JP2009127595A - Abnormality diagnostic system of air-fuel ratio sensor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suitably detect abnormality of an individual characteristic of an air-fuel ratio sensor. <P>SOLUTION: A system up to the air-fuel ratio sensor 17 from a fuel injection valve 12 is modeled by a primary delay element, and a parameter in the primary delay element is identified based on the input air-fuel ratio and the output air-fuel ratio to this model. Abnormality of a predetermined characteristic of the air-fuel ratio sensor is determined based on the identified parameter. When identifying the parameter, a correction is made for removing a bias between the input air-fuel ratio and the output air-fuel ratio. The abnormality of the predetermined characteristic of the air-fuel ratio sensor, not mere abnormality of the air-fuel ratio sensor can be specified and detected. Identification can be performed after removing the bias between the input air-fuel ratio and the output air-fuel ratio, and identification accuracy and diagnostic accuracy can be improved. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an apparatus for diagnosing abnormality of an air-fuel ratio sensor that detects an air-fuel ratio of exhaust gas of an internal combustion engine.

  In an internal combustion engine equipped with an exhaust gas purification system using a catalyst, in order to effectively remove harmful components of exhaust gas by the catalyst, the mixture ratio of air and fuel in the air-fuel mixture burned in the internal combustion engine, that is, the air-fuel ratio is reduced. Control is essential. In order to perform such control of the air-fuel ratio, an air-fuel ratio sensor for detecting the air-fuel ratio is provided in the exhaust passage of the internal combustion engine based on the concentration of a specific component of the exhaust gas, and the detected air-fuel ratio is set to a predetermined target air-fuel ratio. Feedback control is carried out so that it can approach.

  By the way, if an abnormality such as deterioration or failure occurs in the air-fuel ratio sensor, accurate air-fuel ratio feedback control cannot be performed, and exhaust gas emission deteriorates. Therefore, it has been conventionally performed to diagnose abnormality of the air-fuel ratio sensor. In particular, in the case of an engine mounted on an automobile, it is requested by the laws and regulations of each country to detect an abnormality in the air-fuel ratio sensor in an on-board state (onboard) in order to prevent traveling in a state where exhaust gas has deteriorated. ing.

  Patent Document 1 discloses an abnormality of an air-fuel ratio sensor that detects an abnormality of the air-fuel ratio sensor based on a trajectory length or area of an air-fuel ratio sensor output that periodically increases and decreases by open-loop control. A detection device is disclosed.

JP 2005-30358 A

  However, with the technique described in Patent Document 1, although the abnormality of the air-fuel ratio sensor itself can be specified, the details of the abnormality cannot be specified. That is, the air-fuel ratio sensor includes a plurality of characteristics, but it is impossible to specify which of these is abnormal.

  Therefore, the present invention has been made in view of such circumstances, and an object of the present invention is to provide an abnormality diagnosis device for an air-fuel ratio sensor that can preferably detect abnormality of individual characteristics included in the air-fuel ratio sensor. It is in.

According to one aspect of the invention,
An air-fuel ratio sensor abnormality diagnosis device for detecting an air-fuel ratio of exhaust gas of an internal combustion engine,
Identifying means for modeling a system from a fuel injection valve to an air-fuel ratio sensor with a first-order lag element, and identifying a parameter in the first-order lag element based on an input air-fuel ratio and an output air-fuel ratio for the model;
An abnormality determining means for determining an abnormality of a predetermined characteristic of the air-fuel ratio sensor based on the parameter identified by the identifying means;
Active control means for executing active control for forcibly oscillating the input air-fuel ratio at the time of identification by the identification means;
Bias correcting means for correcting the input air-fuel ratio and the output air-fuel ratio so as to remove a bias between the input air-fuel ratio and the output air-fuel ratio at the time of identification by the identifying means, the input air-fuel ratio and the output Bias correction means for calculating an average value of a rich side value and a lean side value per vibration period as a bias value for each air-fuel ratio, and subtracting the bias value from actual data to correct the actual data; An abnormality diagnosis device for an air-fuel ratio sensor is provided.

  According to this, the abnormality of the predetermined characteristic of the air-fuel ratio sensor is determined instead of simply determining the abnormality of the air-fuel ratio sensor. Therefore, the abnormality of the characteristic can be specified and detected, and the abnormality diagnosis of the air-fuel ratio sensor can be executed more precisely and in detail. In addition to this, when there is a bias (deviation) between the input air-fuel ratio and the output air-fuel ratio, the identification can be performed after removing the bias, and the identification accuracy and the diagnostic accuracy can be improved.

  Preferably, the bias correction unit obtains an average value of the rich side values by accumulating the rich side values at every sampling interval and dividing the result by the number of samples per vibration period, and also at the lean side. The average value of the values on the lean side is obtained by accumulating the values at every sampling interval and dividing this by the number of samples, and the average value of the average value of these rich side values and the average value of the lean side values is obtained. Thus, the bias value is calculated, and the bias value is updated every half cycle.

According to another aspect of the present invention,
An air-fuel ratio sensor abnormality diagnosis device for detecting an air-fuel ratio of exhaust gas of an internal combustion engine,
Identifying means for modeling a system from a fuel injection valve to an air-fuel ratio sensor with a first-order lag element, and identifying a parameter in the first-order lag element based on an input air-fuel ratio and an output air-fuel ratio for the model;
An abnormality determining means for determining an abnormality of a predetermined characteristic of the air-fuel ratio sensor based on the parameter identified by the identifying means;
An abnormality diagnosis apparatus for an air-fuel ratio sensor is provided, wherein the identification means identifies the parameter including a bias component corresponding to a bias between the input air-fuel ratio and the output air-fuel ratio.

  This also makes it possible to identify and detect an abnormality in a predetermined characteristic of the air-fuel ratio sensor, and to perform an abnormality diagnosis of the air-fuel ratio sensor more precisely and in detail. In addition to this, since parameter identification is performed including a bias component corresponding to a bias between the input air-fuel ratio and the output air-fuel ratio, the parameter identification accuracy can be improved.

  In this case, preferably, the identification unit simultaneously identifies the parameter and the bias component using a Kalman filter.

  In addition, preferably, active control means for executing active control for forcibly oscillating the input air-fuel ratio at the time of identification by the identification means is provided.

  Preferably, the parameter includes a gain and a time constant, and the predetermined characteristic of the air-fuel ratio sensor includes an output corresponding to the gain and a responsiveness corresponding to the time constant.

  This makes it possible to detect abnormalities in output and responsiveness, which are particularly important characteristics of the air-fuel ratio sensor.

  According to the present invention, an excellent effect is exhibited that abnormalities in individual characteristics included in the air-fuel ratio sensor can be suitably detected.

  The best mode for carrying out the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a schematic view of an internal combustion engine according to the present embodiment. As shown in the figure, the internal combustion engine 1 generates power by burning a mixture of fuel and air inside a combustion chamber 3 formed in a cylinder block 2 and reciprocating a piston 4 in the combustion chamber 3. To do. The internal combustion engine 1 of the present embodiment is a vehicular multi-cylinder engine (for example, a four-cylinder engine, only one cylinder is shown), and is a spark ignition internal combustion engine, more specifically a gasoline engine.

  In the cylinder head of the internal combustion engine 1, an intake valve Vi for opening and closing the intake port and an exhaust valve Ve for opening and closing the exhaust port are provided for each cylinder. Each intake valve Vi and each exhaust valve Ve are opened and closed by a camshaft (not shown). A spark plug 7 for igniting the air-fuel mixture in the combustion chamber 3 is attached to the top of the cylinder head for each cylinder.

  The intake port of each cylinder is connected to a surge tank 8 serving as an intake air collecting chamber via a branch pipe for each cylinder. An intake pipe 13 that forms an intake manifold passage is connected to the upstream side of the surge tank 8, and an air cleaner 9 is provided at the upstream end of the intake pipe 13. An air flow meter 5 for detecting the intake air amount and an electronically controlled throttle valve 10 are incorporated in the intake pipe 13 in order from the upstream side. An intake passage is formed by the intake port, the branch pipe, the surge tank 8 and the intake pipe 13.

  An injector (fuel injection valve) 12 that injects fuel into the intake passage, particularly into the intake port, is provided for each cylinder. The fuel injected from the injector 12 is mixed with intake air to form an air-fuel mixture. The air-fuel mixture is sucked into the combustion chamber 3 when the intake valve Vi is opened, compressed by the piston 4, and ignited and burned by the spark plug 7. It is done.

On the other hand, the exhaust port of each cylinder is connected to an exhaust pipe 6 forming an exhaust collecting passage through a branch pipe for each cylinder. An exhaust passage is formed by the exhaust port, the branch pipe, and the exhaust pipe 6. Catalysts 11 and 19 made of a three-way catalyst are attached to the exhaust pipe 6 on the upstream side and the downstream side. Air-fuel ratio sensors 17 and 18 for detecting the air-fuel ratio of the exhaust gas, that is, pre-catalyst sensors and post-catalyst sensors 17 and 18 are installed at positions before and after the upstream catalyst 11, respectively. These pre-catalyst sensor and post-catalyst sensors 17 and 18 are air-fuel ratio sensors that detect the air-fuel ratio based on the oxygen concentration in the exhaust gas. The pre-catalyst sensor 17 is a so-called wide-range air-fuel ratio sensor, can continuously detect an air-fuel ratio over a relatively wide range, and generates an output proportional to the air-fuel ratio. On the other hand, the post-catalyst sensor 18 is a so-called O ₂ sensor, and has a characteristic that the output changes suddenly at the theoretical air-fuel ratio.

  The spark plug 7, the throttle valve 10, the injector 12, and the like described above are electrically connected to an electronic control unit (hereinafter referred to as ECU) 20 as control means. The ECU 20 includes a CPU, a ROM, a RAM, an input / output port, a storage device, and the like, all not shown. In addition to the air flow meter 5, the pre-catalyst sensor 17, and the post-catalyst sensor 18, the ECU 20 includes a crank angle sensor 14 that detects the crank angle of the internal combustion engine 1 and an accelerator that detects the accelerator opening, as shown in the figure. The opening sensor 15 and other various sensors are electrically connected via an A / D converter or the like (not shown). The ECU 20 controls the ignition plug 7, the throttle valve 10, the injector 12, etc. so as to obtain a desired output based on the detection values of various sensors, etc., and the ignition timing, fuel injection amount, fuel injection timing, throttle opening. Control the degree etc. The throttle opening is normally controlled to an opening corresponding to the accelerator opening.

  The catalysts 11 and 19 simultaneously purify NOx, HC and CO when the air-fuel ratio A / F of the exhaust gas flowing into the catalyst 11 and 19 is the stoichiometric air-fuel ratio (stoichiometric, for example, A / F = 14.6). In response to this, the ECU 20 controls the air-fuel ratio so that the air-fuel ratio A / F of the exhaust gas flowing into the catalysts 11 and 19 becomes equal to the stoichiometric air-fuel ratio during normal operation of the internal combustion engine (so-called stoichiometric). control). Specifically, the ECU 20 sets a target air-fuel ratio A / Ft that is equal to the stoichiometric air-fuel ratio, and makes the basic injection amount such that the air-fuel ratio of the air-fuel mixture flowing into the combustion chamber 3 matches the target air-fuel ratio A / Ft. Is calculated. Then, the basic injection amount is feedback-corrected according to the difference between the actual air-fuel ratio detected by the pre-catalyst sensor 17 and the target air-fuel ratio A / Ft, and the injector 12 is operated for the energization time corresponding to the corrected injection amount. Energize (turn on). As a result, the air-fuel ratio of the exhaust gas supplied to the catalysts 11 and 19 is maintained near the stoichiometric air-fuel ratio, and the maximum purification performance is exhibited in the catalysts 11 and 19. Thus, the ECU 20 feedback-controls the air-fuel ratio (fuel injection amount) so that the actual air-fuel ratio detected by the pre-catalyst sensor 17 approaches the target air-fuel ratio A / Ft. The post-catalyst sensor 18 is provided to correct the air-fuel ratio deviation in such air-fuel ratio feedback control.

  Next, abnormality diagnosis of the air-fuel ratio sensor in the present embodiment will be described. The object of diagnosis in this embodiment is an air-fuel ratio sensor installed on the upstream side of the upstream catalyst 11, that is, a pre-catalyst sensor 17. However, the present invention can also be applied to an air-fuel ratio sensor installed downstream of the upstream catalyst 11, that is, a post-catalyst sensor 18.

  In the abnormality diagnosis, the system from the injector 12 to the pre-catalyst sensor 17 is modeled by a first-order lag element, and parameters in the first-order lag element are identified (estimated) based on the input air-fuel ratio and the output air-fuel ratio for the model. . Based on the identified parameters, an abnormality in a predetermined characteristic of the pre-catalyst sensor 17 is determined.

  As the input air-fuel ratio, a ratio Ga / Q between the fuel injection amount Q calculated based on the energization time of the injector 12 and the intake air amount Ga calculated based on the output of the air flow meter 5 is used. Hereinafter, the input air-fuel ratio is simply referred to as input and is represented by u (t) (u (t) = Ga / Q). On the other hand, the air-fuel ratio converted from the output value of the pre-catalyst sensor 17 is used as the output air-fuel ratio. Hereinafter, the output air-fuel ratio is simply referred to as output, and is represented by y (t).

  As shown in FIG. 2, in the present embodiment, active control for forcibly oscillating the input u (t) is executed when the parameter is identified. In this active control, the target air-fuel ratio A / Ft, and hence the input u (t), is vibrated at a constant period so as to swing by the same amplitude on the lean side and the rich side with a predetermined center air-fuel ratio A / Fc as a boundary. Along with this, the air-fuel ratio detected by the pre-catalyst sensor 17, that is, the output y (t) is vibrated so as to follow the vibration of the input u (t). The center air-fuel ratio A / Fc in the vibration of the target air-fuel ratio A / Ft and the input u (t) is made equal to the theoretical air-fuel ratio, and the amplitude of the vibration is larger than that in the normal air-fuel ratio control. 5 etc.

  The reason for executing this active control is to make it easier to diagnose the abnormality of the pre-catalyst sensor 17 by changing the air-fuel ratio more drastically than usual. In addition, since the active control is executed during the steady operation of the engine, each control amount and each detected value is stabilized and diagnostic accuracy is improved. However, abnormality diagnosis may be executed during normal air-fuel ratio control.

  As shown in the figure, the input u (t) is a stepped waveform, whereas the output y (t) is a waveform with a first-order lag. In the figure, L is a dead time based on a transport delay from the input u (t) to the output y (t). This dead time L corresponds to the time difference from the time when the air-fuel mixture existing in the combustion chamber 3 in the cylinder burns to the time when the exhaust gas based on this combustion reaches the pre-catalyst sensor 17. Practically, the start point of the dead time L can be set, for example, at the time of ignition or when the exhaust valve is opened. However, since the time from ignition to opening the exhaust valve is extremely short compared to the entire dead time, there is no particular problem in terms of accuracy even if it is set at any time.

  Assuming that the dead time L is zero for simplification, the first-order lag element is represented by G (s) = k / (1 + Ts). Here, k is the gain of the pre-catalyst sensor 17, and T represents the time constant of the pre-catalyst sensor 17. The gain k is a value related to output among the characteristics of the pre-catalyst sensor 17, while the time constant T is a value related to responsiveness among the characteristics of the pre-catalyst sensor 17. In FIG. 2, a solid line representing the output y (t) indicates a case where the pre-catalyst sensor 17 is normal. On the other hand, when an abnormality occurs in the output characteristics of the pre-catalyst sensor 17, the gain k becomes larger than normal, and the sensor output increases (expands) as indicated by a, or the gain k becomes smaller than normal. As shown by b, the sensor output decreases (reduces). Therefore, an increase abnormality or a decrease abnormality of the sensor output can be specified by comparing the identified gain k with a predetermined value. On the other hand, if an abnormality occurs in the responsiveness of the pre-catalyst sensor 17, in most cases, the time constant T becomes larger than normal, and the sensor output comes out with a delay as shown by c. Therefore, the responsiveness abnormality of the sensor can be specified by comparing the identified time constant T with a predetermined value.

  In the present embodiment, at least two parameters of gain k and time constant T are identified at the same time, and two abnormalities corresponding to the sensor output abnormality and sensor response abnormality are simultaneously and individually determined. In this way, it is possible to preferably detect abnormality of individual characteristics included in the pre-catalyst sensor 17.

  Next, a method for identifying the gain k and time constant T executed by the ECU 20 will be described.

Equation (20) is a function of the values at the current sample time t and the previous sample time t−1, and this equation means that b ₁ and b ₂ are based on the current value and the previous value, that is, T and k are updated every time. Thus, the time constant T and the gain k are identified sequentially and simultaneously by the sequential least square method. In this sequential identification method, a large number of sample data is acquired and temporarily stored, and the calculation load can be reduced as compared with the method of performing identification, and the capacity of the buffer for temporarily storing data can be reduced. It is suitable for mounting on an ECU (particularly an automotive ECU).

  Using the gain k and time constant T thus identified, the ECU 20 performs an abnormality determination as follows. First, when the identified time constant T is larger than a predetermined time constant abnormality determination value Ts, a response delay occurs, and it is determined that the pre-catalyst sensor 17 is responsive abnormality. On the other hand, when the identified time constant T is equal to or less than the time constant abnormality determination value Ts, the pre-catalyst sensor 17 is determined to be normal with respect to responsiveness.

  If the identified gain k is greater than the predetermined gain increase abnormality determination value ks1, the pre-catalyst sensor 17 is determined to have an output increase abnormality, and the identified gain k is the predetermined gain reduction abnormality determination value ks2 (< If it is smaller than ks1), it is determined that the pre-catalyst sensor 17 is in an output decrease abnormality. When the identified gain k is not less than the gain reduction abnormality determination value ks2 and not more than the gain increase abnormality determination value ks1, it is determined that the pre-catalyst sensor 17 is normal with respect to the output.

  As described above, according to this embodiment, it is possible to individually and simultaneously detect abnormality of individual characteristics (output and responsiveness) of the air-fuel ratio sensor. Therefore, it is extremely suitable as an abnormality diagnosis device for an air-fuel ratio sensor.

  3 and 4 show the results of sequentially identifying the time constant T and the gain k by the above-mentioned sequential least square method in the case of the normal pre-catalyst sensor 17 and the case of the abnormal pre-catalyst sensor 17. FIG. 3 shows the case of the normal pre-catalyst sensor 17, and FIG. 4 shows the case of the abnormal pre-catalyst sensor 17. FIG. 3A and FIG. 4A show the state of vibration between the input (broken line) and the output (solid line).

  FIGS. 3B and 4B show transitions of the time constant T (broken line) and the gain k (solid line) from the start of active control. The time constant T and the gain k, which are identification values, are updated every sampling time and gradually converge to a constant value. The time constant T and the gain k are acquired at a time point (determination time) t1 after a predetermined time (for example, 5 seconds) has elapsed from when the active control starts (at the start of identification) t0, and the values almost converge. The obtained time constant T and gain k are compared with the abnormality determination values Ts1, ks1, and ks2, and abnormality determination of responsiveness and output is made.

  When a test was performed using an abnormal pre-catalyst sensor 17 having a response almost equal to that of the normal pre-catalyst sensor 17 and an output of ½, the time constant T at the determination time t1 is It was almost the same as 0.18 for the normal sensor and 0.17 for the abnormal sensor. On the other hand, the gain k at the determination time t1 is 1 for the normal sensor and 0.5 for the abnormal sensor. As a result, it was confirmed that the same result as the actual sensor could be obtained.

  By the way, the actual input / output data may be biased (shifted) toward the lean side or the rich side with respect to one side due to factors such as load fluctuation, learning shift, and sensor value shift. FIG. 5 shows test results showing the state of this bias. In the figure, u (t) c and y (t) c represent values obtained by passing the input u (t) and the output y (t) through a low-pass filter, or their moving averages. Since the air-fuel ratio detected by the pre-catalyst sensor 17 is controlled to be close to the theoretical air-fuel ratio (A / F = 14.6), the output y (t) that is the detection value of the pre-catalyst sensor 17 is theoretically. It fluctuates around the air-fuel ratio, and the value passed through the low-pass filter or the moving average y (t) c is also kept near the theoretical air-fuel ratio. On the other hand, the input u (t) is biased to the lean side in the illustrated example for the reason described above.

  Since it is difficult to obtain an accurate identification value when identification is performed in such a bias state, in this embodiment, correction is performed so as to remove the bias for input / output data at the time of identification. As a result, accurate identification can be performed, and robustness against load fluctuation, learning deviation, and the like can be improved.

  Here, as a reference example, a method for correcting bias removal of input / output data using a value that has passed through a low-pass filter or a moving average is shown. As shown in FIG. 6, the data of the input u (t) and the output y (t) are passed through the low-pass filter, or the moving average is calculated, and the bias values u (t) c and y (t) c are sequentially obtained. Is calculated automatically. Then, sequentially, the difference Δu (t) (= u (t) −u (t) c) between the input u (t) and its bias value u (t) c, the output y (t) and its bias value A difference Δy (t) (= y (t) −y (t) c) with respect to y (t) c is calculated, and these differences Δu (t) and Δy (t) are corrected inputs u (t) and The value of output y (t) is used. These differences Δu (t) and Δy (t) are collectively displayed as ΔA / F. In this way, the bias is removed, and the input / output values are converted to zero reference values Δu (t) and Δy (t), and the centers of the fluctuations thereof are set to zero.

  For example, when the input u (t) as shown in FIG. 7 is corrected using the value u (t) c that has passed through the low-pass filter, the value Δu (t) as shown in FIG. 8 is obtained.

  However, the method according to this reference example has a problem that the bias cannot be sufficiently removed within a certain time immediately after the start of the active control, as shown in a broken-line circle in FIG. Within this time, the value passed through the low-pass filter or the moving average is not sufficiently converged, so that it is estimated that the influence has occurred. Therefore, in the present embodiment, bias removal is performed as follows by a method different from the reference example.

First, the first aspect of bias removal will be described. In the first aspect, for each of the input u (t) and the output y (t), the average value of the rich side value and the lean side value per vibration period is set as the input bias value u _b (t) and the output. Calculated as the bias value y _b (t) and corrects the actual data by subtracting the bias values u _b (t) and y _b (t) from the actual data u (t) and output y (t). .

  FIG. 9 shows how the bias value is calculated. Since the calculation method is the same for input and output, only the input case is shown in the figure.

As shown in the figure, the input u (t) is vibrated simultaneously with the start of the active control, and the input u (t) has a rich value u _R (t) and a lean value u _L (t) every predetermined time. And come to take alternately. First, in the first one vibration period τ ₁ , the rich value u _R (t) is integrated at every sampling interval Δ during the first rich side control (half cycle). This integration is executed from the time when the target air-fuel ratio is switched to the rich side to the time when the target air-fuel ratio is switched to the lean side simultaneously with the start of the active control. Similarly, during the next lean side control (half cycle), the lean side value u _L (t) is integrated at every sampling interval Δ. This integration is also performed from when the target air-fuel ratio is switched to the lean side to when it is switched to the rich side. At this time, each rich side value u _R (t) and lean side value u _L (t), the number of samples N _R , N _L , integrated values Σu _R (t), Σu _L (t) Is temporarily stored in the buffer of the ECU 20.

Next, the ECU 20 calculates the average value obtained by dividing the rich-side integrated value Σu _R (t) by the sample number N _R and the lean-side integrated value Σu _L (t) from the sample number N. _An average value with the average value obtained by dividing by _L is calculated as the bias value u _b (t) in the one period τ ₁ . That is, the input and output bias values u _b (t) and y _b (t) per vibration period are calculated according to the following equation (21).

However, in order to improve identification accuracy, the input / output data acquired in the first one cycle τ ₁ is not corrected and is not used for identification. The use for identification begins from the half period τ ₂₁ after the first period τ ₁ . Similarly, in this half cycle τ ₂₁ , the rich value u _R (t) is integrated at every sampling interval Δ. The lean-side value u _L (t), the sample number N _L and the integrated value Σu _L (t) in the latter half cycle τ ₁₂ in the first one cycle τ ₁ , and the rich-side value in the subsequent half cycle τ ₂₁ Using the value u _R (t), the number of samples N _R and the integrated value Σu _R (t), the ECU 20 calculates the bias value u _b (t) in the half cycle τ ₂₁ from the equation (21). Then the ECU 20, the bias value _u b in a half period tau ₂₁ (t) is replaced with the bias value in the first one period τ _{1 u} b (t), the bias value _u b (t) is updated.

Next, the ECU 20 corrects each input data u _R (t) by subtracting the bias value u _b (t) from each input data u _R (t) in the half cycle τ ₂₁ . That is, if the actual input / output data is u (t), y (t), and the corrected input / output data is u ′ (t), y ′ (t) without distinction between the rich side and the lean side, The input / output data is expressed by the following equation (22).

Identification in the half cycle τ ₂₁ is performed using the corrected input / output data u ′ (t), y ′ (t). Similarly, the bias values u _b (t) and y _b (t) are updated every half cycle (τ ₂₂ , τ ₃₁ ...), And the bias values u _b (t) and y _b ( The actual input / output data u (t) and y (t) are corrected using t). In this way, the input / output data u (t) and y (t) can be bias-removed and replaced with a zero reference value, so that identification accuracy and robustness can be improved and diagnostic accuracy can also be improved. In particular, since bias removal correction is performed based on actual input / output data immediately after the start of active control, it is possible to sufficiently remove the bias immediately after the start of active control.

  Next, a second mode of bias removal will be described. In the second mode, the identification is performed including the bias component c corresponding to the bias between the input and output when the parameters T and k of the first-order lag element are identified.

  In the present embodiment, a method using a Kalman filter (or Kalman filter method) as described below, instead of the method using the sequential least square method as described above, for simultaneous identification of the parameters T and k of the first-order lag element and the bias component c. Indicates. However, the sequential least square method may be used. Such identification is performed by the ECU 20.

  First, as already described, the transfer function G (s) of the model of the system from the injector 12 to the pre-catalyst sensor 17 is expressed by the following equation.

  Further, when the Laplace transform of the input u (t) and the output y (t) for this model is U (s), Y (s), the following equation (23) is established.

  When the equation (23) is subjected to inverse Laplace transform, the following equation (24) is obtained.

  Here, assuming that the input is constant during the sampling interval Δ (seconds) in the digital processing of the ECU 20 (zero-order hold), the equation (24) is discretized, and the bias component c is further considered. It can be arranged as shown in equation (25). Note that ω is an error due to discretization.

  By this formula (25), the linear model for identification can be expressed as the following formula (26) after all.

  This is rewritten (defined) as the following equation (26) '.

  Hereinafter, the parameter vector θ (t) (a, b, c) is estimated for the expression (26) ′ by the Kalman filter, thereby obtaining a, b, c, and the gain k and the time constant T using a, b. Ask for.

  Here, in the equation (25), by including the bias component c, it is not necessary to perform a bias removal process (for example, correction as shown in the above example) on the actual input u (t) and output y (t). Become. Each value is calculated assuming a bias between input and output. The gain k and the time constant T are values irrelevant to the bias, but are sequentially identified at every sampling interval Δ on the assumption of a bias between input and output. On the contrary, when the bias is not assumed, the identification accuracy of the gain k and the time constant T is deteriorated. In the case where there is no actual bias between input and output, a bias component c close to zero is identified.

  The parameter vector θ (t) in the equation (26) ′ can be regarded as a time-varying parameter that is updated every sampling interval Δ. Therefore, the estimated value of θ (t)

Is represented by the following state equation with state variable x (t).

  Here, v (t) and w (t) are system noise and observation noise, respectively, and are average white noise. V and W are assumed to be covariance matrices of v (t) and w (t), and applied to a Kalman filter of the following format to estimate the parameter vector θ (t). Here, V and W are fixed values. The initial values are as follows.

  Step 1: A Kalman filter gain F (t) is calculated by the following equation (28).

  Step 2: Using the calculated Kalman filter gain F (t), the estimated value is updated as in the following equation (29).

Here, the term on the left side (referred to as “x hat (t | t)” or the like for convenience) means an estimated value (current estimated value) at a certain sampling time, and the first term on the right side (x hat (t | t−t− 1)) means the previous estimated value one sampling interval before. In the right parenthesis, one item y (t) means the current observed value (output), and the product of two items φ ^T (t) and x hat (t | t−1) is calculated from the previous estimated value. Means the expected observation (output). Equation (29) can be said to indicate that the current estimated value is calculated using the previous estimated value and the difference between the current observed value and the observed value predicted from the previous estimated value.

  Step 3: Update the estimated error covariance matrix X as shown in the following equation (30).

  Step 4: Update parameters for next time as follows.

  Step 5: Repeat Step 1 to Step 4.

  The estimated value is sequentially updated at every sampling interval according to the update rule as described above. By doing so, the parameter vector θ (t) (a, b, c) is estimated at each sampling interval, and a, b, c are calculated. Among these, a, b are used to obtain the gain k and the time constant T. At the same time, the bias component c is identified. The gain k and time constant T after the elapse of a predetermined time from the start of identification are compared with the above-described deterioration determination value, and output abnormality and responsiveness abnormality of the pre-catalyst sensor 17 are individually and simultaneously determined.

  According to this, even when there is a bias in the actual input / output data, the identification accuracy and robustness of the parameters k and T can be improved, and the diagnostic accuracy can be improved. In particular, since the parameters k and T including the bias component c are identified immediately after the start of active control, it is possible to obtain an accurate parameter identification value immediately after the start of active control.

  In addition, when the method using the sequential least square method is adopted instead of the method using the Kalman filter described here, for example, the bias component c is included in the previous equation (7) and the following equation (7) ′ is modified. Further, k, T, and c may be identified by the sequential least square method based on the equation (7) ′.

The preferred embodiment of the present invention has been described in detail above, but various other embodiments of the present invention are conceivable. For example, although the above-mentioned internal combustion engine is a multi-cylinder spark ignition internal combustion engine for vehicles, the use, type, etc. of the internal combustion engine are not particularly limited. Although the application example to the so-called wide-range air-fuel ratio sensor has been described in the above embodiment, the present invention can be applied to a so-called O ₂ sensor such as the post-catalyst sensor 18. A sensor for detecting the air-fuel ratio of exhaust gas, including such an O ₂ sensor, is referred to as an air-fuel ratio sensor according to the present invention. In the above-described embodiment, the abnormality of two typical characteristics of the output of the air-fuel ratio sensor and the responsiveness is determined. However, the present invention is not limited to this. For example, only the abnormality of one characteristic may be determined. In addition to the output and responsiveness, abnormality of other characteristics may be determined. In the above embodiment, a plurality of parameters are identified at the same time, and abnormality of a plurality of characteristics is determined at the same time. However, the present invention is not limited to this, and the identification and abnormality determination may be performed with a time difference.

  The embodiment of the present invention is not limited to the above-described embodiment, and includes all modifications, applications, and equivalents included in the concept of the present invention defined by the claims. Therefore, the present invention should not be construed as being limited, and can be applied to any other technique belonging to the scope of the idea of the present invention.

It is the schematic of the internal combustion engine which concerns on this embodiment. It is a figure which shows schematically the mode of a change of the input and output at the time of active control. This is a test result showing a change in time constant and gain with respect to a change in input air-fuel ratio and output air-fuel ratio, and is a case of a normal sensor. This is a test result showing a change in time constant and gain with respect to a change in input air-fuel ratio and output air-fuel ratio, and is a case of an abnormal sensor. It is a test result which shows the mode of a change with an input air fuel ratio and an output air fuel ratio, and is a state with a bias between both. It is the schematic for demonstrating the method of the bias removal correction | amendment as a reference example. It is a test result which shows the mode of a change of an input air fuel ratio, and is the state before bias removal correction | amendment of a reference example. It is a test result which shows the mode of the change of an input air fuel ratio, and is the state after the bias removal correction | amendment of a reference example. It is a figure for demonstrating the calculation method of the bias value which concerns on this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Internal combustion engine 3 Combustion chamber 5 Air flow meter 6 Exhaust pipe 7 Spark plug 12 Injector 14 Crank angle sensor 15 Accelerator opening sensor 17 Pre-catalyst sensor 18 Post-catalyst sensor 20 Electronic control unit (ECU)
A / F Air-fuel ratio u (t) Input (input air-fuel ratio)
y (t) output (output air-fuel ratio)
k gain T time constant τ period u _b (t) input bias value y _b (t) output bias value Δ sampling interval N number of samples c bias component

Claims (6)

An air-fuel ratio sensor abnormality diagnosis device for detecting an air-fuel ratio of exhaust gas of an internal combustion engine,
Identifying means for modeling a system from a fuel injection valve to an air-fuel ratio sensor with a first-order lag element, and identifying a parameter in the first-order lag element based on an input air-fuel ratio and an output air-fuel ratio for the model;
An abnormality determination means for determining an abnormality of a predetermined characteristic of the air-fuel ratio sensor based on the parameter identified by the identification means;
Active control means for executing active control for forcibly oscillating the input air-fuel ratio at the time of identification by the identification means;
Bias correcting means for correcting the input air-fuel ratio and the output air-fuel ratio so as to remove a bias between the input air-fuel ratio and the output air-fuel ratio at the time of identification by the identifying means, the input air-fuel ratio and the output Bias correction means for calculating an average value of a rich side value and a lean side value per vibration period as a bias value for each air-fuel ratio, and subtracting the bias value from actual data to correct the actual data; An abnormality diagnosis apparatus for an air-fuel ratio sensor, comprising: The bias correction means calculates the average value of the rich side value by sampling the rich side value by integrating the rich side value at every sampling interval and dividing this by the number of samples per vibration period. Bias by calculating the average value of the lean side values by accumulating at intervals and dividing this by the number of samples, and calculating the average value of these rich side values and the average value of the lean side values The abnormality diagnosis device for an air-fuel ratio sensor according to claim 1, wherein the value is calculated and the bias value is updated every half cycle. An air-fuel ratio sensor abnormality diagnosis device for detecting an air-fuel ratio of exhaust gas of an internal combustion engine,
Identifying means for modeling a system from a fuel injection valve to an air-fuel ratio sensor with a first-order lag element, and identifying a parameter in the first-order lag element based on an input air-fuel ratio and an output air-fuel ratio for the model;
An abnormality determining means for determining an abnormality of a predetermined characteristic of the air-fuel ratio sensor based on the parameter identified by the identifying means;
The abnormality diagnosis device for an air-fuel ratio sensor, wherein the identification means identifies the parameter including a bias component corresponding to a bias between the input air-fuel ratio and the output air-fuel ratio. The abnormality diagnosis apparatus for an air-fuel ratio sensor according to claim 3, wherein the identification unit simultaneously identifies the parameter and the bias component using a Kalman filter. The abnormality diagnosis device for an air-fuel ratio sensor according to claim 3 or 4, further comprising active control means for executing active control for forcibly oscillating the input air-fuel ratio at the time of identification by the identification means. 6. The parameter according to claim 1, wherein the parameter includes a gain and a time constant, and the predetermined characteristic of the air-fuel ratio sensor includes an output corresponding to the gain and a responsiveness corresponding to the time constant. An abnormality diagnosis device for an air-fuel ratio sensor according to claim 1.

Images