristics of the fuel injection valve based on a difference between a detected value of oxygen concentration in exhaust gas and the predicted value.   The said deviation learning means performs the said learning based on the difference of the said detected value and the said predicted value using the information regarding the mass of the exhaust_gas | exhaustion at the time of the prediction of the oxygen concentration in the said exhaust_gas | exhaustion. Fuel injection control device. Further comprising error learning means for learning the error of the prediction means based on the detected value of the oxygen concentration in the exhaust gas and the predicted value;
The deviation learning means performs the deviation learning based on the detected value and the predicted value through the error learned by the error learning means when the error learning by the error learning means is a predetermined number of times or more. The fuel injection control device according to claim 1 or 2.   The error learning means learns the error for each region divided by the rotational speed and load of the output shaft of the internal combustion engine, and stores information on the mass of the exhaust gas when learning the error. The fuel injection control device according to claim 3, wherein   The deviation learning means converts an area in which an error is learned by the error learning means into an area divided by a fuel pressure, and the conversion is performed based on the learned error and information on the mass of the exhaust gas. The fuel injection control device according to claim 4, wherein a deviation in injection characteristics in the region is learned.   6. The apparatus according to claim 3, further comprising: a unit that corrects a learning result of the error learning unit according to the learned deviation of the injection characteristic when learning by the deviation learning unit is performed. The fuel injection control device described.   7. The fuel injection control apparatus according to claim 6, wherein the deviation learning means filters the error learned by the error learning means, and then learns the deviation using a value after the filtering process. The internal combustion engine includes an exhaust gas recirculation passage for recirculating exhaust gas discharged to an exhaust system to an intake system, and an area variable means for changing a flow passage area in the exhaust gas recirculation passage.
8. The fuel injection control apparatus according to claim 1, further comprising an area adjusting unit that operates the area varying unit based on a predicted value of the predicting unit. Target value setting means for setting a target value of oxygen concentration in the exhaust;
Feedback control means for performing feedback control of the predicted value to the target value;
The fuel injection control device according to claim 1, further comprising: a unit that corrects a predicted value used for the learning based on the aspect of the feedback control.

Images