JP2001123879A - Combustion state detecting device for internal combustion engine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately detect the combustion state, even when the operation state of an internal combustion engine is changed variously. SOLUTION: A CPU 31 in an ECU 30 determines the maximum value and the minimum value of a rotation speed at every one combustion of an engine 10 and calculates a rotation fluctuation amount by a difference between the values. The CPU 31 performs statistics processing on the rotation fluctuation amount of a prescribed sampling parameter content parameter content and calculates a combustion roughness value serving as a parameter indicating a combustion state through a result of statistics processing. Actually, a standard deviation is calculated from a rotation fluctuation amount of the given sampling parameter and decides the combustion roughness value by the calculated standard deviation. Further, the CPU 31 varies the sampling parameter according to a deviation between the combustion roughness value and its target value, the transient state of the engine 10, and the operation region of the engine 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an apparatus for detecting a combustion state of an internal combustion engine, and more particularly to an apparatus for suitably detecting a combustion roughness value as a parameter indicating a combustion state.

[0002]

2. Description of the Related Art Conventionally, there has been a control apparatus for an internal combustion engine which detects a combustion state from a rotation fluctuation amount for each combustion of the internal combustion engine and reflects the detected result in air-fuel ratio control or the like. Further, there is a conventional technique for calculating a combustion roughness value as one method of detecting a combustion state. More specifically, a rotation fluctuation amount for each combustion is sampled for each cylinder,
There is a method in which a standard deviation of a rotation fluctuation amount for a predetermined sampling parameter is determined and the standard deviation is used as a combustion roughness value.

[0003]

However, the combustion state of the internal combustion engine changes sequentially according to the operating state of the engine. For example, when the throttle opening is changed by operating the accelerator, or when the air-fuel ratio is close to the stoichiometric air-fuel ratio (stoichiometric). When the state shifts from the controlled state to the state controlled in the lean region, the combustion state greatly changes. In this case, with the existing configuration in which the combustion roughness value is calculated based on the standard deviation while the sampling parameter is fixed, the combustion roughness value cannot be calculated with good responsiveness to a large change in the combustion state. If the response of the combustion roughness value is delayed and a detection error of the roughness value occurs, the air-fuel ratio control performed according to the combustion roughness value (combustion state) is adversely affected.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has a combustion state detecting device for an internal combustion engine capable of accurately detecting the combustion state even when the operating state of the internal combustion engine changes variously. The purpose is to provide.

[0005]

According to the present invention, the torque variation calculating means calculates the torque variation for each combustion of the internal combustion engine. The roughness value calculating means performs a statistical process on the calculated torque fluctuation amount for the predetermined sampling parameter, and calculates a combustion roughness value as a parameter indicating a combustion state based on a result of the statistical process. The parameter changing means changes the sampling parameter in accordance with the operating state of the internal combustion engine during the statistical processing by the roughness value calculating means.

According to the above configuration, the response and convergence of the combustion roughness value calculated by the statistical processing can be arbitrarily adjusted by appropriately changing the sampling parameter in accordance with the operation state of the internal combustion engine. That is, as an outline, responsiveness of the combustion roughness value is emphasized by reducing the sampling parameter, and conversely, convergence of the combustion roughness value is emphasized by increasing the sampling parameter. As a result, even when the operating state of the internal combustion engine fluctuates in various ways, the combustion state can be accurately detected.

The torque fluctuation amount calculating means determines the maximum value and the minimum value of the rotation speed for each combustion of the internal combustion engine, and calculates the torque fluctuation amount based on the difference between these values. A means for calculating the amount of rotation fluctuation of can be applied.

It is preferable that the parameter changing means changes the sampling parameter as described in claims 3 to 5. That is, in the invention of claim 3, when the deviation between the combustion roughness value and its target value is large, the sampling parameter is reduced.
In this case, even if the deviation of the roughness value increases, the inconvenience that the response of the combustion roughness value deteriorates is solved.

According to the fourth aspect of the present invention, the sampling parameter is reduced during the transient operation of the internal combustion engine. In this case, the inconvenience that the response of the combustion roughness value deteriorates during the transient operation of the internal combustion engine is solved.

According to the fifth aspect of the present invention, the sampling parameter is changed according to the operating range of the internal combustion engine. In this case, the combustion roughness value can be calculated in any operation range while considering the responsiveness and the convergence.

Further, in the invention described in claim 6, the roughness value calculating means performs statistical processing in units of the reference value using a fixed sampling parameter as a reference value, and sequentially accumulates the results of the statistical processing. The combustion roughness value is calculated from the average value of the integrated values. More specifically, as described in claim 7, when the sampling parameter is equal to or greater than a predetermined value, statistical processing is performed in units of the reference value, and the results of the statistical processing are sequentially integrated, and the integrated results are calculated. It is preferable to calculate the combustion roughness value from the average value.

For example, if a relatively small parameter is used as a reference value, the roughness value can be calculated in small parameter units.
Therefore, even when the sampling parameter is large and it takes a long time to calculate the roughness value, it is possible to grasp the trend of the roughness value during the calculation.

The invention in which the sampling parameter is changed according to the operating state of the engine as described above is also effective when calculating the standard deviation and the unbiased variance as statistical processing. That is, in the invention according to claim 8, a standard deviation or an unbiased variance is calculated from the torque fluctuation amount for a predetermined sampling parameter, and a combustion roughness value is determined based on the calculated standard deviation or the unbiased variance. Alternatively, when calculating the unbiased variance, the sampling parameter is changed according to the operating state of the internal combustion engine. In this case, by calculating the standard deviation or the unbiased variance as the statistical processing, the detection accuracy of the combustion roughness value (combustion state) is further improved. At this time, the convergence and response of the roughness value can be arbitrarily adjusted by appropriately changing the sampling parameter.

Further, by performing the air-fuel ratio control using the detected value of the combustion state according to the present invention, the air-fuel ratio control can be performed with high accuracy. That is, in the invention described in claim 9,
A target value of the combustion roughness value is set based on the engine operating state (target roughness value setting means), and a lean target air-fuel ratio is learned according to a deviation between the combustion roughness value and the target value during the air-fuel ratio feedback control (target air-fuel ratio). Fuel ratio learning means).
According to the tenth aspect, the learned value of the lean target air-fuel ratio is backed up in the memory.

In short, in the lean region of the air-fuel ratio, there is a correlation between the air-fuel ratio and the combustion state (combustion roughness value).
As described above, if the detection accuracy of the combustion state is improved, the learning accuracy is also improved by learning the lean target air-fuel ratio using the detected value. Further, if the learned value of the lean target air-fuel ratio is backed up in the memory, the air-fuel ratio control corresponding to individual differences and aging can be realized.

When the process of learning the air-fuel ratio is performed as described above, the vehicle-mounted battery is newly connected to the memory for backing up the learning value of the lean target air-fuel ratio. It is preferable to reduce the sampling parameter at the time or when the lean target air-fuel ratio is not learned. Further, as described in claim 12, it is preferable to increase the sampling parameter as the number of times of learning the lean target air-fuel ratio increases. Even in such a case, the combustion roughness value (combustion state) can be properly detected by giving priority to either the response or the convergence.

Further, when the combustion roughness value is calculated by smoothing, the degree of smoothing may be changed according to the operating state of the internal combustion engine. As an overview, for example, during the transient operation of the internal combustion engine, the responsiveness of the combustion roughness value is emphasized by reducing the smoothing degree, and conversely, during the steady operation of the internal combustion engine, the combustion is increased by increasing the smoothing degree. Emphasis is placed on the convergence of roughness values. As a result, even when the operating state of the internal combustion engine fluctuates variously, the smoothed value of the combustion roughness value can be properly detected.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS (First Embodiment) A first embodiment of the present invention will be described below with reference to the drawings. In the present embodiment, the present invention is embodied as an air-fuel ratio control system that performs feedback (F / B) control of the air-fuel ratio of an air-fuel mixture supplied to an internal combustion engine for a vehicle to a target value. It is an object of the present invention to detect a combustion roughness value and appropriately perform air-fuel ratio control in a lean region according to the roughness value. Hereinafter, the configuration and operation of the present system will be described in detail.

FIG. 1 is an overall configuration diagram showing an outline of an air-fuel ratio control system according to the present embodiment. In FIG. 1, an engine 10 is a six-cylinder spark ignition type internal combustion engine, and intake air sucked from an air cleaner 11 passes through an intake pipe 12 and a throttle valve 13 and is provided at an intake port for each cylinder at an intake port. After being mixed with the fuel injected by the fuel injection 14, the fuel is supplied to each cylinder of the engine 10. Further, exhaust gas discharged from each cylinder after combustion is discharged to the atmosphere via an exhaust manifold 15, an exhaust pipe 16, and the like.

An intake pipe 12 is provided with an intake air temperature sensor 17 for detecting the temperature of the intake air and an air flow meter 18 for detecting the amount of the intake air. A throttle sensor 19 for detecting is provided. The exhaust pipe 16 is provided with an A / F sensor 20 for detecting an air-fuel ratio (A / F) from the oxygen concentration in the exhaust gas. In addition, as a sensor employed in the present system, a water temperature sensor 21 is provided in a cylinder block and detects the temperature of engine cooling water. Crank angle sensor 22
Is provided, for example, on a crankshaft, and outputs a Ne pulse signal every predetermined crank angle (every 10 ° CA in the present embodiment).

The ECU 30 includes a CPU 31, a ROM 32,
A well-known microcomputer having a RAM 33, a backup RAM 34, etc., comprises the aforementioned intake air temperature sensor 17, air flow meter 18, throttle sensor 19, A
/ F sensor 20, water temperature sensor 21, crank angle sensor 2
2, etc., the detection signals of various sensors are taken in, the fuel injection amount by the injector 14 is adjusted, and the air-fuel ratio is controlled as desired. Here, the CPU 31 sequentially executes each control routine described later according to a control program stored in the ROM 32 in advance. The backup RAM 34 is a memory that retains stored contents by power supply from a vehicle battery (not shown). The backup RAM 34 has basic map data for setting a target air-fuel ratio in accordance with, for example, the engine speed Ne and the intake air amount Qa. Is stored.

Next, the operation of the air-fuel ratio control system configured as described above will be described. FIG. 2 is a flowchart showing an air-fuel ratio control routine executed by the CPU 31.

In step 101 of FIG. 2, the engine speed Ne calculated from the detection result of the crank angle sensor 22 and the intake air amount Qa calculated from the detection result of the air flow meter 18 are fetched. ,
Using a map (not shown), the basic injection amount Tp is calculated based on the acquired Ne and Qa.

Thereafter, in step 103, a change in the cooling water temperature Thw calculated from the detection result of the water temperature sensor 21, a change in the intake air temperature Ta calculated from the detection result of the intake air temperature sensor 17, and a change in the intake air amount in the preceding and following combustion cycles in the same cylinder. Quantity ΔQa
(A change amount of Qa between 720 ° CA) is fetched, and in step 104, a correction amount K1 is calculated based on the fetched Thw, Ta, and ΔQa. The correction amount K1 is a publicly known fuel amount correction amount for performing a fuel increase / decrease amount control during a cold state or a transient operation.

In step 105, the backup RAM
The target air-fuel ratio λtg is calculated based on the engine operating state (Ne, Qa) at each time by using the search map stored in the storage 34. In the next step 106, the air-fuel ratio correction amount K2 is calculated based on the deviation of the air-fuel ratio using the actual air-fuel ratio λre detected by the A / F sensor 20 and the target air-fuel ratio λtg. The air-fuel ratio correction amount K2 is a well-known feedback correction value calculated according to the air-fuel ratio deviation amount. Finally, in step 107, the final injection amount TAU is calculated using the calculated basic injection amount Tp, the correction amounts K1, K2, and the like, and this processing ends (TAU = Tp).
・ K1 ・ K2).

On the other hand, the CPU 31
2 is interrupted and activated at every 20 ° CA based on the Ne pulse signal from the control unit 2. In step 201, a crank angle counter (not shown) is used to determine Ne that is 20 ° CA away.
Counting is performed between pulse signals. Subsequently, at step 202, the required time Tne of the predetermined crank angle (20 ° CA)
In step 203, a rotation fluctuation amount ΔNe for each combustion is calculated based on the required time Tne. Here, in the case of a six-cylinder engine, the required time Tne changes at 120 ° CA as one combustion cycle as shown in FIG. 4, and the rotation fluctuation ΔNe is calculated from the difference between the maximum value Tmax and the minimum value Tmax. The data of the ΔNe value is stored in the RAM 33 for each cylinder.

Next, the calculation process of the roughness value will be described with reference to FIG.
And the flowchart of FIG. This process is executed by the CPU 31 every time fuel is injected into each cylinder. In particular, in this processing, the roughness value (actual roughness value Rre) representing the actual combustion state is calculated based on the standard deviation, and the actual roughness value Rre is calculated based on the standard deviation of the rotation fluctuation of the sampling parameter n.

First, at step 301 in FIG. 5, it is determined whether or not an execution condition for calculating the actual roughness value Rre is satisfied. The execution conditions include, for example, that the air-fuel ratio is controlled in a lean region, that the cooling water temperature Thw is 80 ° C. or higher (warm-up condition), and that the fluctuation of the engine speed Ne within 180 ° CA It is not more than a predetermined value (steady operation condition).

The process proceeds to step 302 on condition that step 301 is YES, and it is determined whether or not the sampling parameter n has already been set. Then, if the sampling parameter n has not been set yet, steps 303 to
At 311, the sampling parameter n is set.

That is, in step 303, it is determined whether or not the absolute value of the previous ΔR is larger than a predetermined value Kr by referring to the previous value of the deviation ΔR between the target roughness value Rtg and the actual roughness value Rre. And | previous ΔR |> Kr
, The process proceeds to step 304, where the deviation ΔR of the roughness value
Is set based on the sampling parameter n. in this case,
Basically, the larger | ΔR |
Is set to a small value. Specifically, as shown in FIG. 8, if | ΔR | <R1, n = N1 and | ΔR |
= R1 to R2, then n = N2, and | ΔR |> R2
Then, n = N3.

On the other hand, if | prev. ΔR | ≦ Kr, the routine proceeds to step 305, where engine operation information such as the engine speed Ne and the intake air amount Qa is input. Subsequent step 306
Then, the basic value n1 of the sampling parameter is calculated based on the engine speed Ne and the intake air amount Qa at each time using a search map of FIG. 9 prepared in advance. According to FIG.
The basic value n1 is set to a smaller value as the engine speed is higher and the load is higher, and the basic value n1 is set to a larger value as the engine speed is lower and the load is lower.

In step 307, it is determined whether the battery has just been replaced or the lean air-fuel ratio has not been learned. If step 307 is YES, step 3
At 08, the basic value n1 is set as the sampling parameter n,
In step 309, the number of times of learning g is set to “0”.

If NO in step 307, the learning number g is fetched in step 310, and the sampling parameter n is calculated in step 311 according to the learning number g. At this time, a coefficient f (g) using the number of times of learning g as a parameter is calculated according to the relationship of FIG.
And the basic value n1 of the sampling parameter to calculate the sampling parameter n. According to FIG. 10, the coefficient f (g) is calculated such that the sampling parameter n increases as the number of times of learning g increases.

After setting the sampling parameter n, it is determined in step 312 in FIG. 6 whether fuel injection corresponding to the sampling parameter n has been performed for each cylinder since the previous calculation of the roughness value. Initially, step 312 is NO, but after n fuel injections, step 312 is YES and the process proceeds to step 313.

In step 313, the actual roughness value Rre is calculated based on the standard deviation of the calculated rotation fluctuation amount ΔNe for each cylinder. Specifically, an actual roughness value Rre is calculated using the following equation (1), with the current value of the rotation fluctuation amount for each cylinder being ΔNe (i) and the average value being ΔNeav.

[0036]

(Equation 1) Thereafter, in step 314, a target roughness value Rtg is calculated based on the engine speed Ne and the intake air amount Qa using a preset search map of FIG. According to FIG. 11, the target roughness value Rtg is set to a smaller value as the rotation speed is higher and the load is lower, and the target roughness value Rtg is set to a larger value as the rotation speed is lower and the load is higher. Step 31
At 5, the deviation ΔR between the target roughness value Rtg and the actual roughness value Rre is calculated (ΔR = Rtg−Rre).

Thereafter, in step 316, the actual air-fuel ratio λ
It is determined whether or not re has converged to the target air-fuel ratio λtg. If the difference between the actual air-fuel ratio λre and the target air-fuel ratio λtg is equal to or less than a predetermined value and it is determined that the air-fuel ratio is in a convergence state (YES in step 316), steps 317 to 317 are performed.
At 320, a learning process of the target air-fuel ratio λtg is performed. That is, in step 317, it is determined whether the deviation ΔR of the roughness value is larger than a positive predetermined value (+ KLA), and in step 318, whether the deviation ΔR of the roughness value is smaller than a negative predetermined value (−KLA). It is determined whether or not.

If ΔR> + KLA, the target air-fuel ratio λtg is shifted to a predetermined amount lean side in step 319, and if ΔR <−KLA, the target air-fuel ratio λtg is shifted to a predetermined amount rich side in step 320. Let it. The basic map data in the backup RAM 34 is updated by the lean or rich target air-fuel ratio λtg. If −KLA ≦ ΔR ≦ + KLA, the process is terminated without changing the target air-fuel ratio λtg.

In short, if the actual roughness value Rre is smaller than the target roughness value Rtg, the target air-fuel ratio λtg is learned to a value on the lean side. Conversely, if the actual roughness value Rre is larger than the target roughness value Rtg, the target air-fuel ratio λtg is increased. Air-fuel ratio λt
Learn g to a value on the rich side. Incidentally, at this time, the deviation of the roughness value is determined for each cylinder, but the target air-fuel ratio λtg
If the map is common to all cylinders, the map data is learned from the average value of all cylinders. When a map is provided for each bank, for example, as in the case of a V-type engine, map data is learned based on an average value for each bank.

Finally, in step 321, the number of times of learning g
Is incremented by “1”, and then the present process is temporarily terminated. The data of the learning number g is stored in the backup RAM 34.
Is stored.

FIG. 7 is a time chart for explaining the operation in the present embodiment. As an example, the air-fuel ratio F
4 shows a process in which the / B control shifts from stoichiometric control to lean control.

Before time t1 in FIG. 7, the air-fuel ratio is F / B controlled at the stoichiometric air-fuel ratio (stoichiometric), and the actual air-fuel ratio λre shown by the solid line is changed to the target air-fuel ratio λtg shown by the dotted line. Has converged. The actual roughness value Rre illustrated by the solid line converges to the target roughness value Rtg illustrated by the dotted line.

At time t1, the target air-fuel ratio λtg is switched from the stoichiometric air-fuel ratio to a predetermined value in the lean region. Also,
At this time t1, the target roughness value Rtg is changed to a relatively large value with the switch to the lean air-fuel ratio control.

After time t1, at time t2, the actual air-fuel ratio λ
re substantially converges to the target air-fuel ratio λtg. At that time, the actual roughness value Rre is smaller than the target roughness value Rtg,
Further, since the deviation ΔR (= Rtg−Rre) of the roughness value is equal to or more than a predetermined value, the target air-fuel ratio λtg is updated to the lean side.

After updating the target air-fuel ratio λtg, at time t3, the actual air-fuel ratio λre substantially converges to the target air-fuel ratio λtg again. At this time, the actual roughness value Rre is still smaller than the target roughness value Rtg, and the deviation Δ
Since R (= Rtg-Rre) is equal to or greater than the predetermined value, the target air-fuel ratio λtg is updated again to the lean side. from then,
The actual roughness value Rre converges to the target roughness value Rtg, and the actual air-fuel ratio λre converges to the target air-fuel ratio λtg.

At times t2 and t3, the target air-fuel ratio λtg is learned to a lean value based on the deviation ΔR of the roughness value, and the basic map data in the backup RAM 34 is updated.

On the other hand, when the actual roughness value Rre substantially converges to the target roughness value Rtg, the sampling parameter n
Is set to a relatively large value. On the other hand, immediately after the target air-fuel ratio λtg is switched, the deviation Δ
Since R (= Rtg-Rre) increases, TA, T
In the period B, the sampling parameter n according to the relationship shown in FIG.
Are changed to “N3” and “N2”, which are relatively small.

Therefore, at the beginning when the air-fuel ratio is switched from stoichiometric to lean, the response of the actual roughness value Rre is emphasized by reducing the sampling parameter n, and thereafter, the actual roughness value Rre is increased by increasing the sampling parameter n. The convergence of is important.

In this embodiment, the processing in FIG. 3 corresponds to the torque fluctuation calculating means, and the rotation fluctuation ΔNe calculated in the processing corresponds to the torque fluctuation. Step 313 in FIG. 6 corresponds to roughness value calculating means, and steps 303 to 311 in FIG. 5 correspond to parameter changing means. Further, step 314 in FIG. 6 corresponds to target roughness value setting means, and steps 317 to 320 in FIG. 6 correspond to target air-fuel ratio learning means.

According to the present embodiment described in detail above, the following effects can be obtained. (A) In the present apparatus for calculating the actual roughness value Rre based on the standard deviation, the sampling parameter n is appropriately changed according to the engine operating state, so that the responsiveness and the convergence of the actual roughness value Rre can be arbitrarily adjusted. As a result, the combustion state can be accurately detected even when the engine operation state fluctuates in various ways.

(B) Actually, deviation ΔR of roughness value,
Since the engine operation area, whether the battery initial state or the lean target air-fuel ratio is not learned, and the sampling parameter n is variably set according to each factor such as the number of times of learning g,
The combustion state can be properly detected by selectively giving priority to either responsiveness or convergence as needed at each time.

(C) Since the combustion state (actual roughness value Rre) can be properly detected as described above, the controllability of the lean air-fuel ratio control performed according to the combustion state is also improved. That is, since the air-fuel ratio and the combustion state (actual roughness value Rre) have a correlation in the lean region of the air-fuel ratio, if the detection accuracy of the combustion state is improved as described above, the lean target air-fuel ratio is obtained using the detected value. By learning, the learning accuracy is also improved. Further, by learning the lean target air-fuel ratio, it is possible to realize air-fuel ratio control corresponding to individual differences and aging.

(Second Embodiment) Next, a second embodiment of the present invention will be described. However, the following description focuses on the differences from the first embodiment. In the present embodiment, the standard deviation is calculated in units of the reference value using a fixed sampling parameter as a reference value, the calculation results are successively integrated, and the combustion roughness value is calculated based on the average value of the integrated values.

FIG. 12 is a flowchart showing the procedure for calculating the actual roughness value in the present embodiment. This processing is performed in step 313 of FIG. In the processing of FIG. 12, a reference value kc of the sampling parameter for calculating the roughness value in a relatively small parameter unit is set, and the standard deviation of the sampling parameter n at each time is calculated in kc unit.

In step 401, the counter k is counted up, and in subsequent step 402, it is determined whether or not the counter k has reached the reference value kc. If step 402 is YES, the process proceeds to step 403, where a roughness value r1 in small parameter units is calculated using the following equation (2).

[0056]

(Equation 2) Then, in step 404, the counter k is cleared to "0". In the following step 405, the roughness values r1 in small parameter units are integrated, and the integrated value is set as r2 (r2 =
r2 + r1).

In step 406, the sampling parameter n at that time is fetched. In step 407, it is determined whether or not fuel injection corresponding to the sampling parameter n has been performed since the previous calculation of the actual roughness value Rre. The process proceeds to step 408 on condition that the result is YES. In this embodiment, the sampling parameter n is a number larger than the reference value kc.

In step 408, the actual roughness value Rre is calculated based on the integrated value r2 of the calculated roughness value, the sampling parameter n, and the reference value kc using the following equation (3).

[0059]

(Equation 3) Thereafter, in step 409, the integrated value r2 of the roughness value
Is cleared to "0", and this processing ends. FIG.
May be performed only when the sampling parameter n is large.

FIG. 13 is a time chart showing the actual operation of the processing of FIG. In FIG. 13, the counter k
Is counted by kc, the roughness value r1 in small parameter units is integrated, and the integrated value r2 of the roughness value is calculated. When the number of injections for each cylinder reaches the sampling parameter n at time t10, at time t11 thereafter,
The actual roughness value Rre is calculated by the above equation (3).

According to the second embodiment, the roughness value can be calculated in small parameter units. Therefore, even when the sampling parameter n is large and it takes a long time to calculate the roughness value, it is possible to grasp the trend of the roughness value during the calculation.

The present invention can be embodied in the following modes other than the above. Immediately after calculating the actual roughness value Rre,
It is configured to perform an averaging operation on re. More specifically, the averaging rate sm is set in advance, and averaging value = previous averaging value + (Rre-previous averaging value)
An average value is calculated as sm. As an example, the smoothing rate sm
Is set to 1/64. Then, in the processing of steps 315 to 320 in FIG. 6, the actual roughness value Rre
The deviation between the smoothed value and the target roughness value Rtg is calculated, and learning of the target air-fuel ratio λtg is performed according to the deviation.

In this case, the smoothing rate sm (smoothing degree) may be changed according to the engine operating state. For example, when the responsiveness of the actual roughness value Rre is to be emphasized, such as during transient operation, the smoothing rate sm may be changed. And the actual roughness value Rr
Reduce the degree of smoothing of e. Conversely, when emphasis is placed on the convergence of the actual roughness value Rre during steady operation, the smoothing rate sm is reduced, and the smoothing degree of the actual roughness value Rre is increased. As a result, the combustion state can be accurately detected even when the engine operation state fluctuates variously, and the controllability of the air-fuel ratio is also improved. Note that the CPU 31 corresponds to the averaging means.

In the above embodiment, in the roughness value calculation processing of FIGS. 5 and 6, (1) the deviation Δ
R, (2) engine operating range, (3) whether the battery initial state or the lean target air-fuel ratio has not been learned, (4)
Although the sampling parameter n is variably set in accordance with each factor such as the number of times of learning g, any one of them may be omitted to be embodied. For example, the present invention may be embodied by only one or both of the above (1) and (2).

Further, it is determined whether the engine is in a transient operation based on the amount of change in the target air-fuel ratio λtg, the amount of load change such as the intake air amount, the throttle opening, and the like. You may comprise. in this case,
The response of the combustion roughness value during the transient operation is improved.

In the above embodiment, the standard deviation was calculated in calculating the combustion roughness value by the statistical processing. However, instead of this, the standard deviation was calculated as the statistical processing, and the average deviation was calculated. The combustion roughness value may be calculated. In addition, as means for calculating the torque fluctuation amount for each combustion of the engine, in addition to the above-described rotation fluctuation amount, means for calculating based on the combustion pressure for each combustion, means for calculating based on combustion light, or the like may be applied. . In any case, by appropriately changing the sampling parameter n according to the engine operating state, the combustion roughness value (combustion state) can be properly detected.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an outline of an air-fuel ratio control system for an internal combustion engine according to an embodiment of the invention.

FIG. 2 is a flowchart showing an air-fuel ratio control routine.

FIG. 3 is a flowchart showing Ne interruption processing.

FIG. 4 is a time chart showing a state of rotation fluctuation for each combustion.

FIG. 5 is a flowchart showing a process of calculating a roughness value.

FIG. 6 is a flowchart showing a roughness value calculating process following FIG. 5;

FIG. 7 is a time chart for explaining the operation.

FIG. 8 is a diagram for setting a sampling parameter n.

FIG. 9 is a map for setting a basic value n1 of a sampling parameter.

FIG. 10 is a diagram for setting a coefficient f (g) using the number of times of learning as a parameter.

FIG. 11 is a map for setting a target roughness value Rtg.

FIG. 12 is a flowchart illustrating a procedure for calculating a roughness value according to the second embodiment;

FIG. 13 is a time chart showing a process of calculating a roughness value in the second embodiment.

[Explanation of symbols]

10: engine (internal combustion engine), 14: injector, 3
0 ... ECU, 31 ... Torque variation calculation means, roughness value calculation means, parameter changing means, target roughness value setting means, target air-fuel ratio learning means, CPU as smoothing means, 34 ...
Backup RAM.

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (reference) F02D 45/00376 F02D 45/00 376D 41/14 310 41/14 310H F-term (reference) 3G084 AA04 BA09 DA04 EA04 EB08 EB12 EB20 EB26 EC04 FA02 FA07 FA10 FA19 FA20 FA21 FA23 FA29 FA32 FA33 FA34 FA38 3G301 HA01 HA06 HA15 JA20 LB02 MA01 MA12 NA01 NA08 NA09 NB03 NC02 ND02 ND22 NE14 NE15 NE23 PA01Z PA02Z PA10Z PA11Z PC01Z PC03Z PC07ZZZ PCZZ

Claims (13)

[Claims] 1. A torque variation calculating means for calculating a torque variation for each combustion of an internal combustion engine, and performing statistical processing on the calculated torque variation for a predetermined sampling parameter. A roughness value calculation unit that calculates a combustion roughness value as a parameter indicating a combustion state according to a result, and a parameter changing unit that changes a sampling parameter according to an operation state of the internal combustion engine during statistical processing by the roughness value calculation unit. And a combustion state detection device for an internal combustion engine. 2. The torque fluctuation amount calculating means calculates a maximum value and a minimum value of the rotation speed for each combustion of the internal combustion engine, and calculates a rotation fluctuation amount as a torque fluctuation amount based on a difference between these values. Item 2. The combustion state detection device for an internal combustion engine according to Item 1. 3. The internal combustion engine according to claim 1, wherein said parameter changing means decreases the sampling parameter when the deviation between the combustion roughness value calculated by said roughness value calculating means and its target value is large. Engine combustion state detector. 4. The combustion state detecting device for an internal combustion engine according to claim 1, wherein said parameter changing means reduces a sampling parameter during a transient operation of the internal combustion engine. 5. A combustion state detecting device for an internal combustion engine according to claim 1, wherein said parameter changing means changes the sampling parameter in accordance with an operation range of the internal combustion engine. 6. The roughness value calculating means performs statistical processing in units of the reference value using a fixed sampling parameter as a reference value, sequentially accumulates the results of the statistical processing, and calculates an average value of the integrated values. The combustion state detection device for an internal combustion engine according to any one of claims 1 to 5, which calculates a combustion roughness value. 7. The combustion state detecting device according to claim 6, wherein when the sampling parameter is equal to or more than a predetermined value, statistical processing is performed in units of the reference value, and the results of the statistical processing are sequentially integrated. A combustion state detection device for an internal combustion engine that calculates a combustion roughness value from an average value of integrated values. 8. The roughness value calculating means calculates a standard deviation or an unbiased variance from a torque variation amount for a predetermined sampling parameter, and determines a combustion roughness value based on the calculated standard deviation or the unbiased variance. The combustion of the internal combustion engine according to any one of claims 1 to 7, wherein the parameter changing means changes a sampling parameter in accordance with an operation state of the internal combustion engine when calculating the standard deviation or the unbiased variance by the roughness value calculating means. State detection device. 9. A target roughness, which is applied to an air-fuel ratio control system for feedback-controlling an air-fuel ratio with respect to a target air-fuel ratio set in a lean region of the air-fuel ratio, and sets a target value of a combustion roughness value based on an engine operating state. The fuel supply system according to any one of claims 1 to 8, further comprising: a value setting unit; and a target air-fuel ratio learning unit configured to learn a lean target air-fuel ratio in accordance with a deviation between a combustion roughness value and a target value during the air-fuel ratio feedback control. A combustion state detection device for an internal combustion engine. 10. The combustion state detecting device according to claim 9, wherein the learning value of the lean target air-fuel ratio by said target air-fuel ratio learning means is backed up in a memory. 11. The combustion state detecting device according to claim 10, wherein said parameter changing means is configured to: when a vehicle-mounted battery is newly connected to a memory for backing up a learning value of a lean target air-fuel ratio. Alternatively, a combustion state detection device for an internal combustion engine that reduces the sampling parameter when the lean target air-fuel ratio has not been learned. 12. The combustion state detecting device according to claim 9, wherein said parameter changing means increases the sampling parameter as the number of times of learning of the lean target air-fuel ratio increases. State detection device. 13. A smoothing means for smoothing the calculated combustion roughness value, wherein the smoothing means changes the degree of smoothing according to the operating state of the internal combustion engine. A combustion state detection device for an internal combustion engine of the present invention.