JP2010096019A - Abnormality diagnostic device of air-fuel ratio control device for internal combustion engine - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately diagnose abnormality of an air-fuel ratio control device by using an air-fuel ratio behavior model estimating behavior of an air-fuel ratio from the amount of fuel injected represented by using an ARX (Autoregressive with exogenous input) model. <P>SOLUTION: The A/F behavior models for the amount of fuel injection are individually constructed with respect to a transient state and a steady state based on the past time-series data on the amount of fuel injected and the actual A/F. By using the models, determination to predict whether it is the transient state or the steady state, and estimation of predicted A/F, are performed. Meanwhile, it is actually determined whether it is the transient state or the steady state based on the actual operating condition of an engine, thereby acquiring the actual A/F from an A/F sensor. When a predictive determination result agrees with the actual determination result and a difference between the predicted A/F and the actual A/F is a threshold or below, it is determined that the air-fuel ratio control device is normal, and when the predictive determination result does not agree with the actual determination result or the difference between the predicted A/F and the actual A/F exceeds the threshold, it is determined that the air-fuel ratio control device is abnormal. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an abnormality diagnosis device for an air-fuel ratio control device for an internal combustion engine. Hereinafter, the internal combustion engine may be referred to as an “engine”.

  Conventionally, by adjusting the amount of fuel (fuel injection amount) injected from the fuel injection valve (injector) based on the detection result of the air-fuel ratio sensor disposed in the exhaust passage of the engine, the air-fuel ratio (in the engine) 2. Description of the Related Art An air-fuel ratio control apparatus that controls an air-fuel ratio of an air-fuel mixture to be supplied is widely known.

In such an air-fuel ratio control apparatus, when an abnormality occurs in a component used for air-fuel ratio control, such as an air flow meter that measures the flow rate of air sucked from an air-fuel ratio sensor, an injector, or an intake passage, the air-fuel ratio is appropriately controlled. become unable. For this reason, it is necessary to diagnose abnormality of the air-fuel ratio control device. Various apparatus for diagnosing abnormality of an air-fuel ratio control apparatus having a function of determining whether there is an abnormality in the air-fuel ratio control apparatus have been proposed (for example, see Patent Document 1).
Japanese Patent Laid-Open No. 8-232721

  An object of the present invention is to express an ARX (Auto-Regressive with eXogenous input) model (autoregressive model with external input) widely known in the field of system identification as one of abnormality diagnosis devices for an air-fuel ratio control device. An object of the present invention is to provide an apparatus capable of accurately diagnosing an abnormality in an air-fuel ratio control apparatus by using the “air-fuel ratio behavior model for estimating the air-fuel ratio behavior from the fuel injection amount”.

  An abnormality diagnosis apparatus for an air-fuel ratio control apparatus according to the present invention includes a storage means, a classification means, a plane estimation means, a model construction means, a prediction determination means, an actual determination means, an air-fuel ratio prediction means, and an abnormality determination. Means. Hereinafter, these will be described in order.

  In the storage means, for each of a plurality of different times, the past (a plurality of) time-series data (based on each time) of the air-fuel ratio detected by the air-fuel ratio sensor, and (the command) the fuel injection amount (each time The past (a plurality of) time-series data is obtained and stored.

  The classifying means obtains observation data for each time including time-series data of the detected air-fuel ratio and time-series data of the fuel injection amount for each time. Here, it is assumed that the probability distribution in the regression vector space of the observation data for each time follows a distribution having two distribution tendencies, a distribution trend corresponding to the transient operation state and a distribution trend corresponding to the steady operation state. . As the “distribution having two distribution tendencies”, for example, a mixed normal distribution can be adopted. Under this assumption, “a distribution having two distribution tendencies” is obtained using the maximum likelihood estimation method or the like based on observation data at each time. Then, based on the obtained distribution, the observation data for each time is classified into “group corresponding to the transient operation state” and “group corresponding to the steady operation state”.

  In this classification, the observed data at each time is the observed data among the two distribution tendencies of the obtained distribution: “distribution trend corresponding to the transient operation state” and “distribution trend corresponding to the steady operation state”. The distribution tendency with the larger probability of belonging to is selected and the observation data for each time is classified into a group corresponding to the driving state corresponding to the selected distribution tendency.

  In the plane estimation means, based on the observation data for each classified time (based on the classification result), the “distribution region of data in the group corresponding to the transient operation state” and the “steady operation” in the regression vector space. A hyperplane (a parameter representing the separation hyperplane) that divides the “data distribution area in the group corresponding to the state” is estimated. This estimation of parameters (to be identified) can be achieved, for example, by solving a quadratic programming problem (secondary minimization problem, quadratic optimization problem).

  In the model construction means, as the air-fuel ratio behavior model that predicts the (future) air-fuel ratio from the time-series data up to the present of the air-fuel ratio and the time-series data of the fuel injection amount up to the present, The “behavior model” and the “air-fuel ratio behavior model for the steady operation state” are individually expressed using different local ARX models. The parameters (to be identified) in the “air-fuel ratio behavior model for the transient operation state” are based on the observation data classified into the “group corresponding to the transient operation state” and the “air-fuel ratio behavior model for the steady operation state”. Parameters to be identified (to be identified) are individually identified based on observation data classified into “groups corresponding to steady operating states”. Thereby, the “air-fuel ratio behavior model for the transient operation state” and the “air-fuel ratio behavior model for the steady operation state” are individually constructed.

  The prediction determination means determines whether the state is a transient operation state or a steady operation state based on the estimated separation hyperplane and the time series data of the detected air-fuel ratio (from the past) to the present (prediction determination). For example, when the time series data of the detected air-fuel ratio (from the past) to the present is substituted as each component of the regression vector in the equation representing the estimated separation hyperplane, the value of the equation is positive. It can be achieved by determining whether it is negative or negative.

  The actual determination means determines whether the engine is in a transient operation state or a steady operation state based on the actual operation state of the internal combustion engine (actual determination). As an actual driving state, for example, an accelerator pedal opening degree (change speed thereof) and the like can be mentioned.

  In the air-fuel ratio prediction means, among the constructed air-fuel ratio behavior models, the air-fuel ratio behavior model corresponding to the result of the prediction determination, the time-series data of the detected air-fuel ratio (from the past) to the present, and the (past To) a (future) air-fuel ratio is predicted based on time-series data up to the present.

  The abnormality determination means determines whether the air-fuel ratio control apparatus is abnormal based on a comparison between the prediction determination result and the actual determination result and a comparison between the predicted air-fuel ratio and the detected air-fuel ratio. Preferably, when the result of the prediction determination matches the result of the actual determination and the difference between the predicted air-fuel ratio and the detected air-fuel ratio is equal to or less than a threshold value, it is determined that “the air-fuel ratio control device is normal”, and the result of the prediction determination When the actual determination result does not match or when the difference between the predicted air-fuel ratio and the detected air-fuel ratio exceeds the threshold value, it is determined that “the air-fuel ratio control device is abnormal”.

  Hereinafter, the operation and effect of the above configuration will be described. In general, when an air-fuel ratio behavior model constructed based on data in either a transient operation state or a steady operation state is used, the estimation accuracy of the air-fuel ratio behavior in either the transient operation state or the steady operation state is improved. Lower. On the other hand, according to the above configuration, as the air-fuel ratio behavior model, the “air-fuel ratio behavior model for the transient operation state” and the “air-fuel ratio behavior model for the steady operation state” are individually constructed, and the transient operation state and The air / fuel ratio is estimated using the “air-fuel ratio behavior model for the transient operation state” while the prediction determination is being made, and the “air-fuel ratio behavior model for the steady operation state” while being predicted and determined to be the steady operation state. Is used to estimate the air / fuel ratio. Therefore, it is possible to stably and accurately estimate the air-fuel ratio behavior regardless of the transient operation state or the steady operation state.

  Further, in general, the accuracy of transient / steady state prediction determination is greatly influenced by aging of the air-fuel ratio control system, individual errors of the engine, and the like. On the other hand, according to the above configuration, transient / steady state prediction determination is performed using the separated hyperplane in the regression vector space estimated based on actual observation data. Therefore, it is possible to perform the transient / steady state prediction determination with high accuracy without being affected by the secular change of the air-fuel ratio control system, the individual error of the engine, or the like.

  Further, when the abnormality diagnosis of the air-fuel ratio control device is performed based only on the comparison between the predicted air-fuel ratio and the detected air-fuel ratio, the accuracy of the abnormality diagnosis is not necessarily high. On the other hand, according to the above configuration, the abnormality diagnosis of the air-fuel ratio control device not only compares the predicted air-fuel ratio with the detected air-fuel ratio, but also compares the result of the prediction determination with respect to the transient / steady state and the result of the actual determination. Also done based on. Accordingly, the abnormality diagnosis is performed based on the plurality of pieces of comparison information, and thus the accuracy of the abnormality diagnosis is high.

  For example, if the difference between the predicted air-fuel ratio and the detected air-fuel ratio is very small and the prediction determination result for transient / steady state is different from the actual determination result, the abnormality diagnosis is performed when the predicted air-fuel ratio and the detected air-fuel ratio are different. When the determination is made only based on the comparison with the fuel ratio, it is determined that “the air-fuel ratio control device is normal”. On the other hand, in the above configuration, it can be determined that “the air-fuel ratio control device is abnormal”.

  Embodiments of an abnormality diagnosis apparatus for an air-fuel ratio control apparatus for an internal combustion engine according to the present invention will be described below with reference to the drawings.

(Constitution)
FIG. 1 shows a schematic configuration of a system in which an abnormality diagnosis device for an air-fuel ratio control device according to an embodiment of the present invention is applied to a spark ignition type multi-cylinder (four-cylinder) internal combustion engine 10 (fuel is gasoline). The internal combustion engine 10 includes a cylinder block 20 including a cylinder block and an oil pan, a cylinder head 30 fixed on the cylinder block 20, and a gasoline mixture supplied to the cylinder block 20. An intake system 40 and an exhaust system 50 for releasing the exhaust gas from the cylinder block 20 to the outside are included.

  The cylinder block unit 20 includes a cylinder 21, a piston 22, a connecting rod 23, and a crankshaft 24. The heads of the cylinder 21 and the piston 22 form a combustion chamber 25 together with the cylinder head portion 30.

  The cylinder head portion 30 includes an intake port 31 communicating with the combustion chamber 25, an intake valve 32 that opens and closes the intake port 31, an intake camshaft that drives the intake valve 32, and continuously changes the opening and closing timing of the intake valve 32. A variable intake timing device 33, an actuator 33a of the variable intake timing device 33, an exhaust port 34 communicating with the combustion chamber 25, an exhaust valve 35 for opening and closing the exhaust port 34, an exhaust camshaft 36 for driving the exhaust valve 35, an ignition plug 37, An igniter 38 including an ignition coil that generates a high voltage to be applied to the spark plug 37 and a fuel injection valve 39 for injecting fuel into the intake port 31 are provided.

  The intake system 40 is provided in an intake pipe 41 including an intake manifold that communicates with the intake port 31 and forms an intake passage together with the intake port 31, an air filter 42 provided at an end of the intake pipe 41, and the intake pipe 41. A throttle valve 43 for changing the opening cross-sectional area of the intake passage, a throttle valve actuator 43a for driving the throttle valve 43, a swirl control valve (SC valve) 44 for changing the flow rate of the intake air flowing from the intake passage into the combustion chamber 25, An SC valve actuator 44a for driving the SC valve 44 is provided.

  The exhaust system 50 includes an exhaust manifold 51 that communicates with the exhaust port 34, and an exhaust pipe (exhaust pipe) that is connected to the exhaust manifold 51 (actually, a collection portion of the exhaust manifolds 51 that communicate with each exhaust port 34). ) 52, a three-way catalyst 53 and an EGR gas passage 54 disposed (interposed) in the exhaust pipe 52. The exhaust port 34, the exhaust manifold 51, and the exhaust pipe 52 constitute an exhaust passage.

  The EGR gas passage 54 is configured to communicate the exhaust passage upstream of the three-way catalyst 53 and the intake passage downstream of the throttle valve 43. In the EGR gas passage 54, an EGR gas cooler 55, an EGR valve 56, and an actuator 56a of the EGR valve 56 are interposed. The opening area of the EGR valve 56 can be adjusted by the actuator 56a of the EGR valve 56.

  On the other hand, this system includes an air flow meter 61, a throttle valve opening sensor 62, a cam position sensor 63, a crank position sensor 64, a water temperature sensor 65, an exhaust passage upstream of the three-way catalyst 53 (in this example, each of the exhaust manifolds described above). 51, an A / F sensor 66, an EGR valve opening sensor 67, and an accelerator opening sensor 68 are provided.

  The air flow meter 61 detects the flow rate (mass flow rate) of fresh air flowing through the intake passage and outputs a signal representing the fresh air flow rate Ga. The throttle valve opening sensor 62 detects the opening of the throttle valve 43 and outputs a signal representing the throttle valve opening TA. The cam position sensor 63 detects the opening / closing timing of the intake valve 32 and outputs a signal representing the opening / closing timing VVT. The crank position sensor 64 detects the rotational speed of the crankshaft 24 and outputs a signal representing the engine rotational speed NE. The water temperature sensor 65 detects the temperature of the cooling water of the internal combustion engine 10 and outputs a signal representing the cooling water temperature THW. The A / F sensor 66 detects the air-fuel ratio of the exhaust gas and outputs a signal representing the air-fuel ratio. The EGR valve opening degree sensor 67 detects the opening degree of the EGR valve 56 and outputs a signal representing the EGR valve opening degree Aegr. The accelerator opening sensor 68 detects the operation amount of the accelerator pedal 81 operated by the driver, and outputs a signal representing the operation amount Accp of the accelerator pedal 81.

  The electric control device 70 includes a CPU 71 connected to each other by a bus, a routine (program) executed by the CPU 71, a table (map), and a ROM 72, a RAM 73, a backup RAM 74, and an interface 75 including an AD converter. It is the microcomputer which consists of.

  The interface 75 is connected to the sensors 61 to 68, supplies signals from the sensors 61 to 68 to the CPU 71, and in response to instructions from the CPU 71, the actuator 33a, the igniter 38, and the fuel injection valve 39 of the variable intake timing device 33. A drive signal is sent to the throttle valve actuator 43a, the SC valve actuator 44a, and the actuator 56a of the EGR valve 56.

(Air-fuel ratio control)
The air-fuel ratio control apparatus in this example uses one of well-known methods based on the output of the A / F sensor 66 (actual A / F) and the output of the air flow meter 61 (new air flow rate Ga). Thus, the feedback control is performed on the (command) fuel injection amount (accordingly, the amount of fuel actually injected) instructed to the injector 39 so that the air-fuel ratio of the air-fuel mixture supplied to the engine matches the stoichiometric air-fuel ratio. It has come to be. Therefore, if an abnormality occurs in the air-fuel ratio control device (specifically, an abnormality in the A / F sensor 66, the air flow meter 61, the injector 39, etc.), the air-fuel ratio cannot be matched with the theoretical air-fuel ratio with high accuracy. Therefore, it is necessary to diagnose abnormality of the air-fuel ratio control device.

  In this example, when performing abnormality diagnosis of the air-fuel ratio control device, an “air-fuel ratio behavior model for estimating the behavior of the air-fuel ratio from the fuel injection amount” (hereinafter referred to as “A / F behavior model”) is used. The Hereinafter, the A / F behavior model will be described first.

(A / F behavior model)
In general, the air-fuel ratio behavior differs between the transient operation state and the steady operation state. Therefore, in this example, the “A / F behavior model for the transient operation state” and the “A / F behavior model for the steady operation state” are separately prepared.

  The (command) fuel injection amount (accordingly, the actual fuel injection amount) at time k is u (k), and time series data of past fuel injection amounts with respect to time k are u (k-1), u (k-2 ), ... The real A / F at time k is y (k), and the time series data of the past real A / F with respect to time k is y (k-1), y (k-2),. Here, (k−j) (j: 1, 2,...) Represents the time of the sample timing j times before the time k. The time interval of the sample timing corresponds to the time interval between a certain fuel injection time and the next fuel injection time, for example. In the case of a four-cylinder internal combustion engine as in this example, this corresponds to the time (180 ° CA) during which the crankshaft 24 rotates by 180 °.

  The “A / F behavior model for the transient operation state” and the “A / F behavior model for the steady operation state” used in this example are respectively expressed by the following formula (1) using different local ARX models: It is expressed by equation (2).

In the above formulas (1) and (2), a _** , b _** , and c _* are identification parameters in the A / F behavior model identified as described later. f _* , g _* , and h are parameters representing separation hyperplanes identified as described later. ε (k) is an error. The natural numbers n and m respectively represent the number of past time series data of actual A / F and fuel injection amount necessary for describing the A / F behavior model. The expressions (1) and (2) can be expressed by the following expressions (3) and (4), respectively.

In the above formulas (3) and (4), x (k) is a regression vector and is expressed according to the following formula (5) using the past fuel injection amount and the time series data of the actual A / F. . θ _* is a vector using the identification parameter in the A / F behavior model, and is expressed according to the following equations (6) and (7). As will be described later, in the regression vector space, the distribution region of observation data (described later) corresponding to the transient operation state and the distribution region of observation data (described later) corresponding to the steady operation state are expressed by the following equation (8). It is divided by a hyperplane (separated hyperplane). Each of a and b is a vector and a scalar identified as described later, and is represented by the following equations (9) and (10). “T” attached to the right shoulder of the letter means a transposed matrix.

  Hereinafter, referring to FIG. 2, the above-described separation hyperplane required for constructing the “A / F behavior model for the transient operation state” and the “A / F behavior model for the steady operation state” is described. The estimation (identification of the parameter representing the separation hyperplane) and the identification parameter identification in the A / F behavior model will be described.

<Clustering> (Step 205 in FIG. 2)
An observation data vector z _k for time k is defined by the following equation (11). In this example, time series data u (k−1), u () of past fuel injection amounts with respect to each time k for N different times k (k = 1, 2,..., N), respectively. k-2), ... and time series data y (k-1), y (k-2), ... of the past actual A / F with respect to each time k, and the actual A / Fy at each time k (k) is measured and stored in a predetermined database (RAM 73). The observation data vector z _k is acquired for each of N different times k (k = 1, 2,..., N) by reading data from the database. The N pieces of observation data z _k are classified into a group (cluster C ₁ ) corresponding to the transient operation state and a group (cluster C ₂ ) corresponding to the steady operation state as described below.

It is assumed that the probability distribution in the regression vector space of the N observation data vectors z _k follows a distribution having two distribution trends, a distribution trend corresponding to the transient operation state and a distribution trend corresponding to the steady operation state. In this example, a mixed normal distribution is adopted as the distribution having the two distribution tendencies. That is, it is assumed that the probability distribution in the regression vector space of N observation data vectors z _k is expressed by the following equation (12).

In the equation (12), Φ is a vector using a parameter (identified as described later) representing a mixed normal distribution, and is represented by the following equation (13). In the equation (12), P ₁ and P ₂ are normal distributions corresponding to the transient operation state and the steady operation state, and are expressed by the following equations (14) and (15). μ ₁ and μ ₂ are averages of P ₁ and P ₂ , respectively, and Σ ₁ and Σ ₂ are variances of P ₁ and P ₂ , respectively. α ₁ and α ₂ are weighting factors of P ₁ and P ₂ , respectively, and have a relationship shown in the following equation (16).

In this example, the parameter Φ representing the mixed normal distribution is identified using the maximum likelihood estimation method. That is, the parameter Φ is identified so that the likelihood function L (Φ) expressed by the following equation (17) is maximized. Thereby, the mixed normal distribution is obtained. FIG. 3 is a diagram showing an example of the mixed normal distribution (p = α ₁ · p ₁ + α ₂ · p ₂ ) obtained in this way, simplified in two dimensions. In this procedure, the non-convex optimization problem is solved, but the non-convex optimization problem can be solved by iteratively calculating using a well-known EM algorithm.

Using the identified parameter Φ (therefore, the obtained mixed normal distribution p = α ₁ · p ₁ + α ₂ · p ₂ ), N observation data z _k are converted into a group (cluster C) corresponding to the transient operation state. ₁ ) and a group (cluster C ₂ ) corresponding to the steady operation state. In this classification, for example, N pieces of observation data z _k have two distributions, a normal distribution (α ₁ · p ₁ ) corresponding to the transient operation state and a normal distribution (α ₂ · p ₂ ) corresponding to the steady operation state. The normal distribution is classified into a group corresponding to the driving state corresponding to the normal distribution with the higher probability that the observation data belongs (see FIG. 3).

For example, in the case of the observation data zp shown in FIG. 3, the probability attributed to the normal distribution (α ₁ · p ₁ ) corresponding to the transient operation state is the normal distribution (α ₂ · p ₂ ) corresponding to the steady operation state. Greater than the probability of belonging. Therefore, the observation data zp is classified into a group (cluster C ₁ ) corresponding to the transient operation state. Hereinafter, among the N observation data z _k (k = 1, 2,..., N), N ₁ observation data z _k (k = k ₁₁ , k ₁₂ ,... K _1N1 ) is in a transient operation state. It is classified into a corresponding group (cluster C ₁ ), and N ₂ observation data z _k (k = k ₂₁ , k ₂₂ ,... K _2N2 ) is classified into a group (cluster C ₂ ) corresponding to the steady operation state. Shall. Here, N = N ₁ + N ₂ is established.

<Estimation of Separation Hyperplane> (Step 210 in FIG. 2)
Next, in the regression vector space, estimation of a hyperplane (separated hyperplane) that divides the distribution region of observation data in the group corresponding to the transient operation state and the distribution region of observation data in the group corresponding to the steady operation state Will be described.

As described above, this separated hyperplane is expressed according to the above equation (8). Therefore, the estimation of the separation hyperplane is specifically achieved by identifying the parameters a and b shown in the equations (9) and (10). In this example, the parameters a and b are represented by the following equation (18) based on N pieces of observation data z _k classified into _two groups (cluster C ₁ and cluster C ₂ ) as described above. It is identified by solving the next planning problem (secondary minimization problem, second optimization problem).

<Identification of Identification Parameter in A / F Behavior Model> (Step 215 in FIG. 2)
Next, identification of identification parameters θ ₁ and θ _{2 in} the A / F behavior model shown in equations (6) and (7) will be described. In this example, the identification parameter θ ₁ corresponding to the transient operation state has N ₁ observation data z _k (k = k ₁₁ , k ₁₂ ,...) Classified into a group (cluster C ₁ ) corresponding to the transient operation state. k _1N1 ) is identified using the least square method as shown in the following equation (19), and the identification parameter θ ₂ corresponding to the steady operation state is assigned to the group (cluster C ₂ ) corresponding to the steady operation state. Based on the classified N ₂ observation data z _k (k = k ₂₁ , k ₂₂ ,... K _2N2 ), identification is performed using the least square method as shown in the following equation (20).

By the above procedure (steps 205, 210, and 215 in FIG. 2), the parameters a and b shown in the equations (9) and (10), and the identification parameters θ ₁ , θ ₂ is identified. Using these identified parameters, “A / F behavior model for transient operation state” and “A / F for steady operation state” which are local ARX models corresponding to equations (3) and (4), respectively. The “behavior model” is individually constructed as shown in equations (21) and (22). ξ (k) is a regression vector and is expressed according to the following equation (23) using the fuel injection amount from the past to the present and the time series data of the actual A / F. Here, (k) represents the present, and (kj) (j: 1, 2,...) Represents the time of the sample timing j times before the present. (k + 1) represents the time (that is, the future) of the next sample timing with the present as a reference. The time interval of the sample timing corresponds to the time interval between a certain fuel injection time and the next fuel injection time, as described above. In the case of a four-cylinder internal combustion engine as in this example, this corresponds to the time (180 ° CA) during which the crankshaft 24 rotates by 180 °.

  That is, in the transient operation state (when it is determined as the transient operation state in the prediction determination described later), the fuel from the past to the present is expressed by using the “A / F behavior model for the transient operation state” shown in Equation (21). A future air-fuel ratio (predicted A / F) is estimated based on the injection amount and time series data of the actual A / F. Further, in the steady operation state (when it is determined as the steady operation state in the prediction determination described later), the fuel from the past to the present is obtained using the “A / F behavior model for the steady operation state” shown in Equation (22). A future air-fuel ratio (predicted A / F) is estimated based on the injection amount and time series data of the actual A / F.

(Abnormality diagnosis of air-fuel ratio control device)
Next, “abnormality diagnosis of the air-fuel ratio control device” using the A / F behavior model expressed by the above-described equations (21) and (22) will be described with reference to FIG.

  In step 220 of FIG. 2, the internal combustion engine 10 is in a transient operation state or a steady operation state based on the separation hyperplane estimated as described above and the actual A / F time-series data from the past to the present. It is determined whether it is present (steady state / transient prediction determination). Specifically, in this prediction determination, when the following equation (24) is satisfied, it is determined as “transient operation state”, and when the following equation (25) is satisfied, it is determined as “steady operation state”. Is done.

  In step 225 of FIG. 2, the A / F behavior model corresponding to the driving state determined by the above prediction determination among the A / F behavior models represented by the above equations (21) and (22), and The predicted A / F (described above) is estimated based on the fuel injection amount from the past to the present and the time series data of the actual A / F.

  In step 230 of FIG. 2, it is determined whether the internal combustion engine 10 is in a transient operation state or a steady operation state based on the actual operation state of the internal combustion engine 10 (actual determination of steady / transient). Specifically, in this actual determination, when the change speed (dTA / dt) of the throttle valve opening TA obtained from the throttle valve opening sensor 62 is equal to or greater than a threshold value, it is determined as “transient operation state”, and the throttle valve When the change rate (dTA / dt) of the opening degree TA is less than the threshold value, it is determined as “steady operation state”. Instead of the changing speed (dTA / dt) of the throttle valve opening TA, the changing speed (dAccp / dt) of the accelerator pedal operation amount Accp obtained from the accelerator opening sensor 68 may be used.

  In step 235 of FIG. 2, based on the comparison between the above-described prediction determination result and the above-described actual determination result, and comparison between the above-described prediction A / F and the actual A / F obtained from the A / F sensor 66. Thus, abnormality diagnosis of the air-fuel ratio control device is performed.

  Specifically, when the result of the prediction determination matches the result of the actual determination and the difference between the prediction A / F and the actual A / F is equal to or less than a threshold value, it is determined that “the air-fuel ratio control device is normal” and the prediction When the result of the determination does not match the result of the actual determination, or when the difference between the predicted A / F and the actual A / F exceeds the threshold value, it is determined that “the air-fuel ratio control device is abnormal”.

As described above, according to the abnormality diagnosis device for an air-fuel ratio control apparatus for an internal combustion engine according to the above-described embodiment, N pieces of observation data z _k each including time-series data regarding the fuel injection amount and the actual A / F are converted into transient operation. It is classified into a group corresponding to the state and a group corresponding to the steady operation state. In the regression vector space, a hyperplane (separated hyperplane) that divides the distribution region of the observation data in the group corresponding to the transient operation state and the distribution region of the observation data in the group corresponding to the steady operation state is estimated. In addition, parameters in the “A / F behavior model for the transient operation state” are identified based on the observation data in the group corresponding to the transient operation state, and the local ARX model “A / F for the transient operation state” is identified. The “F behavior model” is constructed. Similarly, the parameters in the “A / F behavior model for the steady operation state” are identified based on the observation data in the group corresponding to the steady operation state, and “A / F for the steady operation state” which is the local ARX model. The “F behavior model” is constructed. In the state where the two A / F behavior models for the fuel injection amount are individually constructed as described above, based on the separation hyperplane estimated as described above and the time series data of the actual A / F from the past to the present. Thus, steady / transient prediction determination is performed, based on the A / F behavior model corresponding to the operation state determined by the prediction determination, the fuel injection amount from the past to the present, and the time series data of the actual A / F. The prediction A / F is estimated. Similarly, steady / transient actual determination is performed based on the actual operating state of the internal combustion engine 10, and the actual A / F is acquired from the A / F sensor. In the abnormality diagnosis of the air-fuel ratio control device, when the result of the prediction determination matches the result of the actual determination and the difference between the prediction A / F and the actual A / F is less than or equal to the threshold value, When the result of the prediction determination does not match the result of the actual determination, or when the difference between the prediction A / F and the actual A / F exceeds the threshold, it is determined that the air-fuel ratio control device is abnormal.

  Hereinafter, the operation and effect of the above configuration will be described. Normally, when an air-fuel ratio behavior model constructed based on data in a transient operation state is used, the estimation accuracy of the air-fuel ratio behavior in a steady operation state is lowered (and vice versa). On the other hand, according to the above-described embodiment, the “A / F behavior model for the transient operation state” and the “A / F behavior model for the steady operation state” are individually constructed, and the transient operation state is predicted and determined. The predicted A / F is estimated using the “A / F behavior model for the transient operation state” while the “A / F behavior model for the steady operation state” is predicted while being predicted to be the steady operation state. Is used to estimate the predicted A / F. Therefore, it is possible to estimate the predicted A / F stably with high accuracy regardless of the transient operation state or the steady operation state.

Usually, the accuracy of transient / steady state prediction determination is greatly influenced by aging of the air-fuel ratio control system, individual errors of the engine, and the like. On the other hand, according to the above embodiment, the separation hyperplane in the regression vector space is estimated based on the actual N observation data z _k , and transient / steady state prediction determination is performed using this separation hyperplane. Done. Therefore, it is possible to perform the transient / steady state prediction determination with high accuracy without being affected by the secular change of the air-fuel ratio control system, the individual error of the engine, or the like.

  Furthermore, when the abnormality diagnosis of the air-fuel ratio control device is performed based only on the comparison between the predicted A / F and the actual A / F, it cannot be said that the accuracy of the abnormality diagnosis is necessarily high. On the other hand, according to the above-described embodiment, the abnormality diagnosis of the air-fuel ratio control apparatus is not only a comparison between the predicted A / F and the actual A / F, but also the result of the prediction determination and the result of the actual determination regarding transient / steady state. It is also based on the comparison. Accordingly, the abnormality diagnosis is performed based on the plurality of pieces of comparison information, and thus the accuracy of the abnormality diagnosis is high.

  For example, both the predicted A / F and the actual A / F are in the vicinity of the theoretical air-fuel ratio, the difference between them is very small, and the result of the prediction determination for the transient / steady state is different from the result of the actual determination. In this case, if the abnormality diagnosis is performed only based on the comparison between the predicted A / F and the actual A / F, it is determined that the air-fuel ratio control device is normal. On the other hand, in the above embodiment, it can be determined that “the air-fuel ratio control apparatus is abnormal”.

The present invention is not limited to the above-described embodiment, and various modifications can be employed within the scope of the present invention. For example, in the above embodiment, after the ignition is turned on, when N pieces of observation data z _k are obtained, the A / F behavior model is constructed only once by the procedure of steps 205, 210, and 215 in FIG. May continue to use the A / F behavior model constructed until the ignition is turned off, or may repeatedly update the A / F behavior model with a predetermined period during the ignition on.

  In the above embodiment, two types of A / F behavior models, “A / F behavior model for transient operation state” and “A / F behavior model for steady operation state”, are constructed. The “operating state” is further divided into “decelerated operating state” and “accelerated operating state”, “A / F behavior model for decelerating operating state”, “A / F behavior model for accelerated operating state”, and “steady state” Three types of A / F behavior models of “A / F behavior model for operating state” may be constructed.

  In addition, in the above embodiment, Equation (1), Equation (21), etc. correspond to the transient operation state, Equation (2), Equation (22), etc. correspond to the steady operation state, (1) Equations (21) and the like may correspond to the steady operation state, and equations (2) and (22) and the like may correspond to the transient operation state.

1 is a schematic configuration diagram of a system in which an abnormality diagnosis device for an air-fuel ratio control device according to an embodiment of the present invention is applied to a spark ignition type multi-cylinder internal combustion engine. 2 is a flowchart showing a procedure for constructing an A / F behavior model and diagnosing an abnormality of an air-fuel ratio control device, which is executed by the abnormality diagnosis device shown in FIG. 1. It is the figure which simplified and expressed an example of the mixed normal distribution used for classification of observation data.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Spark ignition type multi-cylinder internal combustion engine, 39 ... Fuel injection valve, 61 ... Air flow meter, 62 ... Throttle valve opening sensor, 66 ... A / F sensor, 68 ... Accelerator opening sensor, 70 ... Electric control apparatus, 71 ... CPU, 73 ... RAM

Claims (3)

Air-fuel ratio control means is provided for controlling the air-fuel ratio by adjusting the amount of fuel injected from the fuel injection valve based on the detection result of the air-fuel ratio sensor for detecting the air-fuel ratio of the exhaust gas in the exhaust passage of the internal combustion engine. An abnormality diagnosis device applied to an air-fuel ratio control device for an internal combustion engine,
Storage means for storing past time-series data of the detected air-fuel ratio and past time-series data of the fuel injection amount for each of a plurality of different times;
The probability distribution in the regression vector space of the observation data for each time including the time series data of the detected air-fuel ratio and the time series data of the fuel injection amount for each time corresponds to the transient operation state. Under the assumption that a distribution trend and a distribution trend corresponding to a steady operation state follow a distribution having two distribution trends, a distribution having the two distribution trends is obtained based on observation data for each time, Classification means for classifying observation data for each time based on the obtained distribution into a group corresponding to the transient operation state and a group corresponding to the steady operation state;
Based on the observation data for each classified time, the distribution region of data in the group corresponding to the transient operation state and the distribution of data in the group corresponding to the steady operation state in the regression vector space Plane estimation means for estimating a separated hyperplane that is a hyperplane that divides the region;
An air-fuel ratio behavior model for the transient operation state and an air-fuel ratio behavior model for the steady operation state for predicting the air-fuel ratio from the time-series data of the air-fuel ratio to the present and the time-series data of the fuel injection amount to the present; Are individually expressed using different local ARX models, the parameters in the air-fuel ratio behavior model for the transient operation state and the parameters in the air-fuel ratio behavior model for the steady operation state are represented by the transient operation. By individually identifying the observation data classified into the group corresponding to the state and the observation data classified into the group corresponding to the steady operation state, respectively, the air-fuel ratio behavior model for the transient operation state and the Model construction means for individually constructing an air-fuel ratio behavior model for a steady operation state;
A prediction determination means for determining whether the internal combustion engine is in the transient operation state or the steady operation state based on the estimated separation hyperplane and time-series data of the detected air-fuel ratio up to the present time; ,
Actual determination means for determining whether the internal combustion engine is in the transient operation state or the steady operation state based on an actual operation state of the internal combustion engine;
Of the constructed air-fuel ratio behavior model, the air-fuel ratio behavior model corresponding to the operating state determined by the prediction determination means, the time-series data of the detected air-fuel ratio up to the present, and the time of the fuel injection amount up to the present Air-fuel ratio predicting means for predicting the air-fuel ratio based on the series data;
Based on the comparison between the determination result by the prediction determination means and the determination result by the actual determination means, and the comparison between the predicted air-fuel ratio and the detected air-fuel ratio, the presence or absence of abnormality of the air-fuel ratio control device is determined. An abnormality determination means to perform,
An abnormality diagnosis device for an air-fuel ratio control device for an internal combustion engine comprising: The abnormality diagnosis apparatus for an air-fuel ratio control apparatus for an internal combustion engine according to claim 1,
The abnormality determining means includes
When the determination result by the prediction determination unit matches the determination result by the actual determination unit and the difference between the predicted air-fuel ratio and the detected air-fuel ratio is equal to or less than a threshold value, the air-fuel ratio control apparatus determines that it is normal. When the determination result by the prediction determination means and the determination result by the actual determination means do not match or when the difference between the predicted air-fuel ratio and the detected air-fuel ratio exceeds the threshold, the air-fuel ratio control device An abnormality diagnosis device for an air-fuel ratio control device for an internal combustion engine configured to determine an abnormality. In the abnormality diagnosis device for an air-fuel ratio control device for an internal combustion engine according to claim 1 or 2,
The classification means includes
The observation data for each time is attributed to two of the two distribution trends, the distribution trend corresponding to the transient operation state and the distribution trend corresponding to the steady operation state of the obtained distribution. An abnormality diagnosis device for an air-fuel ratio control device for an internal combustion engine configured to be classified into a group corresponding to an operating state corresponding to a distribution tendency having a higher probability of performing.

Images