JP5482532B2 - Fuel injection control device - Google Patents

Description

  The present invention relates to a fuel injection control device that performs learning injection to learn a correction amount of an injection amount of a fuel injection valve.

  Conventionally, it is known to learn the amount of deviation of the actual injection amount from the command injection amount of the fuel injection valve caused by machine difference or change over time, and calculate the correction amount of the injection amount so that the actual injection amount becomes the command injection amount. It has been. For example, in Patent Document 1, in a diesel engine that performs a small amount of pilot injection before main injection in order to reduce NOx and combustion noise, injection amount learning for correcting a small injection amount by pilot injection is performed. .

  Further, in Patent Document 2, a drive pulse width shift when the preliminary injection is started by gradually extending the drive pulse width with respect to the preliminary injection with a small injection amount executed before the main injection is learned, and the preliminary injection is performed. The drive pulse width for executing the injection is adjusted.

  When learning the injection amount in this way, the relationship between the command injection amount and the fluctuation amount of the engine operating state caused by executing the learning injection of the command injection amount is measured in advance as a basic injection characteristic for each model, Based on the basic injection characteristics, the actual injection amount is estimated from the fluctuation amount of the engine operation state caused by the learning injection.

Japanese Patent Laid-Open No. 2005-140046 Japanese Patent Laid-Open No. 3-100350

  However, since the basic injection characteristics differ from model to model, especially when many models such as commercial vehicles, agricultural machinery, and construction machines are prepared, the basic injection characteristics are measured according to the fuel pressure for each model and mapped. There is a problem that a lot of man-hours are required to create the above.

  The present invention has been made to solve the above-described problem, and an object of the present invention is to provide a fuel injection control device that estimates basic injection characteristics based on actual injection amount characteristics detected by executing learning injection. To do.

According to the first to fourth aspects of the present invention, the injection command means instructs the fuel injection valve to perform learning injection with different command injection quantities, and the fuel injection valve instructs the fuel injection valve to instruct the fuel injection valve. The actual injection characteristic detecting means detects the actual injection characteristic that represents the correlation with the actual fluctuation amount of the operating state of the internal combustion engine caused by executing the learning injection, and the actual injection characteristic detected by the actual injection characteristic detecting means is translated. Thus, the basic injection characteristic estimating means estimates the basic injection characteristic representing the correlation between the command injection quantity and the fluctuation amount of the operating state, and the command injection quantity between the basic injection characteristic estimated by the basic injection characteristic estimating means and the actual injection characteristic is calculated. The learning value calculation means calculates the difference in the increase / decrease direction as a learning value for correcting the injection amount.

  In this way, even if there is a deviation between the actual injection characteristic representing the correlation between the command injection amount and the actual fluctuation amount of the operating state and the basic injection characteristic, the difference is in the offset direction, and the learning injection By executing the above and detecting the actual injection characteristics, the basic injection characteristics can be estimated by translating the actual injection characteristics.

  Thereby, it is possible to estimate the basic injection characteristic while operating the internal combustion engine without measuring the basic injection characteristic in advance for any type of vehicle. Thereby, the difference in the increase / decrease direction of the command injection amount between the basic injection characteristic and the actual injection characteristic can be calculated as a learning value for correcting the injection amount of the fuel injection valve.

Further, according to the first to fourth aspects of the present invention, the actual injection characteristic detecting means is applicable when the square of the correlation coefficient between the plurality of command injection amounts and the actual variation amounts having different injection amounts is equal to or less than a predetermined value. The actual fluctuation amount to be used is not adopted as learning data.
In this way, when the square of the correlation coefficient between the plurality of command injection amounts having different injection amounts and the actual variation amount is equal to or less than a predetermined value, the actual injection characteristic representing the correlation between the command injection amount and the actual variation amount is detected. But reliability is low. Therefore, when the correlation coefficient between the command injection amount and the actual fluctuation amount is equal to or less than a predetermined value, it is desirable not to adopt the corresponding actual fluctuation amount as the learning data. In particular, when learning injection is performed in a combustion state in which the correlation between the command injection amount and the actual variation amount is large during traveling, this is effective in preventing a decrease in learning accuracy.
According to the second aspect of the invention, the injection command means commands learning injection to the fuel injection valve in the fuel injection state during traveling.
In this way, the learning frequency of the injection amount can be increased by executing the learning injection in the fuel injection state during traveling.

  According to the third aspect of the present invention, the actual injection characteristic detecting means learns the corresponding actual variation when the variation in the actual variation when the learning injection is executed with the same command injection amount exceeds a predetermined range. Not adopted as data for use.

  In this way, when the variation in the actual fluctuation amount when the learning injection is executed with the same command injection amount exceeds the predetermined range, it is desirable not to employ the learning data because the reliability of the learning data is low. In particular, it is effective in preventing a decrease in learning accuracy when the actual fluctuation amount varies due to a factor other than the learning injection as a result of executing the learning injection in the fuel injection state during traveling.

  According to the fourth aspect of the present invention, the actual injection characteristic detecting means, when the actual fluctuation amount does not exceed the predetermined fluctuation amount even when the command injection amount is increased to the predetermined injection amount and the learning injection is executed, The amount of variation is not used as learning data.

  Thus, if the actual fluctuation amount does not exceed the predetermined fluctuation amount even when the command injection amount is increased to the predetermined injection amount and the learning injection is executed, the rate of change of the actual fluctuation amount with respect to the command injection amount, in other words, the actual injection Since the characteristic sensitivity is low, the detection accuracy of the actual injection characteristic is low. Even if the correction amount of the injection amount is calculated based on the difference between the basic injection characteristic estimated from the actual injection characteristic with low detection accuracy and the actual injection characteristic, the correction accuracy is low, so the corresponding actual fluctuation amount is used as learning data. It is desirable not to adopt. In particular, when learning injection is performed in a combustion state in which the sensitivity of the actual injection characteristics is low during traveling, this is effective in preventing a decrease in learning accuracy.

  The functions of the plurality of means provided in the present invention are realized by hardware resources whose functions are specified by the configuration itself, hardware resources whose functions are specified by a program, or a combination thereof. The functions of the plurality of means are not limited to those realized by hardware resources that are physically independent of each other.

The block diagram which shows the fuel-injection system by this embodiment. Explanatory drawing which shows the increase in work equivalent amount by learning injection. The characteristic view which shows the shift | offset | difference of an injection characteristic. The distribution map which shows the dispersion | variation in the work equivalent amount by learning injection. The characteristic view which shows the sensitivity of the work equivalent amount learned. The characteristic view which shows the correlation with instruction | command injection amount and work equivalent amount.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
A fuel injection system according to this embodiment is shown in FIG.
(Fuel injection system 10)
The fuel injection system 10 is for injecting fuel into, for example, a four-cylinder diesel engine (hereinafter also simply referred to as “engine”) 2 for an automobile. The fuel injection system 10 includes a fuel supply pump 14, a common rail 20, a fuel injection valve 30, and an electronic control unit (ECU) 40.

  The fuel supply pump 14 incorporates a feed pump that pumps fuel from the fuel tank 12. The fuel supply pump 14 is a known pump that pressurizes the fuel sucked into the pressurizing chamber from the feed pump when the plunger reciprocates as the cam of the camshaft rotates.

  The metering valve 16 serving as a metering actuator is installed on the suction side of the fuel supply pump 14 and controls the amount of fuel sucked by each plunger of the fuel supply pump 14 in the suction stroke by current control. . By adjusting the fuel intake amount, the fuel discharge amount from each plunger of the fuel supply pump 14 is adjusted. The amount of fuel discharged from each plunger of the fuel supply pump 14 may be measured by a metering valve installed on the discharge side of the fuel supply pump 14.

  The common rail 20 is a hollow member that accumulates fuel discharged from the fuel supply pump 14. The common rail 20 is provided with a pressure sensor 22 that detects the internal fuel pressure (common rail pressure), and a pressure limiter 24 that opens to discharge the fuel in the common rail 20 when the common rail pressure exceeds a predetermined pressure. .

  The engine 2 is provided with a rotational speed sensor 32 that detects an engine rotational speed (NE) as a sensor that detects an operating state. Further, as other sensors for detecting the driving state, an accelerator sensor for detecting an accelerator opening (ACCP) that is an operation amount of an accelerator pedal by a driver, a temperature of cooling water (water temperature), a temperature of intake air (intake air temperature). The fuel injection system 10 is provided with a temperature sensor or the like for detecting each of these.

  The fuel injection valve 30 is installed in each cylinder of the engine 2 and injects fuel accumulated in the common rail 20 into the cylinder. The fuel injection valve 30 is, for example, a known electromagnetic valve that controls the lift of the nozzle needle that opens and closes the nozzle hole with the pressure in the control chamber. The injection amount of the fuel injection valve 30 is controlled by the pulse width of the injection command signal commanded from the ECU 40. As the pulse width of the injection command signal increases, the injection amount increases.

  The ECU 40 is mainly configured by a microcomputer centering on a CPU, RAM, ROM, flash memory and the like. The ECU 40 executes various control of the fuel injection system 10 based on output signals taken from various sensors including the pressure sensor 22 and the rotation speed sensor 32 by the CPU executing a control program stored in the ROM or flash memory. Run.

  For example, the ECU 40 controls the energization amount to the metering valve 16 so that the common rail pressure detected by the pressure sensor 22 becomes the target pressure, and regulates the discharge amount of the fuel supply pump 14. The ECU 40 sets a current value for controlling the metering valve 16 based on a characteristic map representing a correlation between the current value for controlling the metering valve 16 and the discharge amount.

  Further, the ECU 40 controls the fuel injection amount of the fuel injection valve 30, the fuel injection timing, and the pattern of the multistage injection in which the pilot injection, the pre-injection before the main injection, the after injection, the post injection after the pilot injection, etc. To do.

  The ECU 40 stores a so-called TQ map indicating the correlation between the pulse width (T) of the injection command signal for instructing the fuel injection valve 30 and the injection amount (Q) into a ROM or flash memory for each predetermined pressure range of the common rail pressure. I remember it. Then, when the injection amount of the fuel injection valve 30 is determined based on the engine speed and the accelerator opening, the ECU 40 refers to the TQ map of the corresponding pressure range according to the common rail pressure detected by the pressure sensor 22, The pulse width of the injection command signal that commands the determined injection amount to the fuel injection valve 30 is acquired from the TQ map.

(Injection amount learning)
The ECU 40 instructs the fuel injection valve 30 to perform pilot injection for learning when the following injection amount learning conditions (1) and (2) are satisfied.
(1) Every predetermined traveling distance, for example, every 5000 km, or every predetermined driving time, for example, every 500 hours.
(2) When the engine operating state is stable. For example, fluctuations in the rotation speed, common rail pressure, and injection amount are within predetermined ranges.

  The pilot injection for learning is different from the pilot injection in the normal multi-stage injection commanded for the purpose of reducing combustion noise, and detects the fluctuation amount of the operating state of the engine 2 when there is no pilot injection and when there is no pilot injection. To be commanded.

  When the learning condition is satisfied, the ECU 40 commands the pre-injection and the main injection a predetermined number of times without pilot injection, for example, as shown in FIG. As shown in (B) of 2, multi-stage injection in which pilot injection is added to pre-injection and main injection is commanded a predetermined number of times. Then, based on the fluctuation amount of the engine speed that fluctuates due to the addition of the pilot injection (the solid line portion in FIG. 2C), the work amount of the engine 2 as the actual fluctuation amount of the engine operating state that fluctuates due to the pilot injection. The actual work equivalent amount corresponding to is calculated.

  When calculating the work equivalent amount from the fluctuation amount of the engine speed, until the filter action of the bandpass filter (BPF) for removing noise from the detection signal of the engine speed is stabilized, as shown by a dotted line 200 in FIG. Ignoring the first few injection data of pilot injection with learning injection and without pilot injection.

  Then, after the filter function of the BPF is stabilized, the pilot injection injected at the command pilot injection amount from, for example, the difference between the average values of the work equivalent amounts of the injections for each of the 10 injections when there is no pilot injection and when there is pilot injection. The work equivalent amount (ΔW) is calculated.

The calculation of the work equivalent amount is repeated, for example, five times for each command pilot injection amount, and the average value is set as the work equivalent amount for the corresponding command pilot injection amount.
Then, the command injection quantity of the pilot injection is increased from 0 mm ³ to each 1 mm ³ for example 5 mm ³ as the predetermined injection amount, relative to the learning data representing the correlation between the command pilot injection quantity and work substantial amount by the least square method Then, an actual injection characteristic representing the correlation between the command pilot injection amount and the actual work equivalent amount is calculated. The correlation between the command pilot injection amount and the work equivalent amount is expressed by a linear expression.

As shown in FIG. 3, when the actual injection characteristic 100 obtained in this way does not pass through the origin of the command pilot injection amount = 0 mm ³ and the work equivalent amount = 0, the command pilot injection amount and the work equivalent amount are It is determined that the deviation of the actual injection characteristic 100 caused by the machine difference or the change with time with respect to the basic injection characteristic 110 representing the correlation is caused by the deviation in the offset direction rather than the inclination of the characteristic.

Thus, the basic injection characteristic 110 is changed by translating the actual injection characteristic 100 in the offset direction, that is, the increase / decrease direction of the command pilot injection quantity so as to pass through the origin of the command pilot injection quantity = 0 mm ³ and the work equivalent quantity = 0. Can be estimated. For example, when the primary expression of the actual injection characteristic calculated by the least square method is expressed by the following expression (1), the basic injection characteristic 110 is expressed by the following expression (2).

ΔW = α × Q + β (1)
ΔW = α × Q (2)
Q: Command pilot injection amount.
ΔW: work equivalent when pilot injection is present.
α: Slope of injection characteristics.
β: intercept of actual injection characteristics.

Therefore, the injection amount ΔQ, which is a difference between the actual injection characteristic 100 and the basic injection characteristic 110, can be calculated from the following equation (3).
α × Q2 + β = α × Q1
α (Q2−Q1) = − β
(Q2−Q1) = − β / α
ΔQ = −β / α (3)
The ECU 40 executes the above-described injection amount learning when the learning condition is satisfied, and stores the injection amount ΔQ learned for each common rail pressure at that time. In this case, the ECU 40 learns the injection amount ΔQ from the basic injection amount characteristic and the actual injection amount characteristic for each predetermined pressure range by adjusting the discharge amount of the fuel supply pump 14 to increase or decrease the common rail pressure. Also good.

  The ECU 40 corrects the minute injection amount in the normal injection by correcting the command injection amount at the corresponding common rail pressure based on the injection amount ΔQ calculated as a learning value for correcting the injection amount by the equation (3).

  Note that the minute injection amount correction may be performed based on the common injection amount ΔQ calculated as the learning value regardless of the engine operation region, or may be performed based on the injection amount ΔQ learned for each engine operation region. Good.

(Reliability of learning data)
Next, the reliability of the learning data measured by the learning injection will be described with respect to (1) variation, (2) sensitivity, and (3) correlation.

(1) Variation As shown in the upper part of FIG. 2, when there is no pilot injection shown in FIG. 4 in the work equivalent amount for each command pilot injection amount obtained by learning injection performed with no pilot injection and with pilot injection. When the relationship between the standard deviation σ _nx of the work equivalent amount and the standard deviation σ _ax of the work equivalent amount with pilot injection is the following expression (4), the ECU 40 corresponds to the work at the command pilot injection amount at that time The quantity is not adopted as learning data, and the learning injection is retried.

σ _ax > 2σ _nx (4)
This is because the variation of the work equivalent amount when the pilot injection is performed with respect to the absence of pilot injection is too large exceeding the predetermined range, and the reliability of the data is low.

If the equation (4) is not satisfied and the equation (4) is not satisfied even if the learning injection is retried a predetermined number of times, the ECU 40 stops the injection amount learning.
(2) As shown in sensitivity Figure 5 by a white circle, the pilot injection quantity increases from 0 mm ³ by 1 mm ^3, even at 5 mm ^3, and the work substantial amount of time without the pilot injection, when there pilot injection When the test statistic u satisfies the following equation (5) as a result of the one-sided test with a significance level of 1%, the ECU 40 obtains the work equivalent in the current command pilot injection amount as learning data. Retry the injection amount learning.

u <2.326 (5)
This is because the work equivalent amount has not increased sufficiently despite the pilot injection, but the sensitivity for calculating the primary line of the actual injection characteristics is low, and the accuracy of calculating the actual injection characteristics is low. . If the equation (5) is not satisfied even after repeating the injection amount learning for a predetermined number of times, the ECU 40 stops the injection amount learning.

In contrast, the pilot injection amount, when going is increased from 0 mm ³ by 1 mm ^3, the change from the relationship of formula (5) satisfies the following equation (6), ECU 40 sensitivity required injection amount learning It is determined that the data satisfying the conditions has been acquired.

u ≧ 2.326 (6)
The actual injection characteristic is calculated from the relationship between the command pilot injection amount that first satisfies the equation (6), the command pilot injection amount immediately before the command pilot injection amount, and a total of five different command pilot injection amounts and work equivalent amounts thereafter. calculate. For example, in a black circle shown in FIG. 5, is satisfied equation (6) to the beginning 3 mm ^3, comprises a 2 mm ³ immediately before, 2 mm ³ in total in ～6Mm ³ 5 pieces of different command pilot injection amount and work substantial amount Therefore, the actual injection characteristic 100 is calculated by the least square method.

However, when the command pilot injection amount is increased to 5 mm ³ and the equation (6) is not satisfied, the ECU 40 stops the injection amount learning.
On the other hand, as indicated by the dotted line 120 when the command pilot injection amount is 0 mm ³ , the relationship between the standard deviation σ _nx of the work equivalent amount without pilot injection and the work equivalent amount ΔW satisfies the following equation (7). In this case, since the sensitivity of the work equivalent amount to the command pilot injection amount is too high, the ECU 40 stops the injection amount learning.

ΔW> 5σ _nx (7)
(3) Correlation Even if the actual injection characteristic 100 is calculated by the least square method from a plurality of learning data representing the relationship between the command pilot injection amount and the work equivalent amount, as shown in FIG. When the distance from the learning data is large, it is determined that the correlation between the detected command pilot injection amount and the work equivalent amount is low and the reliability of the calculated actual injection characteristic 100 is low. As a standard representing the degree of correlation between the command pilot injection amount and the work equivalent amount, the square of the correlation coefficient (R) shown in the following equation (8) is adopted.

R ² = {(Σ (x _i −x _ave ) (y _i −y _ave )} ²
/ {Σ (x _i −x _ave ) ² Σ (y _i −y _ave ) ² }
= (Σx _i y _i −nx _ave y _ave ) ²
/ {(Σx _i ² −nx _ave ² ) (Σy _i ² −ny _ave ² )} (8)
n: Number of data.
x _i : Each data of the command pilot injection amount.
x _ave : Average of command pilot injection amount.
y _ave : Average of work equivalent amount.
y _i : Each data of work equivalent amount.

The closer the correlation coefficient (R) is to 1 or -1, the higher the correlation between the detected command pilot injection amount and the work equivalent amount. Therefore, it is determined whether R ² is larger than a predetermined value. For example, when R ² > 0.7 is satisfied, it is determined that the correlation between the command pilot injection amount learned this time and the work equivalent amount is high, and learning data Adopt as. The magnitude relationship with the predetermined value may be determined directly based on the value of the correlation coefficient (R) instead of the square of the correlation coefficient (R).

On the other hand, when R ² is equal to or less than a predetermined value, for example, R ² ≦ 0.7, the ECU 40 determines that the correlation between the command pilot injection amount learned this time and the work equivalent amount is low, and does not adopt it as learning data. Retry the injection amount learning. If R ² ≦ 0.7 even if the injection amount learning is repeated a predetermined number of times, the ECU 40 stops the injection amount learning.

  In the above-described embodiment, the basic injection characteristic is estimated from the actual injection characteristic detected in the fuel injection state during traveling. Thereby, since it is not necessary to measure the basic injection characteristics in advance, the man-hour for measuring the basic injection characteristics for each model can be omitted. In any model, the difference in the command injection amount between the basic injection characteristic and the actual injection characteristic can be calculated as a learning value for correcting the injection amount of the fuel injection valve.

  Then, by correcting the command injection amount based on the calculated learning value, it is possible to execute a minute injection within a desired error range when performing a minute injection such as pilot injection or post injection during normal injection. . Thereby, it can prevent that very small amount injection becomes non-injection.

  Further, since the learning injection is executed in the fuel injection state during traveling, the execution frequency of the learning injection can be increased. Furthermore, the learning injection can be executed in a wide fuel pressure range in the fuel injection state during traveling.

  In addition, the reliability of the learning data measured by the learning injection is determined with respect to variation, sensitivity, and correlation, and it is determined whether or not to use the learning data and cancel the injection amount learning. When performing the learning injection, it is effective in preventing a decrease in learning accuracy when the actual variation varies due to factors other than the learning injection.

In this embodiment, the engine 2 corresponds to the internal combustion engine of the present invention, the fuel injection valve 30 corresponds to the fuel injection valve of the present invention, and the ECU 40 corresponds to the fuel injection control device of the present invention.
The ECU 40 also functions as an injection command means, actual injection characteristic detection means, basic injection characteristic estimation means, and learned value calculation means of the present invention.

[Other Embodiments]
In the above embodiment, the learning injection is executed in the fuel injection state during traveling. In addition to this, as long as the noise, vibration, drivability, and the like due to the learning injection are within the allowable range, the learning injection may be executed in the deceleration no-injection operation state and the idle operation state, for example.

  Further, in the above embodiment, in order to ensure the reliability of the learning data measured by the learning injection, it is determined about the variation, sensitivity, and correlation of the learning data, and whether to accept the learning data and cancel the injection amount learning are determined. did. On the other hand, depending on the required accuracy for the reliability of the learning data, the measured learning data may be used as it is, and the injection amount learning may be executed.

  In the above embodiment, the functions of the injection command means, the actual injection characteristic detection means, the basic injection characteristic estimation means, and the learning value calculation means are realized by the ECU 40 whose functions are specified by the control program. On the other hand, at least a part of the functions of the above means may be realized by hardware whose function is specified by the circuit configuration itself.

  As described above, the present invention is not limited to the above-described embodiment, and can be applied to various embodiments without departing from the gist thereof.

2: diesel engine (internal combustion engine), 30: fuel injection valve, 40: ECU (fuel injection control device, injection command means, actual injection characteristic detection means, basic injection characteristic estimation means, learning value calculation means)

Claims (4)

Injection command means for instructing the fuel injection valve to perform learning injection with different command injection amounts;
An actual injection characteristic representing a correlation between the command injection amount commanded by the injection command means and an actual variation amount that is an actual variation amount of the operating state of the internal combustion engine caused by the fuel injection valve executing the learning injection An actual injection characteristic detecting means for detecting
A basic injection characteristic estimation unit that translates the actual injection characteristic detected by the actual injection characteristic detection unit and estimates a basic injection characteristic that represents a correlation between the command injection amount and the fluctuation amount of the operating state;
Learning value calculating means for calculating a difference in the increase / decrease direction of the command injection amount between the basic injection characteristic estimated by the basic injection characteristic estimating means and the actual injection characteristic as a learning value for correcting the injection amount;
Equipped with a,
The actual injection characteristic detecting means adopts the corresponding actual variation amount as learning data when the square of the correlation coefficient between the plurality of command injection amounts having different injection amounts and the actual variation amount is a predetermined value or less. do not do,
A fuel injection control device.   The fuel injection control device according to claim 1, wherein the injection command means commands the learning injection to the fuel injection valve in a fuel injection state during traveling.   The actual injection characteristic detection means does not adopt the corresponding actual variation amount as learning data when the variation in the actual variation amount when the learning injection is executed with the same command injection amount exceeds a predetermined range. The fuel injection control device according to claim 1, wherein the fuel injection control device is a fuel injection control device.   The actual injection characteristic detection means increases the command injection amount to a predetermined injection amount, and if the actual fluctuation amount does not exceed the predetermined fluctuation amount even when the learning injection is executed, the corresponding actual fluctuation amount is used for learning. The fuel injection control device according to any one of claims 1 to 3, wherein the fuel injection control device is not adopted as data.

Images