JPH06101517A - Evaporated fuel treatment device for engine - Google Patents

Abstract

PURPOSE:To prevent an air-fuel ratio from leaning immediately after purging by considering transferring delay of purge fuel flow. CONSTITUTION:A purge fuel density is estimated by means of an estimation means 32, and a flow rate through a purge valve is calculated by means of a calculation means 33 based on an operation signal. A purge fuel flow rate is calculated by means of a calculation means 34 from the purge valve flow rate and purge fuel density. Weighted mean application is carried out in respect to the purge fuel flow rate two times sequentially by means of an estimation means 35. The resultant value is delayed by a specified time in its output timing, which is estimated as a cylinder intake amount of the purge fuel flow rate. A calculation means 36 corrects a reference injection rate according to an operation condition for calculating the fuel injection rate during purging. The resultant rate of injected fuel is supplied to an intake pipe.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an improved engine fuel vapor treatment system.

[0002]

2. Description of the Related Art In order to prevent the fuel evaporated from a fuel tank from diffusing into the atmosphere, the fuel vapor is adsorbed to an activated carbon canister, and the fuel accumulated in the activated carbon canister is purged into an intake pipe under predetermined operating conditions (by the atmosphere, It is burnt in the cylinder together with the fuel injected from the injector by ejecting it from the canister.

However, when a purge gas which is not measured by an air flow meter is added, it affects the control air-fuel ratio, so that there is a valve that opens the purge valve (purge control valve) during air-fuel ratio feedback control. Although the air-fuel ratio shifts to the rich side at the beginning of the purge valve opening, the air-fuel ratio feedback correction coefficient α shifts to the lean side from the control center (1.0) and eventually settles to a certain value (for example, 0.8). As a result, the air-fuel ratio can be kept within the catalyst window (a predetermined width around the theoretical air-fuel ratio) even during purging.

However, even if the accelerator pedal is depressed during the purging, the supplied fuel amount cannot be increased to the required value corresponding to the depression amount of the accelerator pedal at once, so-called breathing occurs and the drivability deteriorates. Since α has to start from the above value (0.8) deviating to the lean side and becomes large, and α increases only at a constant rate, the fuel amount cannot be rapidly increased.

Therefore, in Japanese Patent Application Laid-Open No. 2-19631, the difference between α when the value becomes equal to or less than a predetermined value after the start of purging and α immediately before the start of purging is calculated, and the operating condition is adjusted by the reduction correction amount according to this difference. While the basic injection amount corresponding to the above is subtracted, α is forcibly returned to the value immediately before the start of the purge from the time when α becomes equal to or less than the predetermined value.

When α is settled to a lean value by the purge and becomes equal to or less than a predetermined value (0.8), the fuel is supplied by correcting the fuel injection amount from the basic injection amount by the fuel increase amount by the purge. Since the amount of fuel is kept the same before and after purging, and when the accelerator pedal is depressed during purging, α is increased from the value before purging (usually 1.0) to increase α as soon as possible. is there.

[0007]

By the way, in the above apparatus, since the difference between α before and after the change is observed after the change of α is completed by the purge, the air-fuel ratio error (theoretical air gap) Deviation from the fuel ratio) is large,
There is room for improvement in exhaust performance and driving performance.

On the other hand, the poor response of α can be compensated by the so-called basic air-fuel ratio learning value αm.
However, the basic air-fuel ratio learning value αm does not originally take into consideration the air-fuel ratio error due to purging, but takes into consideration the steady-state error of the air-fuel ratio caused by the air flow meter or the injector due to aging. Therefore,
If the air-fuel ratio error during purging is also incorporated into the basic air-fuel ratio learning value αm, a learning error will occur. Therefore, learning is prohibited during purging even with the conventional device, and learning is restarted after α has settled down. . That is, the basic air-fuel ratio learning value cannot deal with the air-fuel ratio error during purging.

Therefore, there has been proposed an apparatus which starts updating the purge learning value WC introduced separately from the basic air-fuel ratio learning value αm at the time of switching the purge ON and OFF. As will be described later, rather than waiting for the change in α to be completed by turning ON and OFF the purge, the learning frequency is increased by starting the purge learning immediately after the switching, and the learning is rapidly advanced.

For example, if purge learning has never been performed, α (= ALPHA) when switching to purge ON.
Is shifted to the smaller side as shown in Fig. 36.
V (average value of α) is ALPST (average value of α before switching)
The purge learning value (corresponding to the fuel concentration of purge gas) WC is rewritten from the time point when it becomes smaller (point A). In this way, if the purge learning is started immediately after the switching, the air-fuel ratio is set to the catalyst window from the beginning regardless of the fuel concentration of the purge gas (hereinafter referred to as the purge fuel concentration) during the next purge, unless the fuel concentration of the purge gas (hereinafter referred to as the purge fuel concentration) is significantly changed. Even if the purge fuel concentration has changed from the previous time, the purge learning value WC is updated at that time, and the air-fuel ratio is immediately stored in the catalyst window according to the updated value.

When the purge learning value WC corresponding to the purge fuel concentration is obtained in this way, the value obtained by multiplying the purge learning value WC by the purge valve flow rate (the sum of air and fuel) is the purge fuel flow rate (purge valve flow rate). Therefore, the fuel injection amount during purging can be calculated by subtracting the purge fuel flow rate from the basic injection amount according to the operating condition. Since an extra amount of purge fuel is supplied during purging, the air-fuel ratio during purging can be made the same as that before purging by subtracting this fuel flow amount from the basic injection amount.

However, there is a transportation delay until the purge fuel flow rate reaches the cylinder from the purge valve position,
When this fuel transportation delay is not taken into consideration, the accuracy of correction of the basic injection amount decreases, and the air-fuel ratio fluctuates before and after purging. During the delay of fuel transportation, the amount of supplied fuel is rather short by the amount of the purge fuel flow rate, and the air-fuel ratio shifts to the lean side.

Therefore, an object of the present invention is to prevent the air-fuel ratio from becoming lean immediately after purging by taking into account the transportation delay of the purge fuel flow rate.

[0014]

As shown in FIG. 1, the present invention relates to an evaporated fuel processing apparatus for an engine equipped with a purge valve for adjusting the amount of purge gas introduced from a canister into an intake pipe. Prediction means 32
A means 33 for calculating the flow rate of the purge valve by receiving an operating condition signal, a means 34 for calculating the purge fuel flow rate based on the purge valve flow rate and the purge fuel concentration, and a weighted average of the purge fuel flow rate. Means 35 for predicting the value obtained by delaying the output timing of the value obtained by performing the weighted average twice and delaying the output by a predetermined time as the cylinder intake amount of the purge fuel flow rate, and the operating condition based on the cylinder intake amount predicted value. The means 36 for correcting the basic injection amount according to the above and calculating the fuel injection amount during purging, and the device 37 for supplying this fuel injection amount to the intake pipe are provided.

[0015]

[Operation] When all the gaseous fuel (light components) in the purge gas has the same size, the fuel transport delay time synchronized with the rotation of the engine occurs even though the fuel particles are transmitted from the purge valve position to the cylinder. is there.

Further, considering the difference in the fuel particle diameter, the diffusion speed of the fuel particles flowing along with the air flow also changes due to the difference in the fuel particle diameter, so that the purge fuel flow rate is transmitted to the cylinder with a temporal spread.

In this case, when the weighted average is performed twice on the purge fuel flow rate in the present invention, the diffusion distribution of the fuel particles is approximated by the value obtained by performing the weighted average twice, and the weighted average is calculated twice. When the output timing of the performed value is delayed for a predetermined time, this delayed value has a fuel transport delay until it reaches the cylinder, and the purge fuel flow rate is temporally distributed due to the difference in the fuel particle size. It is in good agreement with the actual fuel behavior of reaching the cylinder.

When the cylinder intake amount of the purged fuel flow rate is accurately predicted in this way, the correction accuracy of the basic injection amount based on this predicted value is improved, and the air-fuel ratio immediately before the purge determined by the basic injection amount is the purge amount. Immediately after that, he does not lean to the lean side.

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In FIG. 2, a control unit 2 comprising a microcomputer (for example, a 16-bit microcomputer) is provided for engine control.

The exhaust pipe 3 is provided with a three-way catalyst 4 for treating three harmful components such as CO, HC and NOx discharged from the engine. The three-way catalyst 4 can process the harmful three components at the same time only when the air-fuel ratio of the air-fuel mixture supplied to the engine is within a narrow range (catalyst window) centered on the theoretical air-fuel ratio. If the air-fuel ratio deviates from this catalyst window to the rich side, CO, H
When the amount of C emission increases and, on the contrary, it shifts to the lean side, a large amount of NOx is emitted.

For this reason, the control unit 2 feedback-controls the fuel injection amount based on the actual air-fuel ratio signal from the O ₂ sensor 5 so that the three-way catalyst 4 can fully exert its capacity.

The output of the O ₂ sensor 5 provided upstream of the three-way catalyst 4 changes abruptly at the stoichiometric air-fuel ratio (the output is approximately 1 V on the rich side and about 0 V on the lean side of the stoichiometric air-fuel ratio). Therefore, it is determined that the air-fuel ratio is on the rich side when the output of the O ₂ sensor is higher than the slice level (approximately 0.5 V), and is on the lean side when it is lower than the slice level. If such a determination is made in synchronization with the engine rotation, it can be determined whether the air-fuel ratio has just changed to the rich side (or the lean side) or whether the air-fuel ratio is on the rich or lean side.

From these judgment results, the step amount P is subtracted from the air-fuel ratio feedback correction coefficient α immediately after the air-fuel ratio is reversed to the rich side, and the integral amount I is subtracted from α until just before the air-fuel ratio is next reversed to the lean side. (Conversely, P is added to α immediately after the actual air-fuel ratio is reversed to the lean side, and I is added until just before the actual air-fuel ratio is next reversed to the rich side). Immediately after the air-fuel ratio is reversed, a large value of P is given in a stepwise manner to change the response to the opposite side, and after the step change, a small value of I is used to slowly change the air-fuel ratio to the opposite side. It stabilizes the feedback control.

Even if the engine operating conditions are different,
Ratio of intake air amount measured by the air flow meter 7 located upstream of the throttle valve 6 and fuel amount supplied from the injector 8 toward the cylinder (that is, air-fuel ratio)
The control unit 2 determines the valve opening pulse width (injection pulse width) of the injector 8 that is intermittently opened in synchronization with the engine rotation so that the ratio becomes approximately the stoichiometric air-fuel ratio.

Reference numeral 9 is a signal corresponding to the engine speed and Re
The crank angle sensor 10 for outputting the f signal (crank angle reference position signal) is a throttle valve opening (T
A sensor for detecting VO), a water temperature sensor 11 and a vehicle speed sensor 12 are also input to the control unit 2.

By the way, if the injector 8 is clogged due to secular change, the supplied fuel amount decreases even if the injector 8 is driven with the same pulse width. Therefore, each time the engine is started, the air-fuel ratio feedback control is started for a while. Has a leaner air-fuel ratio. To avoid this,
The control unit 2 performs basic air-fuel ratio learning.

After a while after entering the air-fuel ratio feedback control, the average value of α settles at a value larger than the control center (1.0) (α itself swings around this value), so this value is basically used. If the battery is backed up as the air-fuel ratio learning value αm, the air-fuel ratio can be kept in the catalyst window from the beginning of the feedback control by increasing the injection pulse width by the learning value αm from the next engine operation. Of.

On the other hand, the fuel evaporated from the fuel tank 15 when the engine is stopped and adsorbed by the activated carbon in the canister 16 is released from the activated carbon when the atmosphere is introduced from the outside of the canister 16 while the engine is operating, and contains the released fuel. Clean air (purge gas) is sucked into the intake passage.

A purge valve 21 is provided in a passage 18 for connecting the activated carbon canister 16 and the collector portion 17a of the intake manifold 17 to adjust the inflow amount of the purge gas. The purge valve 21 is a valve driven by a linear solenoid, and is driven by a pulse signal having a constant cycle (for example, a cycle of 6.4 ms) from the control unit 2, and the valve opening increases as the ON duty (ON time ratio) increases. Increase.

If the purge valve 21 sticks when it is fully open, the engine may stall (engine stall) or the idle speed may increase due to the purge. To prevent this, a VC negative pressure valve (diaphragm valve) is provided. 22 is provided in the passage 18 in series with the purge valve 21. The VC negative pressure is a negative pressure that rises with respect to the throttle opening TVO as shown in FIG. 3. If the accelerator pedal is released and the throttle valve 6 is closed, the VC negative pressure becomes close to the atmospheric pressure and the VC is reduced. The negative pressure valve 22 is closed. As a result, the passage 18 is closed regardless of whether the purge valve 21 is opened or closed. By the way, although the purging is performed during the air-fuel ratio feedback control, the basic air-fuel ratio learning value αm is also maintained during the purging.
Therefore, the control unit 2 prohibits the learning value αm from being updated during the purge.

However, if the fluctuation of the air-fuel ratio due to the purging is dealt with only by chasing α, since α changes only at a constant rate, it is possible that the air-fuel ratio shifts to the rich side until the change of α ends. .

For this reason, the flow charts shown in FIGS. 5 to 22 are assembled.

In a nutshell, this control system is for absorbing the air-fuel ratio error due to the purge.
The concept of the control system will be briefly described using, and then it will be divided into terms and outlined. Since the present invention forms a part of this control system, it will be described in detail at the end.

Quantitatively looking at fuel and air using the symbols shown in FIG. 4, the purge valve flow rate (total of fuel flow rate and air flow rate) Qpv is calculated from the purge valve duty (EVAP) and the differential pressure across the purge valve. Qef = Qpv · WC ... [A] However, WC; the purge fuel flow rate Qef can be obtained from the purge fuel concentration.

On the other hand, the purge air flow rate Qea is Qea = Qpv-Qef.KFQ # ... [B] where KFQ # is a constant for converting the fuel flow rate into the air flow rate, so the purge gas inflow portion of the intake pipe is The air flow rate immediately downstream of the air flow rate is the sum of Qea and the air flow rate Qs measured by the air flow meter 7.

When the air amount (Qs + Qea) flowing at the upstream position away from the injector 8 is determined in this way, a known manifold-cylinder filling model can be applied, and Qc = (Qs + Qea) .Fload + Qc. (1-Fload) ... [C] However, the cylinder air amount (air amount flowing into the cylinder) Qc is obtained as a first-order lag of (Qs + Qea) by the Fload; weighting coefficient.

On the other hand, the fuel injection amount Qf from the injector 8 provided at the port is Qf = Qc · K # −Qefc ... [D] where K # is a constant for keeping the air-fuel ratio constant Qefc; The purged fuel amount (Qefc) is subtracted from the cylinder intake amount.

Note that Qefc is determined with respect to Qef in consideration of the fact that the fuel gas is propagated while being diffused and a simple time delay.

That is, if the purge gas fuel concentration is accurately obtained by learning, it becomes clear how much the air amount and the fuel amount should be corrected during purging. However, in the past, the purge fuel concentration was not measured, and only an appropriate value was adopted from the empirical value, etc., and there was room for improvement in exhaust performance and operating performance when switching to purge ON and OFF. Of.

Next, the control system itemization will be described.
In the following, in principle, capital letters are used for symbols that indicate quantities, and sometimes the symbols and statements used for operators are diverted from those used in programming languages.

(1) Purge cut condition (1-1) When any of the following conditions <1> to <5> is satisfied, the purge valve duty (EVAP) is set to 0 to turn off the purge valve. Close in steps (cut immediately). Since the VC negative pressure valve 22 is closed under these conditions, the purge valve 21 is also closed stepwise in accordance therewith. On the contrary, when all of these conditions are cancelled, the valve is opened stepwise as in the case of canceling the conditions <6> to <11> described later. <1> When the ignition switch is off (step 23 in FIG. 7). <2> At the time of engine stall determination (step 24 in FIG. 7). <3> When the starter switch is ON (step 25 in FIG. 7). <4> When the idle switch is ON (step 26 in FIG. 7). <5> When the vehicle speed (VSP) is lower than the predetermined value (VCPC #) (step 27 in FIG. 7).

These conditions are checked, and if either of them is satisfied, the zero cut flag = 1 and the cut flag = 1 are set (steps 23 to 27 in FIG. 7, step 30 in FIG. 8). The cut flag = 1 indicates that the purge cut is to be performed, and the zero cut flag = 1 indicates that the purge cut is to be performed stepwise. Therefore, the purge cut is performed stepwise by the zero cut flag = 1 and the cut flag = 1. .

However, since it is sufficient to set these flags once, in step 29 of FIG. 8, if the flag (# F1STGKZ) passed once in step 29, it is determined that the flag was previously set, and step 30 follows. You are exiting the routine without proceeding. For the first time, the other two flags that passed once (# F1STGKP and # F1STGKY) =
0, continuous purge ON time counter value (PONREF) =
By setting it to 0, it is prepared for the next time (steps 30 and 28 in FIG. 8).

(1-2) When any of the following conditions <6> to <11> is satisfied, the purge valve is closed stepwise. This is because purging under these conditions adversely affects operating performance and exhaust performance. Therefore, when all of these conditions are released, the purge valve is opened stepwise.

<6> When the load is too small (step 33 in FIG. 7). For example, a cylinder air amount equivalent pulse width (described later) TP is compared with a purge lower limit allowable value (TPCPC), and if TP <TPCPC, it is determined that the load is too small (steps 32 and 33 in FIG. 7). TPCPC
With respect to, the table having the contents of FIG. 31 is looked up from the engine speed NE (with interpolation calculation).
Since all the table lookups have an interpolation calculation, they will be simply referred to as table lookups below.

<7> When the load is too large (step 34 in FIG. 7). For example, the air flow rate QH0 is compared with the purge upper limit allowable value (EVPCQH #), and QH0 ≧ EVPCQ
It is judged that the load is too large in H #. Note that QH0 is an air flow rate (volume flow rate) at the throttle valve portion, which will be described later, and is determined from the throttle opening TVO and the engine speed NE.

<8> When the air-fuel ratio feedback control is not in progress (step 35 in FIG. 7). This is because the air-fuel ratio cannot be contained in the catalyst window unless the air-fuel ratio feedback control is being performed. For example, the flag (#FCLS
Since 1) = 0, it is determined that feedback control is not in progress.

<9> During clamping (while air-fuel ratio feedback control is stopped), an artificial selection flag introduced corresponding to the following various clamps (provided as an option) =
When 0 (steps 37 to 44 in FIG. 7). Temin clamp (flag is #FPGTEM). O ₂ sensor initialization clamp (flag is #FPGCL
C). High load KMR clamp (flag is #FPGKM
R). KHOT clamp (flag is #FPGKH).

The clamping condition of is the effective pulse width T
When e (which is a value obtained by subtracting the invalid pulse width Ts from TI described later) is less than or equal to the minimum value, the clamp condition of is during initialization of the O ₂ sensor, the clamp condition of is in the high load range,
Clamping conditions are when the engine is overheated and the water temperature is high.

Here, the artificial selection flag is used because the purging speed demand differs depending on the vehicle type (fuel tank system), and it is desired to adjust the purging region. This allows the developer to artificially select the value of the flag in order to be compatible with the inability to correct the error. Therefore, the value of the flag is determined by the specifications at the time of development.

FIG. 32 shows what happens to the purge area under the purge OFF conditions of <1> to <9> above.
When the condition of <6> is satisfied, the purge cut is performed in the area shown by the TP cut arrow in the figure. Similarly, when the condition of <7> is satisfied, in the region indicated by the arrow of QH0 cut in the figure,
When the condition of <9> is satisfied, purge cutting is performed in each of the regions shown by KMR cutting in the figure. Therefore, the remaining area is the area to be purged. However, it is shown that even in the purge region, purge cut may be performed by KHOT cut (heat resistant cut) or the like.

<10> When the cut flag for purge learning = 1. Purge learning (W in the figure
Cut flag (#FWCCU) for C learning
T) = 1 (step 60 in FIG. 9). The purge learning is learning for absorbing the air-fuel ratio error due to the purge and will be described later. EONREF # ≠ FFFF (step 51 in FIG. 9). This is so that it is possible to select by EONREF # (described later) whether or not the purge cut for the purge learning is artificially made. By artificially inserting FFFF (maximum hexadecimal value) into EONREF #. If so, it is possible not to perform the purge cut for the purge learning. Offset learning reservation flag (#FOFGKGO) = 1
If not (step 52 in FIG. 9). The offset learning is learning for absorbing variations in the purge valve and will be described later. When the purge learning permission flag (#FWCGKOK) is not 1 (step 54 in FIG. 9). If purge learning is enabled, PONREF (continuous purge ON time counter value)
= 0 (steps 54 and 61 in FIG. 9). This is to count the continuous purge ON time from the end of purge learning. When the continuous purge ON time counter value (PONREF) is greater than or equal to the predetermined value (#EONREF) (step 55 in FIG. 9). When the air-fuel ratio feedback control is in progress (# FCLS1 = 1) and the clamp is not in progress (# FCLMP1 = 0) (step 56 in FIG. 9). The basic duty (EVAP0) described below is the lower limit value (W
CGDTY #) or more (step 5 in FIG. 9)
7). When the load (QH0) is less than or equal to the upper limit value (WCGQH #) (step 58 in FIG. 9). When a certain delay time has passed after the conditions (1) to (3) have been satisfied (CONTWCJ ≧ WCGDLY #) (step 59 in FIG. 9).

In particular, the reason why the purge cut is performed when the condition is satisfied is that if the purge is performed for a long time, the amount of fuel leaving the activated carbon canister 16 is reduced, and the purge fuel concentration is lowered. This is because there is a gap between the purge learning value WC) and the purge learning value WC. Therefore, even under the purging conditions, the purge learning is performed while intermittently performing the purge cut.

<11> When the cut flag for offset learning = 1. When all of the following conditions are satisfied, the cut flag (#FOFCUT) for offset learning is set to 1 (step 67 in FIG. 9). When the offset learning reservation flag = 1 (step 52 in FIG. 9). This reservation is reserved when the purge learning value is clamped and the purge learning is completed as described later. When the air-fuel ratio feedback control is being performed (# FCLS1 = 1) and the clamp is not being performed (# FCLMP1 = 0) (step 64 in FIG. 9). The basic duty (EVAP0) described later is the upper limit value (O
FGDTY #) or less (step 6 in FIG. 9)
5). When a certain delay time has passed after the conditions (1) to (3) have been satisfied (CONTOFJ ≧ OFGDLY #) (step 66 in FIG. 9).

When any of the above conditions <6> to <11> is satisfied, the cut flag = 1 and the slow flag = 1 are set (step 47 in FIG. 8). Since the slow flag = 1 indicates that the purge valve is opened and closed in stages, the purge valve is gradually closed by the slow flag = 1 and the cut flag = 1. On the other hand, <6>
~ When the conditions of <11> are all cleared, purge O
At the time of switching to N, the purge flag is opened stepwise, so that the cut flag = 0 and the slow flag = 1 are set (FIG. 8).
Step 49).

When the purge valve is opened and closed stepwise, it suffices to set the flag once and to clear the continuous purge ON time counter (PONREF). This is the same as 30, 28 (steps 46, 47, 45, steps 48, 49 in FIG. 8).

(2) Purge valve opening characteristic (2-1) Connection with purge cut condition When any of the above conditions <6> to <11> is satisfied, until (EVAP = EVPCUT #), (EVAP
The purge valve duty (EVAP) is reduced at a speed of (T-EVPCUT #) * SPECUT # (steps 91 to 95, steps 91 to 94, 96, 97 in FIG. 10).

When all of the above conditions <1> to <11> are released, once EVAP = EVPCUT # is set, and (EVAPT-
EVPCUT #) * SPEOON # at a speed of increasing the purge valve duty EVAP (step 91 in FIG. 10,
92, 98-100, steps 91, 92, 98, 9
9, 101, 102).

Here, EVPCUT # is the purge valve duty under the purge OFF condition, EVAPT is the purge valve target duty, SPECTUT # is the closing speed of the purge valve, SP
EON # is the opening speed of the purge valve.

FIG. 33 shows a control waveform of EVAP (purge valve duty) with a solid line. From purge OFF to purge O.
When switching to N, EVAP is once set to EVPCUT # and then gradually increased toward EVAPT,
By switching from purge ON to purge OFF, this time E
It is gradually reduced from VAPT to EVPCUT #. In FIG. 33, the switching to the immediate cut is also shown by overlapping with a broken line, and only in this case, the EVAP is set to 0 stepwise.

On the other hand, the value of EVAP is also given to the background job shown in FIG. 5 (steps 12 to 1).
7). This is 100m as shown in Fig. 10 in all cases.
If the value of EVAP is given by the job every sec, a response delay occurs in the change of EVAP at the time of transition (for example, a response delay occurs when it is desired to immediately cut immediately). To switch
The calculation is performed in the background job except when the opening and closing is performed stepwise at a cycle of 100 msec.

(2-2) Purge Valve Target Duty The target duty EVAPT of the purge valve is EVAPT = EVAP0 + OFSTPV + VBOFPV ... [1] where EVAP0: Basic duty of the purge valve OFSTPV; Learning value of the purge valve rising duty VBOFPV; It is determined by the duty battery voltage correction rate (step 9 in FIG. 5). The calculated EVAPT is limited to the upper limit value (EVPMAX #) (steps 10 and 11 in FIG. 5).

Here, OFSTPV in the equation [1] is a learning value (simply referred to as an offset learning value) corresponding to the purge valve rising duty, which will be described later.

The basic duty EVAP0 of the equation [1] is
Basically, EVAP0 = constant * TQPV / (KPVQH * KPVVB) ... [a] However, TQPV; purge valve target flow rate KPVQH; purge valve flow rate negative pressure correction rate KPVVB; purge valve flow rate battery voltage correction rate, or The table having the characteristics shown in FIG. 27 is looked up to obtain.

Further, the purge valve target flow rate TQP of the formula [a] is used.
V is TQPV = Qs * PAGERT ... [b] where Qs is the intake air amount of the airflow meter section PAGERT is the target purge rate.

The target purge rate PAGERT in the equation [b] is
From the purge learning value WC corresponding to the purge fuel concentration, FIG.
A table having the characteristics shown in FIG. 5 is looked up to obtain (step 3 in FIG. 5).

As shown in FIG. 23, PAGERT is reduced when WC is large, and PAGE is reduced when WC is reduced.
RT is increased. In order to prevent the fuel vapor of the purge gas from leaking to the outside of the vehicle, the purge rate may be increased, but in this case, if the purge fuel concentration is high, the error in the air-fuel ratio A / F will be large. . Therefore, when it is determined that WC is large (purge fuel concentration is high), the target purge rate PAGERT is decreased while W is decreased in order to prevent the A / F error from increasing.
When it is determined that C is small (the purge fuel concentration has become thin), the purge is rapidly performed at a large target purge rate.

PAGERT is the air flow meter flow rate Qs
As a general rule, even if WC changes, the characteristics are horizontal regardless of whether WC is large or small as shown in the upper part of FIG. 34 (shown by a solid line). However, since the flow rate of the purge valve becomes maximum when the purge valve is fully opened, the purge rate gradually decreases after the maximum flow rate of the purge valve.

If any of the sensors (O ₂ sensor, air flow meter, throttle sensor) is abnormal (indicated by NG in the figure), RAGERT = NGPGRT
# (Steps 1 and 2 in FIG. 5). NGPGRT
# Is the purge rate (constant rate) when the sensor is abnormal.

By the way, as shown in equations [a] and [b], EV
If the number of variables used to calculate AP0 increases (TQ
PV, KPVQH, KPVVB, Qs, and PAGERT), and the precision of EVAP0 depends on the precision of the number of bytes and table given to these variables.

Therefore, here, the purge valve target flow rate TQ is set.
PV is calculated by TQPV = (Qs * PAGERT * coefficient) / KPVQH ... [2] (step 7 in FIG. 5), and EVAP0 is obtained from EVAP0 = table value / KPVVB ... [3] (step 8 in FIG. 5). ). Since these methods have better correction accuracy than the above equations [a] and [b], they are used.

The table value of the equation [3] is a value obtained by looking up a table having the characteristics shown in FIG. 28 from the purge valve target flow rate TQPV.

The KPVQH in the equation [2] is a correction factor for changing the flow rate due to the differential pressure across the purge valve even if the flow passage area of the purge valve section is constant, and a table containing the characteristics of FIG. 24 is looked up from the flow rate QH0. Up and ask (step 4 in FIG. 5). As the differential pressure across the flow decreases, it becomes more difficult to flow, so
When the front-rear differential pressure is small (when QHO is large), the target flow rate is corrected to be large.

Since the differential pressure is not actually detected,
Here, QHO is used as the differential pressure equivalent amount. QH
0 is a known volume flow rate of the throttle valve portion determined by the engine speed NE and the throttle opening TVO.

In terms of the phase during transition, since the cylinder air amount equivalent pulse width TP is closer to the differential pressure across the purge valve than QH0, it is desirable to use TP, but TP is different in atmospheric pressure and intake air temperature. Therefore, QHO is used here because the relationship with the differential pressure across the purge valve deviates. In addition,
TP is also publicly known, and in consideration of the fact that even if the air flow meter section measures the air amount as will be described later, it actually flows into the cylinder with a first-order lag, and the cylinder air amount that flows in with the first-order lag is considered. On the other hand, the fuel amount is given in a fixed proportional relationship.

KPVVB in the equation [3] is the battery voltage V
A table containing the characteristics of FIG. 25 is looked up from B (step 5 of FIG. 5), and the battery voltage correction factor VBOF of the purge valve rising duty of the formula [1] is calculated.
PV is obtained by looking up a table having the characteristics of FIG. 26 from the battery voltage VB (step 6 of FIG. 5).

KPVVB is a correction factor for the relationship between the basic duty EVAP0 of the purge valve and the flow rate of the purge valve, which varies depending on the voltage applied to the purge valve, and VBOFPV is also the purge valve duty when the flow path of the purge valve starts to open. The correction rate varies depending on the voltage applied to the valve, and the correction rate varies depending on the type of purge valve. 25 and 26 are examples when the purge valve is driven by a linear solenoid.

(2-3) Purge valve flow rate predicted value Purge valve flow rate predicted value QPV is QPV = EVAPQ * KPVQH ... [4] where EVAPQ: basic flow rate of purge valve KPVQH: negative pressure correction rate of purge valve flow rate (Step 19 in FIG. 6).

The EVAPQ in the equation [4] is (EVAP-O
FSTPV-VBOFPV) * KPVVB is obtained by looking up a table having the characteristics shown in FIG. 32 (step 18 in FIG. 6). In FIG. 29, the horizontal axis is EV
Not purging AP0 * KPVVB is purge ON / OFF
This is because the value of EVAP-OFSTPV-VBOFPV (that is, the value at the time of transition) and the value of EVAP0 (the value at the time of equilibrium) do not match when switching to.

(3) Purge learning control In addition to the basic air-fuel ratio learning value αm, a purge learning value WC (which is also a learning value of the mixture ratio of purge gas) corresponding to the purge fuel concentration is introduced. Differentiating from αm is the purpose of introducing αm, which is different from the air-fuel ratio error that changes very slowly (due to variations in the characteristics of the air flow meter and the injector). Therefore, the control accuracy of the air-fuel ratio is increased by separating the purge learning value WC and the basic air-fuel ratio learning value αm.

Now, the purge fuel concentration is the purge ON, OF
Prediction can be made from α that changes due to switching to F as follows.

If only the purged fuel concentration becomes richer than the previous time (there is no air-fuel ratio error due to an air flow meter, etc.), the air-fuel ratio leans toward the rich side, so that the value of α is returned to the lean side. (Or its average value) shifts to the side smaller than the control center (1.0).
Therefore, when α deviates to the smaller side, the purge learning value W
When C is updated to the larger side, the updated WC corresponds to the purged fuel concentration that is thicker than the previous time. On the contrary, when the purge fuel concentration is lower than the previous time, α deviates to the larger side from the control center. Therefore, if WC is updated to the smaller side at this time, the updated WC becomes thinner than the previous time. Equivalent to.

By predicting the purge fuel concentration in this way, the purge ON / OFF can be performed without providing a sensor.
It is possible to prevent the air-fuel ratio error immediately after the switching to.

(3-1) Battery Backup Although the purge learning value WC is a battery battery, WC = INWC # is set when the control unit 2 is first energized (steps 101 and 102 in FIG. 11). INWC # is an initial value of WC for the first energization.

In other cases, when the control unit 2 is energized, WC = WC hold value + WCST #, where WCST # is the added value of WC at the time of starting (steps 101 and 105 in FIG. 11). WCST
# Considers the increase in fuel stored in the activated carbon canister while the vehicle is stopped. If there is no time between the last engine stop and this engine start, WC
Although the air-fuel ratio error does not occur during the purge during the current operation depending on the held value, over a period of time, evaporated fuel accumulates in the activated carbon canister, and this amount is the air-fuel ratio error when starting the engine this time. Appears as an error.
Therefore, this amount (that is, the increased fuel amount while the vehicle is stopped) is WCS
It is estimated by T #.

(3-2) Purge learning permission condition Purge learning is permitted at the time of switching to purge ON or purge OFF. Therefore, after setting the flag instructing switching of purge ON and OFF, that is, step 30, FIG. It is performed following 47 or 49. In FIG. 8, if offset learning is not reserved, the WC learning permission flag (#FWC
GKOK) = 1 (steps 82 and 83 in FIG. 8).

The reason why learning is permitted at the time of switching to purge ON and OFF without waiting for the change of α to end is as follows.
This is to increase the frequency of learning.

(3-3) Discontinuation Conditions for Purge Learning Discontinue purge learning when the following conditions are satisfied (FIG. 1).
2 steps 116-119, 113). When the purge learning permission flag = 0 (step 116 in FIG. 12). This is to interrupt the purge learning when the purge ON condition is switched to the purge OFF condition or vice versa while the purge learning condition is satisfied (steps 81 and 8 in FIG. 8).
5). Under the conditions other than the condition that the air-fuel ratio feedback control is being performed and the clamping is not being performed (step 117 in FIG. 12). This is because the learning condition is a condition during the air-fuel ratio feedback control and not during the clamp. Basic duty (EVAP0) is a predetermined value (WCGDT)
Y #) smaller (step 118 in FIG. 12). This is because when the basic duty is small, it is confused with the air-fuel ratio error due to the variation in the purge valve rising duty, which is to be avoided. As shown in FIG. 34,
In terms of the purge valve flow rate, the high flow rate range is the purge learning condition and the low flow rate range is the offset learning condition. When the load (QH0) is too high above the predetermined value (WCGQH #) (step 119 in FIG. 12).

However, even if one of the conditions (1) to (4) is satisfied, if the cut flag for purge learning is 1,
Step 113 is skipped (step 12 in FIG. 12).
0,114). In this case, the purge learning permission flag is not set to 0. When the purge learning is interrupted by the purge cut for the purge learning, it is switched to the purge ON again, but the purge learning is not started at that time. This is because

If any of the sensors is abnormal, W
Since C = NGWC # and the purge learning is interrupted,
The purge learning permission flag is set to 0 (step 11 in FIG. 12).
1 to 113), and initialization and post-processing of RAM and flags for purge learning (step 11 in FIG. 12).
4).

In the learning when the purge is ON, the EVAP
Is the purge valve offset (for example, VBOFPV + D
Purge learning is kept waiting until LYWCG # (set to the delay time equivalent amount) is passed (step 12 in FIG. 12).
1, 122).

(3-4) Update of Purge Learning Value The average value ALPAV of the first air-fuel ratio feedback correction coefficient that has entered the purge learning is stored in ALPST of the memory (steps 131 to 133 in FIG. 13). The ALPAV at the start of learning (which is also the ALPAV before the purge ON / OFF switching) is stored in ALPST.

When the purge learning is permitted, the purge learning value WC is updated by WC = WC holding value + ΔWC (step 181 in FIG. 14). WC after update
Is limited between the upper limit value (WCMAX #) and the lower limit value (WCMIN #) (steps 183 to 18 in FIG. 14).
5).

ΔWC is the learning update amount, which is the air-fuel ratio feedback correction coefficient ALPHA (α
The difference between (1) and ALPST is large and the difference is small (± PWCH and ± IWCH when large, ± PWCL and ± IWCL when small), and ΔWC is given as shown in the table. Therefore, FIG. 14 and FIG. 15 are combined.

Here, instead of the description of FIGS. 14 and 15, by explaining how the learning update amount ΔWC is given with the waveform shown in FIG. 36.

FIG. 36 shows the state when switching to purge ON.

When α is moved to the lean side by switching to purge ON (a long integration amount I is acting), α
When the average value of ALPAV (shown by the broken line) also moves to the lean side and ALPAV <ALPST (point A)
The learning value WC is stepwise by PWCL # (step amount)
Then, it is gradually increased by IWCL # (integration amount).

These learning update amounts (PWCL # and IWC)
L #) is not enough and α moves across the predetermined width (DALPH #) toward the lean side, and at this crossing point (point B), the learning value WC is larger than PWCL # above.
The WCH # (this is also the step amount) is increased stepwise, and thereafter the I value of the value larger than the IWCL # is used.
WCH # (integrated amount) is gradually increased. α is ALP
When the width of DALPH # based on ST is deviated, the learning speed is increased by giving a larger step amount PWCH #.

However, in order to avoid overshooting of the learning value WC, the addition (or subtraction) of the step amount PWCH # is performed only once during the purge learning permission.

As a result, the difference between α and ALPST is DALP.
If it falls within the width of H #, IWCL # from point D
And IWCH # are used, and ALPAV is ALPST.
From the point of time (point E) that exceeds, the PWCL # and IWCL # with small values for both the step amount and the integration amount are used.

(3-5) Clamp for Purge Learning EVAP = EVAPT or EVAP = EVPCUT #
Then, the average value of the peak values of the latest two learning values WC (that is, WC immediately before PWCL # addition / subtraction) is calculated, and thereafter, WC is clamped to this value to complete the purge learning.

Therefore, ΔWC is + PW in FIG.
Immediately before entering CL # (or -PWCL #), the value stored in OLDWC1 of the memory is changed to OLDWC of the memory.
2, the value stored in WC of the memory is set to the OLDW of the memory
If each is moved to C1 and the PWCL addition count counter value (CONTPWCL) is incremented by 1 (steps 155, 158 and 16 in FIG. 14).
4,166), CONTPWCL is a predetermined value NSWCGK
It is determined that the learning value has converged when # (for example, 3) or more, and the value of (OLDWC1 + OLDWC2) / 2 is reset to WC (steps 125 and 126 in FIG. 12),
The purge learning permission flag = 0 is set (step 128 in FIG. 12).

The reason for clamping the purge learning value is as follows. By separating the purge learning value WC from the basic air-fuel ratio learning value αm, the control accuracy of the air-fuel ratio is increased.
I want to finish the purge learning early. This is because if purge learning is performed forever, the amount of change in α due to changes in operating conditions will be included in the purge learning as an error. In other words, it is desirable to end the purge learning while the operating condition does not change and eliminate the air-fuel ratio error by the basic air-fuel ratio learning when the operating condition changes (except when the purge ON / OFF is switched).

The PWCL added at point Q in FIG.
Since # indicates that the purge valve is being switched (slowly flag = 1), the PWCL addition count counter value (CONTP
WCL) is not counted (steps 124 and 130 in FIG. 12).

(3-6) Air-fuel ratio feedback correction coefficient ALPAV is calculated by a known method often used in basic air-fuel ratio learning. For example, when ALPAV is added with the step amount P (steps 263 and 268 in FIG. 18), ALPAV = (ALPHA + ALPO) / 2, where ALPO; is obtained by α immediately before the addition of the previous P minutes (steps 263 and 2 in FIG. 18).
66).

However, when the feedback control is started during the clamp in the air-fuel ratio feedback control, the ALP is started from the first cycle of the control after the clamp is released as shown in FIG.
In order to calculate AV, the P addition counter value (CO
When (UNTP) is less than 3, ALPAV is 1.0 (steps 265 and 269 in FIG. 18). Of course, during clamping during air-fuel ratio feedback control, A
Both LPAV and ALPO are 1.0 (steps 261 and 262 in FIG. 18).

(4) Learning of Purge Valve Rising Duty When the purge valve 21 is driven by the linear solenoid, the rising duty of the purge valve (duty when the purge valve starts to open) depends on the temperature as shown in FIG.
Changes and the flow rate of the purge valve varies, especially in the low flow rate range. Since the purge valve becomes harder to open as the temperature becomes higher, even if the same basic duty EVAP0 is given, the flow rate of the purge valve becomes substantially smaller at high temperature.

Therefore, a learning value (also referred to as an offset learning value) OFST corresponding to the rising duty of the purge valve is set.
PV is introduced separately from the purge learning value WC.

If the linear flow rate characteristic of the purge valve is considered to move in parallel to the left and right according to the temperature as shown in FIG. 38 (that is, the change in the inclination of the straight line is ignored), target duty = basic duty The target duty may be given by the + purge valve rising duty.

For example, if the rising duty of the purge valve becomes larger than the previous one due to the temperature rise of the purge valve (see FIG. 38), the purge rate (ratio of the flow rate of the purge valve to the intake air amount) decreases due to the valve opening delay, so that the air-fuel ratio is decreased. Leans toward the lean side, and α and ALPAV deviate to a side larger than ALPST in an attempt to return the lean side to the rich side.

At this time, when the offset learning value (corresponding to the rising duty) OFSTPV is updated to the larger side, the updated learning value corresponds to the rising duty with respect to the purge valve temperature at that time. Increase the duty EVAP0 (the value obtained by adding the learning value to the basic duty EVAP0 is the target duty E
By setting VAPT), the purge valve flow rate can be made the same as before the temperature rise.

On the contrary, when the valve opening speed becomes faster due to the decrease of the purge valve temperature, the flow rate of the purge valve increases and the air-fuel ratio becomes rich, so that α and ALPAV deviate from ALPST to the smaller one. At this time, the offset learning value O
By updating FSTPV to a smaller value, the updated learned value corresponds to the rising duty with respect to the purge valve temperature at that time.

In this way, the concept of offset learning is exactly the same as that of purge learning, and the only difference is that the constants and variable names are different and the learning values are updated in the opposite direction. Therefore, only the differences will be briefly described.

(4-1) Offset Learning Permission Condition Offset learning is also permitted when switching to purge ON or purge OFF, but offset learning is reserved after the purge learning value WC is clamped (step of FIG. 12). 126, 127). This is because the air-fuel ratio error (corrected by the learned value WC) caused by the purge and the air-fuel ratio error (learned value OFSTPV that occurs due to the variation in the purge valve).
It is to separate with).

This will be further described with reference to FIG. 34. If the purge rate characteristic is overlapped with the variation of the valve characteristic caused by the temperature characteristic, the variation of the purge rate as shown by the broken line in the small flow rate region (the region where Qs is small) is increased. Expand rapidly. This is because, even if the purge valve has the same flow rate variation, when converted into the purge rate, the smaller the Qs, the larger the rate of the variation amount with respect to Qs.

An air-fuel ratio error occurs due to such valve variations and the purge itself. In other words, even though an air-fuel ratio error is a superposition of two air-fuel ratio errors,
By learning (purge learning) the large flow rate range that is not affected by valve variation as the purge learning condition, first eliminate the air-fuel ratio error caused by the purge, and then learn the small flow rate range where large valve variation occurs as the offset learning condition (offset learning). By performing the learning, the air-fuel ratio error due to the variation in the rising duty is eliminated.

(4-2) Interrupt learning condition for offset learning As can be seen by comparing step 118 in FIG. 12 and step 194 in FIG. 15, when the basic duty (EVAP0) is larger than the predetermined value (OFGDTY #) in offset learning. Stop learning (steps 194, 1 in FIG. 15).
96). The range where EVAP0 is small (small flow rate range) is the offset learning condition, and the range where EVAP0 is large (large flow rate range) is the purge learning condition.

(4-3) Update of Learning Value As can be seen by comparing FIG. 39 with FIG. 35, the method of giving the positive / negative of the learning update amount ΔOFSTPV is opposite to the case of ΔWC. Therefore, the change of the learning value OFSTPV at the time of switching to the purge ON is as shown in FIG. In addition,
In FIG. 40, OFSTPV is negative on the upper side and positive on the lower side.

(5) Basic air-fuel ratio learning (5-1) Learning prohibition condition When any of the following conditions is satisfied, the basic air-fuel ratio learning value αm is not updated (steps 281 to 28 in FIG. 18).
4, 285). When the purge learning is not performed even once (step 281 in FIG. 18). Slowly flag = 1 (step 28 in FIG. 18)
2). That is, it is the time of switching to purge ON or purge OFF. When the purge learning permission flag = 1 (step 283 in FIG. 18). When the offset learning reservation flag = 1 or the offset learning permission flag = 1 (step 284 in FIG. 18).

The basic air-fuel ratio learning is prohibited not only during (during) the purge but also during the purge learning and the offset learning (when the and), because the air-fuel ratio error due to the purge gas, which changes relatively quickly, is prevented. , .Alpha.m is introduced for the purpose of preventing the influence of an air-fuel ratio error that changes very slowly (due to variations in the characteristics of the air flow meter and the injector).

The basic air-fuel ratio learning αm is αm = αm retention value + Δαm where Δαm is updated by the learning update amount, and the learning update amount Δαm is Δαm = (ALPAV-1.0) · GAIN where ALPAV; ALPHA It goes without saying that the average value GAIN is calculated by the update ratio (value of 1 or less).

(6) Characteristic Expression of Fuel Injection Pulse Width (6-1) Fuel Injection Pulse Width Fuel injection pulse width CTIn for each cylinder is CTIn = TI + CHOSn + ERACIn ... [5] where n: injector number TI; common to all cylinders Fuel injection pulse width CHOSn; cylinder-by-cylinder increase / decrease amount ERACIn; calculated by the transition pulse width from interrupt injection to synchronous injection (step 323 in FIG. 21). This formula itself is known.

Here, the fuel injection pulse width TI of the equation [5]
Is during purging Simultaneous injection: TI = (TP-TEFC + KATHOS) * TFBYA * (α + αm) + Ts ... [6] Sequential injection: TI = (TP-TEFC + KATHOS) * TFBYA * (α + αm) * 2 + Ts ... [7] , TP; cylinder air amount equivalent pulse width TEFC; purge fuel equivalent pulse width KATHOS; wall flow correction amount TFBYA; target fuel air ratio α; air fuel ratio feedback correction coefficient αm; basic air fuel ratio learning value Ts; invalid pulse width ( Step 322 in FIG. 21).

The difference from the conventional method is that in the equations [6] and [7], TEFC is subtracted from TP when purging is in progress. This is because when the purge gas is introduced into the intake pipe, the fuel component (TEFC) of the purge gas is additionally added and flows into the cylinder, so that the same air-fuel ratio is maintained during purging as when not purging. In order to do so, the fuel amount obtained by subtracting this purge gas fuel amount may be supplied from the injector 8 for each cylinder.

Cylinder air amount equivalent pulse width TP
Is TP0 = Qs * KCONST # * KTRM / NE ... [8] TP = TP0 * FLOAD + TP * (1-FLOAD) ... [9] where TP0: Airflow meter air amount equivalent pulse width Qs; Airflow meter air amount KCONST #; Constant KTRM; Trimming coefficient used for correction of air amount error NE; Engine speed FLOAD;
12, 313). These expressions are also known and are for phase matching to correspond to the cylinder intake air amount.

(6-2) Purge fuel equivalent pulse width The purge fuel equivalent pulse width TEF is TEF = QEF * KCONST # / NE ... [10-2] where QEF: Purge fuel cylinder intake amount predicted value KCONST #; NE: Calculated by the engine speed. The expression [10-2] is the same as the expression [8], and is a unit conversion of the cylinder intake amount predicted value (QEF) of the purge fuel into a unit equivalent to the injection pulse width.

The purge fuel flow rate QEF in the equation [10-2] is obtained by QEF = WC * QPV * KQPV ... [13] where WC is the purge learning value QPV is the purge valve flow rate predicted value KQPV is the correction rate of the purge valve flow rate. (Step 21 in FIG. 6). By multiplying the purge valve flow rate estimated value QPV by the fuel concentration equivalent value (WC) of the purge gas, the QEF as the purge fuel amount can be obtained.

The flow rate correction factor KQPV of the equation [13] is obtained by looking up a table having the characteristics of FIG. 30 from the purge valve flow rate predicted value QPV (step 20 of FIG. 6). (7) Intake air amount When purging, the air amount Q used for the injection amount calculation is Q = Qs + QEA ... [14] where Qs: air flow meter air amount QEA; purge air flow rate (excluding fuel) Ask in.

Here, the artificial selection flag (FPQ
If A) = 1, the air amount is corrected by the purge air amount by the equation [14], and if FPQA = 0, it is not corrected (steps 302 and 303, steps 302 and 304 in FIG. 20).

The reason why the formula [14] is adopted during purging is as follows.
Since the amount of air leaking from the purge valve 21 to the intake pipe (intake manifold 17) is not measured by the upstream air flow meter 7, the leaked air due to the purge causes a lean error in the air-fuel ratio. Therefore, during the purging (including the case where the fuel is not adsorbed to the activated carbon canister 16 and only the air leaks), the lean error is prevented by using Q in the formula [14].

The air flow meter 7 also has a measurement delay, which can be dealt with by the method described in JP-A-3-222849.

The purge air flow rate QEA in the equation [14] is QEA = QPV-QEF * KFQ # ... [15] where QPV is the purge gas flow rate (air + fuel) QEF is the purge fuel flow rate KFQ #; Derived correction rate.

Since QEA is synchronized with the output of EVAP (purge valve duty) executed in the Ref signal job (step 291 in FIG. 19), the QEA value obtained in the background job is stored in QEAB of the memory ( (Temporary storage) (Step 22 in FIG. 6), Ref
The signal job transfers the value of QEAB to QEA in the memory (step 292 in FIG. 19).

The KFQ # in the equation [15] corrects the difference because the flow rates of air and fuel vapor are different even in the same flow path.

(8) Wall Flow Correction Amount Equilibrium adhesion amount (M) for each operating condition for the purpose of correcting the low frequency component of the wall flow (the wall flow component that changes relatively slowly).
FH) is stored and a change in the equilibrium adhesion amount due to a transient is added as a total correction amount (KATHOS) to the cylinder air amount equivalent pulse width TP at a predetermined ratio for each fuel injection (subtraction during deceleration). (Japanese Patent Laid-Open No. 63-38656)
No. JP-A-63-38650). Furthermore, for the purpose of correcting the high frequency component of the wall flow (the wall flow component that changes relatively quickly), the wall flow rate (CHOS
In some cases, n: increase / decrease amount for each cylinder, INJSETn: interrupt injection amount for each cylinder, ERACIn; interrupt injection → synchronous injection transition pulse width) are introduced (see Japanese Patent Laid-Open No. 3-116139).

The wall flow correction amount KATHOS takes the fuel supply delay into consideration. Although the injection amount must be increased during acceleration, no matter how good the atomization characteristics of the injector are, a part of the fuel adheres to the intake manifold wall and flows as a liquid along the intake pipe wall (this flow Wall flow), flowing into the cylinder at a slower rate than the fuel carried in the air. That is, the air-fuel mixture sucked into the cylinder is temporarily thinned by the wall-flow fuel, so that the wall-flow correction amount KATHOS is accelerated at the time of acceleration in order to prevent the temporary lean mixture.
Just increase the amount. On the contrary, during deceleration when the manifold pressure suddenly becomes a high negative pressure, the fuel adhering to the manifold wall is vaporized all at once, so the mixture becomes temporarily too rich, and CO and HC increase. To do. Therefore, during deceleration, the vaporized wall flow is reduced.

However, since the gaseous fuel in the purge gas is a light component (a component such as butane that volatilizes at a low temperature) that evaporates from the fuel tank 15, even in the intake pipe, most of it vaporizes to form a flow wall flow. Never. Therefore, when calculating the wall flow correction amount KATHOS, it is necessary to exclude the purge fuel component (TEFC) described later.

Therefore, when purging is performed, the equilibrium adhesion amount MFH is calculated as follows: MFH = MFHTVO * CYLINDR # * (TP-TEFC) ... [16] However, MFHTVO; adhesion ratio CYLINDR #; number of cylinders TP; cylinder air amount equivalent Pulse width TEFC; Purge fuel equivalent pulse width. That is, by subtracting TEFC, which is the fuel amount that does not form the wall flow, from TP, the wall flow correction amount KATH
The prediction accuracy of S is improved, and the air-fuel ratio during transition can be made more appropriate.

However, since some engines do not introduce the wall flow correction amount, the artificial selection flag (FPFHL) =
When the value is 1, the formula [16] is adopted (step 31 in FIG. 21).
5, 316), and when FPFHL = 0, MFH = MFHTVO * CYLINDR # * TP is used as in the conventional case (steps 315 to 31 in FIG. 21).
8), so that it can be applied to any type of engine.

Similarly, CHOSn, INJSET
Also for n and ERACIn, the artificial selection flag (F
When PFHS) = 1, TP-TEFC (= TPP) is used, and when FPFHS = 0, TP is used as before (steps 319 and 320, steps 319 and 321 in FIG. 21).

(9) Idle speed control When the purging air is drawn into the engine, the output (torque) increases. In other words, even if the accelerator opening is the same, the output fluctuates greatly when the load is low and the drivability deteriorates by switching the purge ON and OFF.

In this case, in the case where a valve (auxiliary air valve) capable of continuously changing the opening degree according to the duty signal is provided in the passage bypassing the throttle valve 6, this auxiliary air valve is used to introduce purge air. If you squeeze according to, you can prevent the driving performance from getting worse.

Therefore, the artificial selection flag (F
When EVISC = 1, the control duty (ISCON) to the auxiliary air valve is ISCON = conventional ISCON-ISCEVP ... [17] where ISCEVP; is determined by the purge correction amount (steps 324 and 326 in FIG. 22), F
If EVISC = 0, ISCON = conventional ISCON is used (steps 324 and 32 in FIG. 22).
7).

Purge correction amount ISC EVP of equation [17]
41 looks up a table containing the characteristics of FIG. 41 from QEA / KPVQH (step 325 of FIG. 22).

The conventional ISCON of the formula [17] is
For example, ISCON = ISCi + ISCp + ISCtr + ISCat +
ISCa + ISCrfn ... [18] However, ISCi: integral part of idle feedback control ISCp: differential part of idle feedback control ISCtr: air increase amount during deceleration (equivalent to dashpot) ISCat; N ← → D range correction part of A / T vehicle ( Large in D range) ISCa: Correction amount when the air conditioner is ON ISCrfn: Correction amount when the radiator fan is ON

This is the end of the description of the control system itemization.

Now, when the purge fuel concentration is predicted by the purge learning value WC (FIGS. 12 to 14) and the purge valve flow rate predicted value QPV is obtained according to the operating conditions (step 19 in FIG. 6), the purge valve The fuel flow rate (purge fuel flow rate) QEF of the flow rate is obtained by QEF = WC * QPV * KQPV ... [13] (step 21 in FIG. 6). In addition, [13]
KQPV in the equation is a flow rate correction factor, and does not need to be considered if the rate of fuel desorption from the canister 16 is the same regardless of whether the purge valve flow rate is large or small.

Since this purge fuel flow rate QEF is sucked into the cylinder excessively, the air-fuel ratio leans toward the rich side. Therefore, during the purge, this purge fuel flow rate is subtracted from the cylinder air amount equivalent pulse width TP (that is, basic injection). By correcting TP corresponding to the amount), the air-fuel ratio during purging can be kept the same as before purging.

However, since the communication passage 18 between the canister and the intake pipe is actually connected to the collector portion 17a, there is a transportation delay until the purge fuel flow rate QEF reaches the cylinder from the purge valve position, When this fuel transportation delay is not taken into consideration, the accuracy of correction of the basic injection amount decreases, and the air-fuel ratio fluctuates before and after purging.

In order to cope with this, in the control unit 2, QEF1 = QEF * EDMP1 # + QEF1 * (1-EDMP1 #) ... [11] QEF2 = QEF1 * EDMP2 # + QEF2 * (1-EDMP2 #) ... [12] , QEF1; Intermediate predicted value of purged fuel flow rate EDMP1 #; Weighted average coefficient 1 QEF2; Intermediate predicted value of purged fuel flow rate EDMP2 #; QEF2 (weighted average series combination value) is calculated by the weighted average coefficient 2 and this QEF2 Ref signal for a predetermined number of times (QEFDLY #) (every 180 ° CA for 4 cylinders, 120 for 6 cylinders)
The value that is delayed by the number of rises for each °) is calculated as the cylinder intake amount predicted value QEFC of the purge fuel flow rate (steps 293 to 295 in FIG. 19).

When storing the calculated QEF2 value in the memory, if a certain number of memories are prepared and shifted to the next memory in sequence, the value QEFDLY # times before is selected from these memories. It can be QEFC (step 295 of FIG. 19).

This is performed twice in succession with the weighted average of the purge fuel flow rate QEF, and the value obtained by further delaying the output timing of the series combination value of the weighted average by a predetermined time (corresponding to the fuel transportation delay time) is used. It is used as the cylinder intake amount predicted value QEFC.

Predicted cylinder intake amount QE thus obtained
FC is converted into the fuel injection pulse width by TEFC = QEFC * KCONST # / NE ... [10] in order to match the unit of TP (step 3 in FIG. 21).
11), value TPP obtained by subtracting this TEF from TP
(= TP-TEFC) is made equivalent to the basic injection amount during purging (step 314 in FIG. 21).

Now, the operation of this example will be described with reference to FIG.

When all the gaseous fuel (light components) in the purge gas has the same size, the fuel particles are transmitted from the purge valve position to the cylinder, as shown in FIG. 42 (B).
The simple delay time shown in (is a value synchronized with the rotation of the engine,
It corresponds to the fuel transportation delay time).

Actually, not all the fuel particles reach the cylinder at the same time, but as shown in FIG. 42 (B), the purge fuel flow rate (actual value) spreads out temporally (left and right in the figure). (Indicated by) is transmitted. The reason is that the gaseous fuel in the purge gas is composed of a collection of fuel vapors with different fuel particle sizes (weights), of which the lighter particles flow faster in the air stream and the heavier particles slower. Therefore, the lighter particles reach the cylinder earlier, and the heavier particles reach the cylinder later. That is, due to the difference in the diffusion speed of the fuel particles, FIG.
It is thought that this is because the purge fuel flow rate spreads over time (in the horizontal direction in the figure) as shown in.

In this case, in this example, the purge fuel flow rate QE
When the predicted flow rate value QEF1 is obtained from F by the first weighted average, this becomes a waveform that rises with a first-order lag with respect to QEF (broken line in FIG. 42 (C)), and the predicted flow rate is calculated from this QEF1 by the second weighted average. When the value QEF2 is further obtained, a waveform considerably close to the distribution waveform due to diffusion of fuel particles due to this QEF2 (solid line in FIG. 42C)
Is given. By optimally setting the two weighted average coefficients of EDMP1 # and EDMP2 #, the actual diffusion distribution can be approximated by QEF2.

Since the diffusion distribution of the fuel particles becomes wider as the intake pipe length from the purge valve position to the cylinder becomes longer, in such a case, a weighted average of two times is required. Also, if the fuel properties are different in summer and winter (fuel that is more volatile than in summer may be sold in winter to help engine performance during cold), the diffusion distribution of fuel particles may be different in summer and winter. Changes. In this case, the weighted average coefficient (EDMP1 #, EDMP2 #) is prepared separately for each season.

On the other hand, in order to measure the simple delay time of FIG. 42 (B), the Ref signal is counted QEFDLY # times, and if the output timing of QEF2 is delayed during this counting, the fuel transportation delay is considered. And then
This delayed value (QEFC) closely matches the actual behavior of the purge fuel flow rate (actual value in FIG. 42 (B)).

In this way, when the cylinder intake amount of the purge fuel flow rate is accurately predicted, this predicted value QEFC
Therefore, the accuracy of TP correction corresponding to the basic injection amount is improved, and there is a fuel transport delay until the purge fuel flow rate reaches the cylinder, and the fuel component (purge fuel flow rate) carried as fuel vapor is equal to the fuel particle size. Due to the difference, the air-fuel ratio immediately before the purge determined by TP does not lean toward the lean side immediately after the purge even if the air-fuel ratio is distributed temporally and reaches the cylinder.

In the embodiment, the linear solenoid driven purge valve has been described, but the present invention is not limited to this, and a rotary valve or a step motor driven valve may be used.

[0162]

According to the present invention, the purge fuel flow rate is calculated based on the purge valve flow rate and the purge fuel concentration, a weighted average is performed twice on the purge fuel flow rate, and a value obtained by performing the weighted average twice. Predict the value obtained by delaying the output timing of for a predetermined time as the cylinder intake amount of the purge fuel flow rate,
Since the fuel injection amount during purging is calculated by correcting the basic injection amount according to operating conditions with this cylinder intake amount predicted value, the correction accuracy of the basic injection amount improves and the purge fuel flow rate reaches the cylinder. There is a delay in fuel transport until the fuel injection rate reaches the cylinder due to the temporal distribution of the purge fuel flow rate due to the difference in the fuel particle size, and the air-fuel ratio immediately before the purge determined by the basic injection amount leans toward the lean side immediately after the purge. Never.

[Brief description of drawings]

FIG. 1 is a diagram corresponding to the claims of the present invention.

FIG. 2 is a system diagram of an embodiment.

FIG. 3 is a characteristic diagram of a VC negative pressure and a VC negative pressure valve lift.

FIG. 4 is a characteristic diagram showing a manifold-cylinder filling model.

FIG. 5 is a flowchart for explaining calculation of the purge valve duty EVAP and setting of the purge valve duty EVAP at the transition time.

FIG. 6 is a flowchart for explaining calculation of a purge air amount QEA.

FIG. 7 is a flow chart for explaining determination of purge ON / OFF conditions.

FIG. 8 is a flow chart for explaining setting of a flag for instructing purge ON and OFF and learning permission (including reservation).

FIG. 9 is a flowchart for explaining determination of purge cut conditions for purge learning and offset learning.

FIG. 10 is a flow chart for explaining calculation of a purge valve duty EVAP during switching between purge ON and OFF.

FIG. 11 is a flowchart for explaining initialization of a learning value.

FIG. 12 is a flowchart for explaining determination of a purge learning interruption condition and clamping of the purge learning value WC.

FIG. 13: ALPAV before switching to purge ON / OFF
5 is a flowchart for explaining sampling and selection of a learning update amount ΔWC.

FIG. 14: Selection of learning update amount ΔWC and purge learning value WC
5 is a flowchart for explaining the update of the.

FIG. 15 is a flowchart for explaining determination of an interruption learning interruption condition and clamping of an offset learning value OFSTPV.

FIG. 16: ALPAV before switching to purge ON / OFF
5 is a flow chart for explaining sampling and selection of a learning update amount ΔOFSTPV.

FIG. 17 is a flowchart for explaining selection of a learning update amount ΔOFSTPV and updating of an offset learning value OFSTPV.

FIG. 18 is a flowchart for explaining calculation of an air-fuel ratio feedback correction coefficient α and determination of an update prohibition condition for basic air-fuel ratio learning.

FIG. 19 is a flowchart illustrating a Ref job.

FIG. 20 is a flowchart for explaining calculation of an air amount Q used for calculating a fuel injection amount.

FIG. 21 is a flowchart for explaining calculation of a fuel injection pulse width CTIn.

FIG. 22: ON-duty ISCO to auxiliary air control valve
6 is a flowchart for explaining calculation of N.

FIG. 23 is a characteristic diagram of a target purge rate PAGERT.

FIG. 24 is a characteristic diagram of a negative pressure correction factor KPVQH of a purge valve flow rate.

FIG. 25: Purge valve flow rate battery voltage correction factor KPVV
It is a characteristic view of B.

FIG. 26 is a characteristic diagram of a battery voltage correction factor VBOFPV of a purge valve rising duty.

FIG. 27 is a characteristic diagram of the basic duty EVAP0 of the purge valve.

FIG. 28 is a characteristic diagram of lookup values.

FIG. 29 is a characteristic diagram of the basic flow rate EVAPQ of the purge valve.

FIG. 30 is a characteristic diagram of a purge fuel flow rate correction rate KQPV.

FIG. 31 is a characteristic diagram of a load purge permission lower limit value TPCPC.

FIG. 32 is a characteristic diagram showing a purge region.

FIG. 33 is a characteristic diagram showing a switching waveform of the purge valve.

FIG. 34 is a characteristic diagram of purge valve flow rate and purge rate with respect to valve characteristics.

FIG. 35 is a table diagram for explaining selection of a learning update amount ΔWC.

FIG. 36 is a waveform diagram of a purge learning value WC when switching to purge ON.

FIG. 37 is a waveform diagram of α after unclamping.

FIG. 38 is a flow rate characteristic diagram of a purge valve driven by a linear solenoid.

FIG. 39 is a table diagram for explaining selection of the learning update amount ΔOFSPV.

FIG. 40: Offset learning value O when switching to purge ON
It is a waveform diagram of FSTPV.

FIG. 41 is a characteristic diagram of a purge correction amount ISCEVP when purge is ON.

FIG. 42 is a waveform diagram for explaining the operation of the embodiment.

[Explanation of symbols]

2 control unit 3 exhaust pipe 4 three-way catalyst 5 O ₂ sensor (air-fuel ratio sensor) 6 throttle valve 7 air flow meter 8 injector (fuel supply device) 15 fuel tank 16 activated carbon canister 21 purge valve 32 purge fuel concentration predicting means 33 purge valve Flow rate calculation means 34 Purge fuel flow rate calculation means 35 Cylinder intake amount prediction means 36 Fuel injection amount calculation means 37 Fuel supply device

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Internal reference number FI technical display location F02M 25/08 K 7114-3G

Claims (1)

[Claims] 1. An evaporative fuel treatment system for an engine, comprising a purge valve for adjusting the amount of purge gas introduced from a canister to an intake pipe, and means for predicting the fuel concentration of the purge gas, and an operating condition signal for the flow rate of the purge valve. And a means for calculating the purge fuel flow rate based on the purge valve flow rate and the fuel concentration, and a weighted average is performed twice for this purge fuel flow rate, and this weighted average is performed twice. Means for predicting a value obtained by delaying the output timing of a predetermined value for a predetermined time as the cylinder intake amount of the purge fuel flow rate, and correcting the basic injection amount according to the operating condition with this cylinder intake amount predicted value, and fuel injection during purging An evaporative fuel treatment system for an engine, characterized in that it is provided with means for calculating the amount and a device for supplying this fuel injection amount to the intake pipe.