JP2554854B2 - Learning control method for automobile engine - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

TECHNICAL FIELD The present invention relates to an actuator for performing engine control, and a learning control method for an automobile engine capable of improving control accuracy by correcting a characteristic change of a sensor that gives information for calculating a control amount of the actuator by learning. .

[Prior art]

As an electronic control system for an automobile engine, learning control is known in which data in a table is rewritten for fuel supply control of an electronic fuel injection system (see, for example, JP-A-55-96339). . Here, the amount of fuel injected into the engine is determined in relation to engine operating variables such as intake air amount, engine speed, engine load. The amount of fuel is determined by the valve opening time (injection pulse width) of the fuel injection valve. The basic fuel injection width Tp is obtained by the following equation. Tp = K × Q / N (1) where Q: intake air amount N: engine speed K: constant. The desired injection pulse width Ti is obtained by modifying the basic injection width Tp with engine operating variables. The following formula is an example of calculating the desired injection pulse width. Ti = Tp × (COFE) × α × Ka (2) where COFE is a correction coefficient obtained by the sum of correction coefficients such as coolant temperature, throttle opening, engine load α: λ correction coefficient (in the exhaust passage The integrated value of the feedback signal of the O ₂ sensor of Ka) is a correction coefficient by learning (hereinafter referred to as a learning control coefficient). Coefficients such as coolant temperature coefficient and engine load are
Obtained by looking up a table in relation to the detected information. The learning control coefficient Ka value is obtained from the learning value table in relation to the engine load and the engine speed. To further explain the calculation of the desirable injection pulse width (Ti in the equation (2)), all the learning values are set to "1" as initial values in the learning value table when the engine is started for the first time. This indicates that the fuel supply system is designed to deliver almost the correct amount without the factor Ka. However, not all automobiles are manufactured to have the desired functions that produce the same result due to variations in use. Therefore, the learning values in the table need to be rewritten by learning when all cars are actually used. If the difference between the initial value "1" and the rewritten value is large, the fuel injection system causes hunting. In order to avoid such hunting, rewriting is incremented or decremented little by little. In addition, since the temperature of the O ₂ sensor body is low at the time of general engine startup, the output voltage of the O ₂ sensor is also low. In such a state, the system sets "1" as the value of α. Therefore, the computer determines the desired injection pulse width Ti by the equation (2) by the intake air amount Q, engine speed N, COFE, α,
Calculate from Ka. When the engine is warmed up and the O ₂ sensor is activated, the integrated value of the O ₂ sensor output voltage at the predetermined time is supplied as the value of α. More specifically, the computer functions as an integrator, and integrates the output voltage of the O ₂ sensor. FIG. 7 (b) outputs an integral output. The system outputs the integrated value at a predetermined interval (for example, 40 ms). For example, in FIG. 7 (b), the time
It provides the integral I ₁ ……… I _n at T ₁ ……… T _n . Therefore, the amount of fuel is controlled according to the integrated feedback signal α from the O ₂ sensor. The problem here is the intake air amount Q / N per revolution as the engine load with the ignition timing preset.
That is, the basic fuel injection pulse width Tp according to the above equation (1)
(Tp = K × Q / N, K = constant, and since Q / N is proportional to the basic fuel injection pulse width, Tp is substantially a parameter as conventionally known from Japanese Patent Publication No. 58-33385. In the case of a method in which the ignition timing control is performed simultaneously by obtaining from a map composed of a value corresponding to (1) and the engine speed N, when the characteristics of the sensor such as the air flow meter change due to changes over time. The value of Q / N changes, the ignition timing is not controlled normally, and exhaust gas emission deteriorates, output torque decreases, and knocking occurs. That is, there are input sensors such as an air flow meter and actuators such as a fuel injection valve as a device related to the control according to the equation (2). If any of these characteristics changes, the result is that the value of α changes. However, the value of Ka is also changed to correct this, and the fuel injection control is normally performed. However, regarding the ignition timing, it is impossible to determine only the characteristic change of the sensor such as the air flow meter, which causes a problem such as knocking. Further, although the characteristic change due to the deterioration of the actuator over time is reflected in the learning value Ka as described above, the learning value Ka is not only learned about the characteristic change of the actuator, so that the control accuracy is improved. Has a limit. In this regard, Japanese Patent Laid-Open No. 58-10126 (first prior art) discloses first, second, and third memories corresponding to engine idling, low load, and high load. Part,
Idle, low load, or high load is detected by the intake air flow rate by the air flow meter, and the value in the corresponding storage unit is corrected based on the deviation of the return air-fuel ratio by the air-fuel ratio sensor feedback signal with respect to the reference air-fuel ratio, When the values of at least two storage units have a difference of a predetermined value or more with respect to the reference value, the altitude correction value is corrected, and the value of the first storage unit corresponding to the idling and the value of the first storage unit corresponding to the high load are corresponded. Third
Is larger than the difference between the value of the second storage section and the value of the third storage section corresponding to the low load, the correction value of the output of the air flow meter is corrected and the fuel injection is performed. By using the altitude correction value and the correction value of the air flow meter output in the amount calculation, it is possible to perform the altitude correction and the correction of the output of the air flow meter, and a technique for significantly reducing the number of nonvolatile memory elements is disclosed. . Further, in Japanese Patent Laid-Open No. 55-134730 (second prior art example), when the air-fuel ratio feedback control is performed and a constant and quasi-static state (engine steady operation state) is detected, the measured air-fuel ratio and theoretical Calculate the deviation from the air-fuel ratio, and based on its plus or minus,
The coefficient for correcting the output of the sensor for measuring the intake air flow rate, which is stored in the storage device of the microcomputer in advance, is subtracted or added by the unit amount for correction, and the air-fuel ratio is calculated based on this corrected coefficient. A technique is disclosed in which the characteristics of the sensor for measuring the intake air flow rate are constantly corrected by controlling the air-fuel ratio to perform good fuel injection control. Further, Japanese Patent Publication No. 58-33385 (third prior art example) discloses that the difference between the average value and the set value of the air-fuel ratio feedback correction amount based on the output of the exhaust gas sensor (O ₂ sensor) is set in a steady operation state. If the value is more than the specified value, the relationship between the air flow sensor output data stored in RAM and the intake air amount data is corrected.If the average value deviates from the specified value by more than the specified value, the basic air-fuel ratio becomes leaner. Therefore, the relationship between the two is corrected so that the intake air amount data becomes larger for the same air flow sensor output. Conversely, if the average value is smaller than the set value by more than the specified value, the basic empty Since the fuel ratio is shifted to the rich side, the relationship between the two is corrected so that the value of the intake air amount data corresponding to the air flow sensor output data becomes smaller, and the intake air amount is corrected based on this corrected RAM data. By performing a positive calculation and calculating the fuel injection amount using this intake air amount, the basic air-fuel ratio itself is constantly controlled to be almost equal to the theoretical air-fuel ratio, and the characteristics of the exhaust gas sensor and engine characteristics are A technique for preventing the air-fuel ratio spike phenomenon in the engine transient operation state regardless of the situation is disclosed.

[Problems to be Solved by the Invention]

However, in the first prior art example described above, the altitude correction value and the correction value of the air flow meter output are corrected by simply using the three values at idling, low load, and high load. Therefore, it is not possible to perform precise correction corresponding to the characteristic change of the air flow meter (sensor) corresponding to the change of the engine operating state, and the characteristic change such as deterioration with time for the injector (actuator) is not taken into consideration. Also, the controllability cannot be improved sufficiently. Furthermore, after correcting the altitude correction value and the correction value of the air flow meter output by the value of the first, second, or third storage unit, the value of each storage unit is cleared and the value of the storage unit is corrected again from the beginning. Since the correction value of the altitude and the correction value of the output of the air flow meter are stored in the non-volatile storage element again, the learning process becomes troublesome, and the learning procedure becomes complicated and the control system becomes complicated. Further, in the second prior example, when the engine steady operation state is detected, the coefficient for simply correcting the output of the sensor for measuring the intake air flow rate is uniquely corrected based on the deviation between the measured air-fuel ratio and the theoretical air-fuel ratio. However, an appropriate sensor correction value for each engine operating state cannot be obtained, and the above first
As in the preceding example, since the characteristic change such as deterioration with respect to the injector (actuator) is not taken into consideration, the control accuracy and controllability cannot be sufficiently improved. Further, in the third prior art example, when the difference between the average value of the air-fuel ratio feedback correction amount and the set value is a predetermined value or more during the engine steady operation state, the relationship between the air flow sensor output data and the intake air amount is uniquely set. I am just fixing
An appropriate correction value cannot be obtained for each engine operating state, and
As in the first and second prior art examples, the change in characteristics of the injector (actuator) such as deterioration with time is not taken into consideration, and therefore there is a limit in control accuracy and controllability. The present invention has been made in view of the above circumstances, and accurately changes the characteristics of an actuator as an engine control target and the characteristics of a sensor that provides information for calculating the control amount of the actuator, depending on the engine operating state. By learning and using the learning value obtained by this learning in the control amount calculation, it is possible to reliably correct the characteristic changes of the actuator and the sensor, and to improve the control accuracy of the engine control. It is an object to provide a learning control method.

[Means for Solving the Problems]

In order to achieve the above object, the learning control method for an automobile engine according to the present invention determines whether the engine is in a steady operating state based on the engine operating state, and when it is determined to be the steady operating state, O ₂
Information based on the sensor output is taken as a learning value into a table constituted by engine control specifications, and in the learning control method for an automobile engine using the learning value as a control variable for engine operation control, the basic control amount of the actuator is calculated. The information based on the output of the O ₂ sensor is taken in as a first learning value according to the engine load in the first learning value table with the engine load as a parameter, and information is given for calculating the control amount of the actuator. In the second learning value table using the sensor output as a parameter,
It is characterized in that information based on the O ₂ sensor output is taken in as a second learning value according to the sensor output, and the first learning value and the second learning value are used for the control amount calculation of the actuator. .

[Action]

In the present invention, when it is determined that the engine is in a steady operation state,
The information based on the O ₂ sensor output is used as a first learning value table having the engine load indicating the basic control amount of the actuator as a parameter, and the sensor output giving information for calculating the control amount of the actuator as a parameter. The learning value table is loaded as the first learning value and the second learning value, respectively. Then, both learning values are used for calculating the control amount of the actuator. Therefore, the characteristic changes of the actuator and the sensor are appropriately corrected when calculating the control amount of the actuator as the engine control target.

【Example】

An embodiment in which the learning control method for an automobile engine according to the present invention is applied to air-fuel ratio control will be specifically described below with reference to the drawings. FIG. 1 shows a schematic view of the entire control system, and reference numeral 1 in the drawing is an engine body. In this engine, the air introduced from the air cleaner 2 is mixed with the fuel injected from the injector 4 in the throttle body 3, and then the mixture is introduced into the intake system via the throttle valve 5. In the exhaust system, exhaust gas purification measures are taken so that harmful components in the exhaust gas are removed in the exhaust gas reactor (three-way catalytic converter) 6. Part of the exhaust gas from the exhaust system is the EGR valve 7
Is recirculated to the intake system through the EGR valve 7, and the EGR valve 7 is opened and closed by a valve 8 provided in a negative pressure pipe communicating with the intake passage.
The opening / closing operation is performed depending on the presence or absence of the negative pressure applied to the diaphragm inside the valve 7 via the negative pressure pipe. Fuel is supplied to the injector 4 from a fuel tank 9 by a fuel pump 10 via a filter 13 and a pressure regulator 11. The fuel pump 10 to the injector 4
A fuel damper 12 is provided in the fuel supply path leading to. Further, an idle speed control valve 14 is provided in a bypass communicating with the throttle body 3 upstream and downstream of the throttle valve 5 to control the engine speed during idling. Further, in FIG. 1, reference numeral 15 is a control device composed of a microcomputer.
In the exhaust system, a voltage signal from an O ₂ sensor 16 installed in the preceding stage of the exhaust gas reactor 6, an electric signal measuring the air flow rate from an air flow meter 17 installed in the intake passage 2a, and a throttle installed in the throttle valve 5 The sensor 18 gives a voltage signal corresponding to the throttle opening, and the engine 1 gives a water temperature sensor 19 an electric signal about the water temperature. Further, the control device 15 is provided with a crank angle reference position detection signal and a pulse signal for each 1 degree of crank angle by a crank angle sensor 21 provided in the distributor 20, and a neutral position switching signal from the transmission 22. Starter switching signals are supplied from the starter 23, respectively. In FIG. 1, reference numeral 24 is a battery, 25 is an injector relay, and 26 is a fuel pump relay. Further, the control device 15 includes a microprocessor unit (hereinafter referred to as MPU) 27 as shown in FIG.
It is connected to ROM 29, RAM 30 and RAM 31 with backup via 8. Also, the above O ₂ sensor 16, air flow meter 1
7, The analog signal from the throttle sensor 18, etc. is digitally converted via the A / D converter 32, and then the MPU27 via the bus 28.
Brought to. Other signals are input to the MPU 27 through the I / O port 33. In the description of the present invention, what is stored in the table is referred to as a learning value, and what is read out by performing interpolation calculation and applied to the equation (2) is referred to as a learning control coefficient. In the system of the present invention, the learning value stored in the learning value table is rewritten with the data calculated during the steady operation period of the engine operation. Therefore, it is necessary to detect the engine steady operation state (stable state). The determination of the engine steady operation state in the system is determined by the engine load and the continuous state of the engine speed. As described above, the engine load uses the intake air amount Q / N per one revolution, that is, the basic fuel injection width Tp indicating the basic control amount of the injector 4 as the actuator. The upper part (a) of FIG. 3 shows a matrix for detecting the engine steady operation state, for example, 5 lines and 5 lines.
It consists of 16 sections, which are divided by a line of steps. The magnitude of the engine load is set at five points L ₀ to L _{4 on} the X axis, and the magnitude of the engine speed is set at five points N ₀ to N _{4 on} the Y axis. Therefore, the engine load is L ₀ L ₁ ,
It is divided into four ranges, L ₁ L ₂ , L ₂ L ₃ , and L ₃ L ₄ , and the engine speed is also divided into four ranges. On the other hand, as shown in FIG. 7 (a), the output voltage of the O ₂ sensor changes cyclically through the reference voltage indicating the stoichiometric air-fuel ratio according to the rich and lean states of the air-fuel ratio. In the system, when the output voltage of the O ₂ sensor repeats the rich and lean cycles three times in one of the 16 compartments in the matrix, the engine is judged to be in the steady operation state. When the engine steady operation state is determined in this manner, the learning value is taken into the learning value table. First, the learning value table will be described. The learning value table of the conventional method is composed of the number of revolutions and the load. For example, the number of revolutions and the load are each divided into 4 to provide 4 × 4 = 16 divided areas (addresses). The learning value is fetched to the corresponding address inside and the previous learning value is rewritten. However,
In this way, the time required for all learning to be performed at least once for each divided area becomes considerable. That is, when the matrix of the four-division area in the number of revolutions and the matrix of the four-division area in the load are filled with the learning values, in the steady operation state, for example, learning probability at low load / low rotation (idling state), high load / high rotation (high speed) The learning probability in the running state) is very high, but the learning probability in the low load / high rotation region should be close to zero, and the learning probability in the high load / low rotation region is similar. Therefore, when a learning probability of 70% or more is plotted, it will be in the form of, for example, FIG. 5 (a) or (b). Also, each time,
There will be areas where learning will be delayed depending on operating conditions and conditions. While these remain, the learned values of the matrix have variations and cannot be used for control. Therefore, in this method, as a learning value table, the load (basic fuel injection width Tp) is used as a parameter in the RAM 31 with backup as shown in the lower side of FIG. 3, that is, L ₀ L ₁ , L ₁ L ₂ ,
A learning control coefficient Ka for correcting the characteristic change of the injector 4 as an actuator for each divided region of L ₂ L ₃ and L ₃ L _4.
For learning the basic fuel injection width Tp (fuel injection width
Ti) In order to correct the characteristic change of the air flow meter 17 as a sensor that gives information for calculation, that is, the intake air amount Q, the air flow meter output voltage is used as a parameter to have four divided regions, and the characteristics of the air flow meter are corrected. the second learning value table to obtain the learning control coefficient Qa for providing addresses a _1, a for the respective table
_2, a _3, a ₄ and b _1, b _2, b _3, first the appropriate address in the b _4, and a second capture learning value is rewritten learning value. Here, the rotational speed is in any region ( N ₀ N ₁ , N ₁ N ₂ , N
₂ N ₃ , N ₃ N ₄ ), the first learning value is stored in a ₁ , a ₂ , a ₃ , a ₄ in correspondence with the load division area, and the change of the control characteristic of the air flow meter 17 That is, the second learning value for correcting the intake air amount Q is stored in b ₁ , b ₂ , b ₃ , b ₄ corresponding to the divided regions of the output voltage of the air flow meter 17. The first, second learned value _{(a 1, a 2, a} 3, a 4 and _{b 1, b 2, b 3} , b 4 to the memory by content) is immediately read out in accordance with the operation state As a control variable, in the MPU 27, the learning control coefficient Ka based on the first learning value is incorporated into the arithmetic expression of the fuel injection pulse width Ti in the above equation (2), and the learning control coefficient Qa based on the second learning value is calculated. It is used as a correction coefficient for the intake air amount Q in the equation (1). Here, the rotation speed does not participate in the actual air-fuel ratio control. However, it is unlikely that the accuracy of the air-fuel ratio control will be so lowered. That is, as described above, the learning probability of the conventional learning value table is very low, and in the case of this method, the first learning value stored in the first learning value table for the same load is There is one for each of the regions L ₀ L ₁ , L ₁ L ₂ , L ₂ L ₃ and L ₃ L ₄ , but under the condition that rewriting is done sequentially and in the steady operation, the control value in the close rotation region is approximate Considering the points to be taken, it is considered that the learning value is sufficiently maintained to be practical. The same applies to the second learning value table corresponding to the divided region of the air flow meter output voltage. As described above, each value in each table is "1" before the first driving of the vehicle. To explain the learning value rewriting, when the steady state of the engine operation is detected, the first learning value table and the second learning value table are rewritten with the values related to the feedback signal from the O ₂ sensor. The first rewriting is performed by the arithmetic mean A of the maximum value and the minimum value in one cycle of integration, such as the values of Imax and Imin in FIG. 7 (b). After that, when α is not “1”, the learning value table is incremented or decremented by the minimum value ΔA (minimum resolution) that can be obtained by the computer. In other words, BC which is the value A of the learning value rewritten in the first learning
One bit is added or subtracted from the D code. Furthermore, in this electronic control method, RA
At the time of reading information from M31, an operation to supplement the unlearned area is performed. That is, when the learning value is taken in to the learning value table, a flag area for determining whether or not the information is taken in after the learning is started is provided for each divided area of the learning value table so that the information is taken in. When information is read for each area for control, if the flag is set, that information is used as the learning control coefficient, and if the flag is not set, the flag is set in the adjacent area. That is, the learning control coefficient is estimated and calculated by using the information obtained from the information. Eg 8
When constructing a learning table in a bit RAM, the table data is configured in bit units (in this case, the learning value resolution is 128), and the most significant 1 bit or the least significant 1
The bit is used as a flag as to whether learning has been performed, this 1 bit is cleared at the start of control, and set to 1 at the time of rewriting the first table value. Next, when the table is read, the bit is checked, and if the flag is set, its value is calculated. If it is not set, the value read from the adjacent table areas on the left and right is calculated by the interpolation calculation method Ka , Or Qa, and use it. If there is no adjacent table area or if the table area is in an unlearned state, the initial value of the area may be used for calculation. In reading from a general table, the learning value is stored in each divided area in the table, but the actual load value fluctuates freely between L ₀ L ₄ and this fluctuation However, if the number of divided regions is increased to increase the memory capacity, the linear interpolation method is used here for MPU2.
The learning control coefficient between each divided area is obtained by the calculation of 7. This linear interpolation method can also be applied to the interpolation calculation method that employs the data in the adjacent table areas described above. Now, as shown in FIG. 4, the first learning value stored in each region L ₀ L ₁ , L ₁ L ₂ , L ₂ L ₃ and L ₃ L ₄ in the first learning value table is y ₁ , y ₂ , y ₃ and y ₄
Assuming that the corresponding load values x ₁ , x ₂ , x ₃ and x ₄ of y ₁ , y ₂ , y ₃ and y ₄ are the midpoints of the respective regions, the value of the learning control coefficient Ka at the load x is It can be calculated from the first learning values y ₁ , y ₂ , y ₃ and y _{4 of} each region by the following equation. Now, assuming that the value of the load x is between x ₃ and x ₄ , the learning control coefficient Ka is Ka = {(x−x ₃ ) / (x ₄ −x ₃ )} × (y ₄ −y ₃ ). Calculate with + y ₃ . The broken line in FIG. 4 indicates the area dividing boundary line of the table. Here, if it is assumed that the first learning value is not yet filled in L ₂ L ₃ (the flag is not set), the first learning value of the adjacent L ₁ L ₂ instead of y ₃ Instead of the learning value y ₂ of the load value x ₃ and the load value x ₃ , the adjacent load value x ₂ can be used instead to perform interpolation calculation. By learning such air-fuel ratio control, for example, operation in an unstable state of the O ₂ feedback signal from the O ₂ sensor 16 (throttle fully open region, O ₂ sensor 16 inactive region)
Also, it is possible to control by analogy using the table value. Further, in the second learning value table corresponding to the divided region of the output voltage of the air flow meter (or the suction pipe pressure sensor), the O ₂ in each region in the grid is also changed.
The acquisition of information from the sensor 16 is realized on the basis of the above-mentioned determination condition, and the matrix area can be used in the second area in the RAM area.
A memory area (table) is secured as a learning value table of and written here. As described above, the first learning value written in the first learning value table with the engine load, that is, the basic fuel injection width Tp indicating the basic control amount of the injector 4 as the actuator as a parameter is The fuel injection width Ti is read by the above equation (2) as the learning control coefficient Ka.
By being used for the calculation of, the characteristic change of the injector 4, such as deterioration over time, is corrected. That is, if the injector 4 deteriorates with time, the lift amount of the injection nozzle valve body decreases, and even if a drive signal of the same fuel injection width Ti is output from the control device 15 due to carbon adhering to the injection nozzle or the like, compared to when the injector is new. fuel injection amount reduced Te, the air-fuel ratio is lean, this O ₂ sensor 16
The learning value is detected based on this information and is learned to correct the air-fuel ratio to the rich side.
A value larger than 1.0), and a basic fuel injection width Tp indicating the basic control amount of the injector 4 as the engine load.
Is taken in as a first learning value in accordance with Tp (load) as an engine operating state in a first learning value table using as a parameter, and the learning control coefficient Ka based on this first learning value is calculated when calculating the fuel injection width Ti. By using the injector 4
The characteristic change due to deterioration with time is corrected. Further, when the air flow meter 17 as a sensor deteriorates, its output value decreases, and even if the amount of air passing through the air flow meter 17 is the same, it is detected as a smaller intake air amount Q than when it was new. This intake air amount Q
Is used to calculate the basic fuel injection width Tp according to the equation (1), and the basic fuel injection width Tp is reduced by Tp = K × Q / N, and the fuel injection width Ti is also reduced accordingly, and the air-fuel ratio is reduced. Becomes lean. This is detected by the O ₂ sensor 16, and the learning value is learned based on the information to correct the air-fuel ratio to the rich side. Therefore, the second learning value table using the intake air amount Q by the air flow meter output value as a parameter. The second learning value learns the characteristic change due to the deterioration with time of the air flow meter by being taken in as the second learning value. Therefore, according to the intake air amount Q by the air flow meter 17, the second learning value is read out from the second learning value table with interpolation calculation and used as the learning control coefficient Qa for correcting the intake air amount Q. It is possible to correct the characteristic change of 17 due to deterioration over time. Further, if the characteristic change of the air flow meter itself is learned from the information from the O ₂ sensor 16, even if the air flow meter 17 deteriorates with time and changes the characteristic, the second learned value is set using the air flow meter output voltage as a parameter. The intake air amount Q is corrected by the learning value control coefficient Qa set with interpolation calculation from the table to not only optimize the air-fuel ratio control but also control the ignition timing using the intake air amount Q from the air flow meter output. In this case, the ignition timing is properly controlled, and a control error such as knocking can be avoided. Next, an example of a learning value and writing program executed by the MPU 27 will be specifically described with reference to the flowchart of FIG. The learning program has a preset time (40
every ms). The engine speed is detected in step 1. If the engine speed is in the controlled range N ₀ and N ₄
If it is in the range between and, the program proceeds to step 2. If the engine speed is out of range, the program jumps from step 1 to EXIT and exits the routine. In step 2, the position of the row containing the detected engine speed in the matrix of FIG. 3 is detected, and the position is stored in the RAM 30. After that, the program proceeds to step 3 and the engine load (basic fuel injection width Tp) is read. If the engine load is within the controlled range L ₀ to L ₄ , the program proceeds to step 4. If the engine load is out of range, the program exits the routine. The column position associated with the detected engine load is then detected in the matrix and that position is stored in RAM 30. Then, the position of the section relating to the engine operating conditions depending on the engine speed and the engine load is determined in the matrix, for example, section D ₁ in FIG. The program proceeds to step 5 and the determined partition position is compared with the partition determined in the previous learning. However, since the first learning is not comparable, the program exits the routine through steps 7 and 11. In step 7 of the first learning, the partition position is RAM30.
Will be stored in. In the learning after the first learning, the detected position is compared with the previously stored partition position in step 5. If the partition position in the matrix is the same as the previous one, the program proceeds to step 6 and the output voltage of the O ₂ sensor is detected. If the output voltage alternates between rich and lean and there is a sign conversion, the program proceeds to step 8 and, if not, the program exits the routine. In step 8, the counter counts the number of rich and lean cycles of the output voltage.
In step 9, if the counter has counted, for example three times, the program proceeds to step 10. If the count has not reached 3, the program exits the routine. In step 10, the counter is cleared and the program steps
Go to 12. On the other hand, if the partition position is not the same as the previous learning in step 5, the program proceeds to step 7 and the old data at the partition position is rewritten with new data. In step 11, the count is cleared. In step 12, the arithmetic mean A of the maximum value and the minimum value of the integrated value of the output voltage of the O ₂ sensor is calculated for, for example, 3 cycles of the output waveform, and the value A is stored in the work area of the RAM 30. Thereafter, the program proceeds to step 13, the first learning value table, for example compartments D ₁ as shown in FIG. 3
An address corresponding to the position of the partition is detected, such as an address a ₂ corresponding to. In step 14, it is detected whether the detected address is flagged. In the first learning, the address is not flagged, so
The program proceeds to step 15. In step 15, the third
The first learned value in the address of the first learned value table in the figure is rewritten with half of the new arithmetic mean value A obtained in step 12, and at the same time, the address is flagged. In the learning after the first data rewrite, if the address detected in the above process is the same as the last rewritten address (the address is flagged), the program proceeds from step 14 to step 16 and the correction Coefficient α
Value (integrated value of O ₂ sensor output) is compared with 1. If the value of α is greater than 1, the program will step
Proceeding to 17, the minimum unit ΔA (1 bit) is added to the first learned value in the associated address. If the value of α is not greater than 1, the program proceeds to step 18 to determine if the value of α is less than 1. If the value of α is less than 1, then in step 19, the minimum unit ΔA is subtracted from the first learned value. If the value of α is less than 1, it means that the value of α is 1 and the program exits the rewrite routine. The program proceeds from step 15, step 17 or step 19 to step 20 to identify the address (b _{1 to} b _{4 in} FIG. 3 (c)) of the second learning value table corresponding to the air flow meter output voltage, This operation is performed in the same manner as in step 13 for the learning value table of No. 2. Thus, the same program as described above rewrites the second learning value corresponding to the airflow meter output voltage in the second learning value table in step 21, step 24, step 25, step 26, step 27 and step 28, and the airflow meter 17
Is performed to correct the control characteristics of the. Then, when the desired injection pulse width Ti is calculated, the learning control coefficient Ka and the learning control coefficient Qa for correcting the characteristics of the air flow meter are the first values corresponding to the value of the engine load L.
Of the learning value table and the second learning value table corresponding to the value of the air flow meter output voltage. When the values of the coefficients Ka and Qa are in the middle of the tables, the interpolation calculation of the learning value table is performed as already described in FIG. The learning control coefficient Qa read from the second learning table based on the output voltage of the air flow meter 17 is used as a correction coefficient for the intake air amount Q in the above equation (1), and based on the engine load (basic fuel injection width Tp). The learning control coefficient Ka read out from the first learning table is used for the calculation of the fuel injection width Ti according to the equation (2), and is used for controlling the injector 4.

【The invention's effect】

As described above in detail, according to the present invention, the information based on the output of the O ₂ sensor is used for the first learning value table having the engine load indicating the basic control amount of the actuator as a parameter and the control amount calculation of the actuator. Are taken into the second learning value table having the sensor output that gives information as a parameter, as the first learning value and the second learning value, respectively.
Since both learning values are used for the control amount calculation of the actuator, the characteristic changes of the actuator and the sensor are accurately learned according to the engine operating state, and the characteristic changes of the sensor and the actuator due to deterioration over time during the control amount calculation. Is properly corrected according to the engine operating state and reflected in the engine control, and the control accuracy and controllability can be sufficiently improved.

[Brief description of drawings]

FIG. 1 is a schematic diagram of an engine control system that employs a learning control method for an automobile engine according to the present invention, FIG. 2 is a schematic configuration diagram of a control device, and FIG. 3 is a region determination matrix and first and second matrixes. FIG. 4 is a diagram showing a learning value table in parallel, FIG. 4 is a diagram visually showing an interpolation calculation method, FIG. 5 is an explanatory diagram for explaining information input probability to a matrix, and FIG. 6 is a book. FIG. 7 is a flowchart showing an example of an electronic control method for an automobile engine of the invention, FIG. 7 (a) is a diagram showing an output voltage of an O ₂ sensor, and FIG. 7 (b) is a diagram showing an output voltage of an integrator. 1 ... Engine, 2 ... Air cleaner, 3 ... Throttle body, 4 ... Injector, 5 ... Throttle valve, 6 ... Exhaust gas reactor, 7 ... EGR valve, 8 ... Valve, 9 ... Fuel Tank, 10 ... Fuel pump, 11 ... Pressure regulator, 12 ... Fuel damper, 13 ... Filter, 14 ... Idle speed control valve, 15 ... Control device, 16 ... O ₂ sensor, 17 ... Air flow Meter, 18 ... Throttle sensor, 19 ... Water temperature sensor, 20 ... Distributor, 21 ... Crank angle sensor, 22 ... Transmission, 23 ... Starter, 24 ... Battery, 25 ... Injector relay, 26 ... Fuel pump relay, 27 …… MPU, 28 …… bus, 29 …… ROM, 30,31 …… RAM, 32 …… A / D converter, 33 …… I / O port.

Claims (1)

(57) [Claims] 1. A steady operating state is determined based on the engine operating state, and when it is determined to be the steady operating state, information based on the O ₂ sensor output is taken as a learning value into a table constituted by engine control specifications, In the learning control method for an automobile engine, which uses the learning value as a control variable for engine operation control, a first learning value table having a parameter of an engine load indicating a basic control amount of an actuator is used as a parameter based on the O ₂ sensor output. Information is taken in as a first learning value according to the engine load, and a second learning table using a sensor output as a parameter for giving information for the control amount calculation of the actuator is set as a second learning table.
It is characterized in that the information based on the O ₂ sensor output is taken in as a second learning value according to the sensor output, and the first learning value and the second learning value are used for the control amount calculation of the actuator. Learning control method for automobile engine.