DE19725567B4 - Fault diagnosis system for an air / fuel ratio control system - Google Patents

Abstract

Air / fuel ratio sensor fault diagnostic system, comprising:
An air / fuel ratio sensor (26) for generating a signal representing a detected air / fuel ratio λ of an internal combustion engine (1),
An air-fuel ratio correction coefficient generating means (42, P3, 103) for generating an air-fuel ratio correction coefficient (FAF) according to a difference between the detected air-fuel ratio λ and a target air-fuel Ratio λTG,
An air-fuel ratio change calculating means (42, 304, 314, 311) for calculating a change speed Δ ² SM of the detected air-fuel ratio λ,
A correction coefficient change calculating means (42, 307, 315, 312) for calculating a change rate Δ ² FAFSM of the air-fuel ratio correction coefficient FAF and
- a fault diagnosis device (42, 203 - 222, 300) for diagnosing a fault of the air-fuel ratio sensor (26) based on a comparison between the air-fuel ratio change speed Δ ² SM and the correction coefficient change speed Δ ² FAFSM.

Description

The The invention relates to an air / fuel ratio control system having a Air / fuel ratio sensor, which its output according to the air / fuel ratio of a Heat engine with internal combustion changes linearly, So that the Air / fuel ratio control on the basis of the detection result of this sensor an electronic control unit, which is a microprocessor or the like, executable is. More particularly, the invention relates to a fault diagnosis system for a such air / fuel ratio control system.

In an air-fuel ratio control system of recent years, an air-fuel ratio sensor (eg, a limit current type oxygen sensor) has been used to linearly adjust the air-fuel ratio of a heat engine according to the oxygen concentration in an exhaust gas capture. A microprocessor receives the detection result of the air-fuel ratio by the sensor to control the fuel injection amount to the internal combustion heat engine. In this case, the microprocessor calculates the air-fuel ratio correction coefficient based on the air-fuel ratio detection result from the air-fuel ratio sensor to correct the fuel injection amount with the air-fuel ratio correction coefficient. Accordingly, in the internal combustion heat engine, the optimum combustion is realized to reduce the content of toxic by-products (eg, CO, HC, NO _x , etc.) in the exhaust gas.

There in the air / fuel ratio control system on the other hand the reliability by the air / fuel ratio sensor decreases to be detected air / fuel ratio deteriorates significantly the control accuracy of the system. Therefore required The prior art discloses a method for performing highly accurate fault diagnosis the air / fuel ratio sensor. At this point, an error of the air / fuel ratio sensor or a control error due to the electronic control unit (eg. The microprocessor) as a failure of the air / fuel ratio control system Are defined.

The characteristic error of the sensor output will be described below with reference to FIG 23A described. In the normal sensor operation, the detected air / fuel ratio (ie, the detected λ value) and the actual air / fuel ratio (ie, the actual λ value) are substantially coincident, as by the solid line in FIG 23A is shown. At the sensor error timing, the detected λ value is smaller or larger than the actual λ value, as shown by the dashed lines. It can be assumed that this characteristic error due to aging, such. By electrode agglomeration or electrode separation or element breakage (eg breakage of a diffused resistive layer or solid electrolyte layer) or by conduction or insulation failure).

The error by which the response of the sensor output decreases becomes with reference to 23B described. As the response of the sensor output decreases, the delay in the transition of the detected λ value from that of the actual λ value occurs as the fuel injection quantity changes from the lean side to the rich side, as shown , It is believed that this reduction in response due to the occlusion of a cover or a porous electrode layer due to the contamination of the sensor or due to insufficient activity due to a Heizelementstörung occurs.

As described above Problems related to the air / fuel ratio sensor output be caused by a number of factors, and a procedure to run a highly accurate fault diagnosis of the air / fuel ratio sensor is sought. Finally It should be noted that a System for diagnosing the sensor error based on the behavior the detected air / fuel ratio through the air / fuel ratio sensor or the behavior of the value FAF has been developed.

at the regulation of the air / fuel ratio is in general e.g. the vaporized fuel (or vaporized gasoline) in the Fuel tank to the heat engine intake system and the value FAF may vary erroneously or in a transitional period deviate from a set value (e.g., a reference = 1.0). Even if, furthermore, a learned air / fuel ratio value deleted is, immediately after the battery was changed, can Error of value FAF according to the peculiarity of Heat engine occur. Consequently, if the diagnosis of the present sensor error is executed at different times of the value FAF, there is a problem that a faulty diagnosis can take place.

Furthermore, documents describing the prior art, such as the patent specification DE 3710154 C2 A failure diagnosis method for detecting a failure of a connection system in the sensor for determining the oxygen concentration of the limit current type according to an applied voltage and a detected current.

Of the The prior art described above may fail in the art Circuitry such as e.g. a break or short circuit of the Connection system, capture. However, when the air / fuel ratio decreases, is the sign of a resulting error not detectable. In other words, it's impossible the accuracy of the detected by the air / fuel ratio sensor Air / fuel ratio to determine (if the sensor output is normal or abnormal), i. e. the reliability of Determine sensor output. Further, it is impossible to reliability of the electronic air-fuel ratio control system using the To detect the detection result of the air / fuel ratio sensor.

Another prior art form the publications DE 4301968 A1 . EP 0652358A2 and EP 0549566 B1 ,

The publication DE 4301968 A1 is directed to a fault diagnosis method and a fault diagnosis apparatus in stereo ram control. The fault diagnosis method uses codes for errors of the two oxygen probes of the stereo lambda control. If an error is detected in one of the two probes, it is examined whether there is also an error of the other probe. If so, a combination error indicating this case is accordingly output. It can also be indicated by the combination error code that only one of the two probes has an error.

The publication EP 0652358 A2 deals with a method for checking the functionality of lambda probes, in particular with the dynamic functionality of lambda probes, namely by measuring the residence times, within which the lambda probe signal indicates a rich or lean mixture, and by comparing these times with operating point dependent stored limits. Thus, for example, the lambda probe is classified as functioning if both the fat and lean residence times are less than predetermined limit values assigned to the individual residence times.

Finally, the publication EP 0549566 B1 to an apparatus for detecting an abnormality of an oxygen sensor and controlling the air-fuel mixture ratio. An abnormality of the oxygen sensors is detected by an abnormality detecting means when an oxygen concentration signal is outside normal values, the abnormality means measuring the maximum and minimum of the open loop oxygen concentration signal, switching the air / fuel mixture ratio between lean and rich. At this time, the abnormality detecting means determines that the oxygen sensor is abnormal when at least one of the maximum and minimum values is within a range between predetermined first and second threshold values.

The The present invention has been made in view of the above-mentioned problems of the prior art, and an object of the invention is therein, a fault diagnosis system for an air-fuel ratio control system which can perform a highly accurate diagnosis of a fault of the air / fuel ratio sensor.

The The above object is achieved by the air / fuel ratio sensor fault diagnosis system according to claim 1 solved.

Further advantageous embodiments are the subject of the dependent claims.

Accordingly, the air-fuel ratio sensor failure diagnosis system of the present invention includes, among other things, an air-fuel ratio sensor that linearly changes an output with respect to the air-fuel ratio of an internal combustion engine, and the air / Fuel ratio, as detected by the air / fuel ratio sensor, is supplied to an electronic control unit having a microprocessor or the like. The electronic control unit determines an air-fuel ratio correction coefficient according to the difference between the detected air / fuel ratio and a target air / fuel ratio and corrects the amount of fuel to be supplied to the internal combustion heat engine to the air / fuel ratio control (as air / fuel ratio correction coefficient determining means and air / fuel ratio control means).

In particular, according to one aspect of the invention, the above-described object is to achieve improved sensor error detection using the CPU 42 is satisfied to compare the ratio between a sum value of the second derivative of the air-fuel ratio and a sum value of the second derivative of the correction coefficient with a predetermined error decision value to perform the fault diagnosis.

The Control systems according to the state of technology the proper air / fuel ratio control do not continue if an error as a result of injury or the break of the air / fuel ratio sensor occurs or when an error of controllability by the microprocessor occurs. According to one Example of the present invention will therefore be the error of Air / fuel ratio sensor and the control error of the microprocessor together as a fault the air / fuel ratio control system defined so that the error can be detected with high precision. this happens using the CPU, not just the difference between the Air / fuel ratio and calculate the desired air / fuel ratio, but also the difference between the air / fuel ratio correction coefficient and its average value.

Other Aspects and features of the present invention are disclosed in US Pat Course of the description below.

SUMMARY THE DRAWINGS

Further Advantages of the present invention will become apparent from the following detailed Description of the preferred embodiments with reference to the attached Drawings clearly.

1 13 shows the overall structure of an air-fuel ratio control system of an internal combustion engine according to a first preferred embodiment of the present invention;

2 FIG. 12 is a cross-sectional view of an air-fuel ratio sensor according to the first embodiment; FIG.

3 shows a graph of the voltage-current characteristic of the air / fuel ratio sensor in 2 .

4 10 is a block diagram for explaining the principle of an air-fuel ratio control system according to the first embodiment;

5 FIG. 12 is a flowchart of a fuel injection amount calculation routine according to the first embodiment; FIG.

6 and 7 show a flowchart of a sensor failure diagnostic routine in the first embodiment;

8th and 9 show a flowchart of a routine for calculating the sum values Σ | Δ ² λSM | and Σ | Δ ² FAFSM | in the first embodiment,

10 FIG. 10 is a graph showing a correlation between a failure diagnosis parameter and a sensor responsiveness in the first embodiment; FIG.

11A - 11H show graphs of the operations of the fault diagnosis,

12 shows a flowchart with the in the middle of the flowchart in 6 operations to be inserted,

13 shows a graph for setting an error decision value of a variable in the ers th embodiment,

14 FIG. 12 is a flowchart of a sensor failure diagnostic routine according to a second preferred embodiment of the present invention; FIG.

15 FIG. 12 is a flowchart of a sensor failure diagnostic routine according to a third preferred embodiment of the present invention; FIG.

16 FIG. 12 is a flowchart of a sensor failure diagnostic routine according to a fourth preferred embodiment of the present invention; FIG.

17 and 18 show a flowchart of a routine for calculating the sum values ΣΔλLR and ΣΔFAFLR,

19 shows a schedule from to the middle of 16 to be inserted according to a fifth preferred embodiment of the present invention,

20 10 is a graph for setting the error decision values δ1 and δ2 in the fifth embodiment;

21 11 is a graph for setting the error decision value α in other embodiments of the present invention;

22 11 is a graph for setting the error decision values δ1 and δ2 as variables in other embodiments;

23 FIG. 12 is a graph showing the states of the characteristic error and the reduction in the responsiveness of the air-fuel ratio sensor; FIG.

24 FIG. 12 is a flowchart of a main routine for deciding a failure of the air-fuel ratio control system of a non-claimed example; FIG.

25 FIG. 12 is a flowchart of a λ sum value calculation subroutine of the unclaimed example; FIG.

26 FIG. 12 is a flowchart of an FAF sum calculation subroutine of the unclaimed example; FIG.

27 FIG. 12 is a flowchart of an error decision subroutine of the unclaimed example; FIG.

28A and 28B show graphs of the change of the air-fuel ratio λ with respect to a target air-fuel ratio λTG and the change of an air-fuel ratio correction coefficient FAF with respect to an air-fuel ratio correction coefficient average value FAFAV of FIG unclaimed example,

29 FIG. 10 is a graph showing the areas defined by the threshold values A, B, C and D for deciding the normal state / error state of the λ sum value and the FAF sum value of the unclaimed example; FIG.

30A - 30H show graphs which more specifically illustrate the error decision action of the air-fuel ratio control system;

31 FIG. 12 is a flowchart of a threshold learning routine of the unclaimed example; FIG.

32 FIG. 12 is a flowchart of an error decision subroutine of another unclaimed example; FIG.

33 shows a normal state / error state decision graph which is divided into small regions according to the λ sum value and the FAF sum value, according to the other not bean example,

34 FIG. 10 is a graph showing a decision area learning routine of the further unclaimed example; FIG.

35A and 35B show normal state / error state decision graphs with the result obtained by the routine in FIG 34 is learned

36 FIG. 12 is a graph for determining the summation time duration of the values Tλ and TFAF according to the λ sum value and the FAF sum value of the further unclaimed example; and FIG

37 FIG. 10 is a graph for determining the number of error decisions according to the λ-sum value and the FAF-sum value of the further unclaimed example.

DESCRIPTION THE PREFERRED EMBODIMENTS

1 shows an internal combustion heat engine equipped with an air / fuel ratio control system and peripheral devices of the internal combustion heat engine. As 1 shows is a heat engine with internal combustion 1 a four-stroke heat engine with four cylinders and spark ignition. The intake air flows on the intake side through an air intake filter 2 , an intake pipe 3 , a throttle 4 , a tranquilizer 5 and an intake manifold 6 and comes with the of each fuel injector 7 injected fuel in the intake manifold 6 mixed, so that each cylinder is supplied with a mixture having a predetermined air / fuel ratio. A high voltage supplied by an ignition circuit 9 , is distributed and a spark plug 8th supplied to each cylinder of the heat engine with internal combustion 1 is arranged so that the spark plug 8th ignites the air / fuel mixture of each cylinder at a predetermined timing. The spent exhaust gas flows through an exhaust manifold 11 and an exhaust pipe 12 and is through a three-way catalyst 13 which is in the exhaust pipe 12 is cleaned of the toxic components (eg, CO, HC and NO _x ) and is discharged into the atmosphere.

The intake pipe 3 is with an intake air temperature sensor 21 and with an intake air pressure sensor 22 equipped, the temperature sensor 21 the temperature (ie, the intake air temperature Tam) of the intake air and the intake air pressure sensor 22 the pressure detected (ie, the intake air pressure PM) of the intake air downstream of the throttle 4 , The throttle 4 is with a throttle sensor 23 for detecting the opening degree (ie, the throttle opening TH) of the throttle valve 4 fitted. This throttle valve sensor 23 Not only outputs an analog signal according to the throttle opening TH, but also a detection signal indicating that the throttle valve 4 is substantially completely closed. The engine block of the heat engine with internal combustion 1 has a water temperature sensor 24 for detecting the temperature (that is, the cooling water temperature Thw) of the cooling water in the internal combustion heat engine 1 on. The distribution device 10 is with a speed sensor 25 for detecting the number of revolutions (ie, the engine speed Ne) of the internal combustion heat engine 1 fitted. This speed sensor outputs twenty-four pulse signals at an equal interval for every two revolutions, ie 720 degrees CA, of the internal combustion heat engine 1 out.

An air / fuel ratio sensor 26 formed by a limit current type oxygen sensor for outputting a linear air-fuel ratio signal over a wide range in proportion to the oxygen concentration in the exhaust gas emitted from the internal combustion heat engine 1 is discharged, is in the exhaust pipe 12 Access side of the three-way catalyst 13 arranged. A downstream O ₂ sensor 27 for outputting a voltage VOX2 according to the state whether the air-fuel ratio λ is "rich" or "lean" with respect to the stoichiometric air-fuel ratio (λ = 1), is on the downstream side of the three-way catalyst 13 arranged. The air / fuel ratio is expressed by the excess air ratio "λ", while the stoichiometric air / fuel ratio (= 14.7) is expressed as λ = 1.

Further, a container loaded with activated charcoal 52 for adsorbing the from a fuel tank 50 evaporated fuel via a fuel tank connection line 51 with the fuel, z. As gasoline, filled fuel tank 50 connected, wherein the fuel of the heat engine with internal combustion 1 is to be supplied. The container 52 is with a ventilation hole 53 provided for the admission of ambient air. The container 52 is via an exit line 54 on the outlet side of a throttle valve 4 with a suction pipe 3 connected so that the container 52 supplied vaporized gasoline in the connecting line section (access side of the intake manifold, not shown) of the intake manifold 3 is admitted.

A vacuum purge switching valve (VSV) 55 for controlling the amount of purge flow in accordance with one of an electronic control unit (ECU) 41 incoming control signal is in the middle of the outgoing line 54 arranged. In other words, the control signal is from the ECU 41 the vacuum flushing valve (VSV) 55 fed to thereby the connection between the container 52 and the intake pipe 2 over the departure line 54 so that fresh air from the atmosphere through the ventilation hole 53 is introduced. Because the fresh air in this way the inside of the container 52 refreshed and in the intake pipe 2 the heat engine with internal combustion 1 arrives; the container wins 52 its adsorption function again. The purge flow amount based on the amount of supplied fresh air at this time is changed by changing the duty ratio of one of the ECU 41 to the vacuum purge switching valve 55 controlled to be supplied pulse signal. In short, the degree of opening of this vacuum purge switching valve 55 is determined by the duty cycle signal based on the pulse width modulation from the electronic control unit 41 regulated to regulate the purge flow rate of the air, which the evaporated fuel from the container 52 having.

2 shows a cross-sectional view of the air / fuel ratio sensor 26 , In 2 indicates the air / fuel ratio sensor 26 a cover 31 , a sensor body 32 and a heater 33 on. The cover 31 generally has a U-shaped section, and its outer wall has a plurality of pores 31a on, which establish the connection between the inside and outside. The sensor body 32 generates a cutoff current corresponding to either the oxygen concentration in a "lean" air / fuel ratio range or the concentration of unburned gasoline (eg, CO, HC, and H ₂ ) in a "rich" range of the air / fuel ratio. ratio.

In the sensor body 32 is an exhaust-side electrode layer 36 firmly on the outer surface of a solid electrolyte layer 34 arranged, which is formed as a cup-shaped portion, and an outside air-side electrode layer 37 is fixed on the inner surface. An immigrated resistance layer 35 is on the outside of the exhaust-gas-side electrode layer by a plasma spray coating method 3b educated. The solid electrode layer 34 is made of a sintered substance of an oxygen ion conductive oxide, which is produced by solid dissolving CaO, MgO, Y ₂ O ₃ or Yb ₂ O ₃ as a stabilizer in ZrO ₂ , HfO ₂ , ThO ₂ or Bi ₂ O ₃ . The immigrated resistance layer 35 is made of a refractory inorganic substance, such as. Alumina, magnesia, quartzite, spinel or mullite. Both the exhaust side electrode layer 36 and the outside air-side electrode layer 37 are made of a noble metal which has a high catalytic activity, such as. As platinum, and are applied chemically on their surfaces with a porous material. The exhaust-gas-side electrode layer 36 has an area of about 10 to 100 mm ² and a thickness of about 0.5 to 2.0 μm, and the outside air side electrode layer 37 has an area of 10 mm ² or larger and a thickness of about 0.5 to 2.0 μm.

The heater 33 is in the outside air side electrode layer 37 arranged and heated the sensor body 32 (Including the outside air-side electrode layer 37 , the solid electrolyte layer 34 , the exhaust gas side electrode layer 36 and the immigrant resistance layer 35 ) by their heat energy. The heater 33 has sufficient heating power to the sensor body 32 to activate.

In the thus constructed air / fuel ratio sensor 26 generates the sensor body 32 According to the oxygen concentration, in a "leaner" region than the stoichiometric air / fuel ratio, a limiting current. In this case, the limiting current is the oxygen concentration through the surface of the exhaust gas side electrode layer 36 , the thickness of the immigrated resistance layer 35 , which determines porosity and average pore size. The sensor body 32 can detect the oxygen concentration linearly. However, since a high temperature of about 650 ° C or higher is required to the sensor body 32 to activate and there the activation temperature range of the sensor body 32 is relatively narrow, the activation range by heating can only with the exhaust gas of the internal combustion engine 1 not be regulated. Therefore, the heater will 33 by an ECU described below 41 regulated, so that the sensor body 32 is maintained at a predetermined activation temperature. Within a "rich" range with respect to the stoichiometric air / fuel ratio, the temperature of the unburned gas, such as carbon monoxide (CO), changes substantially linearly with respect to the air / fuel ratio, such that the sensor body 32 generates a limiting current according to the concentration of CO or the like.

The voltage-current characteristics of the sensor body 32 will be described below with reference to 3 described. The current-voltage characteristics in 3 show that in the solid electrolyte layer 34 whose sorkörpers 32 flowing current is proportional to the oxygen concentration (or the air / fuel ratio), which is determined by the air / fuel ratio sensor 26 and with respect to the solid electrolyte layer 34 applied voltage is linear. When the sensor body 32 at a temperature T = T1 in the activated state, a stable state is indicated by a characteristic line L1 represented by a solid line in FIG 3 , In this case, the straight portions of the characteristic line L1 which are parallel to a voltage axis V indicate the limit currents of the sensor body 32 , The change in the limit current corresponds to the change in the air / fuel ratio (ie, "lean" or "rich") so that the cutoff current increases as the air / fuel ratio shifts to the "lean" side and decreases when the air / fuel ratio shifts to the "rich" side.

The range of the voltage-current characteristic in which the voltage is lower than the levels corresponding to the straight portions, in parallel with the voltage axis V, is a resistance-determined range. The slope of the characteristic L1 within the resistance-determined region is determined by the internal resistance of the solid electrolyte layer 34 in the sensor body 32 expressed. Since the internal resistance of the solid electrolyte layer 34 changes with temperature, so that the resistance increases when the temperature of the sensor body 32 decreases, the increase is lower. When the temperature T of the sensor body 32 T ₂ is lower than T ₁ , the current-voltage characteristic is defined by a characteristic L2 as indicated by the broken line in FIG 3 is shown. The straight portions of the characteristic L2 parallel to the voltage axis V specify the limit currents of the sensor body 32 at T = T ₂ . The limit currents are substantially equal to those determined by the characteristic L1.

As the characteristic line L1 continues to show, the through the sensor body 32 flowing current a limit current Ipos (as in point Pa in 3 when a positive voltage Vpos is applied to the solid electrolyte layer 34 of the sensor body 32 is created. When a negative voltage Vneg to the solid electrolyte layer 34 of the sensor body 32 is applied, on the other hand, by the sensor body 32 flowing current does not depend on the oxygen concentration, but only proportional to the temperature (as in point Pb in 3 designated).

An electronic control unit ("ECU") for controlling the operation of the internal combustion heat engine 1 in 1 has a logic operation circuit with a CPU (Central Processing Unit) 42 , a ROM (read-only memory) 43 , a RAM (Random Access Memory) 44 and a backup RAM 45 on. The CPU 41 is by a bus 48 with an input port 46 for inputting the detection signals of the above-mentioned individual sensors and having an output port 47 connected to output the control signals to the individual actuators. The ECU 41 takes the intake air temperature Tam, the intake air pressure PM, the throttle opening TH, the cooling water temperature Thw, the engine speed Ne, the air / fuel ratio signal, etc. from the individual sensors through the input port 46 on. On the basis of these values, the ECU calculates 41 the control signals relating to a fuel injection amount TAU, a timing of the ignition Ig, etc., and outputs these control signals through the output port 47 to the fuel injection valve 7 , the ignition switch 9 etc. off. The ECU 41 executes the error decision process described below to diagnose whether or not there is an error of the air-fuel ratio feedback control system. If there is an error, the ECU will switch 41 a warning light 49 to alert the driver to the occurrence of the error.

A method which has been developed to carry out the air-fuel ratio control in the aforementioned fuel injection control system will be described below. The following procedure is in the document EP 0312835 81 been described.

(1) Modeling of the control object

A self-repeating first-degree moving average model having a dead time P = 3 is used to control the air-fuel ratio λ of the internal combustion heat engine 1 is used, and an external disturbance d is taken into account for the approximation.

The model of the system using the self-repetitive moving average model for controlling the air-fuel ratio λ can be approximated by the equation (1): λ (k) = a * λ (k-1) + b * FAF (k-3) (1) where FAF is an air / fuel ratio correction coefficient, a and b denote coefficients for determining the responsiveness of the model, and k is the number of control operations that after the Start the first scan to be performed.

Taking into account the external disturbance d, the control system model can be approximated by equation (2): l (k) = a * l (k-1) + b * FAF (k-3) + d (k-1) (2)

For that up this way approximated Model it is easy to use the model constants a and b, d. H. a transfer function G of the system for controlling the air / fuel ratio λ by discrete Scanning in one revolution (360 ° CA) using a step response.

(2) Procedure for indication the state set of variables X (where X: vector set)

und X1(k + 1) = aX1(k) + bX2(k) + d(k) = λ(k + 1) (4a) X2(k + 1) = FAF(k – 2) (4b) X3(k + 1) = FAF(k – 1) (4c) X4(k + 1) = FAF(k) (4d) Using a state variable set X (k) = [X1 (k), X2 (k), X3 (k), X4 (k)] ^T , equation (2) can be rewritten into the matrix equation (3) and then into the equations (4a) - (4d). Here T represents a fallen matrix.

and X1 (k + 1) = aX1 (k) + bX2 (k) + d (k) = λ (k + 1) (4a) X2 (k + 1) = FAF (k-2) (4b) X3 (k + 1) = FAF (k-1) (4c) X4 (k + 1) = FAF (k) (4d)

(3) Structure of the controller

When a controller is constructed based on the equations (3) and (4a) - (4d), the air-fuel ratio correction coefficient FAF can be expressed by the equation (5) using the optimum feedback gain K = [K1, K2, K3, K4] and a state variable set X ^T (k) = [1 (k), FAF (k-3), FAF (k-2), FAF (k-1)]. FAF (k) = K x T (k) = K1 * (k) + K2 * FAF (k-3) + K3 * FAF (k-2) + K4 * FAF (k-2) (5)

By adding an integration expression ZI (k) to the equation (5) for absorbing errors, the air-fuel ratio correction coefficient FAF can be expressed by the equation (6). FAF (k) = K1 * (k) + K2 * FAF (k-3) + K3 * FAF (k-2) + K4 * FAF (k-2) + ZI (k) (6)

The integration term ZI (k) is a value determined by a difference between a target air-fuel ratio λTG and an actual air-fuel ratio λ (k) and by an integration constant Ka, as represented by equation (7) : ZI (k) = ZI (k-1) + Ka · (λTG -λ (k)) (7)

4 FIG. 12 is a block diagram of an air-fuel ratio λ control system for which the above-mentioned model has been designed. As in 4 ¹ , the model uses a Z ^-1 transformation to derive an air-fuel ratio correction coefficient FAF (k) from FAF (k-1). For this operation, the previous air-fuel ratio correction coefficient FAF (k-1) in the RAM 44 stored and then read at the following timing of the controller. The expression "FAF (k-1)" represents the last air-fuel ratio correction coefficient. The term "FAF (k-2)" represents the air-fuel ratio correction coefficient corresponding to the term FAF (k-1). immediately before, and the expression "FAF (k-3)" represents the air-fuel ratio correction coefficient which immediately precedes the expression "FAF (k-2)".

The block P1, which in 4 is enclosed by the dot-and-dash line, shows a state variable amount determining section X (k) while subjecting the air-fuel ratio λ (k) to feedback control to the target air-fuel ratio λTG. The block P2 indicates a section (or summation section) for determining the integration expression ZI (k). The block P3 indicates a section for calculating the present air-fuel ratio correction coefficient FAF (k) based on the state variable amount X (k) as determined by the block P1 and the integration term ZI (k) as determined by the block Block P2 is determined.

(4) Determination of the optimal Feedback gain K and the integration constant Ka

The optimal feedback gain K and the integration constant Ka may, for. By minimizing a weighting function J, as represented by equation (8):

In However, equation (8) is intended that the weighting function J the behavior of the air-fuel ratio correction coefficient FAF (k) limited and the difference between the air / fuel ratio λ (k) and the Target air / fuel ratio λTG minimized. The weighting of the restriction the air-fuel ratio correction coefficient FAF (k) can by change the values of the weight parameters Q and R are tuned. Consequently can the optimal feedback gain K and the integration constant Ka through repeated simulations with changes the weight parameters Q and R until the optimum control characteristic is obtained will be determined.

The optimal feedback gain K and the integration constant Ka are also from the model constants a and b dependent. To provide sufficient stability (or robustness) of the system despite the variations (the parameters) of the system for controlling the actual air-fuel ratio λ, is the change the model constants a and b to estimate the optimal feedback gain K and to determine the integration constant Ka. Thus, the simulation becomes considering the actual potential change the model constants a and b carried out, thereby optimal Feedback gain K and the integration constant Ka to determine which one is sufficient stability guarantee.

The system described above assumes that (1) the model of the controlled object, (2) the method for indicating the state variable set, (3) the design of the controller, and (4) the determination of the optimal feedback gain and the integration constant already exist. The ECU 41 Also, assume that only equations (6) and (7) are used to perform the air / fuel ratio control in the fuel injection control system.

5 shows a flowchart of the by the CPU 42 in the ECU 41 to be executed fuel injection amount calculation routine. This routine becomes synchronous with the rotations of the internal combustion heat engine 1 executed, ie at each 360 ° CA. The in 5 The process shown corresponds to the air / fuel ratio correction coefficient determination means and the air / fuel ratio control means.

The CPU 42 calculated in step 101 a basic fuel injection amount TP based on the intake air pressure PM, the engine speed Ne, etc., and decides in the step 102 Whether the conditions for the return of the air / fuel ratio λ have been satisfied or not. As is well known, the recirculation conditions are satisfied when the cooling water temperature Thw is not lower than a predetermined value and when the engine operation is not performed in a high-speed region or in a high-load region. If the feedback conditions are met, the CPU goes 42 continue to step 103 in which the air-fuel ratio correction coefficient FAF for converting the air-fuel ratio λ into the target air-fuel ratio λTG (for the stoichiometric air-fuel ratio λ = 1) is determined, and then goes to the step 104 , More precisely, in step 103 is calculated using the equations (6) and (7) based on the target air-fuel ratio λTG and the air-fuel ratio λ (k), which by the air / fuel ratio sensor 26 is detected, the air-fuel ratio correction coefficient FAF is calculated.

Are the return conditions in the step 102 not satisfied, the CPU goes 42 continue to step 105 in which the air-fuel ratio correction coefficient FAF is set to "1.0", and proceeds to step 104 , Here, FAF = 1.0 means that there is no correction of the air-fuel ratio λ, so that the so-called "open loop control" is executed.

In step 104 determines the CPU 42 the fuel injection amount TAU based on the basic fuel injection amount TP, the air-fuel ratio correction coefficient FAF, and another correction coefficient FALL according to the equation (9): TAU = TP · FAF · CASE (9)

Then, based on the fuel injection amount TAU, a control signal is sent to the fuel injection valve 7 output to control the valve opening duration, ie, the actual fuel injection duration of the fuel injection valve 7 so that the air / fuel ratio λ is tuned with the target air / fuel ratio λTG.

The determination of the air-fuel ratio correction coefficient FAF according to the difference between the air-fuel ratio λ and the target air-fuel ratio λTG (ie, λTG = 1.0) has been described above. However, the air-fuel ratio control system does not operate normally when the air-fuel ratio sensor 26 is abnormally effective or if a control deviation (or an error of the feedback gain) by the CPU 42 caused. Whether or not the air-fuel ratio control system has an error is diagnosed on the basis of an error decision element which is the air-fuel ratio λ determined by the air-fuel ratio sensor 26 , and the air-fuel ratio correction coefficient FAF is derived.

First Embodiment

A first preferred embodiment of the present invention will be described below in connection with 6 - 13 described. As in 6 shown when the routine is initiated, determines the CPU 42 in step 201 the activation state of the air / fuel ratio sensor 26 , When the element temperature of the air / fuel ratio sensor 26 (ie the temperature of the sensor body 32 ) is not lower than 650 ° C or the resistance of the element of the air-fuel ratio sensor 26 is not greater than 90 Ω, it is decided that the air / fuel ratio sensor 26 is activated. The CPU 42 decides in step 202 whether a fault is different from that of the air / fuel ratio sensor 26 is detected or not. This decision is made to ensure that another fault is the fault diagnosis of the air / fuel ratio sensor 26 unaffected.

Subsequently, the process calculates the sum values Σ | Δ ² λSM | and Σ | Δ ² FAFSM | in step 300 as described in more detail below. Then the CPU decides 42 in step 203 Whether or not a summation time counter CDG1 reaches a predetermined value KX1 (ie, a numerical value corresponding to a summation period of 30 seconds in the present embodiment). The CPU 42 terminate the present routine as it is when CDG1 ≠ KX1 and proceed to the step 204 if CDG1 = KX1. The CPU 42 compares and decides in the step 204 a fault diagnosis parameter, ie the ratio between the sum values Σ | Δ ² λSM | and Σ | Δ ² FAFSM | and a predetermined error decision value α. Specifically, it decides whether the inequality ( 10 ) is fulfilled or not: Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α (10)

Fulfilling this inequality means that the air / fuel ratio λ and the air / fuel ratio correction coefficient FAF change at satisfactory rates of change while interrelated with each other, and e.g. As the maintenance of the response of the air / fuel ratio sensor 26 correspond. Here shows 10 a characteristic curve La of the relationship of the sensor responsiveness and the fault diagnosis parameter "(Σ | Δ ² FAF |) / (Σ | Δ ² λ |)" (where Σ | Δ ² FAF | and Σ | Δ ² λ | conveniently show the parameter values left unrounded If the reference to the decision of the correctness of the sensor response is made, for example, 500 ms, as in 10 is shown, the error decision value α is set along the characteristic La. According to this characteristic La in 10 For example, the numerical value of "(Σ | Δ ² FAF |) / (Σ | Δ ² λ |)" becomes larger for the low sensor responsiveness, so that the error diagnostic parameters, as in 10 shown on the ordinate, which better reflects sensor responsiveness for the decision of sensor error.

As in the following inequality ( 11 ), it can be determined from the above description that the air / fuel ratio sensor 26 is faultless, though Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α (11) and that the air / fuel ratio sensor 26 is incorrect, as in the inequality ( 12 ) is shown, if Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | ≥ α (12)

If the answer in step 204 of the 6 Yes, the CPU determines 42 in that the air / fuel ratio sensor 26 is faultless and goes to the step 205 , Is the answer in step 204 NO, determines the CPU 42 in that the air / fuel ratio sensor 26 is faulty and continues to step 207 , In step 205 holds the CPU 42 an error decision counter CDG3 and increments in step 206 a deciding the normal state counter CDG4 by "1." In step 207 increments the CPU 42 on the other hand, the error decision counter CDG3 by "1" and set in step 208 the deciding the normal state counter CDG4 back to "0".

Then the CPU continues 42 in step 209 the sum value Σ | Δ ² λSM | back to "0" and set in step 210 Σ | Δ ² FAFSM | back to "0" The CPU 42 resets the summing time counter CDG1 to "0".

Then the CPU goes 42 continue to step 212 in 7 to increment a fault diagnosis execution counter CDG2 by "1", and goes to step 213 to decide whether or not the fault diagnosis execution counter CDG2 reaches a predetermined value KX2 (eg, KX2 = 3 in the present embodiment). In this case, the CPU stops 42 the present routine as it is when CDG2 ≠ KX2 and go to the step 214 when CDG2 = KX2, to decide whether the error decision counter CDG3 at this time is not less than a predetermined value KX3 (eg, KX3 = 2 in the present embodiment).

The operation of this step 214 essentially corresponds to the decision on the occurrence of the error. If CDG3 <KX3, the CPU goes 42 continue to step 215 This error decision flag XDGAF is effected by the final decision upon the occurrence of the error and indicates at XDGAF = 0 that there is no error and at XDGAF = 1 that an error has occurred.

If, on the other hand, in step 214 CDG3 ≥ KX3, the CPU goes 42 continue to step 216 to set the error decision flag XDGAF to "1." According to this flag operation, the CPU shifts 42 in step 217 a warning light 49 one. At this error decision timing, a diagnosis is executable to the air / fuel ratio control in addition to turning on the warning light 49 to stop.

Then the CPU continues 42 in the steps 218 and 219 both the error diagnosis execution counter CDG2 and the error decision counter CDG3 back to "0".

Then the CPU decides 42 in step 220 whether the remaining normal state counter CDG4 has a value not lower than a predetermined value KX4 (eg, KX4 = 4 in the present embodiment) or not, and decides in the step 221 whether the warning light 49 is now turned on or not, ie, whether the error decision flag XDGAF is set or not. Is the answer in step 220 or in the step 221 NO, terminates the CPU 42 the present routine as it is. If both answers in the steps 220 and 221 YES, the CPU is pacing 42 continue to step 222 to the warning light 49 then exit the current routine. At the same time, the error decision flag XDGAF is reset to "0".

The above-mentioned operations of the steps 220 to 222 are executed to indicate that the error of the air / fuel ratio sensor 26 Occurs temporarily, and that the error condition is avoided. These operations are executed again when the air-fuel ratio control is stopped in the fault diagnosis time.

In the 8th and 9 Flowcharts shown describe the detailed process of the step 300 in 6 , ie the processes of calculating the sum values Σ | Δ ² λSM | and Σ | Δ ² FAFSM |.

When the in 8th routine is started, the CPU calculates 42 in the steps 302 to 304 the second derivative Δ ² λSM _i (corresponding to the acceleration of the air / fuel ratio λ) using the air / fuel ratio λ, as determined by the air / fuel ratio sensor 26 is detected and calculated in the steps 305 to 307 the second derivative Δ ² FAFSM _i (corresponding to the acceleration of the air-fuel ratio correction coefficient FAF) using the correction coefficient FAF under the condition that the air-fuel ratio feedback is performed (ie, under the condition that the diagnostic condition of the step 301 is satisfied).

In short, the CPU 42 reads in the step 302 the A / D converted air / fuel ratio λ and filters in step 303 the read air / fuel ratio λ. This filtering operation is carried out to eliminate the influence of the dispersion between the cylinders of the heat engine, and a smoothed value λSM _i is determined by the following equation (13). λSM i = λSM i-1 + (λ - λSM i-1 ) / k (13)

Here, the index "i" corresponds to the number of operations by the CPU 42 , the subscript value of i indicates the present value, where the subscript value of (i-1) indicates the previous value, and k is the number of rule operations.

In step 304 calculates the CPU 42 the second derivative Δ ² λSM _{i of} the smoothing value λSM _i using the following equation (14): Δ 2 λSM i = (2λSM i - 2λSM i-1 ) - (2λSM i-1 - 2λSM i-2 ) (14)

In step 305 reads the CPU 42 the air / fuel ratio correction coefficient FAF and filters in step 306 the previously read air / fuel ratio correction coefficient FAF. In this filtering operation, a smoothing value FAFSM _{i is determined} from the following equation (15): FAFSM i = FAFSM i-1 + (FAF - FAFSM i-1 ) / k (15)

In step 307 calculates the CPU 42 the second derivative Δ ² FAFSM _{i of} the smoothing value FAFSM _i using the following equation (16): Δ 2 FAFSM i = (FAFSM i - FAFSM i-1 ) - (FAFSM i-1 - FAFSM i-2 ) (16)

Then the CPU decides 42 in the steps 308 to 310 whether the conditions (ie the summation conditions) for summing the second derivatives Δ ² λSM _i and Δ ² FAFSM _{i are} satisfied.

In other words, the CPU 42 decides in step 308 whether the warm-up of the heat engine is completed or not. Specifically, the CPU decides 42 Whether or not a cooling water temperature Thw exceeds a predetermined level Y1 (eg, 80 ° C). The CPU 42 decides in step 309 Whether the running conditions meet the predetermined conditions or not. Specifically, it is decided individually whether an engine speed Ne is within a predetermined range Y2 to Y3 (z. B. 600 min ^-1 to 4000) or not, whether a vehicle speed SPD within a predetermined range Y4 to Y5 (eg. B 0 to 120 km / h) and whether or not an intake air pressure PM is within a predetermined range Y6 to Y7 (eg, 25 to 95 kPa). The CPU 42 decides in step 310 Whether the running condition is for a sudden acceleration or not. Specifically, it is decided whether or not the change ΔPM of the intake air pressure is lower than a predetermined value.

In the air-fuel ratio control system of the present embodiment, the air-fuel ratio sensor is 26 and the heat engine with internal combustion 1 imitated in the model. If the above-mentioned summation conditions are satisfied under the adaptation conditions for the model, the inequality ( 17 ) as the behavior of the air-fuel ratio λ and the air-fuel ratio correction coefficient FAF: Δ 2 λSM i <Δ 2 FAFSM i (17)

When all the summation conditions of the steps mentioned above 308 to 310 are fulfilled, go the CPU 42 continue to step 311 in 9 to obtain a new sum value Σ | Δ ² λSM | by adding the absolute value of the second derivative Δ ² λSMi, as calculated at that time, to the previous sum value Σ | Δ ² λSM | to calculate (Σ | Δ ² λSM | = Σ | Δ ² λSM | + Σ | Δ ² λSM _i |). The CPU 42 calculated in step 312 a new sum value Σ | Δ ² λFAFSM | by adding the absolute value of the second derivative Δ ² FAFSM _i as calculated at that time to the previous sum value Σ | Δ ² FAFSM | (Σ | Δ ² FAFSM | = Σ | Δ ² FAFSM | + Σ | Δ ² FAFSM _i |).

Finally, the CPU increments 42 in step 313 the summation time counter CDG1 by "1" and ends the present routine.

If the above-mentioned diagnosis condition in step 301 is not met or if any of the summation conditions of the steps 308 to 310 are not met, the CPU goes 42 continue to the step 314 in 9 , The CPU 42 keeps in the steps 314 and 315 the sum values Σ | Δ ² λSM | and Σ | Δ ² λFAFSM | stuck and keeps in step 316 the summation time counter CDG1 to complete the present routine.

The above-mentioned fault diagnosis routine by the CPU 42 will be related to the in 11A - 11H specifically described curves. In 11A - 11H becomes the error of the air / fuel ratio sensor 26 not before, but detected after a time t1. On the other hand, the summation conditions (ie, the conditions of the above-mentioned steps 308 to 310 ) is usually met repeatedly and not satisfied at time t1 and before time t1. For simplicity, it is assumed here that the summation conditions are always satisfied at time t1 and after time t1. It is also assumed for the sake of simplification that the diagnosis condition (ie the error condition in step 301 of the 8th ) is maintained once it is fulfilled.

In this case, as in 11C - 11E ² , the summation time counter CDG1 and the sum values Σ | Δ ² λSM | and Σ | Δ ² λFAFSM | set individually when the diagnostic conditions and the summation conditions are met, and are held at the values if they are not met.

In the 11A - 11H shown graphs are described sequentially. At the time t1 and before the time t1, the summing time counter CDG1 reaches the predetermined value KX1 at a time t0, and the fault diagnosis is then carried out (at the step t1) 204 of the 6 ), as in 11C is shown using the ratio between the sum values Σ | Δ ² λSM | (in 11D ) and Σ | Δ ² λFAFSM | (in 11E ). At time t1 and before time t1, the inequality ( 18 ) (ie, the answer in step 204 of the 6 is YES): Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α (18)

As a result, the error decision counter CDG3 becomes 11G not incremented. At this time t0, the fault diagnosis execution counter CDG2 is set in 11F is incremented, and this count reaches the predetermined value KX2 (= 3) (ie, the answer in step 213 of the 7 is YES), so that the error diagnosis execution counter CDG2 is reset to "0." In this case, the error decision counter CDG3 = 3, so that the error decision flag XDGAF in FIG 11H is received to "0".

At time t1 and after time t1, the fault diagnosis execution counter CDG2 is incremented at times t2, t3 and t4. At this time t2 the inequality ( 19 ) (ie, the answer in step 204 of the 6 is YES): Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α (19)

On the other hand, at times t3 and t4, the inequality ( 20 ) (ie, the answer in step 2O4 of the 6 is NO) Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | ≥ α (20) so that the error decision counter CDG3 is incremented. As a result, the error decision counter CDG3 reaches the predetermined value KX3 (= 2) at time t4 (ie, the answer at step 214 of the 7 is YES), so that the error decision flag XDGAF is set (in step 216 of the 7 ). According to this flag operation, the warning light becomes 49 switched on.

If the error symptom is temporary, so that the answer in step 204 of the 6 again If YES, their number is counted by the persistent normal state counter CDG4 (not shown). When the normal condition of the air / fuel ratio sensor 26 so that the count value of the same counter CDG4 becomes KX4 (= 4) (ie, the answer in step 220 of the 7 is YES), the error decision flag XDGAF is reset, and the warning light 49 is switched off.

• (a) Das durch den Luft/Kraftstoff-Verhältnis-Sensor 26 erfaßte Luft/Kraftstoff-Verhältnis λ und der Luft/Kraftstoff-Verhältnis-Korrekturkoeffizient FAF sind unterschiedlich, wenn jeweils ihre Empfindlichkeit verglichen wird, und dieser Unterschied ist herausragend, wenn der Sensor fehlerhaft wird und sein Ansprechvermögen geringer wird. Wenn daher das Verhältnis zwischen der Änderungsgeschwindigkeit des erfaßten Luft/Kraftstoff-Verhältnisses λ und die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF als der Fehlerdiagnoseparameter verwendet wird, so daß der Parameter verglichen wird, ist der Fehler des Luft/Kraftstoff-Verhältnis-Sensors 26 genau diagnostizierbar. Selbst wenn in diesem Fall der Luft/Kraftstoff-Verhältnis-Korrekturkoeffizient FAF verändert wird, bei Anwendung auf ein reelles System, durch die Veränderung des Verdampfungsspül- oder des Luft/Kraftstoff-Verhältnis-Lernwerts, ist der Berechnungsfehler des Fehlerdiagnoseparameters verminderbar, um die fehlerhafte Erfassung der Fehlerdiagnose zu verhindern. Da weiterhin die Verdampfungsspülung den Fehlerdiagnoseparameter nicht ohne weiteres beeinflußt, kann die Ausführung der Fehlerdiagnose weniger beschränkt werden, um die Ausführungshäufigkeit der Fehlerdiagnose zu erhöhen. Demzufolge ist ein hochzuverlässiges Luft/Kraftstoff-Verhältnis-Regelungssystem realisierbar, um Nachteile zu unterdrücken, wie z. B. die Verschlechterung der Qualität der Emissionssteuerung. In der vorliegenden Ausführungsform werden die Diagnosebedingungen und die Summierungsbedingungen durch den vor stehen erwähnten Ablaufplan in 8 wirklich eingestellt, aber diese Bedingungen weisen relativ lockere Beschränkungen auf, so daß sie die Diagnosehäufigkeit des Sensorfehlers nicht wesentlich vermindern.
• (b) Besonders in der vorliegenden Ausführungsform werden die zweiten Ableitungen Δ²λSM des Luft/Kraftstoff-Verhältnisses λ für die vorbestimmte Zeitdauer summiert, so daß ihr Summenwert Σ|Δ²λSM| als die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF verwendet wird, und die zweiten Ableitungen Δ²FAFSM des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF für die vorbestimmte Zeitdauer summiert werden, so daß ihr Summenwert Σ|Δ²FAFSM| als die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF verwendet wird. Weiterhin wird das Verhältnis zwischen dem Summenwert der Zustandsvariablenmenge X(k) der zweiten Ableitung von λ und dem Summenwert der zweiten Ableitung von FAF als der Fehlerdiagnoseparameter verwendet, und dieser Parameter wird mit dem vorbestimmten Fehlerentscheidungswert α verglichen, so daß der Fehler des Luft/Kraftstoff-Verhältnis-Sensors 26 aus dem Vergleichsergebnis diagnostiziert wird. In diesem Fall ist die Summierung der zweiten Ableitungen von λ und von FAF für die vorbestimmte Zeitdauer ein Verfahren zur Klärung des Unterschieds zwischen dem Einzelverhalten, während damit zusammenhängende Erscheinungen, wie z. B. der charakteristische Fehler des Luft/Kraftstoff-Verhältnis-Sensors 26 und der Fehler des Ansprechvermögens sowie der Sensorfehlers (z. B. der charakteristische Fehler oder der Fehler des Ansprechvermögens), leicht diagnostizierbar.
• (c) In der vorliegenden Ausführungsform wird weiterhin diagnostiziert, daß der Luft/Kraftstoff-Verhältnis-Sensor 26 normal ist, wenn die folgende Beziehung erfüllt wird, wenn das Verhältnis zwischen dem Summenwert Σ|Δ²λSM| der zweiten Ableitung von λ und dem Summenwert Σ|Δ²FAFSM| der zweiten Ableitung von FAF mit dem vorbestimmten Fehlerentscheidungswert α verglichen wird, wie die Ungleichung (21) zeigt: Σ|Δ2FAFSM|/Σ|Δ2λSM|/ < α (21) In anderen Worten, beim Abfall des Ansprechvermögens des Luft/Kraftstoff-Verhältnis-Sensors 26 wird die Änderungsgeschwindigkeit des erfaßten λ erheblich vermindert, doch die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF wird weniger erniedrigt, wie vorstehend beschrieben worden ist. Gemäß der Diagnose des vorliegenden Aufbaus ist der Fehler, das Vorliegen der Verminderung des Sensoransprechvermögens, sachgemäß diagnostizierbar.
• (d) Außerdem werden die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnisses λ und die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF der Filteroperation als eine Glättungsoperation unterzogen. Wird diese Diagnose auf eine Vier-Zylinder-Wärmekraftmaschine mit innerer Verbrennung angewendet, kann die Streuung der einzelnen Elemente dieser Zylinder ausgeschlossen werden, um die Fehlerdiagnosegenauigkeit weiter zu erhöhen.
• (e) In der vorliegenden Ausführungsform wird der den fortbestehenden Normalzustand entscheidende Zähler CDG4 verwendet, um die Fehlerentscheidung durch die Zähloperation des Zählers CDG4 zurückzusetzen, selbst nachdem die Fehlerentscheidung erfolgt ist. Demzufolge kann die Luft/Kraftstoff-Verhältnis-Regelung in geeigneter Weise wieder aufgenommen werden, wenn der Sensorfehler zeitweise auftritt, so daß der Fehlerzustand des Sensors vermieden wird oder wenn der Fehler einmal fälschlich entschieden worden ist.

The following effects can be achieved according to the present embodiment, which has been described in detail so far.

• (a) The air / fuel ratio sensor 26 detected air / fuel ratio λ and the air / fuel ratio correction coefficient FAF are different when each of their sensitivity is compared, and this difference is outstanding when the sensor is faulty and its response is lower. Therefore, if the ratio between the rate of change of the detected air / fuel ratio λ and the rate of change of the air / fuel ratio correction coefficient FAF is used as the error diagnosis parameter so that the parameter is compared, the error of the air / fuel ratio is sensor 26 accurately diagnosable. In this case, even if the air-fuel ratio correction coefficient FAF is changed when applied to a real system by the variation of the vapor purge or the air-fuel ratio learning value, the calculation error of the fault diagnosis parameter is reducible to the erroneous one To prevent detection of the fault diagnosis. Further, since the evaporation purge does not readily affect the fault diagnosis parameter, the execution of the fault diagnosis can be less limited to increase the execution frequency of the fault diagnosis. As a result, a highly reliable air / fuel ratio control system is feasible to suppress disadvantages such. B. the deterioration of the quality of emissions control. In the present embodiment, the diagnosis conditions and the summation conditions by the above-mentioned flowchart in 8th Really set, but these conditions have relatively loose restrictions, so that they do not significantly reduce the diagnosis frequency of the sensor error.
• (b) Especially in the present embodiment, the second derivatives Δ ² λSM of the air-fuel ratio λ are summed for the predetermined period of time so that their sum value Σ | Δ ² λSM | is used as the rate of change of the air-fuel ratio correction coefficient FAF, and the second derivatives Δ ² FAFSM of the air-fuel ratio correction coefficient FAF are accumulated for the predetermined period of time so that their sum value Σ | Δ ² FAFSM | is used as the rate of change of the air-fuel ratio correction coefficient FAF. Further, the ratio between the sum value of the state variable quantity X (k) of the second derivative of λ and the sum value of the second derivative of FAF is used as the error diagnosis parameter, and this parameter is compared with the predetermined error decision value α, so that the error of the air / fuel ratio sensor 26 is diagnosed from the comparison result. In this case, the summation of the second derivatives of λ and FAF for the predetermined period of time is a method for clarifying the difference between the individual behaviors, while related phenomena such. B. the characteristic error of the air / fuel ratio sensor 26 and the error of the response and the sensor error (eg the characteristic error or the error of the response), easily diagnosable.
• (c) In the present embodiment, it is further diagnosed that the air-fuel ratio sensor 26 is normal when the following relationship is satisfied when the ratio between the sum value Σ | Δ ² λSM | the second derivative of λ and the sum value Σ | Δ ² FAFSM | the second derivative of FAF is compared with the predetermined error decision value α, like the inequality ( 21 ) shows: Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | / <α (21) In other words, the drop of the responsiveness of the air-fuel ratio sensor 26 For example, the rate of change of the detected λ is significantly reduced, but the rate of change of the air-fuel ratio correction coefficient FAF is less lowered as described above. According to the diagnosis of the present structure, the error, the presence of the reduction of the sensor responsiveness, can be properly diagnosed.
• (d) In addition, the rate of change of the air / fuel ratio λ and the rate of change of the air / fuel ratio correction coefficient FAF are subjected to the filtering operation as a smoothing operation. When this diagnosis is applied to a four-cylinder internal combustion engine, the dispersion of the individual elements of these cylinders can be eliminated to further increase the fault diagnosis accuracy.
• (e) In the present embodiment, the persistent normal state becomes tough CDG4 is used to reset the error decision by counting counter CDG4, even after the error decision has been made. As a result, the air-fuel ratio control can be appropriately resumed when the sensor failure occurs intermittently, so that the fault condition of the sensor is avoided or once the fault has been erroneously decided.

In a modification of the embodiment described above, the error decision value α is variably set according to the running state of the heat engine. In this case, the in 12 shown operations between the above-mentioned operations of the steps 203 and 204 in 6 inserted.

As in 12 shown is in the step 250 a sum value ΣPM of the intake air pressure PM is read. This sum value ΣPM is a parameter indicating the rate of change of the intake air pressure for a predetermined period of time. In step 251 The error decision value α is variably set on the basis of the sum value ΣPM of the intake air pressure PM. Here, the error decision value α becomes z. B. using an in 13 determined relationship determined. In 13 For example, the relationship is set so that the error decision value α is larger for larger values of the sum value ΣPM of the intake air pressure.

Even if the running conditions of the heat engine change, according to this structure, the associated fault diagnosis is successively executable to further increase the sensor fault diagnosis accuracy. In the 13 The relationship shown also is achievable even for an abscissa of the rate of change of the engine speed Ne or the rate of change of a throttle opening TH. Thus, the parameter for setting the variable α is variable to these rates of change.

Further embodiments The present invention will be described below with reference to FIGS attached Drawings described. In the following trainings of the individual embodiments the description of the sections is omitted with which the above explained first embodiment match. At this point, the differences from the first embodiment will be pointed out.

Second Embodiment

14 FIG. 12 shows a flowchart of a fault diagnosis routine according to a second preferred embodiment of the present invention. FIG. In the 14 shown routine replaces the in 6 shown flowchart in the first embodiment and is by the CPU 42 for a predetermined period of time (or in synchronization with the injections of the fuel) as described above.

Only the parts of in 14 The error detection routine shown are described which differ from those in FIG 6 differentiate shown. In step 300 in 6 the sum value "Σ | Δ ² λSM |" of the second derivative Δ ² λSM of the air-fuel ratio λ is calculated as the rate of change of the air-fuel ratio λ, and the sum value Σ | Δ ² FAFSM | of the second derivative Δ ² FAFSM of the air-fuel ratio correction coefficient FAF is calculated as the rate of change of the value FAF 204 of the 6 the fault diagnosis of the air / fuel ratio is carried out, depending on whether the inequality ( 22 ) is fulfilled or not: Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α

In the in 14 On the other hand, the routine shown in FIG 410 the sum value "Σ | ΔλSM |" a derivative ΔλSM (= ΔλSM _i - ΔλSM _i-1 ) is calculated as the rate of change of the air-fuel ratio λ, and the sum value "Σ | Δ ² FAFSM |" of the second derivative Δ ² FAFSM is calculated as the rate of change of the value FAF In other words, the rate of change of the air-fuel ratio λ is expressed by "Σ | Δ ² SM |" In this case, the sum value Σ | ΔλSM | to be used in the present routine assumes the value which remains the predetermined value KX1 (Σ | ΔλSM | = Σ | ΔλSM | + ΔλSM) is summed.

In step 420 continues to be the fault of the air / fuel ratio sensor 26 is diagnosed using a predetermined error decision value β1, depending on whether the inequality ( 22 ) is fulfilled or not: Σ | Δ 2 FAFSM | / Σ | ΔλSM | <β1 (22)

In step 430 after the fault diagnosis, the sum value Σ | ΔλSM | reset to 0. The operations represented by the same step numbers as those in 6 are designated, and the operations in step 211 and after the step 211 are with those in 6 and 7 shown and their description is omitted.

Also in the present embodiment, the object of the present invention is achieved as in the first embodiment described above. As with respect to 12 and 13 has been explained, the error diagnosis accuracy can also be increased by setting the error decision value β1 which is variable by the ΣPM value.

Third Embodiment

15 FIG. 12 shows a flowchart of a fault diagnosis routine according to a third preferred embodiment of the present invention. FIG. This in 15 shown routine replaces the schedule in 6 in the first embodiment and can by the CPU 42 for a predetermined period of time (in synchronization with the injection of the fuel).

Only the points in 15 which is different from those in 6 will be described below. In the in 15 shown routine is in step 450 the sum value "Σ | Δ ² λSM |" of a second derivative Δ ² λSM is calculated as the rate of change of the air-fuel ratio λ, and the sum value "Σ | ΔFAFSM |" The derivative ΔFAFSM (= FAFSM _i - FAFSM _i-1 ) is calculated as the rate of change of the value FAF. In other words, the rate of change of the value FAF becomes "Σ | Δ ² FAFSM |" to "Σ | ΔFAFSM |" changed. In this case, the sum value "Σ | FAFSM |" to be used in the present routine assumes the value which is accumulated during the period until the summing time counter CDG1 reaches the predetermined value KX1 (Σ | ΔFAFSM | = Σ | ΔFAFSM | + ΔFAFSM) reached.

In step 460 continues to be the fault of the air / fuel ratio sensor 26 is diagnosed using a predetermined error decision value β2, depending on whether the inequality ( 23 ) is fulfilled or not: Σ | ΔFAFSM | / Σ | Δ 2 λSM | <β2 (23)

In step 470 after the fault diagnosis, the sum value Σ | ΔFAFSM | reset to "0".

Also in the present embodiment, the object of the present invention is achieved as in the first embodiment described above. As with respect to 12 and 13 has been described, the error diagnosis accuracy can also be increased by variably setting the error decision value β2 according to the ΣPM value.

Fourth Embodiment

A fourth preferred embodiment of the present invention will be described below with reference to FIG 16 - 18 described. In the present embodiment, the amplitude ΔλLR of the rate of change Δλ of the air-fuel ratio λ is determined, and the amplitude ΔFAFLR of the rate of change ΔFAF of the value FAF is determined so that the rates of change are the sum values "ΣΔλLR" and "ΣΔFAFLR" of the individual amplitude values be determined for predetermined times. The fault diagnosis of the air / fuel ratio sensor 26 is performed on the basis of these values ΣΔλLR and ΣΔFAFLR.

In short, the summation values Σ | Δ ² λSM | and Σ | Δ ² FAFSM | as used in the above-described individual embodiments, and the sum values ΣΔλLR and ΣΔFAFLR used in the present embodiment are substantially the same (ie, the sum of the second derivatives is equal to that of FIG Is the sum of the amplitude ") when the summation times (summation time duration) of the individual values are longer than half the wavelength of the fluctuation period of the air-fuel ratio λ or the air-fuel ratio correction coefficient FAF ΣΔλLR and ΣΔFAFLR are used as the rates of change to perform the fault diagnosis.

16 FIG. 12 is a flowchart of a fault diagnosis routine of the present embodiment. FIG. This routine in 16 replaces the in 6 shown flowchart of the first embodiment and is by the CPU 42 for a predetermined period of time (or in synchronism with the injection of the fuel).

When the in 16 routine is initiated, determines the CPU 42 in step 501 the active state of the air / fuel ratio sensor 26 and determined in the step 502 whether a fault is different from that of the air / fuel ratio sensor 26 is detected or not. If both answers of the steps 501 and 502 YES, the CPU goes 42 continue to step 600 to calculate the sum values ΣΔλLR and ΣΔFAFLR by the subroutines described below, which are described in 17 and 18 are shown to calculate.

Then the CPU decides 42 in step 503 Whether both of the summation time counters CDG0 and CDG1 exceed the predetermined value KX1 (for example, a numerical value corresponding to the summation time of 30 seconds in the present embodiment). If CDG0 <KX1 or CDG1 <KX1, the CPU stops 42 the present routine as it is. If CDG0 ≥ KX1 and CDG1 ≥ KX1, the CPU goes 42 continue to step 504 , In step 504 uses the CPU 42 the relationship between the sum values ΣΔλLR and ΣΔFAFLR as the error diagnosis parameter to compare the parameter with a predetermined error decision value γ. In other words, it is decided whether the inequality ( 24 ) is fulfilled or not: ΣΔFAFLR / ΣΔλLR <γ (24)

In this case, satisfying this inequality means that the rates of change of the air / fuel ratio λ and the value FAF are in a satisfactory state in correlation with each other and correspond to the fact that the response of the air / fuel ratio sensor 26 is obtained. In other words, an in 10 The relationship shown between the error decision parameter (ΣΔFAFLR / ΣΔλLR) and the error decision value γ (ie, the ordinate and the abscissa in FIG 10 are replaceable by the error decision parameter and the error decision value γ).

It follows that the air / fuel ratio sensor 26 then considered normal if, as in the inequality ( 25 ), ΣΔFAFLR / ΣΔλLR <γ (25) is, and that the air / fuel ratio sensor 26 can be considered erroneous if, as in the inequality ( 26 ), ΣΔFAFLR / ΣΔλLR ≥γ (26) is.

If the answer in step 504 YES, the CPU stops 42 in step 503 the error decision counter CDG3 and increments in step 506 On the other hand, the answer in step is the counter CDG4 which decides the continuing normal state 504 NO, the CPU increments 42 in step 507 Set the error decision counter CDG3 by "1" and set in step 508 the deciding the normal state counter CDG4 back to "0".

Then the CPU continues 42 in step 509 returns the sum value ΣΔλLR to "0" and sets in step 510 the sum value ΣΔFAFLR back to "0." The CPU 42 put in step 511 both summation time counters CDG0 and CDG1 are reset to 0. The routine in step 511 and after the step 511 is with the in 7 and the further description is therefore omitted.

The process of calculating the sum values ΣΔλLR and ΣΔFAFLR or the process of the step shown in detail 600 in 16 will be described below with reference to the in 17 and 18 described flowchart described.

When the in 17 routine is started, the CPU calculates 42 in the steps 601 to 603 a derivative ΔλSM _i using the by the air / fuel ratio sensor 26 detected air / fuel ratio λ and calculated in the steps 604 to 606 using the air-fuel ratio correction coefficient FAF, a derivative ΔFAFSM _i . Specifically, the CPU reads 42 in the step 601 the A / D converted air / fuel ratio λ and filters in the step 602 the read air / fuel ratio λ, to calculate the smoothed value λSM _i . In step 603 calculates the CPU 42 from the present value and the previous value of the previously calculated smoothed value λSM (ΔλSM _i = λSM _i -λSM _i-1 ), the derivative ΔλSMi.

On the other hand, the CPU reads 42 in step 604 the value FAF and filters in step 605 the read value FAF to calculate the smoothed value FAFSM _i . In step 606 calculates the CPU 42 from the present value and the previous value of the calculated, smoothed value FAFSM the derivative ΔλFAFSM _i (ΔλFAFSM _i = FAFSM _i - FAFSM _i-1 ).

Then the CPU decides 42 in the steps 607 to 609 whether or not the conditions (ie the summation conditions) for summing the derivatives ΔλSM _i and ΔFAFSM _{i are} fulfilled (these operations being the same as those of steps 308 to 310 in 8th to match). Specifically, in step 607 On the basis of the cooling water temperature Thw decided whether the warm-up of the heat engine is completed or not, and in the step 608 is decided on the basis of the engine speed Ne, the vehicle speed SPD and the intake air pressure PM, whether the running conditions meet the predetermined conditions or not. In step 609 is decided on the basis of the rate of change of the intake air pressure PM, whether the vehicle is suddenly accelerated or not.

If the summation conditions of the steps 607 to 609 are not met, the CPU goes 42 continue to step 610 and holds both sum values ΣΔλLR and ΣΔFAFLR. The CPU 42 keeps in step 611 the summation time counters CDG0 and CDG1 and ends the present routine.

On the other hand, are all the summation conditions of the steps 607 to 609 fulfilled, the CPU goes 42 continue to step 612 in 18 , In step 612 decides the CPU 42 whether the difference between the present value and the previous value of the difference ΔλSM exceeds "0", as in the inequality ( 27 ) is shown: ΔλSM i - ΔλSM i-1 > 0 (27)

In both steps 613 and 614 decides the CPU 42 whether the difference between the previous value and the previous value of the difference ΔλSM exceeds "0", as in the inequality ( 28 ) is shown: ΔλSM i-1 - ΔλSM i-2 > 0 (28)

If the answer in step 612 YES and the answer is in the step 613 NO is, the CPU goes 42 continue to step 615 to set the previous value ΔλSM _i-1 to "ΔλR", assuming that the previous value (ΔλSM _i-1 ) of the value ΔλSM corresponds to a "rich" peak ΔλR. On the other hand, is the answer in the step 612 NO and the answer in step 614 YES, the CPU goes 42 continue to step 616 to set the previous difference ΔλSM _i-1 to "ΔλL" on the assumption that the previous value (ΔλSM _i-1 ) of the value ΔλSM corresponds to the "lean" peak ΔλL.

In other cases (ie, if the answer in step 613 YES or if the answer is in step 614 NO is) goes the CPU 42 continue to step 617 , The CPU 42 keeps in step 617 the sum value ΣΔλLR and keeps in step 618 the summation time counter CDG0.

After the operation in step 615 or 616 calculates the CPU 42 in step 619 the Δλ amplitude ΔλLR by subtracting the "rich" peak ΔλR from the "lean" peak ΔλL (ΔλLR = ΔλL-ΔλR). Then the CPU updates 42 in step 620 the sum value ΣΔλLR by adding the sum value ΣΔλLR, which is summed until then, and the calculated Δλ amplitude ΔλLR (ΣΔλLR = ΣΔλLR + ΔλLR). The CPU 42 incremented in step 621 the summation time counter CDG0 by "1" and then proceed to the step 622 ,

In step 622 decides the CPU 42 whether the difference between the present value and the previous value of the derivative ΔFAFSM exceeds "0" or not, as shown in the inequality (29): ΔFAFSM i - ΔFAFSM i-1 > 0 (29)

In both steps 623 and 624 decides the CPU 42 whether the difference between the preceding the value and the previous value of the derivative ΔFAFSM exceeds "0", as in the inequality ( 30 ) is shown: ΔFAFSM i-1 - ΔFAFSM i-2 > 0 (30)

If the answer in step 622 YES and the answer is in the step 623 NO is, the CPU goes 42 continue to step 625 to set the previous derivative ΔFAFSM _i-1 to "ΔFAFR", assuming that the previous value (ΔFAFSM _i-1 ) of the ΔFAFSM value corresponds to the "rich" peak ΔFAFR. On the other hand, is the answer in the step 622 NO and the answer in step 624 YES, the CPU goes 42 continue to step 626 to set the previous derivative ΔFAFSM _i-1 to "ΔFAFL", assuming that the previous value (ΔFAFSM _i-1 ) of the ΔFAFSM value corresponds to the "lean" peak ΔFAFL.

In other cases (ie, if the answer in step 623 YES or if the answer is in step 624 NO is) goes the CPU 42 continue to step 627 , The CPU 42 keeps in step 627 the sum value ΣΔFAFLR and holds in step 628 the summation time counter CDG1.

After the operation of the step 625 or 626 calculates the CPU 42 in step 629 the ΔFAF amplitude ΔFAFLR by subtracting the "rich" peak ΔFAFR from the "lean" peak ΔFAFL (ΔFAFLR = ΔFAFL-ΔFAFR). Then the CPU updates 42 the sum value ΣΔFAFLR by adding the calculated ΔFAF amplitude ΔFAFLR to the sum value ΣΔFAFLR until then (ΣΔFAFLR = ΣΔFAFLR + ΔFAFLR). The CPU 42 incremented in step 631 the summation time counter CDG1 by "1" and then ends the present routine.

Also in the present embodiment, the object of the invention is fulfilled as in the individual embodiments described above. Especially in the present embodiment, the sum value ΣΔλLR of the Δλ amplitude is used as the rate of change of the air-fuel ratio λ, and the sum value ΣΔFAFLR of the ΔFAF amplitude is used as the rate of change of the FAF value. The ratio between the sum values ΣΔLR and ΣΔFAFLR is used as the error diagnosis parameter, and this parameter is compared with the error decision value γ, so that the error of the air-fuel ratio sensor 26 is diagnosed from the comparison result. In this case, summing the Δλ amplitude and ΔFAF for the predetermined period of time is a method for clarifying the difference between the individual behaviors, while correlating phenomena such as those of FIG. B. the characteristic error of the air / fuel ratio sensor 26 and the error of the response, and the sensor error can be easily diagnosed by determining the ratio (= ΣΔFAFLR / ΣΔλLR) of these sum values.

Also in the present embodiment, the error diagnosis accuracy can be increased by variably setting the error decision value γ according to the ΣPM value as described with reference to FIG 12 and 13 has been described.

Fifth Embodiment

In each of the foregoing first to fourth embodiments, in the fault diagnosis, the sensor failure is diagnosed depending on whether the fault diagnosis parameter, as determined from the ratio of the air-fuel ratio λ and the FAF, is lower than the error decision value (α, β1, β2 or γ) is or not. On the other hand, in the present embodiment, the sensor error is diagnosed depending on whether or not the failure diagnosis parameter, as determined by the ratio of the air / fuel ratio λ and FAF change rates, is within a predetermined range (δ1 to δ2) , Further, in the present embodiment, the relationship between the sum values Σ | Δ ² λSM | and Σ | Δ ² FAFSM | is used as the error diagnosis parameter, and the error decision values δ1 and δ2 to be used for the error diagnosis are set to be changeable. In this case, the in 19 shown operations after the step 203 in 6 added, and the operation in step 204 is modified.

In 19 the sum value ΣPM of the intake air pressure is read. This sum value ΣPM is a parameter indicating the rate of change of the intake air pressure for a predetermined period of time. In step 271 For example, the error decision values δ1 and δ2 are variably set on the basis of the sum value ΣPM of the intake air pressure PM. Then the values δ1 and δ2 can be z. B. from the in 20 relationships are determined. In 20 For example, the relationships are set so that the values δ1 and δ2 become larger for larger values of ΣPM.

Then in step 272 the fault diagnosis is executed, depending on whether the inequality ( 31 ) is fulfilled or not: δ1 <(Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM |) <δ2 (31)

If the answer in step 272 If YES, the routine continues to step 205 in 6 , In the in 6 shown steps 205 and 206 the error decision counter CDG3 is stopped, and the persistent normal state counter CDG4 is incremented by "1." On the other hand, the answer in step is 207 NO, the routine continues to step 207 in 6 , In the steps 207 and 208 of the 6 The error decision counter CDG3 is incremented by "1", and the remaining normal state counter CDG4 is reset to "0".

Of course, according to the structure of the present embodiment, the object of the invention as in the above-described individual embodiments can be fulfilled, and the following effects are also attainable. If the structure diagnoses the sensor error due to the fact that the error diagnosis parameter is within the predetermined normal range (δ1 to δ2), the characteristic error can be diagnosed properly even if it is included in the in-field 23A shown mode occurs.

In the 20 shown ratios can also be achieved, even if the abscissa in 20 shows the rate of change of the engine speed Ne or the rate of change of the throttle opening TH, so that the parameter for setting the values δ1 and δ2 is replaceable by these rates of change.

(First unclaimed Example)

A first unclaimed example of the diagnostic process calculates the sum value (the "λ sum value Tλ") between the air / fuel ratio λ as it is within a predetermined period of time by the air / fuel ratio sensor 26 is detected, and the target air-fuel ratio λTG, and a sum value (the "FAF sum value TFAF") of the difference between the air-fuel ratio correction coefficient FAF within a predetermined time period and the average value FAFAV of the air / fuel Ratio correction coefficient FAF. When the air / fuel ratio λ changes relative to the target air / fuel ratio λTG, and when the air / fuel ratio correction coefficient FAF changes relative to its average value FAFAV, the λ sum value Tλ and the FAF Sum value TFAF the areas of the hatched sections in 28A and 28B , In the unclaimed example, the λ sum value Tλ corresponds to a first error decision element, and the FAF sum value TFAF corresponds to a second error decision element.

The error of the air-fuel ratio control system is detected depending on whether these values Tλ and TFAF are within the normal state decision area or within the error state decision area. When the value Tλ is within a range between the threshold values A and B and the value TFAF is within a range between the threshold values C and D, as in FIG 29 is shown (in an intermediate area excluding the hatched portions), it is diagnosed that the air-fuel ratio control system is in the normal state. On the other hand, when the values Tλ and TFAF fall within the hatched areas, it is diagnosed that the air-fuel ratio control system is normal. In this case, the range of Tλ ≦ A and TFAF ≦ C is substantially hard to decide as a faulty one. However, in the unclaimed example, this area is not reliable for the fault diagnosis and may possibly be misleading in fault diagnosis, so that this fault decision is prohibited (since it belongs to the fault decision area).

24 FIG. 12 is a flowchart showing an air-fuel ratio control system failure decision routine in the example not claimed. FIG. This routine is done by the CPU 42 eg executed for a period of 4 ms. In the unclaimed example, the routine corresponds to 24 the fault diagnosis device.

• – dass der Luft/Kraftstoff-Verhältnis-Sensor 26 in einem aktiven Zustand ist (in welchem die Elementtemperatur des Sensorkörpers 32 nicht geringer als 650°C ist oder der Elementwiderstand des Luft/Kraftstoff-Verhältnis-Sensors 26 nicht geringer als 90 Ω ist),
• – dass der Ansaugluftdruck nicht geringer als ein vorbestimmter Druck ist,
• – dass die Drehzahl Ne der Wärmekraftmaschine nicht geringer als ein vorbestimmter Wert ist,
• – dass die Drosselklappenöffnung TH nicht kleiner als ein vorbestimmter Wert ist,
• – dass die Wärmekraftmaschine im Leerlauf ist und
• – dass eine vorbestimmte Zeitdauer seit dem Beginn der Luft/Kraftstoff-Verhältnis-Rückführung vergangen ist.

When the in 24 shown routine is decided by the CPU 42 in step 710 Whether the current running state of the heat engine is in the state in which its failure can be decided. In step 710 a more specific decision is made as to whether or not the following conditions are met:

• - that the air / fuel ratio sensor 26 is in an active state (in which the elements tem temperature of the sensor body 32 is not lower than 650 ° C or the element resistance of the air-fuel ratio sensor 26 not less than 90 Ω),
• That the intake air pressure is not less than a predetermined pressure,
• That the rotational speed Ne of the heat engine is not less than a predetermined value,
• That the throttle opening TH is not smaller than a predetermined value,
• - that the heat engine is idling and
• - That a predetermined time has elapsed since the beginning of the air / fuel ratio feedback.

If the answer in step 710 YES is, to allow the error decision, the CPU goes 42 continue to step 720 in which the λ sum value Tλ is calculated. Here, the λ sum value Tλ is represented by an in 25 which corresponds to the first element calculating means as described below. The CPU 42 calculated in step 721 of the 25 the difference between the air / fuel ratio λ, as determined by the air / fuel ratio sensor 26 is detected, and the target air / fuel ratio λTG. In the following step 722 the absolute value | λ-λTG | the difference between the air / fuel ratio λ and the target air / fuel ratio λTG to the previous value Tλi-1 of the λ sum value for calculating the present value Tλi of the λ sum value (ie, Tλi = Tλi - 1 + | λ - λTG |) is added.

In step 730 of the 24 calculates the CPU 42 the FAF sum value TFAF. Here, the FAF sum TFAF is given by a subroutine in 26 which corresponds to a second element calculating means described below. The CPU 42 reads in the step 731 of the 26 the air-fuel ratio correction coefficient FAF determined according to the above-mentioned process and in the subsequent step 732 the air-fuel ratio correction coefficient FAF is set to the present value FAFi of the air-fuel ratio correction coefficient.

The CPU 42 calculated in step 733 the average value FAFAVi of the air-fuel ratio correction coefficient FAF using a well-known rounding operation. Specifically, the average value FAFAVi is calculated by the equation (32): FAFAVi = {FAFAVi - 1 · (n - 1) + FAFi} / n (32)

In Equation (32) is z. Eg n = 64.

Then the CPU calculates 42 in step 734 the difference between the air-fuel ratio correction coefficient FAFi and its average value FAFAVi and adds in the subsequent step 735 the absolute value | FAFi - FAFAVi | between the air-fuel ratio correction coefficient FAFi and its average value FAFAVi to the previous value TFAFi-1 of the FAF sum value to calculate the current value TFAFi of the FAF sum value (ie, TFAFi = TFAFi-1 + | FAFi-FAFAVi | ).

After the processes (to calculate the values Tλ and TFAF) of the steps 720 and 730 in 24 decides the CPU 42 on the other hand in the step 740 Whether or not a predetermined time t (eg 1280 ms) has elapsed since the previous error decision time. If this answer is NO, the present routine is ended as it is. In other words, for the duration (of 1280 ms), while the answer in step 740 Is NO, the summation of the difference in the air-fuel ratio and the summation of the difference of the air-fuel ratio correction coefficient FAF are repeatedly executed.

On the other hand, in step 740 the answer is YES, the CPU goes 42 continue to step 750 in which she makes the error decision. The error decision is made according to the in 27 shown subroutine executed.

This error decision (the in 27 2) is used to decide the error of the air / fuel ratio control system using the calculated λ sum value Tλ and the FAF sum value TFAF. After this decision, the CPU continues 42 in step 780 of the 24 returns the λ sum value Tλ and the FAF sum value TFAF to "0" and ends the present routine.

In 27 increments the CPU 42 in step 751 a processing counter CDG for indicating the number of executions of the error decision by "1" and decides in the subsequent step 752 Whether the processing counter CDG exceeds "5" or not At the early processing stage, the answer is in step 752 NO, and the CPU 42 continue to step 753 ,

The CPU 42 decides in step 753 whether the λ sum value Tλ is within the normal state range defined by the predetermined threshold values A and B. These thresholds A and B are on the ordinate in FIG 29 applied. If A <Tλ <B, the CPU leaves 42 a λ-error decision counter CAFDG as it is. If Tλ <A or Tλ ≥ B, the CPU increments 42 in step 755 the λ-error decision counter CAFDG by "1".

The CPU 42 decides in step 756 Whether or not the λ error decision counter CAFDG is not less than "3." The CPU 42 put in step 757 a λ-error decision flag XDGAF returns to "0" when CAFDG <3, but sets in step 758 the λ-error decision flag XDGAF is "1" when CAFDG ≥ 3.

Then the CPU decides 42 in step 759 whether the FAF sum value TFAF is within the normal state range defined by the predetermined threshold values C and D. These thresholds C and D are on the abscissa in FIG 29 applied. If C <TFAF <D, the CPU will fail 42 in step 760 an FAF error decision counter CFAFDG as it is. If TFAF ≤ C or TFAF ≥ D, the CPU increments 42 in step 761 the FAF error decision counter CFAFDG by "1".

The CPU 42 decides in step 762 Whether or not the FAF error decision counter CFAFDG is not less than "3." If CFAFDG <3, the CPU resets 42 in step 763 On the other hand, if CAFDG ≥ 3, the CPU resets an FAF error decision flag XDGFAF to "0" and terminates the present routine 42 in step 764 the FAF error decision flag XDGFAF to "1" and ends the present routine.

When the routine is in 27 is repeated, so that the answer in step 752 YES, the CPU goes 42 continue to step 765 to decide whether either the λ-error decision flag XDGAF or the FAF-error decision flag XDGFAF is set to "1." If both flags XDGAF and XDGFAF are set to "0", the CPU goes 42 continue to step 766 to reset a final fault decision flag XDG to "0." If at least one of the flags XDGAF or XDGFAF is set to "1", the CPU goes 42 continue to step 767 to set the final error decision flag XDG to "1." According to the setting operation of the final error decision flag XDG, although not shown, both of the error decision flags XDGAF and XDGFAF are reset to "0".

Here, XDG = 0 means that the λ sum value Tλ and the FAF sum value TFAF within the normal state region (middle region) in 29 and XDG = 1 means that the λ sum value Tλ and the FAF sum value TFAF are within the error state range (hatched area) in 29 are located.

Then the CPU continues 42 in step 768 the processing counter CDG back to "0" and sets in the subsequent step 769 the λ error decision counter CAFDG back to "0." The CPU 42 put in step 770 The FAF error decision counter CFAFDG returns to "0" and ends the present routine. When the final error decision flag XDG is determined, the CPU executes 42 the diagnostic processing by switching on the warning light 49 or stopping the air / fuel ratio feedback.

The error decision processing as implemented in the aforementioned manner will be described below with reference to FIGS 30A - 30H described curves. Here t1, t2, t3, t4, t5 and t6 designate times at which the in 27 shown error decision routine is executed. In 30 the processing counter CDG is incremented every 1280 ms and is reset to "0" when CDG = 5 is reached.

During the period from t1 to t2 and during the period from t3 to t5, the air / fuel ratio λ substantially changes with respect to the air / fuel ratio λTG, so that the λ sum value Tλ changes from the allowable range (A to B ) or the normal condition range. At the times t2, t4 and t5, therefore, the λ-error decision counter CAFDG is successively incremented. At the time t5 for CAFDG = 3, the λ-error decision flag XDGAF is set to "1" (ie, at the time t5, the answer at the step 756 of the 27 YES).

During the time t1 or during the period before time t1, during the period from t1 to t3 and during the period from t4 to t5, the air-fuel ratio correction coefficient FAF substantially changes with respect to the average value FAFAV thereof FAF sum value TFAF deviates from the permissible range (C to D) or the normal state range. Therefore, the FAF error is is successively incremented at times t1, t2, t3 and t5, and the FAF error decision flag XDGFAF is set to "1" at time t3 for CFAFDG = 3 (ie, the answer at step S3) 762 in 27 is YES).

To the Time t6 for CDG = 5 become the λ-error decision flag XDGAF and the FAF error decision flag XDGFAF are set to "1" so that the final error decision flag XDG is set to "1".

In connection with in 30A - 30H In the graphs shown, the case where both the λ error (the air-fuel ratio sensor 26 ) as well as the FAF error (the feedback gain by the CPU 42 ) occur substantially simultaneously. In an actual application, however, it is more common for one error to occur and one to cause the other error.

Therefore, in the air-fuel ratio control system of the unclaimed example, after detecting either the λ error or the FAF error, the threshold values A, B, C and D for deciding the λ error and the FAF error "are learned as described in detail below. In this construction, although not shown, the number of settings of the λ-error decision flag XDGAF becomes "1" (ie, the number of approvals in the step 756 of the 27 ) and the number of settings of the FAF error decision flag XDGFAF to "1" (ie, the number of approvals in step 762 of the 27 ) are constantly counted, and the counts are in the RAM 44 saved.

The in 31 The flowchart shown is a threshold learning routine for learning the threshold values A, B, C and D to determine the normal state / error state of the λ sum value Tλ and the FAF sum value TFAF on the basis of the operation states of the λ-error decision flag XDGAF and the FAF error decision flag To decide XDGFAF. This routine is done by the CPU 42 for a period of 10 minutes. In the unclaimed example, the in 31 shown routine of the decision area learning device.

When the in 31 shown routine is decided by the CPU 42 in step 801 which of the two flags, the error decision flag XDGAF and the FAF error decision flag XDGFAF, is earlier determined for the times of the previous processing and the present processing. If none of the flags are determined, the CPU stops 42 the routine as it is (although not shown).

If the λ-error decision flag XDGAF is determined earlier, the CPU goes 42 continue to step 802 to read the number of determinations of the λ-error decision flag XDGAF and proceed to the step 803 to decide whether or not the determined number reaches a predetermined value K1. In step 804 decides the CPU 42 whether the threshold values C and D reach a predetermined protection value or not, that is, whether the threshold values C and D are excessively "learned" or not.

Is the answer in step 803 YES and is the answer in the step 804 NO, the CPU goes 42 continue to step 805 , If, on the other hand, the answer in step 803 NO is or if the answer is in the step 804 YES, bypasses the CPU 42 step 805 and continue to step 806 ,

In step 805 leads the CPU 42 learning the thresholds C and D. When the air / fuel ratio λ (detected by the air / fuel ratio sensor 26 ) becomes erroneous, an error indication may occur in the air-fuel ratio correction coefficient FAF, which should be substantially normal. In order to prevent this occurrence, therefore, the normal state area defined by the thresholds C and D is widened. Specifically, the threshold C is decreased while the threshold D is increased, as in FIG 29 is shown. Then, the learning of the thresholds C and D for extending the normal state area may be performed only for one of the thresholds C or D.

Then the CPU continues 42 in step 806 returns the number of determinations of the error decision flags XDGAF and XDGFAF and ends the present routine.

On the other hand, if the FAF error decision flag XDGAF is determined earlier, the CPU goes 42 continue to step 80 7 to read the number of determinations of the fault decision flag XDGFAF and decides in the step 808 Whether the number of determinations reaches a predetermined value K2 or not. The CPU 42 decides further in the step 809 Whether or not the thresholds reach a predetermined protection value, that is, whether the thresholds A and B are excessively "learned" or not.

Is the answer in step 808 YES and is the answer in the step 809 NO, the CPU goes 42 continue to step 810 , If, on the other hand, the answer in step 808 NO is or if the answer is in the step 809 YES, bypasses the CPU 42 the step 810 and continue to step 806 ,

The CPU 42 leads in the step 810 Learning the thresholds A and B from. If the air-fuel ratio correction coefficient FAF (ie, the feedback gain) becomes erroneous, an error indication may occur in the air-fuel ratio λ, which should be substantially normal. Therefore, to prevent this phenomenon, the normal state area as defined by the thresholds A and B is widened. More specifically, the threshold A is decreased while the threshold B is increased, as in FIG 29 is shown. Then, learning the thresholds C and D to expand the normal state area is executable only for a threshold C or D.

• (a) Bei dem nicht beanspruchten Beispiel wird die Differenz (d.h, der λ-Summenwert Tλ) zwischen dem Luft/Kraftstoff-Verhältnis λ, welches durch den Luft/Kraftstoff-Verhältnis-Sensor 26 erfasst wird, und dem Soll-Luft/Kraftstoff-Verhältnis λTG berechnet, und die Differenz (d.h, der FAF-Summenwert TFAF) zwischen dem Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF und dessen Durchschnittswert FAFAV wird ebenfalls berechnet. Der Fehler des Luft/Kraftstoff-Verhältnis-Regelungssystems wird diagnostiziert, indem jene berechneten Fehlerentscheidungselemente (d.h. der λ-Summenwert Tλ und der FAF-Summenwert TFAF) veranlasst werden, den Normalzustand/Fehlerzustand-Entscheidungsbereichen (wie in 29 gezeigt), die durch die zwei Schwellenwerte einzeln definiert sind, zu entsprechen. Somit ist es möglich, ein System zu schaffen, welches eine leichte und eindeutige Fehlerentscheidung ausführen kann, im Gegensatz zu dem bestehenden Fehlerentscheidungssystem, durch Vornehmen eines Vergleichs zwischen dem Fehlerentscheidungselement und dem Normalzustand/Fehlerzustand-Entscheidungsbereich. Demzufolge ist der Fehler des Luft/Kraftstoff-Verhältnis-Regelungssystems, einschließlich des Fehlers des Luft/Kraftstoff-Verhältnis-Sensors 26 und des Regelfehlers des Mikroprozessors (ECU 41) genau diagnostizierbar, um zu einer Erhöhung der Regelgenauigkeit des Regelungssystems beizutragen.
• (b) Zur Fehlerdiagnose werden weiterhin die Differenz zwischen dem Luft/Kraftstoff-Verhältnis λ, welches durch den Luft/Kraftstoff-Verhältnis-Sensor 26 erfasst wird, und dem Soll-Luft/Kraftstoff-Verhältnis λTG sowie die Differenz zwischen dem Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF und dessen Durchschnittswert FAFAV nacheinander summiert, und diese Summierungsergebnisse (d.h. der λ-Summenwert Tλ, und der FAF-Summenwert TFAF) werden als die Fehlerentscheidungselemente verwendet (d.h., Tλ, ist das erste Fehlerentscheidungselement und TFAF ist das zweite Fehlerentscheidungselement). Somit ist es möglich, die zweckentsprechenden Luft/Kraftstoff-Verhältnisdaten und die Luft/Kraftstoff-Verhältnis-Korrekturdaten zu erhalten. Da die Fehlerdiagnose auf der Grundlage der Summenwerte der einzelnen Daten ausgeführt wird, ist sie kaum durch die äußeren Störungen beeinflusst (z.B. die zeitweiligen Störungen der Sensorausgabe und des Korrekturkoeffizienten).
• (c) Wenn bei dem nicht beanspruchten Beispiel weiterhin die Fehlerentscheidungselemente eine vorbestimmte Anzahl (z.B. 3mal) oder mehrfach entschieden werden, zu den Fehlerentscheidungsbereichen zu gehören, innerhalb einer vorbestimmten Anzahl von Fehlerdiagnosen (z.B. CDG = 5mal), wird schließlich diagnostiziert, dass das Luft/Kraftstoff-Verhältnis-Regelungssystem fehlerhaft ist. In diesem Fall kann die Zuverlässigkeit der Fehlerdiagnose weiter er höht werden als in dem Fall, wenn die Fehlerdiagnose für jede Routine als das Endergebnis verwendet wird.
• (d) In dem Bereich, in welchem der λ-Summenwert Tλ und der FAF-Summenwert TFAF kleiner als die vorbestimmten Werte sind (d.h. Tλ ≤ A und TFAF ≤ C), unter Bezugnahme auf 29, wird eine Bestimmung, dass das Luft/Kraftstoff-Verhältnis-Regelungssystem im Normalzustand ist, unterbunden. D.h., in dem Bereich, in welchem die Veränderung des Luft/Kraftstoff-Verhältnisses λ, und die Veränderung des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF gering sind, ist die Zuverlässigkeit der Fehlerdiagnose unter Verwendung dieser Daten so eingeschränkt, dass der Fehler möglicherweise übersehen werden kann. Wenn die vorstehend erwähnte Entscheidung des Normalzustands unterbunden wird, kann die fälschliche Erfassung der Fehlerdiagnose verhindert werden, um die Zuverlässigkeit zu erhöhen.
• (e) Bei dem nicht beanspruchten Beispiel wird weiterhin der Normalzustand/Fehlerzustand-Entscheidungsbereich "erlernt", um den Normalzustand-Entscheidungsbereich zu vergrößern (d.h. den Bereich von C bis D in 29), wenn eine Erfassung eines Fehlers auf der Grundlage des λ-Summenwerts Tλ erfolgt. Wird andererseits ein Fehler auf der Grundlage des FAF-Summenwerts TFAF erfasst, wird der Normalzustand/Fehlerzustand "erlernt" , um den Normalzustand-Entscheidungsbereich (d.h. den Bereich von A bis B in 29) mit Bezug auf das Luft/Kraftstoff-Verhältnis λ zu vergrößern.

The following effects can be achieved according to the unclaimed example, which is described in detail.

• (a) In the unclaimed example, the difference (ie, the λ sum value Tλ) between the air / fuel ratio λ, which is determined by the air / fuel ratio sensor 26 is detected, and the target air-fuel ratio λTG is calculated, and the difference (ie, the FAF sum value TFAF) between the air-fuel ratio correction coefficient FAF and its average value FAFAV is also calculated. The error of the air-fuel ratio control system is diagnosed by causing those calculated error decision elements (ie, the λ sum value Tλ and the FAF sum value TFAF) to change the normal state / error state decision regions (as in FIG 29 shown) individually defined by the two threshold values. Thus, it is possible to provide a system which can make an easy and unambiguous error decision unlike the existing error decision system by making a comparison between the error decision element and the normal state / error state decision area. Consequently, the error of the air / fuel ratio control system, including the error of the air / fuel ratio sensor 26 and the control error of the microprocessor (ECU 41 ) accurately diagnosable to help increase the control accuracy of the control system.
• (b) Further, for fault diagnosis, the difference between the air / fuel ratio λ, which is determined by the air / fuel ratio sensor 26 and the target air-fuel ratio λTG and the difference between the air-fuel ratio correction coefficient FAF and its average value FAFAV are sequentially summed, and these summation results (ie, the λ sum value Tλ and the FAF sum value TFAF ) are used as the error decision elements (ie, Tλ, is the first error decision element and TFAF is the second error decision element). Thus, it is possible to obtain the appropriate air-fuel ratio data and the air-fuel ratio correction data. Since the fault diagnosis is carried out on the basis of the sum values of the individual data, it is hardly affected by the external disturbances (for example, the intermittent disturbances of the sensor output and the correction coefficient).
• (c) In the unclaimed example, further, if the error decision items are decided a predetermined number (eg, 3 times) or more times to be included in the error decision areas within a predetermined number of failure diagnoses (eg, CDG = 5 times), it is finally diagnosed that the air / Fuel ratio control system is faulty. In this case, the reliability of the fault diagnosis can be further increased than in the case where the fault diagnosis is used for each routine as the final result.
• (d) In the range in which the λ sum value Tλ and the FAF sum value TFAF are smaller than the predetermined values (ie, Tλ ≦ A and TFAF ≦ C), with reference to FIG 29 , a determination that the air-fuel ratio control system is in the normal state is inhibited. That is, in the range in which the change in the air-fuel ratio λ and the change in the air-fuel ratio correction coefficient FAF are small, the reliability of the fault diagnosis using these data is limited so that the error may be overlooked can be. When the above-mentioned decision of the normal state is inhibited, the erroneous detection of the fault diagnosis can be prevented to increase the reliability.
• (e) In the unclaimed example, the normal state / error state decision area is further "learned" to increase the normal state decision area (ie, the area from C to D in FIG 29 ) when detection of an error is made on the basis of the λ sum value Tλ. On the other hand, when an error is detected based on the FAF sum value TFAF, the normal state / error state is "learned" to set the normal state deciding range (ie, the range from A to B in FIG 29 ) with respect to the air / fuel ratio λ.

If the error of the air / fuel ratio sensor 26 or the control error in the ECU 41 occurs It is assumed that the occurrence of both errors is less frequent than the occurrence of an error. When the air / fuel ratio sensor 26 or the ECU 41 faulty, this causes an error indication in the other device. In the above-mentioned configuration, therefore, in the case where only one error is detected, the range for deciding the other normal state is increased to relax the error decision conditions. As a result, the error, if it really occurs, can be diagnosed in a manner to directly reflect it. Furthermore, the error content can be specified more precisely.

(Second unclaimed Example)

A second unclaimed example of the present invention will be described below with reference to FIG 32 - 35 described. In the structure of the further unclaimed example, however, the portions equivalent to those of the above-mentioned unclaimed example are denoted by the same reference numerals while omitting their description but emphasizing the differences from the above-mentioned unclaimed example.

In the present unclaimed example, the error decision processing using the in 33 in the normal state / error state decision curve image in which the λ sum value Tλ and the FAF sum value TFAF are used as the error decision elements. More specifically, in 33 the λ sum value Tλ is divided equally by a number of thresholds into a plurality of areas (eg, sixteen areas in 33 ), and the FAF sum value TFAF is divided equally by a number of thresholds into a plurality of areas (eg, sixteen areas in 33 ). As a result, 16 × 16 = 256 smaller areas are generated in the entire graph.

In this case, the individual small areas (ie, 10 × 10 = 100 small areas) arranged within the thick frame of the same graph are determined as the normal state areas, and the individual small areas outside the thick frame are designated as the error state areas. In the present embodiment, furthermore, the error decision is made as to which of the normal state / error state regions belong to the individual values Tλ and TFAF. This in 33 The graph shown is stored and stored in the backup RAM 45 stored, which acts as a memory.

32 FIG. 12 shows a flowchart of an error decision routine of the present unclaimed example. FIG. This routine is explained by the in 27 shown in the previous unclaimed example (ie, the processing in step 750 in 24 ).

When the in 32 routine is started, the CPU increments 42 in step 901 the processing counter CDG which executes the number of executions of the error decision processing by "1" and decides in the step 902 Whether the processing counter CDG exceeds "5" or not. At the beginning of processing, the answer is in step 902 NO, so the CPU 42 continue to step 903 goes.

The CPU 42 decides in step 903 whether the λ sum value Tλ and the FAF sum value TFAF are within the normal state range of the in 33 are shown graph. Specifically, when Tλ = Tλ1 and TFAF = TFAF1 in the in 33 are shown, a small area P of the normal state area (within the thick frame) is selected so that the normal state is decided. When Tλ = Tλ2 and TFAF = TFAF2, a small area Q within the error condition area (outside the thick frame) is selected so that the error state is decided.

Thus, the values Tλ and TFAF are within the normal range and is the answer in step 903 YES, the CPU goes 42 continue to step 904 to leave an error decision counter CMPDG as is. On the other hand, if the values Tλ and TFAF are within the error condition range and if the answer in step 903 NO is, the CPU goes 42 continue to step 905 to increment an error decision counter CMPFDG by "1".

Then the in 32 shown routine repeatedly executed. If the answer in step 902 Accordingly, the CPU goes 42 continue to step 906 to decide whether the error decision counter CMPDG is not smaller than "3" or not. If CMPDG <3 continues, the CPU resets 42 in step 907 the final error decision flag XDG back to "0". If CAFDG ≥ 3 is, puts the CPU 42 in step 908 the final error decision flag XDG at "1".

Then the CPU continues 42 in step 909 the processing counter CDG back to "0". The CPU 42 put in step 910 returns the error decision counter CMPDG to "0" and ends the present routine.

On the other hand shows 34 a flowchart of an equation learning routine for learning the normal state / error state areas in the in 33 shown graph. This routine is done by the CPU 42 executed for a predetermined period of time. In the present unclaimed example, the in 34 shown routine of the decision area learning device.

In 34 decides the CPU 42 in step 1001 Whether or not the conditions (ie, the learning conditions) for learning the normal state / error state deciding regions of the graph are satisfied. These learning conditions include the case that the small regions to be decided that they are normal or defective are very different each time that both the λ sum value Tλ and the FAF sum value TFAF are obtained at a small value, etc If these learning conditions are met, a proper error decision may not be performed, and the answer in the step 1001 is agreeing because the individual small normal state / error state areas in the in 33 must be learned.

Are the learning conditions met, so the answer in the step 1001 YES, the CPU goes 42 continue to step 1002 to increment a counter CT by "1". The CPU 42 decides in the following step 1003 Whether the numerical value of the counter CT is larger than a predetermined decision value KCT or not. Only if the answer in step 1003 YES, the CPU performs 42 in step 1004 the learning routine of the in 33 shown normal state / error state decision curve image. The CPU 42 sets in the following step 1005 reset the counter CT to "0" and then terminate the present routine. If the answers in the steps 1001 and 1003 NO, the CPU stops 42 the present routine without any action and does not execute the equation learning routine.

An example of learning the normal state / error state equation is in 35A and in 35B in which the inside of a thick frame indicates the normal state area and the outside of the thick frame indicates the error state area. These areas can be updated for all small areas or for the numerous small areas as a whole. At this learning time, the number of small areas for the normal state area can be set (eg in 33 to 100) or not set.

• (a) Bei dem vorliegenden nicht beanspruchten Beispiel wird der Fehler des Luft/Kraftstoff-Verhältnis-Regelungssystems auf der Grundlage des Normalzustand/Fehlerzustand-Entscheidungskurvenbilds entsprechend den Fehlerentscheidungselementen (d.h. des λ-Summenwerts Tλ und des FAF-Summenwerts TFAF) entschieden. Somit kann die Entscheidung der Normalzustand/Fehlerzustand-Entscheidungsbereiche leichter realisiert werden.
• (b) In dem in 33 gezeigten Kurvenbild ist andererseits eine Anzahl (d.h. 16 × 16 = 256) von kleinen Bereichen ausgebildet, welche gleichmäßig unterteilt sind, um einzeln den λ-Summenwerten Tλ und den FAF-Summenwerten TFAF zu entsprechen, so dass die Daten, welche den Normalzustand anzeigen, oder die Daten, welche den Fehlerzustand anzeigen, für die einzelnen kleinen Bereiche gespeichert werden. Somit können nähere Normalzustand/Fehlerzustand-Entscheidungen als in dem Fall getroffen werden, wenn die Normalzustand/Fehlerzustand-Entscheidungsbereiche als größere Bereiche (z.B. 4 × 4 = 16 Bereiche) erzeugt sind.
• (c) Bei dem vorliegenden nicht beanspruchten Beispiel werden weiterhin die Normalzustandsdaten oder die Fehlerzustandsdaten der Normalzustand/Fehlerzustand-Entscheidungsgleichung, wenn notwendig, erlernt. Somit ist eine genauere Fehlerdiagnose ausführbar, selbst wenn ein Fehlerzustand der einen Einrichtung (z.B. der Luft/Kraftstoff-Verhältnis-Sensor 26 oder der ECU 41) den Zustand der anderen Einrichtung beeinflusst, dass diese ein Fehleranzeichen ausweist.

The present unclaimed example can achieve the following effects in addition to those described above.

• (a) In the present unclaimed example, the error of the air-fuel ratio control system is decided on the basis of the normal state / error state decision curve image corresponding to the error decision elements (ie, the λ sum value Tλ and the FAF sum value TFAF). Thus, the decision of the normal state / error state decision areas can be more easily realized.
• (b) In the in 33 On the other hand, a number (ie, 16 × 16 = 256) of small areas which are equally divided to correspond individually to the λ sum values Tλ and the FAF sum values TFAF is formed so that the data indicating the normal state, or the data indicating the error condition is stored for the individual small areas. Thus, closer normal state / error state decisions can be made than in the case where the normal state / error state decision regions are generated as larger regions (eg, 4 × 4 = 16 regions).
• (c) In the present unclaimed example, the normal state data or the error state data of the normal state / error state decision equation is further learned, if necessary. Thus, a more accurate fault diagnosis can be performed even if a fault condition of the one device (eg, the air-fuel ratio sensor 26 or the ECU 41 ) influences the state of the other device that it indicates an error.

• (1) In der ersten Ausführungsform werden die zweiten Ableitungen Δ²λSM und Δ²FAFSM, entsprechend den Beschleunigungen des Luft/Kraftstoff-Verhältnisses und des Luft/Kraftstoff-Verhältnis-Korrekturwerts FAF, unter Verwendung der vorstehend erwähnten Gleichungen (14) und (15) berechnet. Für diese Berechnungen der zweiten Ableitungen können die folgenden Grundgleichungen (33) und (34) verwendet werden: Δ2λSMi = (λSMi – λSMi–n) – (λSMi–n – λSMi–m–n) (33) und Δ2FAFSMi = (FAFSMi – FAFSMi–n) – (FAFSMi–n – FAFSMi–m–n) (34)(wobei m = 1, 2, 3 ... und n = 1, 2, 3 ... sind). Was in der ersten Ausführungsform ausgeführt wurde, ist in den Grundgleichungen als m = 1 und n = 1 einzusetzen, und diese m- und n-Werte können geändert werden.
• (2) In der vorhergehend beschriebenen ersten Ausführungsform wird die Änderungsgeschwindigkeit des erfaßten Luft/Kraftstoff-Verhältnisses beispielhaft durch den Summenwert Δ|Δ²λSM| der zweiten Ableitung Δ²λSM des λSM-Werts dargestellt, und die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten wird durch den Summenwert Σ|Δ²FAFSM| der zweiten Ableitung Δ²FAFSM des FAFSM-Werts beispielhaft dargestellt, und der Sensorfehler wird in Abhängigkeit davon diagnostiziert, ob die folgende Ungleichung (35) erfüllt ist oder nicht: Σ|Δ2FAFSM|/Σ|Δ2λSM| < α (35)Dieser Aufbau ist wie folgt abwandelbar. Die zweite Ableitung Δ²λSM (nicht summiert) des λSM-Werts kann als die Änderungsgeschwindigkeit des erfaßten Luft/Kraftstoff-Verhältnisses verwendet werden, und die zweite Ableitung Δ²FAFSM (nicht summiert) des FAFSM-Werts wird als die Änderungsgeschwindigkeit des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten verwendet, so daß der Sensorfehler diagnostiziert werden kann, abhängig davon, ob die folgende Ungleichung erfüllt ist oder nicht: Δ2FAFSM/Δ2λSM < α (36)In der zweiten bis vierten Ausführungsform kann der Sensorfehler ebenfalls unter Verwendung der Werte diagnostiziert werden, welche für die vorbestimmten Zeiten unsummiert verbleiben. Wenn die Fehlerdiagnose in diesen Ausbildungen aus geführt wird, werden die einzelnen numerischen Werte vor dem Summieren zu Änderungsgeschwindigkeiten, und ihr Verhältnis wird der Fehlerdiagnoseparameter, so daß die Aufgabe der vorliegenden Erfindung durch die Ausbildungen ebenfalls erfüllt werden kann.
• (3) Durch die veränderbaren Einstellungen der Fehlerentscheidungswerte (α, β1, β2, γ, δ1 und δ2) in den vorhergehenden einzelnen Ausführungsformen werden die Fehlerentscheidungswerte verändert, um die Belastungs-Zustandsmengen ΣPM und ΣNe zu erhöhen, doch diese Einstellungen sind abwandelbar, wie nachstehend beschrieben wird. Wie z. B. in 21 gezeigt, kann der Fehlerentscheidungswert α (oder β1, β2 oder γ) für eine größere Kühlwassertemperatur Thw kleiner werden. Wie in 22 gezeigt, können weiterhin die Fehlerentscheidungswerte δ1 und δ2 zum Einstellen des Normalzustandsbereichs für eine größere Kühlwassertemperatur Thw kleiner werden.
• (4) In der vorhergehenden ersten bis vierten Ausführungsform werden die Filteroperationen in den Schritten 303 und 306 der 6 ausgeführt, um die Streuung des Luft/Kraftstoff-Verhältnisses λ und des Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF zwischen den Zylindern der Wärmekraftmaschine auszuschließen. Diese Operationen sind in andere Glättungsoperationen abwandelbar, wie z. B. Glättungsoperationen mit n Stufen (n = 16, 64 oder dergleichen). Weiterhin sind diese Filter- (oder Glättungs-)-Operationen nicht unbedingt wesentlich und können ausgelassen werden, während sie noch zu dem vorliegenden Fehlerdiagnosesystem gehören.
• (5) In der in 6 und 7 gezeigten Routine werden die Zähler CDG1 bis CDG4 verwendet, um die Fehlerentscheidung auszuführen, doch sie können vereinfacht werden. In spezifischer Weise kann der den fortbestehenden Normalzustand entscheidende Zähler CDG4 ausgelassen werden, um den Zustand zu erhalten, welcher einmal beim Setzen des Fehlerentscheidungsflags XDGAF begründet wurde. Weiterhin sind die spezifischen Werte der vorbestimmten Werte KX1 bis KX4 zum Entscheiden der Zählwerte der einzelnen Zähler CDG1 bis CDG4 nicht auf die vorstehend spezifizierten Werte begrenzt, sondern können nach Belieben geändert werden. Um die Sensor-Fehlerdiagnosegenauigkeit zu erhöhen, ist es wirkungsvoll, die Zählzeitdauer des Summierzeitzählers CDG1 (oder des Zählers CDG0 in 16) zu verlängern oder die Anzahl der Zählschritte des Fehlerdiagnose-Ausführungszählers CDG2 und des Fehlerentscheidungszählers CDG3 zu erhöhen.
• (6) Weiterhin sind die vorbestimmten einzelnen Werte KX1 bis KX4 jederzeit gemäß dem Laufzustand der Wärmekraftmaschine veränderlich einstellbar, so daß sie für einen gleichmäßigen Lauf der Wärmekraftmaschine auf relativ kleine Werte eingestellt werden, wogegen sie für einen kurzzeitigen Lauf der Wärmekraftmaschine auf relative große Werte eingestellt werden.
• (7) In der Ausführungsform, in welcher die Fehlerdiagnose abhängig davon ausgeführt wird, ob der Fehlerdiagnoseparameter innerhalb des vorbestimmten Normalzustandsbereichs ist oder nicht, wie in der fünften Ausführungsform beschrieben, kann der Fehlerdiagnoseparameter in die Amplitude der Luft/Kraftstoff-Verhältnis-Schwankung oder das Verhältnis der Amplituden der FAF-Schwankung abgewandelt werden.
• (8) In der vorhergehenden ersten bis vierten Ausführungsform wird die Fehlerdiagnose für den Luft/Kraftstoff-Verhältnis-Sensor (d. h. den Luft/Kraftstoff-Verhältnis-Sensor 26), hergestellt als ein Sauerstoffsensor der Grenzstromtype, ausgeführt. Eine Fehlerdiagnose ähnlich jener der vorhergehenden Ausführungsformen ist auf einen Luft/Kraftstoff-Verhältnis-Sensor anwendbar, welcher als ein Sauerstoffsensor der Pumpstromtype hergestellt ist.
• (9) In der vorhergehenden ersten bis vierten Ausführungsform wird die erfindungsgemäße Sensorfehlerdiagnose durch das Luft/Kraftstoff-Verhältnis-Regelungssystem ausgeführt, welches die Luft/Kraftstoff-Verhältnis-Regelung unter Anwendung der modernen Regelungstheorie realisiert. Jedoch kann die vorliegende Erfindung natürlich auch von einem System gemäß einer anderen Regelung ( z.B. der PID-Regelung) ausgeführt werden.
• (10) In den vorhergehenden nicht beanspruchten Beispielen werden der λ-Summenwert Tλ und der FAF-Summenwert TFAF für die Zeitdauer von 1280 ms berechnet, so dass die Fehlerentscheidung unter Verwendung der Berechnungsergebnisse ausgeführt wird. Die Zeitdauer zur Berechnung des λ-Summenwerts Tλ und des FAF-Summenwerts TFAF kann jedoch verkürzt oder verlängert werden. Wie z.B. in 36 gezeigt, ist die Zeitdauer zum Berechnen des λ-Summenwerts Tλ und des FAF-Summenwerts TFAF gemäß diesen Summenwerten veränderbar (wie gezeigt auf 320 ms, 640 ms und 1280 ms). Dann ist es für den in längerer Zeitdauer berechneten λ-Summenwert Tλ und den FAF-Summenwert TFAF wahrscheinlicher, dass der Wärmekraftmaschinen-Laufzustand kurzzeitig ist, so dass die Summierzeitdauer verlängert wird.
• (11) In den vorhergehenden nicht beanspruchten Beispielen ist die Anzahl der Ausführungen der Fehlerentscheidung gleich 5 (wie mit Bezug auf die Routine in 24 gezeigt ist). Wird der Fehler 3mal oder häufiger für die fünf Ausführungen entschieden, erfolgt schließlich die Entscheidung, dass das Luft/Kraftstoff-Verhältnis-Regelungssystem fehlerhaft ist. Die Anzahl der Ausführungen der Fehlerentscheidung ist änderbar, wie in 37 gezeigt ist. In 37 ist die Anzahl der Ausführungen auf 3, 5 und 7 eingestellt, gemäß dem λ-Summenwert Tλ und dem FAF-Summenwert TFAF. Für die größeren Werte gemäß Tλ und TFAF wird eingeschätzt, dass der Wärmekraftmaschinen-Laufzustand kurzzeitiger ist, so dass die Anzahl der Ausführungen größer wird. In dieser Abwandlung kann die Anzahl der Entscheidungen (der Schritte 756 und 762 in 24) des λ-Fehlers und des FAF-Fehlers bei der abschließenden Entscheidung, dass das Luft/Kraftstoff-Verhältnis-Regelungssystem fehlerhaft ist, bei 3 Ausführungen belassen oder verändert werden.
• (12) In den vorhergehenden nicht beanspruchten Beispielen werden der λ-Summenwert Tλ und der FAF-Summenwert TFAF als das erste und das zweite Fehlerentscheidungselement verwendet, doch sie können geändert werden. Z.B. werden die Differenz zwischen dem Luft/Kraftstoff-Verhältnis λ, welches durch den Luft/Kraftstoff-Verhältnis-Sensor 26 erfasst wird, und dem Soll-λTG sowie die Differenz zwischen dem Luft/Kraftstoff-Verhältnis-Korrekturkoeffizienten FAF und dessen Durchschnittswert FAFAV berechnet und jeweils als das erste Fehlerentscheidungselement und als das zweite Fehlerentscheidungselement verwendet. Gemäß diesen einzelnen Fehlerentscheidungselementen kann weiterhin der Fehler des Luft/Kraftstoff-Verhältnis-Regelungssystems diagnostiziert werden. In dieser Abwandlung kann der Fehler gemäß den Schwellenwerten, wie bei dem ersten nicht beanspruchten Beispiel, oder auf der Grundlage der Gleichung, wie bei dem zweiten nicht beanspruchten Beispiel, entschieden werden.
• (13) In den vorhergehenden nicht beanspruchten Beispielen werden die Schwellenwerte zum Entscheiden des Normalzustands/Fehlerzustands als eine Vielzahl für jeweils den λ-Summenwert Tλ, und den FAF-Summenwert TFAF erzeugt, können jedoch auf einen vermindert werden. Dieselben Schwellenwerte werden einzeln für den λ-Summenwert Tλ und den FAF-Summenwert TFAF erzeugt, aber es kann eine unterschiedliche Anzahl von Schwellenwerten vorgesehen werden.
• (14) In dem zweiten nicht beanspruchten Beispiel ist die Anzahl der Schwellenwerte zum Definieren der Bereiche des Normalzustand/Fehlerzustand-Entscheidungskurvenbilds veränderbar, um die Anzahl der zu definierenden Bereiche zu erhöhen oder zu vermindern. Wird z.B. die Anzahl der Bereiche reduziert, nimmt die Entscheidungsgenauigkeit mehr oder weniger ab, doch die Speicherkapazität kann vermindert werden.
• (15) Ist einmal das abschließende Fehlerentscheidungsflag XDG bestimmt, braucht die Fehlerentscheidungsroutine nicht mehr ausgeführt zu werden, um die CPU 42 von Operationen zu entlasten. In diesem Fall wird zu den in 24 ge zeigten Fehlerentscheidungsbedingungen eine Bedingung hinzugefügt, um zu bestätigen, dass das abschließende Fehlerentscheidungsflag XDG nicht bestimmt ist. Wenn dieses abschließende Fehlerentscheidungsflag XDG bestimmt ist, wird die vorliegende Routine ohne jede Handlung beendet, während die nachfolgende Routine umgangen wird.
• (16) In den vorhergehenden nicht beanspruchten Beispielen wird die Sensorfehlerdiagnose der vorliegenden Erfindung durch das Luft/Kraftstoff-Verhältnis-Regelungssystem ausgeführt, welches die Luft/Kraftstoff-Verhältnis-Regelung unter Anwendung der gegenwärtigen Regelungstheorie realisiert. Jedoch kann die vorliegende Erfindung natürlich durch ein anderes Regelungsverfahren (z.B. die PID-Regelung) ausgeführt werden. Derartige Änderungen und Abwandlungen sind als in dem Rahmen der Erfindung fallend anzusehen, welche durch die beigefügten Ansprüche definiert ist.

Although the present invention has been described in detail in connection with the preferred embodiments and with reference to the accompanying drawings, it should be understood that various changes and modifications are suggested to those skilled in the art. For example:

• (1) In the first embodiment, the second derivatives Δ ² λSM and Δ ² FAFSM become the accelerations of the air-fuel ratio and the air-fuel ratio correction value FAF calculated using the aforementioned equations (14) and (15). For these second derivative calculations, the following basic equations (33) and (34) can be used: Δ 2 λSM i = (λSM i - λSM in ) - (λSM in - λSM i-m-n ) (33) and Δ 2 FAFSM i = (FAFSM i - FAFSM in ) - (FAFSM in - FAFSM i-m-n ) (34) (where m = 1, 2, 3 ... and n = 1, 2, 3 ...). What was done in the first embodiment is to be substituted in the basic equations as m = 1 and n = 1, and these m and n values can be changed.
• (2) In the above-described first embodiment, the rate of change of the detected air-fuel ratio is exemplified by the sum value Δ | Δ ² λSM | of the second derivative Δ ² λSM of the λSM value, and the rate of change of the air-fuel ratio correction coefficient is represented by the sum value Σ | Δ ² FAFSM | of the second derivative Δ ² FAFSM of the FAFSM value, and the sensor error is diagnosed depending on whether the following inequality (35) is satisfied or not: Σ | Δ 2 FAFSM | / Σ | Δ 2 λSM | <α (35) This structure can be modified as follows. The second derivative Δ ² λSM (not summed) of the λSM value may be used as the rate of change of the detected air / fuel ratio, and the second derivative Δ ² FAFSM (not summed) of the FAFSM value is taken as the rate of change of the air / Fuel ratio correction coefficient used so that the sensor error can be diagnosed, depending on whether the following inequality is satisfied or not: Δ 2 FAFSM / Δ 2 λSM <α (36) In the second to fourth embodiments, the sensor error may also be diagnosed using the values which remain unsummed for the predetermined times. When the failure diagnosis is performed in these embodiments, the individual numerical values before being added become rates of change, and their ratio becomes the fault diagnosis parameter, so that the object of the present invention can also be accomplished by the embodiments.
• (3) By the changeable settings of the error decision values (α, β1, β2, γ, δ1, and δ2) in the previous individual embodiments, the error decision values are changed to increase the load state quantities ΣPM and ΣNe, but these settings are modifiable will be described below. Such as In 21 2, the error decision value α (or β1, β2, or γ) may become smaller for a larger cooling water temperature Thw. As in 22 Further, the error decision values δ1 and δ2 for setting the normal state region for a larger cooling water temperature Thw can be made smaller.
• (4) In the foregoing first to fourth embodiments, the filtering operations in the steps 303 and 306 of the 6 executed to exclude the dispersion of the air / fuel ratio λ and the air / fuel ratio correction coefficient FAF between the cylinders of the heat engine. These operations are modifiable into other smoothing operations, such as: B. smoothing operations with n stages (n = 16, 64 or the like). Furthermore, these filtering (or smoothing) operations are not necessarily essential and may be omitted while still belonging to the present fault diagnosis system.
• (5) In the 6 and 7 The counters CDG1 to CDG4 are used to execute the error decision, but they can be simplified. Specifically, the persistent normal state counter CDG4 may be skipped to obtain the state which was established once upon setting the error decision flag XDGAF. Further, the specific values of the predetermined values KX1 to KX4 for deciding the counts of the individual counters CDG1 to CDG4 are not limited to the above-specified values, but may be changed as desired. In order to increase the sensor fault diagnosis accuracy, it is effective to set the count period of the summing time counter CDG1 (or the counter CDG0 in FIG 16 ) or to increase the number of counts of the error diagnostic execution counter CDG2 and the error decision counter CDG3.
• (6) Furthermore, the predetermined individual values KX1 to KX4 are changeably adjustable at any time according to the running state of the heat engine, so that they are suitable for smooth running of the thermal power be set to relatively small values, whereas they are set to relatively large values for a short run of the heat engine.
• (7) In the embodiment in which the failure diagnosis is carried out depending on whether the fault diagnosis parameter is within the predetermined normal state range or not, as described in the fifth embodiment, the fault diagnosis parameter may be included in the amplitude of the air-fuel ratio fluctuation Ratio of the amplitudes of the FAF fluctuation are modified.
• (8) In the foregoing first to fourth embodiments, the fault diagnosis is made for the air-fuel ratio sensor (ie, the air-fuel ratio sensor 26 ) made as a limit current type oxygen sensor. A failure diagnosis similar to that of the previous embodiments is applicable to an air-fuel ratio sensor manufactured as a pump-type oxygen sensor.
• (9) In the foregoing first to fourth embodiments, the sensor failure diagnosis according to the invention is carried out by the air-fuel ratio control system that realizes the air-fuel ratio control using the modern control theory. However, the present invention may of course also be implemented by a system according to another scheme (eg PID control).
• (10) In the foregoing unclaimed examples, the λ sum value Tλ and the FAF sum value TFAF are calculated for the period of 1280 ms, so that the error decision is made using the calculation results. However, the period for calculating the λ sum value Tλ and the FAF sum value TFAF may be shortened or lengthened. Like in 36 1, the time period for calculating the λ sum value Tλ and the FAF sum value TFAF is variable according to these sum values (as shown, at 320 ms, 640 ms and 1280 ms). Then, for the λ sum value Tλ calculated in a longer period of time and the FAF sum value TFAF, it is more likely that the heat engine running state is short, so that the summing time period is prolonged.
• (11) In the preceding unclaimed examples, the number of executions of the error decision is 5 (as with reference to the routine in FIG 24 is shown). If the error is decided 3 times or more frequently for the five embodiments, the final decision is made that the air / fuel ratio control system is faulty. The number of executions of the error decision is changeable, as in 37 is shown. In 37 For example, the number of executions is set to 3, 5 and 7 according to the λ sum value Tλ and the FAF sum value TFAF. For the larger values according to Tλ and TFAF, it is estimated that the heat engine running state is shorter, so that the number of executions becomes larger. In this modification, the number of decisions (the steps 756 and 762 in 24 ) of the λ error and the FAF error in the final decision that the air / fuel ratio control system is faulty, left in 3 versions or changed.
• (12) In the foregoing unclaimed examples, the λ sum value Tλ and the FAF sum value TFAF are used as the first and second error decision elements, but they may be changed. For example, the difference between the air / fuel ratio λ, which is determined by the air / fuel ratio sensor 26 is detected, and the target λTG and the difference between the air-fuel ratio correction coefficient FAF and its average value FAFAV are calculated and respectively used as the first error decision element and as the second error decision element. Further, according to these individual error decision elements, the error of the air-fuel ratio control system can be diagnosed. In this modification, the error may be decided according to the threshold values as in the first unclaimed example or based on the equation as in the second unclaimed example.
• (13) In the above unclaimed examples, the thresholds for deciding the normal state / error state are generated as a plurality for each of the λ sum value Tλ, and the FAF sum value TFAF, but may be reduced to one. The same threshold values are generated individually for the λ sum value Tλ and the FAF sum value TFAF, but a different number of threshold values may be provided.
• (14) In the second unclaimed example, the number of thresholds for defining the regions of the normal state / error state decision curve image is changeable to increase or decrease the number of regions to be defined. For example, if the number of areas is reduced, the decision accuracy more or less decreases, but the storage capacity can be reduced.
• (15) Once the final error decision flag XDG is determined, the error decision routine no longer needs to be executed to the CPU 42 to relieve of operations. In this case, the in 24 Error decision conditions showed a condition added to confirm that the final error decision flag XDG is not determined. When this final error decision flag XDG is determined, the present routine becomes without any action finished while bypassing the subsequent routine.
• (16) In the foregoing unclaimed examples, the sensor failure diagnosis of the present invention is performed by the air-fuel ratio control system that realizes the air-fuel ratio control using the current control theory. However, of course, the present invention may be carried out by another control method (eg, PID control). Such changes and modifications are to be considered as falling within the scope of the invention, which is defined by the appended claims.

Claims (10)

Air / fuel ratio sensor failure diagnostic system comprising: - an air / fuel ratio sensor ( 26 ) for generating a signal representing a detected air / fuel ratio λ of an internal combustion engine ( 1 ), an air-fuel ratio correction coefficient generation means (FIG. 42 , P3, 103 ) for generating an air-fuel ratio correction coefficient (FAF) according to a difference between the detected air-fuel ratio λ and a target air-fuel ratio λTG, an air-fuel ratio change calculating means (FIG. 42 . 304 . 314 . 311 ) for calculating a change speed Δ ² SM of the detected air-fuel ratio λ, - a correction coefficient change calculating means (Fig. 42 . 307 . 315 . 312 ) for calculating a rate of change Δ ² FAFSM of the air / fuel ratio correction coefficient FAF, and - a fault diagnosis device ( 42 . 203 - 222 . 300 ) for diagnosing a fault of the air / fuel ratio sensor ( 26 ) based on a comparison between the air / fuel ratio change rate Δ ² SM and the correction coefficient change rate Δ ² FAFSM. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said air-fuel ratio change calculation means (10) 42 304 ) is provided for determining the rate of change Δ ² λSM of the air / fuel ratio based on a second derivative of the air / fuel ratio λ. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said correction coefficient change calculating means (14) comprises: 42 . 307 ) is provided for determining the rate of change Δ ² FAFSM of the air-fuel ratio correction coefficient based on a second derivative of the correction coefficient FAF. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said air-fuel ratio change calculation means (10) 42 . 304 ) is provided for calculating the change speed Δ ² λSM of the air-fuel ratio based on an amplitude of the air-fuel ratio change speed, and the correction coefficient change calculating means for calculating the change speed Δ ² FAFSM of the air-fuel ratio correction coefficient is provided based on an amplitude of the correction coefficient change rate. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said air-fuel ratio change calculation means (10) 42 . 304 314 ) An institution ( 42 . 314 . 311 ) for summing a second derivative Δ ² λSM of the air / fuel ratio λ for a predetermined period of time to obtain the sum value Σ | Δ ² λSM | as the air-fuel ratio change rate, the correction coefficient change calculating means (FIG. 42 . 307 . 315 . 312 ) An institution ( 42 . 315 . 312 ) for summing the second derivative Δ ² FAFSM of the air-fuel ratio correction coefficient FAF for a predetermined time to have the sum value Σ | Δ ² FAFSM | as the correction coefficient change rate, and the fault diagnosis device ( 42 . 203 - 222 . 300 ) An institution ( 42 . 204 ) for comparing a ratio between the summed second derivative Σ | Δ ² λSM | the air / fuel ratio and the summed second Ableitung Σ | Δ ² FAFSM | of the air-fuel ratio correction coefficient having a predetermined error decision value α to detect an error of the air-fuel ratio sensor (FIG. 26 ) to diagnose. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said air-fuel ratio change calculation means (10) 42 . 304 . 314 . 311 ) a device for Summing the amplitude of the air / fuel ratio change for a predetermined period of time to determine the sum value as the air / fuel ratio change rate Δ ² λSM, the correction coefficient change calculating means (FIG. 42 . 307 . 315 . 312 ) comprises means for summing the amplitude of the air / fuel ratio correction coefficient change for a predetermined period of time to determine the sum value as the correction coefficient change rate Δ ² FAFSM, and the fault diagnosis means (14 42 . 202 - 222 . 300 ) comprises means for comparing a ratio between the summed amplitude of the air-fuel ratio change and the summed amplitude of the air-fuel ratio correction coefficient change with a predetermined error decision value to determine an error of the air-fuel ratio sensor (FIG. 26 ) to diagnose. An air-fuel ratio sensor failure diagnostic system according to claim 1, further comprising an error decision value setting means (Fig. 42 . 250 . 251 ) for determining the error decision value β1, β2 which is variable with a running state of the heat engine. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said fault diagnosis means ( 42 . 200 - 222 . 300 ) is provided to diagnose that the sensor ( 26 ) is in the normal state when the ratio between the rate of change Δ ² λSM of the air / fuel ratio and the rate of change Δ ² FAFSM of the air / fuel ratio correction coefficient is less than a predetermined error decision value α. An air-fuel ratio sensor failure diagnostic system according to claim 1, wherein said fault diagnosis means ( 42 . 200 - 222 . 300 ) is provided to diagnose that the sensor ( 26 ) is in the normal state when the ratio between the rate of change Δ ² λSM of the air / fuel ratio and the rate of change Δ ² FAFSM of the air / fuel ratio correction coefficient is greater than a first error decision value δ1 and less than a second error decision value δ2. The air-fuel ratio sensor failure diagnostic system according to any one of claims 1 to 9, wherein the air-fuel ratio change calculation means (Fig. 42 . 304 ) for calculating the rate of change Δ ² λSM of the air-fuel ratio λ by smoothing the air-fuel ratio λ and using the smoothed air-fuel ratio λSM to calculate the air-fuel ratio Δ ² λSM , and the correction coefficient change calculating means (FIG. 42 . 307 ) ² FAFSM the correction coefficient FAF is provided by smoothing the correction coefficient FAF, and for using the smoothed correction coefficient Δ ² FAFSM for calculating the rate of change Δ to the correction coefficient-changing ratio to calculate Δ ² FAFSM.