JPH0694848B2 - Air-fuel ratio learning control method for LPG engine - Google Patents

Description

The present invention relates to an air-fuel ratio learning control method for an engine using liquefied petroleum gas as a fuel, and more particularly to a learning method during feedback control during acceleration. is there.

[Prior Art] An engine using liquefied petroleum gas as fuel (hereinafter simply referred to as LP
Called the G engine. 2), two fuel passages are normally provided in parallel, and one fuel passage is provided with a control valve having a variable opening degree (usually, the opening degree is controlled by a step motor. This is hereinafter simply referred to as a step valve). The other fuel passage is provided with a control valve (hereinafter simply referred to as a power valve) that opens and closes the fuel passage.

The step valve, which is a variable opening control valve, has its opening controlled based on the basic step number determined by the engine speed and the load, for example, the intake pipe pressure. The opening control of the step valve is so-called open control, and the fuel supplied to the engine is controlled to the lean burn side.

On the other hand, the power valve is closed during normal operation (hereinafter referred to as open control) so as not to supply fuel to the engine, but is opened during acceleration to increase the fuel supply amount, Good acceleration response. Further, during acceleration, feedback control is performed on the opening degree of the step valve based on the oxygen concentration detected by the O ₂ sensor provided in the exhaust manifold and the like, and the above-described basic step number and the average opening degree of the step valve during the feedback control are calculated. Learning is performed with the difference from the average number of steps shown as the learning value. This learned value is used to correct the deviation of the air-fuel ratio during open control due to changes over time, changes in fuel components, and so on. The open control during normal operation is for always controlling the air-fuel ratio on the lean burn side, and the feedback control during acceleration is for making the air-fuel ratio close to the stoichiometric air-fuel ratio.

[Problems to be Solved by the Invention] However, in the above method of controlling the amount of fuel supplied to the LPG engine, when the pressure in the intake pipe approaches atmospheric pressure, the number of steps indicating the opening of the step valve is the same. Even if there is, it has a characteristic that the air-fuel ratio rapidly increases due to the influence of intake pulsation. Therefore, if feedback control is performed when the pressure difference between the intake pipe pressure and atmospheric pressure is small,
In order to control the air-fuel ratio to approach the theoretical air-fuel ratio,
The average number of steps is smaller than when the differential pressure is large, and the learning value is learned on the side of decreasing the number of steps.
Therefore, when learning is performed in a place where the atmospheric pressure is low such as in a highland and the learned value is used in a place where the atmospheric pressure is high such as a low ratio, there is a problem that the lean burn side is excessively controlled. That is, even if the pressure in the intake pipe is the same, the pressure difference from the atmospheric pressure is small in high altitudes, so the number of steps is reduced (the fuel amount is decreased) in the region where the air-fuel ratio suddenly becomes rich due to the pulsation of the intake pipe pressure.
The learner learns the learning value, and if this learning value is used as it is during normal driving in lowlands, the amount of fuel supplied is less than the target fuel, unburned fuel remains, and fuel consumption deteriorates. It is thought that problems such as deterioration of the ability and overheating of the three-way catalyst may occur.

Structure of the Invention [Means for Solving Problems] The structure of the air-fuel ratio learning control method for the LPG engine of the present invention in order to solve the above problems is as follows.

The air-fuel ratio learning control method for an LPG engine according to the present invention, as illustrated in FIG. 1, determines the basic fuel amount of the fuel supplied to the engine using liquefied petroleum gas as a fuel through the fuel passage, at least the load of the engine. (P1) based on an operating state including, and when the engine load increases (P2), a predetermined amount of fuel is supplied to the engine through an acceleration fuel increase passage different from the fuel passage (P3), Based on the oxygen concentration in the engine exhaust gas, feedback control is performed on the amount of fuel supplied through the fuel passage so that the air-fuel ratio of the engine is kept constant (P
4) Learn a learning value according to the difference between the average value of the feedback-controlled fuel and the basic fuel amount (P5), and during normal operation (P2), use the learning value to determine the basic fuel amount. In the air-fuel ratio learning control method of the LPG engine, the difference between the intake pipe internal pressure and the atmospheric pressure is corrected (P6) and the fuel supply amount to the engine is controlled based on the corrected basic fuel amount (P7). The point is to perform the above learning (P5) only when the pressure is higher than a predetermined value (P8).

[Operation] According to the air-fuel ratio learning control method for the LPG engine of the present invention, which is based on the above processing, when the load of the LPG engine is increased (P2), the fuel amount for acceleration increase provided separately from the normal fuel passage is provided. Open the passage (P3) and feedback control the normal fuel passage opening based on the oxygen concentration in the exhaust gas (P4) to control the air-fuel ratio of the LPG engine.
4) When the differential pressure between the intake pipe pressure of the LPG engine and the atmospheric pressure is greater than a predetermined value (P8), the basic fuel amount determined based on the engine operating condition including at least the engine load and the acceleration The difference from the average value of the feedback-controlled fuel is learned as a learning value (P5). On the other hand, during normal operation when the load is not increasing (P2), the basic fuel amount is corrected using the learning value (P5) learned during feedback control (P6), and the LPG is corrected based on the corrected basic fuel amount.
Controls the amount of fuel supplied to the engine.

Here, the basic fuel amount is a fuel supply amount to the LPG engine determined based on the operating state including at least the load of the engine, and can be determined by, for example, the engine speed and the intake pressure. is there. This basic fuel amount may be the value itself determined based on the operating state, or may be a value corrected by some factor, such as the cooling water temperature of the LPG engine.

The learning value is the difference between the above basic fuel amount and the average value of the fuel that has been feedback-controlled during acceleration, and is used to correct the deviation of the air-fuel ratio during open control due to changes over time, changes in fuel components, etc. Is. This learning value is learned by gradually increasing or decreasing the learning value during feedback control.

[Examples] Next, examples of the present invention will be described in detail.

FIG. 2 is a schematic configuration diagram of an engine (hereinafter simply referred to as LPG engine) system using a known liquefied petroleum gas to which the air-fuel ratio learning control method for an LPG engine according to an embodiment of the present invention is applied.

The LPG engine 1 communicates with an air cleaner 3 through an intake manifold 2 and takes in outside air from the air cleaner 3, and at the same time, an LPG regulator 6 through an ordinary fuel passage 4 and an acceleration fuel passage 5 provided in the intake manifold 2.
Liquefied petroleum gas (hereinafter simply referred to as LPG) from
After the mixture of the outside air and LPG is exploded and burned to obtain a driving force, the exhaust gas is discharged from the exhaust manifold 7 to the outside. The opening degree of the normal fuel passage 4 is controlled by a step motor 8 provided on the way,
The acceleration fuel passage 5 is opened only by the power valve 9 provided in the middle of the acceleration fuel passage 5 only during acceleration. On the other hand, the intake amount of the air-fuel mixture of outside air and LPG is determined by the opening degree of the throttle 11 provided in the intake manifold 2. Further, the exhaust gas discharged from the exhaust manifold 7 is purified by passing through the three-way catalyst 12, and a part of the exhaust gas is recirculated to the intake system by a so-called exhaust gas recirculation device 13. LPG
The swirl device 14 attached to the top of the engine 1
The swirling flow of the air-fuel mixture is generated in the cylinder of the LPG engine 1. The power valve 9, the exhaust gas recirculation device 13, and the swirl device 14 are turned on / off by the negative pressure switching valves 16, 17, 18 respectively, and the pressure difference between the LPG regulator 6 and the intake manifold 2 is a negative pressure switching valve. Adjusted by 19.
The check valves 21 and 22 respectively connected to the negative pressure switching valves 16 and 19 prevent backflow of the air-fuel mixture. These negative pressure switching valves 16,17,18
And 19 are electrically connected to an electronic control unit (hereinafter simply referred to as an ECU) 23, and their on / off timings are controlled. Further, the ECU 23 includes an intake air temperature sensor 25 for detecting the temperature of the outside air sucked from the air cleaner 3 and a pressure sensor 2 for detecting the pressure in the intake manifold 2.
6, a throttle sensor 27 that detects the opening of the throttle 11,
An O ₂ sensor 28 for detecting the oxygen concentration in the exhaust gas discharged from the exhaust manifold 7, an exhaust temperature sensor 29 for detecting the temperature of the exhaust gas, and a rotation speed sensor attached to the distributor 30 for detecting the rotation speed of the LPG engine. 31 mag is connected. The ECU 23 controls the negative pressure switching valves 16, 17, 18 and 19 according to the output signals output from each of these sensors, and is attached to the step motor 8, injector 32, and LPG engine 1 described above. The distributor 30 and the like are preferably controlled. The injector 32
Is installed in the intake manifold 2 closer to the LPG engine 1 than the throttle 11 and performs fuel injection when the LPG engine 1 is started. Next, the ECU 23 will be described.

FIG. 3 is a block diagram showing the configuration of the ECU 23.

The ECU 23 mainly includes a well-known central processing unit (CPU) 51, read-only memory (ROM) 52, random access memory (RAM) 53, backup RAM 54 for storing stored data, external input circuit 55, external It is configured as a logical operation circuit in which the output circuit 56 and the like are connected by a bus 57.

The external input circuit 55 includes the intake air temperature sensor 25, the pressure sensor 26, the throttle sensor 27, the O ₂ sensor 28, and the exhaust gas temperature sensor 2 described above.
9 and the rotation speed sensor 31 and the like are connected, and the CPU 51 reads the signal output from each sensor or the like as an input value via the external input circuit 55. Based on these input values, the CPU 51 uses the above step motor 8, the negative pressure switching valves 16 to 19, and the distributor 31 connected to the external output circuit 57.
It also controls the injector 32 and the like.

Next, the LP performed in this embodiment having the above configuration
Regarding the air-fuel ratio learning control of the G engine, FIGS. 4 to 6
This will be described with reference to the flowchart shown in the figure.

The "step motor control routine" shown in FIG.
Of the respective processes executed by 3, the process only showing the control of the step motor 8 for operating the opening degree of the normal fuel passage 4 is shown, and is the process periodically executed by the hard interrupt.

First, when the processing shifts to this routine, in step 100, the rotation speed N of the LPG engine 1 detected by the rotation speed sensor 31.
A basic step of the step motor 8 for operating the opening degree of the normal fuel passage 4 from a three-dimensional map or the like stored in advance in the ROM 52 based on T and the intake pressure PM in the intake manifold 2 detected by the pressure sensor 26. Calculate the number S. In the following step 110, the basic step number S is multiplied by the correction coefficient FAF and the learning value KGi to correct the basic step number S to obtain the target step number ST of the step motor 8. The correction coefficient FAF and the learning value KGi are values obtained during feedback control during acceleration and will be described in detail later. In step 120, the current step number SNOW representing the step number of the step motor 8 is read from the backup RAM 54, and in the following step 130, the current step number SNOW and the target step number SOW.
Compare with T. The current step number SNOW is a value written as the current step number SNOW in the backup RAM 54 when the CPU 51 outputs a rotation command to the step motor 8 via the external output circuit 56. In steps 130-170,
Current step number SN indicating the step number of step motor 8
Perform processing to match OW with the target number of steps ST.

(A) First, in step 130, the target number of steps ST
If it is determined that SNOW is equal to the target step number ST, the current step number SNOW of the step motor 8 does not have to be driven, and the "RETURN" is skipped. Finish the routine.

(B) In step 130, the target number of steps ST> SNOW
If it is determined that the current step number SNOW representing the step number of the step motor 8 is smaller than the target step number ST, the CPU 51 issues a forward rotation command to the external output circuit 56 to increment the step number of the step motor 8. Output to the step motor 8 via
After forward rotation by only steps (step 140), the current step number SNOW representing the step number of the step motor 8 is incremented (step 150), and the process returns to "RETURN".

(C) In step 130, the target number of steps ST <SNOW
If it is determined that the current step number SNOW representing the step number of the step motor 8 is larger than the target step number ST, the CPU 51 outputs a reverse rotation command to decrement the step number of the step motor 8 After the motor 8 is reversely rotated by one step (step 160), the current step number SNOW is decremented (step 170).
The process ends to "RETURN" and the routine is finished.

By repeatedly executing the above processes (a) to (c), the number of steps of the step motor 8 is made equal to the target number of steps ST.

Next, the step of the above "step motor control routine"
The correction factor FAF used in 110 will be described.
FIG. 5 is a flowchart showing the "correction coefficient FAF calculation routine". This "correction coefficient FAF calculation routine" is also a subroutine that is periodically executed like the "step motor control routine".

First, in step 200, feedback (hereinafter simply referred to as F / B
Call. ) It is determined whether the control conditions are satisfied. This F / B control condition is satisfied at the time of acceleration and is determined by setting a flag at the time of acceleration in another processing routine. If it is determined that the F / B control condition is satisfied during acceleration, the process proceeds to step 210. Step 210
Then, based on the output signal of the O ₂ sensor 28, it is determined whether the air-fuel ratio is rich. If it is determined to be rich, the next processing of steps 220 to 300 is executed.

In step 220, the flag YOX determines whether or not the air-fuel ratio was lean when this processing routine was executed last time. If the value of the flag YOX is "0", it is assumed that the previous time was lean, and the process proceeds to the next step 230. That is, it is determined that the air-fuel ratio is switched from lean to rich when the process proceeds to step 230 by the determination of steps 210 to 220. Next, in step 230, the arithmetic mean of the current correction coefficient FAF and the old correction coefficient FAFO at the time of shifting from rich to lean the previous time is calculated in order to calculate the average correction coefficient FAFAV during F / B control, and this is calculated. Performs the process of setting the average correction coefficient FAFAV during F / B control.
In steps 240 to 270, which is a subsequent series of processes, the correction coefficient FAF is set to the old correction coefficient FAFO (step 240), and the value obtained by subtracting the skip amount a from the correction coefficient FAF is set to the new correction coefficient FAF (step 250). , Learning timing flag
YKG is set (step 260), the value of the flag YOX is set to "1", and this routine ends. The learning timing flag YKG, which will be described later, is used when determining the timing at which the learning value KGI should be learned.
The value of YOX being "1" means that the air-fuel ratio is rich.

On the other hand, when it is determined in step 220 that the value of the flag YOX is "1", the process executes the processes of steps 280 to 300. Here, when the process proceeds to step 280 by the determination of steps 210 to 220, it means that the air-fuel ratio maintains the rich state. Step 280
Then, it is determined whether the timer counter CNT is greater than or equal to the constant c. This timer counter CNT is incremented in a processing routine by a hardware interrupt having a shorter cycle than this routine. If the timer counter CNT is less than or equal to the constant c, the routine returns to "RETURN" to end this routine. If the timer counter CNT exceeds the constant c, after subtracting the constant b from the correction coefficient FAF in step 290,
This routine, which clears the value of the timer counter CNT to "0", ends. That is, in steps 280 to 300, the value of the correction coefficient FAF is subtracted by the constant b every predetermined time.

The processing of steps 210 to 300 described above is processing when the air-fuel ratio is rich and is processing for reducing the correction coefficient FAF. In contrast to the process of decreasing the correction coefficient FAF, the processes of steps 320 to 400 are processes when the air-fuel ratio is lean and are processes of increasing the correction coefficient FAF.

First, if it is determined in step 210 that the air-fuel ratio is lean, the process proceeds to step 320. In step 320, the above flag
It will be determined whether the value of YOX is "1". If the value of YOX is "1", the process executes steps 330 to 370. When the process proceeds to step 330 based on the determinations in steps 210 and 320, it means that the air-fuel ratio has switched from rich to lean. The processing of steps 330 to 340 is the same as steps 230 to 240 above.
This is the same processing as the processing of, but the average correction coefficient F during F / B control
AFAV is calculated (step 330), and the value of the correction coefficient FAF is set as the old correction coefficient FAFO (step 340). In the following step 350, the skip amount a is added to the correction coefficient FAF to obtain a new correction coefficient FAF, then the learning timing flag YKG is set (step 360), and the value of the flag YOX is reset to "0" ( (Step 370) End this routine.

On the other hand, when it is determined in step 320 that the value of the flag YOX is "0", the process executes steps 380 to 400. Here, when the process proceeds to step 380 according to the determinations in steps 210 and 320, it means that the air-fuel ratio is maintained in the lean state. Step 380
The processes of Nos. 400 to 400 are the processes opposite to the above-mentioned steps 280 to 300, and are the processes of increasing the value of the correction coefficient FAF by a constant b at every predetermined time.

The timing chart of FIG. 7 shows the processing contents of the above steps 200 to 400. As can be seen from FIG. 7, the correction coefficient FAF is increased or decreased according to the air-fuel ratio signal detected by the O ₂ sensor 28, and is controlled so as to approach the stoichiometric air-fuel ratio.

When it is determined in step 200 that the F / B control conditions are not satisfied, the correction coefficient FAF and the old correction coefficient FAFO
The value of each is set to "1" (step 410), and this routine ends.

Next, only when the value of the learning timing flag YKG set in steps 260 and 360 of the "correction coefficient FAF calculation routine" is "1", in other words, the air-fuel ratio is switched from rich to lean or lean to rich. The "learning routine" shown in FIG.

First, at step 500, it is judged if the value of the learning timing flag YKG is "1". When the value of the learning timing flag YKG is not "1", the processing is step 505.
Then, the value of the learning timing flag YKG is reset to "0" and the routine returns to "RETURN" to end this routine. When it is determined that the value of the learning timing flag YKG is “1”, the process proceeds to step 510. In step 510, the pressure sensor 26
It is determined whether the intake pressure PM in the intake manifold 2 detected by is less than a predetermined pressure PMG. This predetermined pressure PMG
Is determined by the "predetermined pressure PMG calculation routine" shown in FIG. 6 (B). Here, first, the predetermined pressure PMG will be described.

The predetermined pressure PMG is a pressure obtained by subtracting a certain pressure e from another predetermined pressure PMVL (step 600). The other predetermined pressure PMVL is the intake pressure PM detected by the pressure sensor 26 when the throttle opening VL detected by the throttle sensor 27 is larger than a predetermined value (50 ° in this embodiment) (step 610). PMVL (Step 62)
0). When the throttle opening VL is larger than 50 °, it is considered that the intake pressure is almost the same as the atmospheric pressure. Therefore, the predetermined pressure PMG (= PMLV-e) is lower than the atmospheric pressure by a certain pressure (e). In step 510 of the "learning routine", the magnitude relationship between the predetermined pressure PMG and the intake pressure PM detected by the pressure sensor 26, which changes every moment, is determined.

When it is determined that the intake pressure PM is less than the predetermined pressure PMG, the process executes steps 520 to 540. Step
At 520, the value of the average correction coefficient FAFAV obtained by the “correction coefficient FAF calculation routine” is determined.

(A) First, when it is determined in step 520 that the average correction coefficient FAFAV = 1, it is considered that the air-fuel ratio has reached the stoichiometric air-fuel ratio, and no processing is performed.

(B) When it is determined that the average correction coefficient FAFAV> 1,
The process proceeds to step 530. In step 530, the learning value KGi corresponding to the intake pressure PM at that time is increased by the constant d.

(C) When it is determined that the average correction coefficient FAFAV <1,
The process proceeds to step 540. In step 540, the learning value KG
i is reduced by a constant d.

The learning value KGi is increased / decreased by ± d by executing the above-described processes (A) to (C) each time lean / rich switching is performed, and eventually becomes the optimum value for the intake pressure PM at that time. This learning value KGI is used to calculate the target step number ST during normal operation (correction coefficient FAF = 1 at this time) other than during acceleration (step 110 of the "step motor control routine").

On the other hand, when it is determined in step 510 that the intake pressure PM is equal to or higher than the predetermined pressure PMG, the learning process of steps 520 to 540 is not executed and the learning timing flag YKG is reset to "0" (step 505) to end this routine. ing. This is because when the intake pressure PM is equal to or higher than the predetermined pressure PMG and is close to the atmospheric pressure, there is a characteristic that the air-fuel ratio rapidly increases due to the influence of intake pulsation, and the learning process is prohibited at this time.

In the present embodiment, the learning value KGI corresponding to the intake pressure PM
When the intake pressure PM is 200 mmHg or more and less than 300 mmHg, KG1,
KG ₂ when the intake pressure PM is 300 mmHg or more and less than 400 mmHg, KG ₃ when the intake pressure PM is 400 mmHg or more and less than 500 mmHg, the intake pressure
When PM is 500 mmHg or more and less than 600 mmHg, KG4, intake pressure PM is
KG5 when the pressure is between 600 mmHg and less than 700 mmHg.

According to the air-fuel ratio learning control method for the LPG engine of the present embodiment described in detail above, the learning value KGi for correcting the basic step number of the step motor 8 determined by the rotational speed NT and the intake pressure PM of the LPG engine during normal operation. The learning area of intake pressure PM
Is less than the predetermined pressure RMG, the intake pressure PM when learning the learning value KGi is always lower than the atmospheric pressure by a predetermined value e. Therefore, a phenomenon that occurs when the intake pressure PM approaches the atmospheric pressure, that is, even if the number of steps of the step motor 8 is the same, the air-fuel ratio rapidly increases due to the influence of the intake pulsation, so the lean side is controlled and the average correction is performed. On the side where the coefficient FAFAV decreases and the target number of steps ST decreases, the learning value KG
There is no occurrence of a phenomenon such as learning i, and fuel can always be favorably supplied to the LPG engine. Therefore, if the learning value KGI learned in a place with low atmospheric pressure such as highlands is used as it is during normal operation in lowlands, the target value is set using the learning value KGI that reduces the target step number ST in lean burn control during normal operation. The problem that the number of steps is determined and the state becomes leaner is sufficiently solved. As a result, the air-fuel ratio does not become excessively lean, and problems such as overheating of the three-way catalyst due to gas, deterioration of fuel efficiency, and deterioration of drivability do not occur.

FIG. 8 is a graph showing the relationship between the intake pressure PM and the air-fuel ratio when the number of steps of the step motor 8 is constant in the lowland and the highland during the feedback control. From this figure, it can be seen that in the highlands, even if the intake pressure PM is the same as in the lowlands, lean control is performed near the atmospheric pressure in the highlands.

According to the air-fuel ratio learning control method of the LPG engine of the present invention,
The learning value is learned only when the pressure difference between the intake pipe internal pressure and the atmospheric pressure is larger than a predetermined value. As a result, the learning value will not be learned on the side that reduces the basic fuel amount in a place where the atmospheric pressure is low, such as in highlands, and fuel can always be favorably supplied to the LPG engine. Therefore, when moving from a highland to a lowland and performing normal operation, the air-fuel ratio becomes excessively lean and problems such as overheating of the three-way catalyst due to gas, deterioration of fuel consumption, and deterioration of drivability are not caused. Absent.

[Brief description of drawings]

FIG. 1 is a flow chart illustrating the basic configuration of the present invention, FIG. 2 is a schematic configuration diagram of an LPG engine system to which the air-fuel ratio learning control method for an LPG engine of one embodiment of the present invention is applied, and FIG. 3 is the same. Block diagram showing the configuration of the electronic control unit (ECU), FIGS. 4 to 6 (A), (B)
7 is a flowchart showing the processing executed by the electronic control unit, FIG. 7 is a timing chart showing the relationship between the air-fuel ratio signal and the correction coefficient FAF at that time, and FIG. 8 is a step motor 8 in the lowland and the highland. 5 is a graph showing the relationship between the intake pressure PM and the air-fuel ratio when the number of steps in is constant. 1 …… LPG engine 2 …… Intake manifold 3 …… Air cleaner 4 …… Normal fuel passage 5 …… Acceleration fuel passage 6 …… LPG regulator 7 …… Exhaust manifold 8 …… Step motor 9 …… Power valve 11 …… Throttle 12 …… Three-way catalyst 13 …… Exhaust gas recirculation device 14 …… School device 16,17,18,19 …… Negative pressure switching valve 21,22 …… Check valve 23 …… Electronic control unit (ECU) 25… Intake air temperature sensor 26 Pressure sensor 27 Throttle sensor 28 O ₂ sensor 29 Exhaust temperature sensor 30 Distributor 31 Rotation speed sensor 32 Injector

Claims (1)

[Claims] 1. A basic fuel amount of fuel supplied to an engine using liquefied petroleum gas as a fuel through a fuel passage is determined based on an operating state including at least a load of the engine, and when the load of the engine increases, A predetermined amount of fuel is supplied to the engine through an acceleration fuel increase passage that is different from the fuel passage, and the fuel passage is provided so that the air-fuel ratio of the engine is constant based on the oxygen concentration in the engine exhaust gas. Feedback control of the amount of fuel supplied via the control unit and learning value according to the difference between the average value of the feedback-controlled fuel and the basic fuel amount, and during normal operation, the learning value is used. In the air-fuel ratio learning control method for an LPG engine, which corrects the basic fuel amount and controls the fuel supply amount to the engine based on the corrected basic fuel amount, Differential pressure between the intake pipe pressure and the atmospheric pressure of the engine and performing the learning only when larger than a predetermined value
Air-fuel ratio learning control method for LPG engine.