JP5402538B2 - Control valve abnormality determination device for internal combustion engine - Google Patents

Description

  The present invention relates to a control valve abnormality determination device applied to an internal combustion engine including a supercharger and a control valve for controlling the supercharger.

  2. Description of the Related Art Conventionally known superchargers (exhaust turbine superchargers) are disposed in an exhaust passage of an internal combustion engine and driven by the energy of exhaust gas, and are disposed in an intake passage of the engine and driven by the turbine. And a compressor driven by the operation. Thereby, the air flowing into the compressor is compressed by the compressor and discharged toward the combustion chamber. That is, supercharging is performed.

  One conventional internal combustion engine includes a first supercharger, a second supercharger connected in series to the first supercharger, and air supplied to the first supercharger and the second supercharger. Alternatively, a plurality of bypass passages for adjusting the flow rate of the exhaust gas and a plurality of control valves disposed in the bypass passages are provided. A control device (hereinafter referred to as “conventional device”) included in the conventional internal combustion engine changes the opening of each control valve in accordance with the operating state of the engine. Thereby, appropriate supercharging according to the operating state of the engine is performed.

  The conventional device is a control valve that adjusts the amount of air supplied to the member of the engine (for example, the compressor of the first supercharger, for the purpose of maintaining the above-described state where appropriate supercharging is performed. It is determined whether or not the intake air switching valve is operating normally. Specifically, the conventional device is determined in advance by experiments or the like, “when the intake air switching valve is operating normally, the engine speed, the throttle opening, and the air pressure at a predetermined location in the intake passage. And the relationship (pressure map). Furthermore, the conventional apparatus obtains the “estimated value of air pressure” at the predetermined location by applying the actual values of the engine speed and the throttle opening to the pressure map. When the “estimated value of air pressure” and the “actual air pressure (actually measured value)” at the predetermined location do not match with each other, the conventional device has a member (for example, an intake switching valve) constituting the engine. It is determined to be abnormal. (For example, see Patent Document 1).

  Hereinafter, the “compressor of the first supercharger” is also referred to as “first compressor” for convenience. Furthermore, “air pressure at a predetermined location in the intake passage” is also simply referred to as “air pressure” for convenience.

Japanese Utility Model Publication No. 2-78734

  As described above, the conventional apparatus expresses the relationship between the engine rotational speed, the throttle opening degree, and the air pressure by a “single pressure map determined in advance by experiments or the like” and uses this pressure map. An air pressure is estimated when a member (for example, an intake air switching valve) constituting the engine is normal. However, if the air pressure is estimated by this pressure map, the estimated air pressure may not be an appropriate value.

  For example, the engine includes an intake air switching valve disposed in a passage (bypass passage) that bypasses the first compressor. When the opening degree of the intake switching valve changes, the amount of air passing through the bypass passage changes without being supplied to the first compressor. More specifically, when the opening of the intake air switching valve is sufficiently small, substantially all of the air introduced into the engine flows into the first compressor. In this case, the air is compressed by the first compressor. On the other hand, when the opening of the intake switching valve is sufficiently large, the front and rear of the first compressor are short-circuited by the bypass passage, so that the air introduced into the engine does not substantially flow into the first compressor. In this case, air is not compressed by the first compressor. Therefore, even if the values of the engine speed and the throttle opening are the same, the air pressure in “when the intake switching valve opening is sufficiently small” and “when the intake switching valve opening is sufficiently large” Air pressure may be different. As described above, the air pressure changes not only due to the influence of the engine speed and the throttle opening, but also due to the influence of the opening of the intake switching valve.

  Therefore, if the air pressure is estimated using a “single pressure map” that does not consider the opening of the intake switching valve, an appropriate air pressure may not be estimated.

  Furthermore, for example, the members constituting the engine have manufacturing variations (differences in dimensions, performance, etc. between members of the same type that occur during manufacturing). Therefore, even if the values of the engine speed and the throttle opening are the same, the air pressure may be different for each individual engine. In other words, individual differences in the “relationship between engine rotational speed, throttle opening, and air pressure” may occur for each individual engine.

  Therefore, if the air pressure is estimated using a “pressure map determined in advance by experiments or the like” that does not take into account individual differences among members constituting the engine, an appropriate air pressure may not be estimated.

  Thus, there is a possibility that the conventional device cannot estimate an appropriate air pressure. As a result, the conventional device has a problem that it may not be possible to accurately determine the abnormality of the intake air switching valve.

  The present invention has been made to address the above problems. That is, one of the objects of the present invention is applied to the “internal combustion engine having a supercharger and a control valve” as described above, and accurately determines whether or not the control valve is operating normally. An object of the present invention is to provide a control valve abnormality determination device capable of

  The control valve abnormality determination device according to the present invention for achieving the above object is applied to an internal combustion engine including a first supercharger and a control valve.

  The first supercharger includes a turbine disposed in an exhaust passage of an internal combustion engine and a compressor disposed in an intake passage of the engine. This turbine is driven by exhaust gas discharged from the combustion chamber of the engine. Further, the compressor is driven by driving the turbine and compresses air in the intake passage.

  The engine may be configured to have a single supercharger (ie, to have only a first supercharger) and to have multiple superchargers (ie, a first supercharger). And one or more other superchargers). Furthermore, when the engine has a plurality of superchargers, the plurality of superchargers may be connected in series or in parallel.

  The control valve is provided in a passage portion that bypasses the compressor. That is, the passage portion may be configured such that one end thereof is connected to the intake passage on the upstream side of the compressor and the other end is connected to the intake passage on the downstream side of the compressor. Further, the control valve is configured to change a ratio between “the amount of air flowing into the compressor” and “the amount of air passing through the passage portion”.

  The control valve abnormality determination device of the present invention is applied to the above-described internal combustion engine. The control valve abnormality determination device includes time-series data acquisition means, data analysis means, operation state estimation means, and abnormality determination means.

The time series data acquisition means includes
The value of “predetermined operation parameter related to the supercharging state by the first supercharger” is acquired as time series data corresponding to the passage of time. As described above, the time series data acquisition means acquires the values of the operation parameters unique to the engine when the engine is actually operated as time series data.

The data analysis means includes
According to the piecewise affine analysis method,
(A) Classifying data vectors including the acquired time-series data as elements into “a plurality of data vector groups”;
(B) Using the plurality of classified data vector groups, estimate “a first function including one or a plurality of functions representing boundaries for dividing the classified plurality of data vector groups in a regression space”. And
(C) Using the time-series data included as elements in each of the plurality of classified data vector groups, for each of the plurality of classified data vector groups, “the dynamic characteristic of the operating parameter is an ARX model The second function expressed as (AutoRegressive eXogeneous model) is estimated.

  In (A) above, the method for classifying the data vectors into a plurality of data vector groups is not particularly limited. For example, well-known data classification methods such as the K-means method and the greedy method can be employed. In addition, for example, under the assumption that the data vector follows a mixed normal distribution, the optimal distribution parameter of the mixed normal distribution is calculated using the maximum likelihood estimation method, and the calculated optimal distribution parameter is used. The probability that a data vector at a given time belongs to each of a plurality of data vector groups (attribution probability) and classifies the data vector at that given time into the data vector group with the highest probability of belonging (See, for example, Hayato Nakada, Kiyotsugu Takada, & Tohru Katayama (2005). Identification of piecewise affine systems based on statistical clustering technique. Automatica, 41, 905-913.) Referred to as literature.).

  In (B) above, the method for estimating the “first function” is not particularly limited. For example, well-known data recognition methods such as a support vector machine (SVM) and a soft margin support vector machine (soft margin SVM) can be adopted.

  In the above (C), the ARX model is a kind of autoregressive model and is a well-known model used when identifying a system. Furthermore, in the above (C), the method for estimating the “second function” is not particularly limited. For example, a well-known approximate function estimation method such as a least square method and a weighted least square method may be employed.

  Hereinafter, for the sake of convenience, “estimating a function representing the dynamic characteristics of the operating parameter” is also referred to as “modeling the dynamic characteristics of the operating parameter”. According to this, for example, “estimating the second function that represents the dynamic characteristic of the driving parameter as an ARX model” is referred to as “modeling the dynamic characteristic of the driving parameter as an ARX model”.

  The “system (dynamic characteristics of operation parameters) as employed in the data analysis means is expressed as a combination of a plurality of ARX models (a plurality of second functions estimated for each of a plurality of data vector groups). The “model” is generally referred to as a piecewise affine autoregressive model (PWARX model, PieceWise affine AutoRegressive eXogeneous model). Furthermore, an analysis method for identifying a system as a piecewise affine autoregressive model is referred to as a piecewise affine analysis method.

  As described above, the data analysis means includes the first function representing the boundary for dividing the plurality of data vector groups in the regression space, and the dynamic characteristics of the operation parameter for each of the plurality of data vector groups. The second function expressed as is estimated.

The operating state estimating means includes
“Regression vector including, as an element, time-series data acquired by the time-series data acquisition unit at a“ first time point ”after the time point when the first function and the second function are estimated by the data analysis unit Is applied to the first function to determine which data vector group of the plurality of data vector groups is classified in the regression space, and
By applying the regression vector to the “second function corresponding to the data vector group in which the determined regression vector is classified in the regression space” of the second function, “after the first time point”. The value of the operation parameter at the “second time point” is estimated.

  As described above, the driving state estimation means uses the first function and the second function estimated by the data analysis means described above, from “a regression vector including time series data at a certain time point (first time point) as an element”. “The value of the operation parameter (estimated value) at a time point (second time point) after that time point” is estimated.

The abnormality determining means includes
By comparing the estimated value (estimated value) of the operating parameter at the second time point with the actual value (actually measured value) of the operating parameter at the second time point, the control valve is abnormal. Whether or not there is is determined.

  More specifically, the abnormality determining means is, for example, that the control valve is abnormal when the “absolute value of the difference between the estimated value of the operating parameter and the measured value of the operating parameter” is equal to or greater than a predetermined threshold. May be configured to determine. Further, the abnormality determining means, for example, when the “state in which the absolute value of the difference between the estimated value of the operating parameter and the measured value of the operating parameter is equal to or greater than a predetermined threshold” continues for a predetermined period or longer, It may be configured to determine that it is abnormal.

  As described above, the control valve abnormality determination device of the present invention models the dynamic characteristic of the operation parameter as a combination of a plurality of ARX models (a piecewise affine autoregressive model). Therefore, the control valve abnormality determination device of the present invention is more accurate in the value of the operation parameter at a predetermined time point (second time point) than the conventional device that models the dynamic characteristic of the operation parameter as a “single model”. It can be estimated well. Furthermore, the control valve abnormality determination device of the present invention models the dynamic characteristics of the operation parameter based on the value of the operation parameter acquired from the engine when the engine is actually operated. Therefore, the control valve abnormality determination device of the present invention can estimate the value of the operating parameter at a predetermined time point (second time point) more accurately than the conventional device using the “map acquired in advance by experiments”. Can do. As a result, the control valve abnormality determination device of the present invention can determine with high accuracy whether or not the control valve is operating normally.

Furthermore, the control valve abnormality determination device of the present invention is
“Each of the plurality of classified data vector groups” may be configured to correspond to “each of a plurality of opening regions dividing the movable range of the opening of the control valve”.

  As described above, the opening degree of the control valve affects the value of the operating parameter (supercharging pressure in the above-described conventional apparatus). Therefore, by configuring so that “each of the plurality of data vector groups” corresponds to “each of the plurality of opening regions of the control valve”, the dynamic characteristics of the operation parameter can be modeled with higher accuracy. . As a result, the control valve abnormality determination device of the present invention can determine whether or not the control valve is operating normally with higher accuracy.

  The number of “a plurality of opening regions of the control valve” and the range of each opening region can be determined in consideration of the influence of the opening of the control valve on the operating parameters. For example, according to experiment and experience, “the dynamic characteristics of the operation parameter when the opening degree of the control valve is less than or equal to the predetermined opening degree, and the dynamic characteristic of the operation parameter when the opening degree of the control valve is larger than the predetermined opening degree ”Are different” in advance, the “plurality of areas” includes two areas: “area where the opening is equal to or less than the predetermined opening” and “area where the opening is greater than the predetermined opening”. It can be defined to consist of two areas.

  On the other hand, when “the number of the plurality of opening regions” and “the range of each opening region” cannot be determined in advance by experiments and experience, the “number of the plurality of regions” is, for example, “Akaike information It can be estimated using a method of estimating the optimal number of regions based on a quantity criterion (CAIC) or a minimum description length (MDL) (see, for example, non-patent literature). Furthermore, the “range of each opening region” refers to, for example, acquiring time series data of operating parameters while associating the values of operating parameters and the opening of the control valve, and a data vector including the time series data as an element After being classified into a plurality of data vector groups by the data analysis means, it can be estimated from the opening degree of the control valve associated with the value of the operation parameter included in each of the classified data vector groups.

  Furthermore, the “plurality of opening regions” may include a region where the opening range is zero. That is, for example, “a plurality of regions” includes “a region where the opening is a fully closed opening (ie, the opening range is zero)” and “a region where the opening is an opening other than the fully closed”. May be included.

  In addition, there is no particular limitation on a method for determining which opening region of each of the “a plurality of data vector groups” corresponds to “a plurality of opening regions”. For example, an “experiment parameter value when the opening degree of the control valve belongs to a predetermined opening range” is acquired as a reference value by an experiment or the like in advance, and this reference value and “any one of the data vector group” By comparing the value of the operation parameter included as an element in the data vector, it is possible to determine which opening region the data vector group corresponds to. Further, for example, time series data is acquired while associating the value of the operation parameter and the opening of the control valve with each other, and “relates to the value of the operation parameter included as an element in any one data vector belonging to a certain data vector group. By referring to the “opening degree of the control valve”, it can be determined which opening range the data vector group corresponds to.

  Furthermore, the control valve abnormality determination device according to the present invention is based on “the control valve abnormality determination based on the operation parameter value at the second time point” described above, and “based on the opening degree of the control valve at the first time point”. To determine whether the control valve is abnormal ”.

More specifically, in the control valve abnormality determination device of the present invention,
The operating state estimating means includes
One of the plurality of opening regions corresponding to “the data vector group into which the regression vector determined by the regression vector and the first function is classified in the regression space” is defined as “the first time point. And is configured to estimate as a control valve estimated affiliation region that is an opening region to which the opening of the control valve belongs,
The abnormality determining means includes
The estimated “the control valve estimated affiliation region” and “the control valve actual affiliation region that is one of the plurality of opening regions to which the actual opening of the control valve at the first time point belongs” and , Can be configured to determine whether the control valve is abnormal.

  More specifically, the abnormality determination means may be configured to determine that the control valve is abnormal when, for example, the control valve estimated affiliation region and the control valve actual affiliation region do not match. Further, the abnormality determination means may be configured to determine that the control valve is abnormal when a state in which the control valve estimated affiliation region does not match the control valve actual affiliation region continues for a predetermined period or longer.

  As described above, the control valve abnormality determination device having the above-described configuration is a control valve abnormality determination based on the operation parameter value at the second time point, and a control valve abnormality determination based on the control valve opening degree at the first time point. Both can be done. Thereby, the control valve abnormality determination device of the present invention can determine whether or not the control valve is operating normally more accurately than when only one of these abnormality determinations is performed. .

Furthermore, in the control valve abnormality determination device of the present invention,
The plurality of opening regions are two regions: a region in which the opening of the control valve is a fully closed opening and a region in which the opening of the control valve is an opening other than the fully closed opening. Can be configured. The “fully closed opening degree” means “an opening degree at which air cannot substantially pass through the passage portion provided with the control valve (that is, the passage portion is blocked)”.

  As described above, in an engine to which the control valve abnormality determination device of the present invention is applied, the control valve is provided in a passage portion that bypasses the compressor. Therefore, if the opening degree of the control valve is “full opening degree”, the passage portion is blocked, so that substantially all of the air introduced into the engine flows into the compressor. On the other hand, if the opening degree of the control valve is “an opening degree other than the fully closed opening degree”, at least a part of the air introduced into the engine passes through the passage portion without flowing into the compressor. Therefore, the dynamic characteristic of the operating parameter when the opening degree of the control valve is a fully closed opening degree is different from the dynamic characteristic of the operating parameter when the opening degree of the control valve is an opening other than the fully closed opening degree. there is a possibility.

  Therefore, by modeling the dynamic characteristics of the operation parameter in each of the two regions, the control valve abnormality determination device of the present invention can be configured to have an opening region to which the opening of the control valve belongs at the first time point, and The value of the operating parameter at the second time point can be estimated with higher accuracy. As a result, the control valve abnormality determination device of the present invention can determine whether or not the control valve is operating normally with higher accuracy.

Furthermore, in the control valve abnormality determination device of the present invention,
The operating parameters are:
It may be configured to include at least one of an air pressure at a predetermined location in the intake passage, an air temperature at a predetermined location in the intake passage, and a rotation speed of the compressor.

As described above, the control valve abnormality determination device of the present invention can be applied to both an engine having a single supercharger and an engine having a plurality of superchargers. For example, when applying the control valve abnormality determination device of the present invention to an engine having a plurality of superchargers, as one aspect of the control valve abnormality determination device,
The control valve abnormality determination device,
“A second supercharger different from the first supercharger”, wherein a turbine of the second supercharger is disposed downstream of the turbine of the first supercharger in the exhaust passage and the same. A compressor of the second supercharger comprises a second supercharger disposed upstream of the compressor of the first supercharger in the intake passage;
The passage portion is branched from the intake passage at a branch portion between the compressor of the first supercharger and the compressor of the second supercharger, and a merging portion downstream of the compressor of the first supercharger It is configured to join the intake passage at
It may be configured to include an “intake throttle valve” that is disposed downstream of the merging portion of the intake passage and whose opening degree is changed.

Furthermore, in the control valve abnormality determination device of the above aspect,
The operating parameters included in the time-series data acquired by the time-series data acquisition means are “the supercharging pressure of the engine”, “the opening degree of the intake throttle valve”, “the compressor of the first supercharger” "Rotational speed" and "rotational speed of the compressor of the second supercharger",
The value of the operating parameter at the second time point estimated by the operating state estimating means is configured to be “the supercharging pressure of the engine”.

1 is a schematic view of an internal combustion engine to which a control valve abnormality determination device according to a first embodiment of the present invention is applied. It is the schematic of the control valve with which the internal combustion engine which concerns on FIG. 1 is provided. It is the schematic which shows the 1st example of the path | route of the intake and exhaust in the internal combustion engine to which the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention is applied. It is the schematic which shows the 2nd example of the path | route of the intake and exhaust in the internal combustion engine to which the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention is applied. It is the schematic which shows the relationship between the engine speed which the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention employ | adopts, fuel injection amount, and turbo mode. It is the flowchart which showed the routine which CPU of the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention performs. It is the flowchart which showed the routine which CPU of the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention performs. It is the flowchart which showed the routine which CPU of the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention performs. It is the flowchart which showed the routine which CPU of the control valve abnormality determination apparatus which concerns on 1st Embodiment of this invention performs. It is the flowchart which showed the routine which CPU of the control valve abnormality determination apparatus which concerns on 2nd Embodiment of this invention performs.

  Embodiments of an internal combustion engine control valve abnormality determination device according to the present invention will be described below with reference to the drawings.

(First embodiment)
<Outline of device>
FIG. 1 shows a schematic configuration of a system in which a control valve abnormality determination device (hereinafter also referred to as “first device”) according to a first embodiment of the present invention is applied to an internal combustion engine 10. The engine 10 is a four-cylinder diesel engine.

  The engine 10 includes an engine main body 20 including a fuel supply system, an intake system 30 for introducing air into the engine main body 20, an exhaust system 40 for discharging exhaust gas from the engine main body 20 to the outside, and an exhaust system 30 EGR device 50 for recirculation to the side, and supercharging device 60 that compresses air that is driven by the energy of exhaust gas and is introduced into engine body 20.

  The engine body 20 includes a cylinder head 21 to which an intake system 30 and an exhaust system 40 are connected. The cylinder head 21 has a plurality of fuel injection devices 22 provided in the upper part of each cylinder so as to correspond to each cylinder. Each fuel injection device 22 is connected to a fuel tank (not shown) and directly injects fuel into the combustion chamber of each cylinder in response to an instruction signal from the electric control device 80.

  The intake system 30 includes an intake manifold 31 communicated with each cylinder via an intake port (not shown) formed in the cylinder head 21, an intake pipe 32 connected to an upstream side assembly portion of the intake manifold 31, and an intake pipe 32. A throttle valve (intake throttle valve) 33 that makes the opening cross-sectional area of the intake passage variable, a throttle valve actuator 33a that rotates the throttle valve 33 in response to an instruction signal from the electric control device 80, and an intake pipe upstream of the throttle valve 33 And an air cleaner 35 disposed at the end of the intake pipe 32 on the upstream side of the supercharger 60 provided upstream of the intercooler 34. . The intake manifold 31 and the intake pipe 32 constitute an intake passage.

  The exhaust system 40 includes an exhaust manifold 41 communicated with each cylinder via an exhaust port (not shown) formed in the cylinder head 21, an exhaust pipe 42 connected to a downstream gathering portion of the exhaust manifold 41, and an exhaust pipe 42. And a well-known exhaust gas purifying catalyst (DPNR) 43 interposed in the exhaust pipe 42 on the downstream side of the supercharging device 60 provided in the engine. The exhaust manifold 41 and the exhaust pipe 42 constitute an exhaust passage.

  The EGR device 50 includes an exhaust gas recirculation pipe 51 constituting a passage (EGR passage) for recirculating exhaust gas from the exhaust manifold 41 to the intake manifold 31, an EGR gas cooling device (EGR cooler) 52 interposed in the exhaust gas recirculation pipe 51, And an EGR control valve 53 interposed in the exhaust gas recirculation pipe 51. The EGR control valve 53 can change the amount of exhaust gas recirculated from the exhaust manifold 41 to the intake manifold 31 in accordance with an instruction signal from the electric control device 80.

  The supercharger 60 has a high-pressure stage supercharger 61 and a low-pressure stage supercharger 62 connected in series to the high-pressure stage supercharger 61. The capacity of the low-pressure stage supercharger 62 is larger than the capacity of the high-pressure stage supercharger 61. That is, the minimum value of the exhaust gas energy required for the high pressure supercharger 61 to perform supercharging is smaller than the minimum value of the exhaust gas energy required for the low pressure stage supercharger 62 to perform supercharging. Thereby, the supercharging device 60 performs supercharging mainly by the high-pressure stage supercharger 61 in the operation region where the load is small, and supercharges mainly by the low-pressure stage supercharger 62 in the operation region where the load is large. It can be performed.

  More specifically, the high pressure supercharger 61 has a high pressure compressor 61a and a high pressure turbine 61b. The high-pressure compressor 61a is disposed in the intake passage (intake pipe 32). The high-pressure turbine 61b is disposed in the exhaust passage (exhaust pipe 42). The high-pressure compressor 61a and the high-pressure turbine 61b are connected to each other so as to be coaxially rotatable by a rotor shaft (not shown). Thus, when the high-pressure turbine 61b is rotated by the exhaust gas, the high-pressure compressor 61a rotates and the air supplied to the high-pressure compressor 61a is compressed (supercharging is performed).

  The low pressure stage supercharger 62 includes a low pressure stage compressor 62a and a low pressure stage turbine 62b. The low pressure stage compressor 62a is disposed upstream of the high pressure stage compressor 61a in the intake passage (intake pipe 32). The low pressure turbine 62b is disposed downstream of the high pressure turbine 61b in the exhaust passage (exhaust pipe 42). The low-pressure compressor 62a and the low-pressure turbine 62b are connected to each other so as to be coaxially rotatable by a rotor shaft (not shown). Thus, when the low-pressure stage turbine 62b is rotated by the exhaust gas, the low-pressure stage compressor 62a rotates and the air supplied to the low-pressure stage compressor 62a is compressed (supercharging is performed).

  Further, the supercharger 60 includes a high-pressure stage compressor bypass passage (bypass pipe) 63, an intake switching valve (ACV) 64, a high-pressure stage turbine bypass passage (bypass pipe) 65, an exhaust switching valve (ECV) 66, a low-pressure stage. A turbine bypass passage (bypass pipe) 67 and an exhaust bypass valve (EBV) 68 are provided.

  One end of the high pressure compressor bypass passage 63 is connected to the intake passage (intake pipe 32) between the high pressure compressor 61a and the low pressure compressor 62a. The other end of the high-pressure stage compressor bypass passage 63 is connected to the intake passage (intake pipe 32) on the downstream side of the high-pressure stage compressor 61a. In other words, the high-pressure stage compressor bypass passage 63 branches from the intake passage (intake pipe 32) at the branch portion between the high-pressure stage compressor 61a and the low-pressure stage compressor 62a, and also joins downstream from the high-pressure stage compressor 61a. Is configured to join the intake passage (intake pipe 32). Thus, the high pressure compressor bypass passage 63 constitutes a path that bypasses the high pressure compressor 61a.

  The intake air switching valve 64 is a butterfly valve disposed in the high-pressure stage compressor bypass passage portion 63. The intake air switching valve 64 is located inside the high-pressure stage compressor bypass passage 63 from the rotation position (full opening degree) shown in FIG. 2 (A) to the rotation position (full opening degree) shown in FIG. 2 (B). It is possible to rotate within the range up to. The intake air switching valve 64 is rotated by an intake air switching valve actuator 64a that is driven in accordance with an instruction from the electric control device 80.

  When the intake air switching valve 64 is in the position shown in FIG. 2A (fully closed opening), the air A cannot pass through the high-pressure compressor bypass passage portion 63. On the other hand, when the intake air switching valve 64 is at the position shown in FIG. 2B (full opening), the air A passes through the high-pressure compressor bypass passage 63 without being substantially affected by the intake air switching valve 64. be able to. That is, when the rotation position (opening degree) of the intake air switching valve 64 changes, the flow rate of the air A passing through the high-pressure compressor bypass passage portion 63 changes.

  Therefore, for example, as shown in FIG. 3, when the opening of the intake air switching valve 64 is a fully closed opening, substantially all of the air In introduced into the engine 10 flows into the high-pressure compressor 61a. This air In is compressed by the high pressure compressor 61a and flows into the combustion chamber CC. On the other hand, as shown in FIG. 4, when the opening degree of the intake air switching valve 64 is an opening degree other than the fully closed opening degree, a part of the air In introduced into the engine 10 flows into the high-pressure compressor 61a. The other part of the air In passes through the high-pressure compressor bypass passage 63 without flowing into the high-pressure compressor 61a. The air that has flowed into the high-pressure compressor 61a is compressed by the high-pressure compressor 61a, then merges with the air that has passed through the high-pressure compressor bypass passage 63, and flows into the combustion chamber CC.

  In this way, the intake air switching valve 64 changes its rotational position (opening degree) in accordance with an instruction from the electric control device 80, and flows into the high-pressure compressor 61a according to the rotational position (opening degree). The ratio between the amount of air and the amount of air passing through the high-pressure stage compressor bypass passage 63 is changed.

  Referring again to FIG. 1, one end of the high-pressure turbine bypass passage 65 is connected to the exhaust gas passage (exhaust pipe 42) on the upstream side of the high-pressure turbine 61b. The other end of the high-pressure turbine bypass passage 65 is connected to the exhaust gas passage (exhaust pipe 42) between the high-pressure turbine 61b and the low-pressure turbine 62b. That is, the high pressure turbine bypass passage 65 constitutes a path that bypasses the high pressure turbine 61b.

  The exhaust gas switching valve 66 is a butterfly valve disposed in the high pressure turbine bypass passage portion 65. The exhaust switching valve 66 has the same structure as the intake switching valve 64. That is, the opening degree of the exhaust gas switching valve 66 is changed by the exhaust gas switching valve actuator 66a that is driven in accordance with an instruction from the electric control device 80. When the opening degree of the exhaust gas switching valve 66 is changed, the flow area of the high-pressure stage turbine bypass passage portion 65 is changed along with the change in the opening degree, whereby the amount of exhaust gas flowing into the high-pressure stage turbine 61b is changed. The ratio of the amount of exhaust gas passing through the high-pressure turbine bypass passage 65 changes. Further, the exhaust gas that has passed through the high-pressure turbine bypass passage 65 is directed to the low-pressure turbine 62b without being introduced into the high-pressure turbine 61b. That is, the exhaust gas switching valve 66 changes the ratio between the amount of exhaust gas supplied to the high-pressure stage supercharger 61 and the amount of exhaust gas supplied to the low-pressure stage supercharger 62.

  One end of the low-pressure turbine bypass passage 67 is connected to the exhaust gas passage (exhaust pipe 42) upstream of the low-pressure turbine 62b and between the high-pressure turbine 61b and the low-pressure turbine 62b. The other end of the low pressure turbine bypass passage portion 67 is connected to the exhaust passage (exhaust pipe 42) on the downstream side of the low pressure turbine 62b. That is, the low-pressure stage turbine bypass passage portion 67 constitutes a path that bypasses the low-pressure stage turbine 62b.

  The exhaust bypass valve 68 is a butterfly valve disposed in the low pressure stage turbine bypass passage portion 67. The exhaust bypass valve 68 has the same structure as the intake switching valve 64. That is, the opening degree of the exhaust bypass valve 68 is changed by the exhaust bypass valve actuator 68a that is driven in accordance with an instruction from the electric control device 80. The exhaust bypass valve 68 changes the flow passage area of the low-pressure stage turbine bypass passage portion 67 in accordance with the change in the opening thereof, whereby the amount of exhaust gas flowing into the low-pressure stage turbine 62b and the low-pressure stage turbine bypass passage portion are changed. The ratio of the amount of exhaust gas passing through 67 is changed.

  Further, the first device includes a hot-wire air flow meter 71, an intake air temperature sensor 72, a throttle valve opening sensor 73, a supercharging pressure sensor 74, a crank position sensor 75, a high pressure supercharger rotational speed sensor 76, a low pressure overpressure. A feeder rotation speed sensor 77, an intake switching valve opening sensor 78, and an accelerator opening sensor 79 are provided.

  The hot-wire air flow meter 71 outputs a signal corresponding to the mass flow rate of intake air flowing through the intake pipe 32 (the mass of air sucked into the engine 10 per unit time, also simply referred to as “flow rate”). It is supposed to be.

  The intake air temperature sensor 72 outputs a signal corresponding to the temperature of intake air flowing through the intake pipe 32.

  The throttle valve opening sensor 73 outputs a signal corresponding to the opening of the throttle valve 33.

  The supercharging pressure sensor 74 is disposed on the downstream side of the throttle valve 33 in the intake pipe 32. The supercharging pressure sensor 74 is a pressure of air in the exhaust pipe 42 where the supercharging pressure sensor 74 is disposed, that is, a pressure of air supplied to the combustion chamber of the engine 10 (the first supercharger 61 and the second supercharger 61). A signal representing the supercharging pressure provided by the feeder 62 is output.

  The crank position sensor 75 outputs a signal having a narrow pulse every time the crankshaft (not shown) rotates 10 ° and a wide pulse every time the crankshaft rotates 360 °. Yes. Based on this signal, the number of rotations of the crankshaft per unit time is calculated.

  The high-pressure supercharger rotational speed sensor 76 outputs a signal corresponding to the rotational speed of the high-pressure supercharger 61 (high-pressure supercharger rotational speed).

  The low pressure turbocharger rotational speed sensor 77 outputs a signal corresponding to the rotational speed of the low pressure turbocharger 62 (low pressure turbocharger rotational speed).

  The intake switching valve opening sensor 78 outputs a signal corresponding to the opening of the intake switching valve 64.

  The accelerator opening sensor 79 outputs a signal corresponding to the opening of the accelerator pedal AP operated by the driver.

  The electric control device 80 includes a CPU 81, a ROM 82, and a RAM 83 that are connected to each other via a bus, a backup RAM 84 that stores data while the power is on, and holds the stored data while the power is shut off, and an AD A microcomputer including an interface 85 including a converter.

  The interface 85 is connected to the sensors and the like, and supplies signals from the sensors and the like to the CPU 81. Further, the interface 85 is connected to the fuel injection device 22 and each actuator (throttle valve actuator 33a, intake switching valve actuator 64a, exhaust switching valve actuator 66a, and exhaust bypass valve actuator 68a) in accordance with an instruction from the CPU 81. A drive signal (instruction signal) is transmitted.

<Outline of device operation>
Next, an outline of the operation of the first device configured as described above will be described.
The first device determines a “turbo mode” that represents the operation mode of the supercharger 60 (the high-pressure stage supercharger 61 and the low-pressure stage supercharger 62) according to the operating state of the engine 10. The first device adjusts the opening degree of the intake switching valve 64, the exhaust switching valve 66, and the exhaust bypass valve 68 according to the turbo mode. On the other hand, when the engine 10 is operated in this way, the first device operates parameters related to the supercharging state of the high-pressure supercharger 61 (low-pressure supercharger rotational speed, high-pressure supercharger rotational speed, The values of the throttle valve opening and the supercharging pressure are acquired as “time series data” corresponding to the passage of time. Then, according to the piecewise affine analysis method, the first device models the “supercharging pressure dynamic characteristics” as a plurality of ARX models corresponding to the opening of the intake switching valve 64 based on the acquired time-series data. .

  The first device estimates the value of the supercharging pressure at a predetermined time using the modeled supercharging pressure dynamic characteristic. Further, the first device acquires the actual value of the supercharging pressure at the predetermined time point based on the output value of the supercharging pressure sensor 74. Then, the first device determines whether or not the intake switching valve 64 is operating normally by comparing the “estimated supercharging pressure value” with the “actual supercharging pressure value”. .

  Further, when it is determined that the intake air switching valve 64 is “abnormal”, the first device notifies the operator of the engine 10 to that effect and performs “retreat operation” with a small burden on the members constituting the engine 10. ”Is executed. On the other hand, when it is determined that the intake air switching valve 64 is “normal”, the first device performs “normal operation” without notifying the operator. The above is the outline of the operation of the first device.

<Determination method of turbo mode>
Next, before explaining the specific operation of the present invention, the turbo mode employed in the first device and the determination method thereof will be described.

  As described above, the energy amount of the exhaust gas that allows the high-pressure stage supercharger 61 to operate is smaller than the energy amount of the exhaust gas that allows the low-pressure stage supercharger 62 to operate. Therefore, the first device controls the exhaust gas switching valve 66 so that the exhaust gas is preferentially supplied to the high-pressure supercharger 61 when the energy of the exhaust gas is small (that is, when the engine load is small). On the other hand, the first device controls the exhaust gas switching valve 66 so that the exhaust gas is preferentially supplied to the low-pressure supercharger 62 when the energy of the exhaust gas is large (that is, when the engine load is large). Further, the first device controls the exhaust bypass valve 68 so that excessive exhaust gas energy is not supplied to the low-pressure stage supercharger 62. In addition, the first device controls the intake air switching valve 64 so that an appropriate amount of air is supplied to the high-pressure compressor 61a.

  That is, the first device is configured so that an appropriate amount of air and exhaust gas are supplied to the high-pressure supercharger 61 and the low-pressure supercharger 62 according to the operating state of the engine 10. The switching valve 66 and the exhaust bypass valve 68 are controlled. As a result, the high pressure supercharger 61 and the low pressure supercharger 62 are appropriately driven according to the operating state of the engine 10. As a result, appropriate supercharging is performed.

  In order to execute such control, the first device divides the operating state of the engine 10 into four regions (operating regions), and an intake switching valve 64 and an exhaust switching valve 66 suitable for each of the four operating regions. And the operating state of the exhaust bypass valve 68 (hereinafter also referred to as “each control valve”). The operating state of each control valve is determined based on the turbo mode.

This turbo mode is determined as follows.
As shown in FIG. 5A, the first device stores a “turbo mode table MapTurbo (NE, Q) in which the relationship among the engine rotational speed NE, the fuel injection amount Q, and the turbo mode is predetermined” ROM82. Is stored. The numbers “1” to “4” shown in FIG. 5A indicate turbo mode numbers. Further, “HP + LP” shown in FIG. 5A indicates that both the high-pressure stage supercharger 61 and the low-pressure stage supercharger 62 operate, and “LP” indicates the low-pressure stage supercharger 62. Indicates that it operates preferentially.

  FIG. 5B shows the operating state of each control valve in each turbo mode. In FIG. 5B, “fully closed” is a control in which the opening degree of the control valve is set to an opening degree that closes the passage in which the control valve is provided, and air or exhaust gas cannot pass through the passage. Indicates the operating state of the valve. On the other hand, “full open” means that the opening degree of the control valve is set to an opening degree that completely opens (to the limit) the passage in which the control valve is provided, and air or exhaust gas substantially affects the passage of the control valve. The operation state of the control valve which can pass without receiving automatically is shown. Furthermore, “open” means that the opening of the control valve is set to an opening between “fully closed” and “fully open”, and the flow rate of air or exhaust gas passing through the passage in which the control valve is provided is controlled. The operation state of the control valve which can be changed according to the opening degree of a valve is shown.

  In FIG. 5B, “ECV” is an abbreviation for the exhaust gas switching valve 66, “ACV” is an abbreviation for the intake air switching valve 64, and “EBV” is an abbreviation for the exhaust gas bypass valve 68.

  The first device determines the turbo mode (the operating state of each control valve) by applying the actual engine speed NE and the fuel injection amount Q to the turbo mode table MapTurbo (NE, Q). And a 1st apparatus adjusts the opening degree of each control valve according to the determined turbo mode. Note that, when the engine 10 is operated in a decelerating state where the required torque for the engine 10 is equal to or less than a predetermined value, the first device is turbocharged regardless of the actual engine speed NE and the fuel injection amount Q. The opening of each control valve is adjusted according to mode 1. The above is the turbo mode employed in the first device and the determination method thereof.

<Control valve abnormality judgment method>
Next, a control valve abnormality determination method in the first device will be described. The first device determines whether or not the intake air switching valve 64 is operating normally according to the following procedure 1 to procedure 4.

(Procedure 1) The value of the operation parameter related to the supercharging state of the high-pressure supercharger 61 is acquired as time series data corresponding to the passage of time.
(Procedure 2) By analyzing the time series data acquired in Procedure 1 according to the piecewise affine analysis method, the dynamic characteristic of the supercharging pressure is modeled as a combination of a plurality of ARX models.
(Procedure 3) Using the dynamic characteristic of the supercharging pressure modeled in Procedure 2, the value of the supercharging pressure at the predetermined time point is estimated from the time series data at the time point before the predetermined time point.
(Procedure 4) Intake switching is performed by comparing the supercharging pressure value (estimated value) at the predetermined time point estimated in step 3 with the actual supercharging pressure value (actual value) at that time point. It is determined whether the valve 64 is operating normally.
Hereinafter, each of the procedure 1 to procedure 4 will be described in more detail.

(Procedure 1) The value of the operation parameter related to the supercharging state of the high-pressure supercharger 61 is acquired as time series data corresponding to the passage of time.
The first device first makes a preliminary determination as to whether or not the intake switching valve 64 is operating normally based on the output value of the intake switching valve opening sensor 78. More specifically, in the first device, when the engine 10 is started, the opening (actual opening) of the intake switching valve 64 obtained from the output value of the intake switching valve opening sensor 78, and the intake switching The opening degree (instruction opening degree) of the intake air switching valve 64 determined by the instruction signal for the valve actuator 64a is compared. The first device operates normally when the absolute value of the difference between the actual opening of the intake air switching valve 64 and the indicated opening of the intake air switching valve 64 is equal to or less than a predetermined value. Preliminary judgment is made.

  When the preliminary determination that the intake air switching valve 64 is operating normally is made, the first device is an operation parameter related to the supercharging state of the high-pressure supercharger 61 “low-pressure supercharger rotation speed”. The high-pressure-stage supercharger rotational speed, the throttle valve opening, and the supercharging pressure ”are acquired every time a predetermined sampling time elapses while corresponding to the elapse of time. The acquired values of these operating parameters are sequentially stored in the ROM 82 as time series data. The first device stores the time-series data until the amount of time-series data (the number of data of the time-series data) reaches a predetermined amount necessary for modeling the dynamic characteristics of the supercharging pressure in the following procedure 2. Continue to get. When the amount of time-series data reaches this predetermined amount, the first device stops acquiring time-series data.

  On the other hand, in the first device, when the preliminary determination that the intake air switching valve 64 is operating normally is not made (that is, when the preliminary determination that the intake air switching valve 64 is abnormal) is made. The time series data is not acquired. Further, at this time, the first device determines that “the intake switching valve 64 is abnormal” without executing steps 2 to 4 described later.

  Hereinafter, it is assumed that the preliminary determination that “the intake switching valve 64 is operating normally” is made in this procedure 1, and the description will be continued. Further, hereinafter, for convenience, the value of the low-pressure supercharger rotational speed at time k is λ (k), the value of the high-pressure supercharger rotational speed at time k is ν (k), and the throttle valve is opened at time k. The value of degree is referred to as e (k), and the value of the supercharging pressure at time k is referred to as ω (k).

(Procedure 2) By analyzing the time series data according to the piecewise affine analysis method, the dynamic characteristic of the supercharging pressure is modeled as a combination of a plurality of ARX models.
When the amount of time-series data acquired during the period in which the preliminary determination that “the intake switching valve 64 is operating normally” is made in step 1 above reaches the predetermined amount, The series data is analyzed according to the piecewise affine analysis method. Specifically, first, the first device determines “regression vector x (k) at time k” as shown in the following equation (1). In the following equation (1), omega value of (k-1) · · · omega (k-n _omega) time k-1 is also in the past than the time k Each · · · k-n boost pressure in _omega , lambda the (k-1) ··· λ than the time k, each of (k-n _lambda) of the low-pressure stage supercharger rotation speed at a past time k-1 ··· k-n _λ value, [nu (k-1) ··· ν values of the high-pressure stage supercharger rotational speed than the time k at past time _k-1 ··· k-n ν each _{(k-n ν), e} (k −1)... E (k−n _e ) represents the value of the throttle valve opening at time k−1... K−n _e past the time k. is n _omega, each _{n λ,} n ν and _{n e,} is a positive integer.

As shown in the above equation (1), the regression vector x (k) is a vector composed of the values of the operation parameters at the “time past the time k”. The regression vectors are sequentially stored in the ROM 82. This regression vector is obtained at each time (n, n−1, n−2,..., K, k−1, k−, where time n is the time when the amount of time series data reaches the predetermined amount. 2... Each time). Therefore, the regression vector, the amount of time-series data (the predetermined amount), as well as, n _omega, n _lambda, the value of n _[nu and n _e, present in an amount determined based on. That is, a plurality of regression vectors are stored in the ROM 82.

  Next, the first device, based on the regression vector x (k) and the supercharging pressure ω (k) at time k, “data vector z (k) at time k” as shown in the following equation (2). Is determined.

  As shown in the above equation (2), the data vector z (k) includes time-series data (“supercharging pressure ω (k)” at time k and “supercharging pressure” at a time point earlier than time k. , Low pressure stage turbocharger rotation speed, high pressure stage turbocharger rotation speed, and throttle valve opening "). This data vector is sequentially stored in the ROM 82. This data vector exists in an amount corresponding to the amount of the regression vector. That is, a plurality of data vectors are stored in the ROM 82.

  As described above, in the engine 10, the intake air switching valve 64 is provided in the high-pressure stage compressor bypass passage portion 63. Therefore, if the opening degree of the intake switching valve 64 is “fully closed opening degree”, substantially all of the air introduced into the engine 10 flows into the high-pressure compressor 61a. On the other hand, if the opening degree of the intake air switching valve 64 is “an opening degree other than the fully closed opening degree”, at least a part of the air introduced into the engine 10 does not flow into the high pressure compressor 61a, but the high pressure compressor. It passes through the bypass passage part 63. Therefore, the dynamic characteristics of the supercharging pressure when “the opening degree of the intake air switching valve 64 is a fully closed opening degree” and the case where “the opening degree of the intake air switching valve 64 is an opening degree other than the fully closed opening degree” are described. It may be different from the dynamic characteristics of the supercharging pressure. Therefore, the first device has a dynamic characteristic of the supercharging pressure when “the opening degree of the intake switching valve 64 is a fully closed opening degree”, and “the opening degree of the intake switching valve 64 is an opening other than the fully closed opening degree”. Each of the dynamic characteristics of the supercharging pressure in the case of “degree” is modeled as an ARX model according to the following procedure 2-1 to procedure 2-3.

(Procedure 2-1) The plurality of data vectors are divided into data vectors in the case where “the opening degree of the intake switching valve 64 is a fully closed opening degree” and the opening of the intake switching valve 64 other than the fully closed opening degree. Data vector in the case of “degree”.
(Procedure 2-2) A function for dividing the two data vector groups classified in the procedure 2-1 in the regression space is estimated.
(Procedure 2-3) The dynamic characteristics of the supercharging pressure are estimated as an ARX model in each of the two data vector groups classified in Procedure 2-1.
Hereinafter, each of the procedures 2-1 to 2-3 will be described in more detail.

(Procedure 2-1) The plurality of data vectors are divided into data vectors in the case where “the opening degree of the intake switching valve 64 is a fully closed opening degree” and the opening of the intake switching valve 64 other than the fully closed opening degree. Data vector in the case of “degree”.
More specifically, the first device first assumes that the probability density p (z; Φ) of the plurality of data vectors described above follows a mixed normal distribution represented by the following equation (3). In the following equation (3), Φ is a parameter of a mixed normal distribution, α is a scalar quantity, μ is a (n _ω + n _λ + n _v + n _e +1) -dimensional average vector, and Σ is (n _ω + n _λ + n _v + N _e +1) × (n _ω + n _λ + n _v + n _e +1) -dimensional dispersion matrix, i represents a variable corresponding to the state of the intake switching valve 64. In the following formulas (3) to (14), the variable i being “1” means that “the opening of the intake switching valve 64 is an opening other than the fully closed opening”, and the variable i is “2” means that “the opening degree of the intake switching valve 64 is a fully closed opening degree”.

In the above equation (3), the parameter Φ of the mixed normal distribution is defined as shown in the following equation (4), the scalar quantity α is defined as shown in the following equation (5), and the function p _i (z; μ _i , Σ _i ) is defined as shown in equation (6) below. In the following formula (6), nz represents a scalar amount shown in the following formula (7).

Then, the first device determines the most suitable parameter Φ _B (hereinafter referred to as “optimal parameter Φ _B ”) as the parameter Φ of the mixed normal distribution shown in the above equation (3) using the maximum likelihood estimation method. To do. More specifically, the first device, the parameter [Phi when the value of the following (8) likelihood function L in the expression ([Phi) is maximum is determined as an optimum parameter [Phi _B. The parameter Φ when the value of the likelihood function L (Φ) is maximized can be calculated by a known expected value maximization method (EM algorithm) or the like (for example, refer to non-patent literature). In the following equation (8), N represents the amount of data vectors (number of data). Furthermore, in the following formulas (8) and (9), the data vector z (k) is simply denoted as “z _k ” for convenience.

Next, when the first device uses the optimum parameter Φ _B acquired as described above, the plurality of data vectors are expressed as “when the opening degree of the intake switching valve 64 is an opening degree other than the fully closed opening degree”. And the data vector group C ₁ corresponding to the case where “the opening degree of the intake switching valve 64 is the fully closed opening degree” and the data vector group C ₂ corresponding to the case where “the opening degree of the intake switching valve 64 is a fully closed opening degree”.

More specifically, the first device is represented by the following equation (9): “a function representing the probability P (kεC _i ) that the data vector z (k) at time k belongs to the data vector group C _i ” By applying the data vector z (k) at time k to the first probability P (kεC ₁ ) that the data vector z (k) belongs to the data vector group C ₁ , and the data vector z The second probability P (kεC ₂ ) that (k) belongs to the data vector group C ₂ is acquired. The optimum parameter Φ _B is applied to α _i , μ _i and Σ _i in the following equation (9).

Then, the first device classifies the data vector z (k) into the data vector group C ₁ when the first probability P (kεC ₁ ) is larger than the second probability P (kεC ₂ ). On the other hand, the first device classifies the data vector z (k) into the data vector group C ₂ when the second probability P (kεC ₂ ) is larger than the first probability P (kεC ₁ ). .

The first device applies all of the plurality of data vectors to the function shown in the equation (9). Thus, the plurality of data vectors is classified into one of the data vector group C ₁ and the data vector group C _2. Hereinafter, for convenience, a set of data vectors belonging to the data vector group C ₁ is referred to as “cluster C ₁ ”, and a set of data vectors belonging to the data vector group C ₂ is referred to as “cluster C ₂ ”.

(Procedure 2-2) A function for dividing the two classified data vector groups in the regression space is estimated.
Next, the first device estimates a function (hereinafter referred to as “first function”) representing a boundary for dividing the cluster C ₁ and the cluster C ₂ in the regression space. More specifically, the first device first assumes that the first function is given by the following equation (10). In the following equation (10), a and b are coefficients of the first function, and x (k) is a regression vector at time k shown in the above equation (1). The plane on the regression space determined by the first function is also referred to as “separation hyperplane H”.

Then, the first device determines the coefficients a and b in the above equation (10) by solving the quadratic optimization problem represented by the following equation (11). In the following equation (11), σ _k is a parameter representing the degree of error (misclassification) that may occur when the data vector z (k) at time k is classified as described above. σ _k is set to an appropriate value as small as possible (see non-patent literature).

(Procedure 2-3) In each of the classified data vector groups, the dynamic characteristic of the supercharging pressure is estimated as an ARX model.
Further, the first device uses the data vector belonging to the cluster C ₁ to model the dynamic characteristic of the supercharging pressure when the “opening degree of the intake switching valve 64 is an opening degree other than the fully closed opening degree” as an ARX model. However, by using the data vector belonging to the cluster C ₂ is modeled as ARX model the dynamic characteristics of the boost pressure when "the opening of the intake air changeover valve 64 a is full-closed." Hereinafter, the modeled function representing the dynamic characteristic of the supercharging pressure is also referred to as a “second function”.

More specifically, the first device assumes that the supercharging pressure ω (k) is represented by the ARX model shown in the following equation (12). In the following equation (12), each of θ ₁ and θ ₂ is a parameter of the ARX model, x (k) is a regression vector at time k shown in the above equation (1), and ε (k) is at time k. Each of S ₁ and S ₂ is an area in the regression space, and each divided area when the regression space is divided by the separation hyperplane H shown in the above equation (10) is an error (noise). Represents. In the first device, the region S ₁ corresponds to a region in the regression space the cluster C ₁ in the case where the projection of the said cluster C ₁ regression space belongs, region S ₂ is obtained by projecting the cluster C ₂ to regression space the cluster C ₂ corresponds to a region in belonging regression space when. In other words, the region S ₁ is “a region in the regression space to which the regression vector belongs when the opening degree of the intake switching valve 64 is an opening other than the fully closed opening degree”, and the region S ₂ is “the intake switching valve 64. Is a region in the regression space to which the regression vector belongs when the opening is fully closed.

Then, the first device estimates the parameter θ ₁ and the parameter θ ₂ in the above equation (12) by the least square method shown in the following equation (13). In the following equation (13), each of x (k _i1 )... X (k _iNi ) is a regression vector value included as an element in the data vector belonging to the cluster C _i , and ω (k _i1 ). Each of (k _iNi ) represents a supercharging pressure value included as an element in a data vector belonging to the cluster C _i , and N _i represents the number of data vectors (number of data) included in the cluster C _i . Further, in the first device, the error ε (k) is set to zero.

Thus, according to the procedure 2 described above, the dynamic characteristic of the supercharging pressure in the case where “the opening degree of the intake air switching valve 64 is an opening degree other than the fully closed opening degree” (the upper stage of the above expression (12)), and Each of the dynamic characteristics of the supercharging pressure (the lower stage of the above expression (12)) in the case where “the opening degree of the intake switching valve 64 is a fully closed opening degree” is modeled as an ARX model. Further, according to this procedure 2, a set of data vectors (cluster C ₁ ) when “the opening degree of the intake air switching valve 64 is an opening degree other than the fully closed opening degree” and “the opening degree of the intake air switching valve 64 is fully closed”. A separation hyperplane H (the first function shown in the above equation (10)) that divides the set of data vectors (cluster C ₂ ) in the case of “the opening degree” in the regression space is estimated.

(Procedure 3) Using the modeled boost pressure dynamic characteristic, the value of the boost pressure at the predetermined time point is estimated from the time series data at the time point before the predetermined time point.
First, the first device calculates the regression vector x (t) at a predetermined time t after performing “modeling of the dynamic characteristic of the supercharging pressure” and “estimating the separation hyperplane H” by the procedure 2 described above. This is applied to the first function (separation hyperplane H) shown in equation (10). Since the first function is a function that divides the region S ₁ and the region S ₂ in the regression space, “the value of the first function when the regression vector x (t) is applied to the first function” is used. , “Whether the regression vector x (t) belongs to the region S ₁ or the region S ₂ ” can be determined. As shown in the above equation (1), the regression vector x (t) is a vector including time-series data at “time t−1 before (past) time t” as an element.

More specifically, when the regression vector x (t) is applied to the first function, if the value of the first function is zero, the regression vector x (t) exists on the separation hyperplane H. . On the other hand, if the value of the first function is greater than zero at this time, the regression vector x (t) is the region of the regression space divided by the separation hyperplane H (ie, the region S ₁ and the region S ₂ ). It exists in one area. On the other hand, if the value of the first function is smaller than zero at this time, the regression vector x (t) exists in another region different from the one region on the regression space. Furthermore, in the engine 10 of the present embodiment, it has been confirmed in advance through experiments or the like that this “one region” corresponds to the region S ₁ and this “other region” corresponds to the region S _2. Yes.

Therefore, when the regression vector x (t) is applied to the first function and the value of the first function is greater than or equal to zero, the first device determines that the regression vector x (t) is in the region S ₁ (intake switching valve). It is determined that it belongs to the region to which the regression vector belongs when the opening of 64 is an opening other than the fully closed opening. Further, when the regression vector x (t) is applied to the first function and the value of the first function is smaller than zero, the first device displays the regression vector x (t) in the region S ₂ (intake switching valve). It is determined that the regression vector belongs to a region where the opening degree of 64 is a fully closed opening degree.

  Next, the first device applies the regression vector x (t) to the second function shown in the equation (12) corresponding to the “region to which the regression vector x (t) belongs” determined as described above. . Since the second function is a function representing the dynamic characteristic of the supercharging pressure modeled as an ARX model, the supercharging pressure at time t is estimated by applying the regression vector x (t) to the second function. be able to. Hereinafter, the estimated supercharging pressure is referred to as supercharging pressure ωest (t).

More specifically, when the first device determines that the regression vector x (t) belongs to the region S ₁ , the first device sets the regression vector x (t) to a second function (above (12) corresponding to the region S _1. Applies to the upper part of the equation. Thereby, the supercharging pressure ωest (t) at time t is estimated. In contrast, the first device, the regression vector x (t) is determined to belong to the area S _2, the lower stage of the second function corresponding regression vector x (t) into the region S ₂ (equation (12) Applies to Thereby, the supercharging pressure ωest (t) at time t is estimated.

  That is, the first device estimates the boost pressure ω (t) according to the value of the first function when the regression vector x (t) is applied to the first function, as shown in the following equation (14). The second function used for the determination is determined. Then, the first device estimates the supercharging pressure ωest (t) at the time t by applying the regression vector x (t) to the determined second function. As described above, the supercharging pressure ωest (t) at “the time t” is estimated from the time-series data at the “time point before the time t” (time-series data included as an element in the regression vector x (t)). Is done.

(Procedure 4) The intake switching valve 64 is normal by comparing the estimated value (estimated value) of the supercharging pressure at a predetermined time point with the actual value (actually measured value) of the supercharging pressure at that time point. It is determined whether or not it is operating.
First, the first device acquires the actual boost pressure ωact (t) at time t based on the output value of the boost pressure sensor 74. Next, the first device compares the actual supercharging pressure ωact (t) at the time t with the supercharging pressure ωest (t) at the time t estimated in the procedure 3.

  More specifically, when the absolute value of the difference between the actual boost pressure ωact (t) and the estimated boost pressure ωest (t) is equal to or greater than a predetermined threshold, It is determined that the intake air switching valve 64 is abnormal. On the other hand, when the absolute value of the difference between the actual supercharging pressure ωact (t) and the estimated supercharging pressure ωest (t) is smaller than a predetermined threshold, the first device “intake switching valve” 64 is normal ”.

  As described above, the first device models the boost pressure dynamic characteristic as an ARX model according to the above-described procedure 1 to procedure 3, and uses the boost pressure dynamic characteristic modeled according to the above-described procedure 4 as an intake switching valve 64. It is determined whether or not is operating normally. The above is the control valve abnormality determination method in the first device.

<Actual operation>
Hereinafter, the actual operation of the first device will be described.
The CPU 81 executes each routine shown in the flowcharts in FIGS. 6 to 9 at predetermined timings. The CPU 81 uses the intake switching valve abnormality flag XACV in these routines.

  When the value of the intake switching valve abnormality flag XACV is “0”, it indicates that the intake switching valve 64 is not determined to be abnormal (normal). On the other hand, when the intake switching valve abnormality flag XACV is “1”, it indicates that the intake switching valve 64 is abnormal.

  The value of this flag is stored in the backup RAM 84. Further, the value of this flag is a predetermined value for the electric control device 80 when it is confirmed that there is no abnormality in the intake air switching valve 64 at the time of factory shipment of the vehicle on which the engine 10 is mounted and at the time of service inspection. Is set to “0”.

  Hereinafter, each routine executed by the CPU 81 will be described in detail. The CPU 81 repeatedly executes the “dynamic characteristic modeling routine” shown by the flowchart in FIG. 6 every time a predetermined time elapses. With this routine, the CPU 81 acquires the operating parameters of the engine 10 as time series data during the period from when the engine 10 is started until the modeling of the dynamic characteristics of the supercharging pressure is completed. Further, according to this routine, the CPU 81 models the dynamic characteristic of the supercharging pressure as an ARX model based on the acquired time series data of the operation parameters.

  Specifically, when the engine 10 is started after shipment from the factory, the CPU 81 starts processing from step 600 in FIG. 6 at a predetermined timing and proceeds to step 610 to model the dynamic characteristics of the supercharging pressure. Whether or not is completed. Since the current time is immediately after the engine 10 is started after shipment from the factory, a sufficient amount of time-series data (predetermined amount in the procedure 1) is not acquired. For this reason, modeling of the dynamic characteristics of the supercharging pressure has not been completed. Therefore, the CPU 81 makes a “No” determination at step 610 to proceed to step 620.

  In step 620, the CPU 81 determines that “the opening degree (actual opening degree) of the intake switching valve 64 obtained from the output value of the intake switching valve opening degree sensor 78 (actual opening degree) Act (t)” at the present time (time t), The absolute value of the difference between “the opening degree (instruction opening degree) Odir (t) of the intake air switching valve 64 determined by the instruction signal sent from the electric control device 80 to the intake air switching valve actuator 64a” at time t) is a predetermined value. It is determined whether or not it is smaller than the threshold value δ1. That is, in step 620, the CPU 81 determines whether or not the intake air switching valve 64 is operating normally in response to an instruction from the electric control device 80. This threshold value δ1 is an appropriate value at which the absolute value of the difference between the actual opening Oact (t) and the indicated opening Odir (t) is smaller than the threshold δ1 when the intake air switching valve 64 is operating normally. Is set to The determination executed in step 620 corresponds to the “preliminary determination” in the procedure 1.

  At this time, if the absolute value of the difference between the actual opening Oact (t) and the command opening Odir (t) is smaller than the threshold value δ1 (that is, if the intake air switching valve 64 is operating normally). The CPU 81 determines “Yes” at step 620 and proceeds to step 630.

  In step 630, the CPU 81 sets “low pressure stage turbocharger rotational speed”, “high pressure stage turbocharger rotational speed”, “throttle valve opening”, and “supercharging pressure” as operating parameters of the engine 10 at the present time. Is acquired in correspondence with the passage of time. The obtained values of these operating parameters are stored in the ROM 82 as time series data (see Procedure 1 described above). Thereafter, the CPU 81 proceeds to step 640. Hereinafter, for convenience, the time series data of the operation parameters is also simply referred to as “time series data”.

  In step 640, the CPU 81 determines whether or not “a condition for modeling the dynamic characteristic of the supercharging pressure (modeling condition)” is satisfied. More specifically, the CPU 81 determines in step 630 that the modeling condition is satisfied when both of the following condition 1 and condition 2 are satisfied. In other words, the CPU 81 determines that the modeling condition is not satisfied when at least one of the condition 1 and the condition 2 is not satisfied.

(Condition 1)
The amount of time series data acquired when the operation (turbo mode 1 or turbo mode 2) in which the opening degree of the intake switching valve 64 becomes a fully closed opening degree is performed is equal to or greater than a predetermined first threshold data amount. .
(Condition 2)
The amount of time-series data acquired when the operation (turbo mode 3 or turbo mode 4) in which the opening degree of the intake switching valve 64 is an opening degree other than the fully closed opening degree is performed is the predetermined second threshold data. More than the amount.

  As described above, the first device has the dynamic characteristics of the supercharging pressure when “the opening degree of the intake switching valve 64 is the fully closed opening degree” and “the opening degree of the intake switching valve 64 is the fully closed opening degree”. Each of the dynamic characteristics of the supercharging pressure in the case of “opening other than” is modeled as an ARX model. In order to perform this modeling, the amount of time-series data when the opening degree of the intake switching valve 64 becomes a fully closed opening degree, and the opening degree of the intake switching valve 64 is an opening degree other than the fully closing opening degree. It is necessary to acquire time-series data until both of the amounts of time-series data in a given case reach an amount necessary for modeling (a predetermined amount in Procedure 1 above). That is, the engine 10 is operated so that the opening degree of the intake switching valve 64 becomes a fully closed opening degree (turbo mode 1 or turbo mode 2), and the opening degree of the intake switching valve 64 is an opening degree other than the fully closed opening degree. It is necessary to fully experience both the operation (turbo mode 3 or turbo mode 4) (see also FIG. 5). In other words, if both condition 1 and condition 2 are satisfied, the dynamic characteristics of the supercharging pressure can be modeled.

  Since the current time is immediately after the engine 10 is started, a sufficient amount of time series data that satisfies the modeling conditions has not been acquired. Therefore, the modeling condition is not satisfied. Therefore, the CPU 81 makes a “No” determination at step 640 to proceed to step 695 to end the present routine tentatively. Thus, when modeling of the dynamic characteristic of the supercharging pressure is not completed, modeling of the dynamic characteristic of the supercharging pressure is not performed unless the modeling condition is satisfied.

  Further, the CPU 81 repeatedly executes the “first abnormality determination routine” shown by the flowchart in FIG. 7 every time a predetermined time elapses. By this routine, the CPU 81 determines whether or not the intake air switching valve 64 is abnormal.

  Specifically, the CPU 81 starts the process from step 700 in FIG. 7 at a predetermined timing and proceeds to step 710 to determine whether or not modeling of the dynamic characteristic of the supercharging pressure has been completed. Since modeling of the dynamic characteristic of the supercharging pressure has not been completed at this time, the CPU 81 makes a “No” determination at step 710 to directly proceed to step 795 to end the present routine tentatively. Accordingly, when modeling of the dynamic characteristic of the supercharging pressure is not completed, the abnormality determination of the intake air switching valve 64 by this routine is not performed.

  Further, the CPU 81 repeatedly executes the “abnormality notification routine” shown by the flowchart in FIG. 8 every time a predetermined time elapses. The CPU 81 notifies the operator of the engine 10 to this effect when the intake air switching valve 64 is abnormal by this routine.

  Specifically, the CPU 81 starts the process from step 800 in FIG. 8 at a predetermined timing and proceeds to step 810 to determine whether or not the value of the intake switching valve abnormality flag XACV is “0”. . As described above, at the present time, the intake air switching valve 64 is operating normally. That is, since the value of the intake switching valve abnormality flag XACV at the present time is “0”, the CPU 81 makes a “Yes” determination at step 810 and proceeds directly to step 895 to end the present routine tentatively. Therefore, when the intake air switching valve 64 is normal (when the value of the intake air switching valve abnormality flag XACV is “0”), the operator is not notified.

  Furthermore, every time the crank angle of an arbitrary cylinder coincides with a predetermined crank angle before compression top dead center (for example, 90 degrees crank angle before compression top dead center) θf, the CPU 81 shows the “fuel” shown in the flowchart in FIG. The “supply control routine” is repeatedly executed. The CPU 81 calculates the fuel injection amount Q and instructs fuel injection by this routine. The cylinder in the compression stroke whose crank angle coincides with the predetermined crank angle θf before the compression top dead center is also referred to as “fuel injection cylinder” hereinafter.

  Specifically, when the crank angle of any cylinder reaches the crank angle θf, the CPU 81 starts the process from step 900 in FIG. 9 and proceeds to step 910, where the value of the intake switching valve abnormality flag XACV is “0”. Is determined. Since the value of the intake switching valve abnormality flag XACV at this time is “0”, the CPU 81 determines “Yes” in step 910 and proceeds to step 920.

  In step 920, the CPU 81 acquires the accelerator pedal opening degree Accp based on the output value of the accelerator position sensor 79, and acquires the engine speed NE based on the output value of the crank position sensor 75. Then, the CPU 81 sets a normal fuel injection amount table MapMain (predetermined “relationship between the accelerator pedal opening Accp, the engine rotational speed NE, and the fuel injection amount Q” when the intake air switching valve 64 is normal. The fuel injection amount Q is acquired by applying the current accelerator pedal opening Accp and the engine speed NE to (Accp, NE). This fuel injection amount Q corresponds to the required torque. Hereinafter, the operation employing the fuel injection amount determined by the normal fuel injection amount table MapMain (Accp, NE) is referred to as “normal operation”.

  Next, the CPU 81 proceeds to step 930 and gives an instruction to the injector 22 to inject the fuel of the fuel injection amount Q from the injector 22 provided corresponding to the fuel injection cylinder. That is, at this time, fuel of the fuel injection amount Q is supplied to the fuel injection cylinder. Thereafter, the CPU 81 proceeds to step 995 to end the present routine tentatively.

  As described above, when the modeling of the dynamic characteristic of the supercharging pressure is not completed and the intake switching valve 64 is “normal”, the fuel injection determined by the normal fuel injection amount table MapMain (Accp, NE). A “normal operation” in which an amount Q of fuel is supplied to the fuel injection cylinder is executed.

  On the other hand, when modeling of the boost pressure dynamic characteristic is not completed and the intake switching valve 64 is “abnormal”, the CPU 81 starts processing from step 600 in FIG. 6 at a predetermined timing. The process proceeds to step 620 via step 610 following step 600. When the intake switching valve 64 is “abnormal”, the absolute value of the difference between the actual opening Oact (t) and the instruction opening Odir (t) of the intake switching valve 64 is equal to or greater than the threshold value δ1, so the CPU 81 In step 620, “No” is determined, and the process proceeds to step 650.

  In step 650, the CPU 81 stores “1” as the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds directly to step 695 to end the present routine tentatively. Therefore, when modeling of the dynamic characteristic of the supercharging pressure is not completed, time series data is not acquired if the intake switching valve 64 is abnormal.

  At this time, when the CPU 81 starts the process from step 800 in FIG. 8 at a predetermined timing, the value of the intake switching valve abnormality flag XACV at the present time is “1”. And proceeds to step 820.

  In step 820, the CPU 81 notifies the operator of the engine 10 that “the intake switching valve 64 is abnormal”. This notification is executed by turning on an alarm lamp (not shown). Thereafter, the CPU 81 proceeds to step 895 to end the present routine tentatively.

  Further, when the crank angle of an arbitrary cylinder matches the crank angle θf, the CPU 81 starts the process from step 900 in FIG. 9 and proceeds to step 910. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 makes a “No” determination at step 910 to proceed to step 940.

  In step 940, the CPU 81 acquires the accelerator pedal opening Accp based on the output value of the accelerator opening sensor 79, and acquires the engine speed NE based on the output value of the crank position sensor 75. Then, the CPU 81 sets the “relationship between the accelerator pedal opening Accp, the engine rotational speed NE, and the fuel injection amount Q” applied when “the intake switching valve 64 is abnormal” when an abnormality occurs in advance. By applying the current accelerator pedal opening degree Accp and the engine speed NE to the fuel injection amount table MapEmg (Accp, NE), the fuel injection amount Q at the time of occurrence of an abnormality is acquired. Hereinafter, the operation employing the fuel injection amount determined by the abnormality-occurring fuel injection amount table MapEmg (Accp, NE) is also referred to as “evacuation operation”.

  The fuel injection amount table MapEmg (Accp, NE) at the time of occurrence of an abnormality indicates that “even if the operation of the engine 10 is continued when the intake switching valve 64 is abnormal, other members of the engine 10 or the entire engine 10 is damaged. FIG. 10 is a table for determining a fuel injection amount “Q” that does not cause any fuel injection. Therefore, as a matter of course, the fuel injection amount determined by the fuel injection amount table MapEmg (Accp, NE) at the time of occurrence of abnormality for any “accelerator pedal opening Accp and engine speed NE” is the “accelerator pedal opening Accp. And the engine speed NE "is smaller than the fuel injection amount determined by the normal fuel injection amount table MapMain (Accp, NE).

  Next, the CPU 81 proceeds to step 930 and injects fuel of the fuel injection amount Q from the injector 22 provided corresponding to the fuel injection cylinder. Thereafter, the CPU 81 proceeds to step 995 to end the present routine tentatively.

  As described above, when the modeling of the dynamic characteristic of the supercharging pressure is not completed, when the intake switching valve 64 is “abnormal” (that is, “the intake switching valve 64 is abnormal in the preliminary determination in the above procedure 1). If it is determined that there is an abnormality), an alarm indicating that the exhaust gas switching valve 66 is abnormal is issued to the operator of the engine 10, and the abnormality occurrence fuel injection amount table MapEmg (Accp, NE) is used. “Evacuation operation” is performed in which fuel of a predetermined fuel injection amount Q is supplied to the fuel injection cylinder.

  Hereinafter, when modeling of the dynamic characteristic of the supercharging pressure is not completed, it is assumed that the intake switching valve 64 is “normal” (that is, “the intake switching valve 64 is normal in the preliminary determination in the above procedure 1). The explanation is continued (assuming that it is determined to be “operating”).

  If the state shown in the above assumption continues, the processing from step 610 to step 640 in FIG. 6 is repeatedly executed, so that the amount of time-series data gradually increases. Furthermore, during the period in which the state shown in the above assumption continues, the engine 10 is operated so that the opening degree of the intake switching valve 64 becomes a fully closed opening degree (turbo mode 1 or turbo mode 2), and the intake switching valve 64. The above-described modeling condition is established if both the operation (turbo mode 3 or turbo mode 4) in which the opening of the valve is an opening other than the fully closed opening are sufficiently experienced. When the modeling condition is satisfied, the CPU 81 executes the process of step 640, determines “Yes” in step 640, and proceeds to step 660.

In step 660, as described with reference to the above formulas (1) to (9), the CPU 81 sets a data vector including time series data as an element “the opening degree of the intake switching valve 64 is fully closed / open. Data vector group (cluster C ₁ ) in the case of “degree”, and data vector group (cluster C ₂ ) in the case of “opening degree of intake switching valve 64 other than the fully closed opening degree”. (See Procedure 2-1 above.) Thereafter, the CPU 81 proceeds to step 670.

In step 670, the CPU 81 classifies the two classified data vector groups (cluster C ₁ and cluster C ₂ ) in the regression space as described with reference to the equations (10) and (11). A first function (separation hyperplane) is estimated (see procedure 2-2 above). Thereafter, the CPU 81 proceeds to step 680.

In step 680, the CPU 81 determines the boost pressure in each of the classified data vector groups (cluster C ₁ and cluster C ₂ ) as described with reference to the expressions (12) and (13). Is estimated as an ARX model (see Procedure 2-3 described above). As described above, the modeled function representing the dynamic characteristic of the boost pressure is also referred to as a “second function”. This second function is stored in the ROM 82.

  Thereafter, the CPU 81 proceeds to step 695 to end the present routine tentatively. Therefore, when modeling of the supercharging pressure dynamic characteristic is not completed, modeling of the supercharging pressure dynamic characteristic is performed if the above modeling condition is satisfied.

  Further, once the modeling of the dynamic characteristics of the supercharging pressure is completed, the CPU 81 determines “Yes” in step 610 when starting the process from step 600 in FIG. 6, and proceeds directly to step 695 to execute this routine. Is temporarily terminated. Therefore, “time-series data for modeling dynamic characteristics” is not acquired after the modeling of the dynamic characteristics of the supercharging pressure is completed.

  When the modeling of the dynamic characteristic of the supercharging pressure is completed, the CPU 81 estimates the supercharging pressure at a predetermined time as described with reference to the above equation (14) (see procedure 3 above).

Specifically, when the CPU 81 starts the process from step 700 in FIG. 7 at a predetermined time t after the time when the supercharging pressure modeling is completed, the CPU 81 determines “Yes” in step 710 following step 700. And proceed to step 720. In step 720, the CPU 81 obtains the regression vector x (t) at the time t and applies the regression vector x (t) to the first function (separation hyperplane). It is determined whether or not the function value (a ^T x (t) + b) is equal to or greater than zero.

  At this time, if the value of the first function is greater than or equal to zero, the CPU 81 makes a “Yes” determination at step 720 to proceed to step 730, and sets the regression vector x (t) at time t to “the intake switching valve 64. This is applied to the second function (ARX model) corresponding to the case where the opening is an opening other than the fully-closed opening (see the upper stage of equation (14) above). Thereby, the CPU 81 estimates the supercharging pressure at time t, and acquires the estimated supercharging pressure as ωest (t). As described above, in step 730, the error ε (t) is set to zero.

  On the other hand, if the value of the first function is smaller than zero at this time, the CPU 81 makes a “No” determination at step 720 to proceed to step 740, and sets the regression vector x (t) at time t to “intake”. This is applied to the second function (ARX model) corresponding to the case where the opening degree of the switching valve 64 is a fully closed opening degree (see the lower part of the above equation (14)). Thereby, the CPU 81 estimates the supercharging pressure at time t, and acquires the estimated supercharging pressure as ωest (t). Also in step 740, the error ε (t) is set to zero.

  In this way, the second function (ARX model) to which the regression vector x (t) is applied according to the value of the first function when the regression vector x (t) is applied to the first function (separation hyperplane). ) Is determined. Further, the boost pressure at time t is estimated by applying the regression vector x (t) to the determined second function. Thereafter, the CPU 81 proceeds to step 750.

  In step 750, the CPU 81 determines whether or not the intake air switching valve 64 is operating normally (see procedure 4 above). More specifically, the CPU 81 determines in step 750 the actual boost pressure ωact () determined by the boost pressure ωest (t) estimated as described above and the output value of the boost pressure sensor 74 at time t. It is determined whether or not the absolute value of the difference between t) is smaller than a predetermined threshold value δ2. This threshold δ2 is such that when the intake air switching valve 64 is operating normally, the absolute value of the difference between the estimated supercharging pressure ωest (t) and the actual supercharging pressure ωact (t) is greater than the threshold δ2. It is set to an appropriate value that decreases.

  If the absolute value of the difference between the estimated supercharging pressure ωest (t) and the actual supercharging pressure ωact (t) is smaller than the threshold value δ2 at the present time (time t) (that is, the intake switching valve 64 is If it is operating normally), the CPU 81 makes a “Yes” determination at step 750 and proceeds to step 760.

  In step 760, the CPU 81 stores “0” in the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds to step 795 to end the present routine tentatively.

  At this time, when the CPU 81 starts processing from step 800 in FIG. 8, the CPU 81 proceeds to step 810. Since the value of the intake switching valve abnormality flag XACV at this time is “0”, the CPU 81 determines “Yes” in step 810, proceeds to step 895, and ends this routine once.

  Further, at this time, when the CPU 81 starts the process from step 900 in FIG. 9, the process proceeds to step 910. Since the current value of the intake air switching valve abnormality flag XACV is “0”, the CPU 81 determines “Yes” in step 910. Thereafter, the CPU 81 proceeds to step 995 via step 920 and step 930, and once ends this routine.

  Accordingly, at this time, the operator is not notified that “the intake switching valve 64 is abnormal”, and “normal operation” is executed.

  On the other hand, if the absolute value of the difference between the estimated supercharging pressure ωest (t) and the actual supercharging pressure ωact (t) is equal to or greater than the threshold δ2 at the present time (time t) (that is, the intake air If the switching valve 64 is abnormal), the CPU 81 makes a “No” determination at step 750 in FIG.

  In step 770, the CPU 81 stores “1” as the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds to step 795 to end the present routine tentatively.

  At this time, when the CPU 81 starts processing from step 800 in FIG. 8, the CPU 81 proceeds to step 810. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 810. Thereafter, the CPU 81 proceeds to step 895 via step 820 and once ends this routine.

  Further, at this time, when the CPU 81 starts the process from step 900 in FIG. 9, the process proceeds to step 910. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 910. Thereafter, the CPU 81 proceeds to step 995 via step 940 and step 930, and once ends this routine.

  Accordingly, at this time, the operator is notified that “the intake switching valve 64 is abnormal” and the “evacuation operation” is executed.

  As described above, the first device acquires the operation parameter as time series data when the intake switching valve 64 is operating normally, and based on the time series data, the dynamic characteristic of the supercharging pressure is determined by the intake switching valve 64. Modeled as a combination of a plurality of ARX models according to the opening. Further, the first device estimates the supercharging pressure at a predetermined time using the modeled dynamic characteristic of the supercharging pressure. In addition, the first device determines whether or not the intake air switching valve 64 is operating normally by comparing the estimated supercharging pressure with the actual supercharging pressure. The first device performs normal operation when it is determined that the intake air switching valve 64 is normal, and to the operator when it is determined that the intake air switching valve 64 is abnormal. And the evacuation operation is executed.

As described above, the first device is
The present invention is applied to an internal combustion engine 10 that includes a first supercharger 61 and a control valve (intake switching valve 64).

This first device is
Time-lapse of predetermined operating parameters (low pressure turbocharger rotation speed, high pressure turbocharger rotation speed, throttle valve opening, and supercharging pressure) related to the supercharging state by the first supercharger 61 6 is provided with time series data acquisition means (step 630 in FIG. 6) for acquiring time series data corresponding to the above.

The first device further comprises:
According to the piecewise affine analysis method, the data vector z (k) including the acquired time series data as an element is classified into a plurality of data vector groups (cluster C ₁ and cluster C ₂ ), and the plurality of classified data vectors A first function composed of a single function or a plurality of functions representing a boundary for dividing the group of the plurality of classified data vectors in the regression space using a group (separation hyperplane H, see the above formula (10)). And using the time-series data included as elements in each of the plurality of classified data vector groups, for each of the plurality of classified data vector groups, “the operation parameter (supercharging pressure) ) Is a data analysis means (step 660 in FIG. 6) for estimating the second function (see the above equation (12)) representing the dynamic characteristics of the It has an optimal step 680).

The first device further comprises:
The time series at a first time point (time point before time t) after the time point when the first function (separation hyperplane H) and the second function (ARX model) are estimated by the data analysis means. By applying a regression vector x (t) including time series data acquired by the data acquisition means as an element to the first function, the regression vector x (t) is converted into the plurality of data vector groups (cluster C ₁ and A data vector group of the cluster C ₂ ) to be classified in the regression space, and the determined same regression vector of the second function is classified into a data vector group to be classified in the regression space. By applying the regression vector x (t) to the corresponding second function, “the value of the operation parameter at the second time (time t) after the first time (time t) And a driving state estimation means for estimating a pressure .omega.est (t)) "(step 720 through step 740 of FIG. 7).

The first device further comprises:
By comparing the estimated value ωest (t) of the operation parameter at the second time point t with the actual value ωact (t) of the operation parameter at the second time point t, the control valve 64 Is provided with an abnormality determination means (step 750 in FIG. 7) for determining whether or not is abnormal.

Furthermore, in the first device:
Each of the classified “plurality of data vector groups (cluster C ₁ and cluster C ₂ )” corresponds to each of “a plurality of opening regions” that divides the movable range of the opening of the control valve 64. It is configured.

More specifically, the plurality of opening regions are:
"(Corresponding to the region S _2.) Area opening is full-closed of the control valve" and the "opening of the control valve is in the region (region S ₁ which is the opening other than the full-closed Corresponding.) ”).

More specifically,
The engine 10 to which the first device is applied is:
The second supercharger 62 is different from the first supercharger 61, and the turbine 62b of the second supercharger 62 is located downstream of the turbine 61b of the first supercharger 61 in the exhaust passage 42. And a second supercharger 62 disposed on the upstream side of the compressor 61a of the first supercharger 61 in the intake passage 32, the compressor 62a of the second supercharger 62 being provided.
The passage portion 63 that bypasses the compressor 61 a of the first supercharger 61 is the intake passage at a branch portion between the compressor 61 a of the first supercharger 61 and the compressor 62 a of the second supercharger 62. 32, and is configured to merge with the intake passage 32 at a merge portion downstream of the compressor 61a of the first supercharger 61,
An intake throttle valve (throttle valve 33) that is disposed downstream of the merging portion of the intake passage 32 and whose opening degree is changed is provided.

The first device applied to the engine 10 is
The operating parameters included in the time-series data acquired by the time-series data acquisition means (step 630 in FIG. 6) are the supercharging pressure ω of the engine, the opening e of the intake throttle valve, and the first supercharging. The rotational speed ν of the compressor 61a of the machine 61 and the rotational speed λ of the compressor 62a of the second supercharger 62,
The value of the operating parameter at the second time point t estimated by the operating state estimating means (steps 720 to 740 in FIG. 7) is configured to be the supercharging pressure ωest (t) of the engine. .

  Thereby, the first device models the dynamic characteristic of the operating parameter (supercharging pressure) of the engine 10 as a combination of a plurality of ARX models (a piecewise affine autoregressive model) according to the opening degree of the intake switching valve 64. To do. Therefore, the first device can estimate the value of the operation parameter (supercharging pressure) at a predetermined time (second time t) with higher accuracy. As a result, the first device can accurately determine whether or not the intake air switching valve 64 is operating normally.

  Further, the first device models the dynamic characteristic of the operation parameter (supercharging pressure) based on the value of the operation parameter acquired when the engine 10 is actually operated. That is, the first device models the dynamic characteristics of the operating parameters for each individual engine. For this reason, the first device reduces the influence of manufacturing variations (individual differences) in the members constituting the engine 10 as compared with the case where an existing dynamic characteristic map acquired in advance through experiments or the like is applied to the abnormality determination device. Can do. Therefore, the first device can determine whether or not the control valve is operating normally with higher accuracy.

(Second Embodiment)
Next, a control valve abnormality determination device (hereinafter also referred to as “second device”) according to a second embodiment of the present invention will be described.

<Overview of device and method for determining turbo mode>
The second device is applied to an internal combustion engine (see FIG. 1) similar to the internal combustion engine 10 to which the first device is applied. Further, the second device determines the turbo mode by the same method as the first device. Therefore, detailed description thereof will be omitted.

<Outline of device operation>
In addition to the abnormality determination of the intake air switching valve 64 in the first device (comparison between the estimated supercharging pressure and the actual supercharging pressure), the second device opens the intake air switching valve 64 at a predetermined time. It differs from the first device in that the abnormality determination based on the degree is performed.

  That is, according to the piecewise affine analysis method, the second device uses the time series data of the operation parameters of the engine 10 as a plurality of ARX models corresponding to the opening degree of the intake switching valve 64 using the “supercharging pressure dynamic characteristics”. In addition to modeling, a “function representing the separation hyperplane” is estimated. Next, the second device applies a regression vector at a predetermined first time point to the dynamic characteristic of the modeled supercharging pressure, thereby obtaining a “supercharging pressure value at a second time point after the first time point. Is estimated. Further, the second device acquires the actual boost pressure value at the second time point based on the output value of the boost pressure sensor 74.

  Furthermore, the second device estimates the “open / close state of the intake air switching valve 64” at the first time point by applying the regression vector at the first time point to the function representing the separation hyperplane. In addition, the second device acquires the actual opening / closing state of the intake air switching valve 64 at the first time point based on the output value of the supercharging pressure sensor 74.

  Then, the second device determines the abnormality of the intake air switching valve 64 by comparing the “estimated supercharging pressure value” with the “actual supercharging pressure value”. Further, the second device determines whether or not the intake switching valve 64 is abnormal by comparing “the estimated opening / closing state of the intake switching valve 64” with “the actual opening / closing state of the intake switching valve 64”.

  When it is determined that the intake device switching valve 64 is normal in both the “abnormality determination based on the value of the boost pressure” and the “abnormality determination based on the open / closed state of the intake air switching valve 64”, A final determination is made that the valve 64 is normal. On the other hand, the second device determines that the intake air switching valve 64 is abnormal in at least one of “abnormality determination based on the value of the boost pressure” and “abnormality determination based on the open / close state of the intake air switching valve 64”. At this time, a final determination is made that the intake air switching valve 64 is abnormal.

  Further, when it is determined that the intake air switching valve 64 is “abnormal”, the second device notifies the operator of the engine 10 to that effect and performs “evacuation operation” with a small burden on the members constituting the engine 10. ”Is executed. In contrast, when it is determined that the exhaust gas switching valve 66 and the exhaust bypass valve 68 are “normal”, the second device performs “normal operation” without notifying the operator. The above is the outline of the operation of the second device.

<Control valve abnormality judgment method>
Hereinafter, a control valve abnormality determination method in the second device will be described.
First, the second device determines whether or not the intake air switching valve 64 is operating normally by the same method as the first device (see Procedure 1 to Procedure 4 described above). A detailed description of this method is omitted. Hereinafter, this “determination of abnormality of the intake switching valve 64 by the same method as that of the first device” will be referred to as “first abnormality determination” for convenience.

  Next, the second device determines whether or not the intake air switching valve 64 is operating normally according to the following procedures 5 to 7.

(Procedure 5) Using the function representing the separation hyperplane, the open / close state of the intake air switching valve 64 at a predetermined time point is estimated.
(Procedure 6) By comparing the opening / closing state of the intake switching valve 64 at the predetermined time point estimated in step 5 with the actual opening / closing state of the intake switching valve 64 at that time point, the intake switching valve 64 is normal. It is determined whether or not it is operating.
(Procedure 7) Based on the result of the abnormality determination of the intake air switching valve 64 in step 4 and the result of the abnormality determination of the intake air switching valve 64 in step 6, it is determined whether or not the intake air switching valve 64 is operating normally. To do.
Each of the procedures 5 to 7 will be described in detail.

(Procedure 5) Using the function representing the separation hyperplane, the open / close state of the intake air switching valve 64 at a predetermined time point is estimated.
First, the second apparatus performs regression at a predetermined time t after performing “modeling of the dynamic characteristic of the supercharging pressure” and “estimating the function (first function) representing the separation hyperplane H” according to the above procedure 2. The vector x (t) is applied to the first function (separation hyperplane H) shown in the above equation (10). As in the above procedure 2, when the regression vector x (t) is applied to the first function, the second device determines that the regression vector x (t) is a region if the value of the first function is greater than or equal to zero. It is determined to belong to the S _1. In contrast, the second device, when regression vector x to the first function (t) is applied, the value of the first function is less than zero, the regression vector x (t) belongs to the area S ₂ Is determined. As shown in the above equation (1), the regression vector x (t) is a vector including time-series data at “time t−1 before (past) time t” as an element.

As described above, the region S ₁ is a region to which the regression vector belongs when the “opening degree of the intake switching valve 64 is an opening other than the fully closed opening degree”, and the region S ₂ is “opening of the intake switching valve 64. This is the region to which the regression vector belongs when the degree is the fully closed opening degree. Therefore, when the regression vector x (t) is applied to the first function and the value of the first function is greater than or equal to zero, the second device indicates that “the opening degree of the intake switching valve 64 at time t−1 is all It is estimated that the opening is other than the opening degree. On the other hand, when the regression vector x (t) is applied to the first function and the value of the first function is smaller than zero, the second device indicates that “the opening degree of the intake switching valve 64 at time t−1”. Is the fully closed opening. "

(Procedure 6) By comparing the estimated opening / closing state of the intake air switching valve 64 at a predetermined time point with the actual opening / closing state of the intake air switching valve 64 at that time point, the intake air switching valve 64 operates normally. It is determined whether or not.
First, the second device acquires the actual opening Oact (t−1) of the intake switching valve 64 at time t−1 based on the output value of the intake switching valve opening sensor 78. Next, the second device compares the actual opening Oact (t-1) at time t-1 with the open / close state of the intake air switching valve 64 at time t-1 estimated in the procedure 5. .

  More specifically, when it is estimated that “the opening degree of the intake air switching valve 64 at the time t−1 is a fully closed opening degree”, the second device determines the actual opening degree Oact (t−1). Is fully closed, it is determined that “the intake switching valve 64 is normal”. On the other hand, at this time, if the actual opening Oact (t−1) is an opening other than the fully closed opening, it is determined that “the intake switching valve 64 is abnormal”.

  On the other hand, when it is estimated that “the opening degree of the intake air switching valve 64 at time t−1 is an opening degree other than the fully closed opening degree”, the second device has an actual opening degree Oact (t−1). If it is an opening other than the fully closed opening, it is determined that “the intake switching valve 64 is normal”. On the other hand, at this time, if the actual opening degree Oact (t−1) is the fully closed opening degree, it is determined that “the intake switching valve 64 is abnormal”.

  Hereinafter, the above-described “abnormality determination of the intake switching valve 64 by the method shown in the procedure 5 and the procedure 6” is referred to as “secondary abnormality determination” for convenience.

(Procedure 7) Based on the result of the abnormality determination of the intake air switching valve 64 in step 4 and the result of the abnormality determination of the intake air switching valve 64 in step 6, it is determined whether or not the intake air switching valve 64 is operating normally. To do.

  The second device determines whether the intake switching valve 64 is abnormal (first abnormality determination) in the procedure 4 and whether the intake switching valve 64 is abnormal (second abnormality determination) in the procedure 6. Is determined as “normal”, a final determination is made that “the intake switching valve 64 is normal”. On the other hand, when the second device determines that “the intake switching valve 64 is abnormal” in at least one of the first abnormality determination and the second abnormality determination, “the intake switching valve 64 is abnormal”. And make the final decision. The above is the control valve abnormality determination method in the second device.

<Actual operation>
Hereinafter, the actual operation of the second device will be described.
The second device is different from the first device only in that the process shown in the flowchart of FIG. 10 is executed instead of the process shown in the flowchart of FIG. 7 described above. Therefore, the description will be continued below centering on this difference.

  The CPU 81 is configured to execute the routines shown in the flowcharts of FIGS. 6 and 8 to 10 at predetermined timings. In these routines, the CPU 81 uses the intake switching valve abnormality flag XACV similar to that of the first device.

  Hereinafter, each routine executed by the CPU 81 will be described in detail on the assumption that modeling of the dynamic characteristic of the supercharging pressure is completed at the present time. When the CPU 81 starts processing from step 600 in FIG. 6 at a predetermined timing, the CPU 81 proceeds to step 610. If the above assumption is followed, since the modeling of the supercharging pressure has been completed at the present time, the CPU 81 makes a “Yes” determination at step 610 to proceed to step 695 to end the present routine tentatively.

  Further, the CPU 81 repeatedly executes the “second abnormality determination routine” shown by the flowchart in FIG. 10 every time a predetermined time elapses. By this routine, the CPU 81 determines whether or not the intake air switching valve 64 is abnormal. The routine shown in FIG. 10 is different from the routine shown in FIG. 7 only in that steps 1010 to 1030 are added. Therefore, in FIG. 10, steps for performing the same processing as the steps shown in FIG. 7 are given the same reference numerals as those given to such steps in FIG. 7. Detailed description of these steps will be omitted as appropriate.

Specifically, when the CPU 81 starts processing from step 1000 in FIG. 10 at a predetermined timing, the CPU 81 proceeds to step 710. If the above assumption is followed, since the modeling of the supercharging pressure is completed at the present time, the CPU 81 determines “Yes” in step 710 and proceeds to step 720. In step 720, the CPU 81 obtains the regression vector x (t) at the current time (time t) and applies the regression vector x (t) to the first function (separation hyperplane). It is determined whether or not the value of one function (a ^T x (t) + b) is greater than or equal to zero.

  At this time, if the value of the first function is greater than or equal to zero, the CPU 81 makes a “Yes” determination at step 720 to proceed to step 730, and sets the regression vector x (t) at time t to “the intake switching valve 64. This is applied to the second function (ARX model) corresponding to the case where the opening is an opening other than the fully-closed opening (see the upper stage of equation (14) above). Thereby, the CPU 81 estimates the supercharging pressure at time t, and acquires the estimated supercharging pressure as ωest (t). As described above, in step 730, the error ε (t) is set to zero. Thereafter, the CPU 81 proceeds to step 1010.

  In step 1010, the CPU 81 stores “1” in an estimated opening index value OIest (t−1) that is an index value indicating the estimated opening (estimated opening) of the intake air switching valve 64. The estimated opening index value OIest (k) indicates that when the value is “1”, the estimated opening of the intake air switching valve 64 at time k is “an opening other than the fully closed opening”. On the other hand, the estimated opening index value OIest (k) indicates that when the value is “0”, the estimated opening of the intake air switching valve 64 at the time k is the “fully closed opening”. Thereafter, the CPU 81 proceeds to step 750.

  On the other hand, if the value of the first function is smaller than zero at this time, the CPU 81 makes a “No” determination at step 720 to proceed to step 740, and sets the regression vector x (t) at time t to “intake”. This is applied to the second function (ARX model) corresponding to the case where the opening degree of the switching valve 64 is a fully closed opening degree (see the lower part of the above equation (14)). Thereby, the CPU 81 estimates the supercharging pressure at time t, and acquires the estimated supercharging pressure as ωest (t). As described above, also in step 740, the error ε (t) is set to zero. Thereafter, the CPU 81 proceeds to step 1020.

  In step 1020, the CPU 81 stores “0” in the estimated opening index value OIest (t−1). Thereafter, the CPU 81 proceeds to step 750.

  Thus, the second function (ARX model) to which the regression vector x (k) is applied according to the value of the first function when the regression vector x (t) is applied to the first function (separation hyperplane). ) Is determined. Then, by applying the regression vector x (k) to the determined second function, the supercharging pressure at time t is estimated. Further, by applying the regression vector x (t) to the first function, the opening degree of the intake air switching valve 64 at the time t−1 is estimated. Then, an estimated opening index value OIest (t−1) is determined according to the estimated opening of the intake air switching valve 64. Thereafter, the CPU 81 proceeds to step 750.

  Next, the CPU 81 performs “primary abnormality determination” for determining abnormality of the intake switching valve 64 using the estimated supercharging pressure, and performs abnormality determination of the intake switching valve 64 using the estimated opening index value. Perform “secondary abnormality determination”. Hereinafter, the cases will be described separately.

(Assumption 1) When the absolute value of the difference between the estimated supercharging pressure and the actual supercharging pressure is equal to or greater than a predetermined threshold value, the CPU 81 determines the primary abnormality of the intake air switching valve 64 (step 750) (See step 4). More specifically, the CPU 81 determines in step 750 the actual boost pressure ωact () determined by the boost pressure ωest (t) estimated as described above and the output value of the boost pressure sensor 74 at time t. It is determined whether or not the absolute value of the difference between t) is smaller than a predetermined threshold value δ2.

  According to Assumption 1, the CPU 81 makes a “No” determination at step 750 to proceed to step 770. In step 770, the CPU 81 stores “1” as the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds to step 1095 to end the present routine tentatively.

  At this time, when the CPU 81 starts processing from step 800 in FIG. 8, the CPU 81 proceeds to step 810. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 810. Thereafter, the CPU 81 proceeds to step 895 via step 820 and once ends this routine.

  Further, at this time, when the CPU 81 starts the process from step 900 in FIG. 9, the process proceeds to step 910. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 910. Thereafter, the CPU 81 proceeds to step 995 via step 940 and step 930, and once ends this routine.

  Therefore, in this case (that is, when it is determined that the intake switching valve 64 is abnormal in the first abnormality determination), the operator is notified that “the intake switching valve 64 is abnormal”. , “Evacuation operation” is executed.

(Assumption 2) The absolute value of the difference between the estimated supercharging pressure and the actual supercharging pressure is smaller than a predetermined threshold, but the estimated opening index value does not match the actual opening index value. If it follows, CPU81 will determine with "Yes" in step 750, and will progress to step 1030.

  In step 1030, the CPU 81 performs a second abnormality determination of the intake air switching valve 64 (see the above procedure 6). Specifically, in step 1030, the CPU 81 obtains the actual opening Oact (t-1) of the intake air switching valve 64 based on the output value of the intake air switching valve opening sensor 78 at time t-1. To do. Further, when the actual opening Oact (t−1) is “an opening other than the fully closed opening”, the CPU 81 provides an actual opening index that is an index value indicating the actual opening of the intake air switching valve 64. “1” is stored in the value OIact (t−1), and the actual opening index value OIact (t−1) is “0” if the actual opening Oact (t−1) is “fully closed opening”. Is stored. Then, the CPU 81 determines whether or not the estimated opening index value OIest (t−1) acquired in the above-described step 1010 or step 1020 matches the actual opening index value OIact (t−1). Determine.

  According to Assumption 2, the estimated opening index value OIest (t-1) and the actual opening index value OIact (t-1) do not match. Therefore, the CPU 81 makes a “No” determination at step 1030 to proceed to step 770. In step 770, the CPU 81 stores “1” as the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds to step 1095 to end the present routine tentatively.

  At this time, when the CPU 81 starts processing from step 800 in FIG. 8, the CPU 81 proceeds to step 810. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 810. Thereafter, the CPU 81 proceeds to step 895 via step 820 and once ends this routine.

  Further, at this time, when the CPU 81 starts the process from step 900 in FIG. 9, the process proceeds to step 910. Since the value of the intake switching valve abnormality flag XACV at the present time is “1”, the CPU 81 determines “No” in step 910. Thereafter, the CPU 81 proceeds to step 995 via step 940 and step 930, and once ends this routine.

  Accordingly, even in this case (when it is determined that the intake switching valve 64 is normal in the first abnormality determination, but the intake switching valve 64 is determined abnormal in the second abnormality determination), the operator Is notified that “the intake switching valve 64 is abnormal” and “evacuation operation” is executed.

(Assumption 3) When the absolute value of the difference between the estimated supercharging pressure and the actual supercharging pressure is smaller than a predetermined threshold, and the estimated opening index value matches the actual opening index value 3, the CPU 81 makes a “Yes” determination at step 750 to proceed to step 1030. Further, according to Assumption 3, the CPU 81 makes a “Yes” determination at step 1030 to proceed to step 760. In step 760, the CPU 81 stores “0” in the value of the intake switching valve abnormality flag XACV. Thereafter, the CPU 81 proceeds to step 1095 to end the present routine tentatively.

  At this time, when the CPU 81 starts processing from step 800 in FIG. 8, the CPU 81 proceeds to step 810. Since the value of the intake switching valve abnormality flag XACV at this time is “0”, the CPU 81 determines “Yes” in step 810, proceeds to step 895, and ends this routine once.

  Further, at this time, when the CPU 81 starts the process from step 900 in FIG. 9, the process proceeds to step 910. Since the current value of the intake air switching valve abnormality flag XACV is “0”, the CPU 81 determines “Yes” in step 910. Thereafter, the CPU 81 proceeds to step 995 via step 920 and step 930, and once ends this routine.

  Therefore, in this case (when it is determined that the intake switching valve 64 is normal in both the first abnormality determination and the second abnormality determination), the operator is informed that “the intake switching valve 64 is abnormal”. Is not notified, and “normal operation” is executed.

  Thus, the second device performs normal operation when it is determined that the intake switching valve 64 is normal in both the primary abnormality determination and the secondary abnormality determination. On the other hand, when it is determined that the intake switching valve 64 is abnormal in at least one of the primary abnormality determination and the secondary abnormality determination, the second device notifies the operator of that fact. At the same time, the evacuation operation is executed.

As described above, in the second device,
The operating state estimating means includes
The data vector group into which the regression vector x (t) determined by the regression vector x (t) and the first function (separated hyperplane H. See equation (10) above) is classified in the regression space. The plurality of opening regions corresponding to (Cluster C ₁ and Cluster C ₂ ) (regions where the opening degree of the intake air switching valve 64 is an opening degree other than the fully closed opening degree) and the opening amounts of the intake air switching valve 64 are all One of the regions that are the opening degree) is a control valve estimated affiliation region (estimated opening index value OIest (t) that is the opening region to which the opening degree of the control valve 64 belongs at the first time point t-1. -1)) (steps 720 to 740 in FIG. 10 and steps 1010 and 1020),
The abnormality determining means includes
The estimated control valve affiliation region OIest (t-1) and one of the plurality of opening regions to which the actual opening of the control valve 64 at the first time point t-1 belongs. It is configured to determine whether or not the control valve 64 is abnormal by comparing the control valve actual affiliation region (actual opening index value OIact (t−1)) (FIG. 10). Step 1030).

  Thereby, the second device can perform both the abnormality determination of the control valve based on the value of the supercharging pressure at the second time point and the abnormality determination of the control valve based on the opening degree of the control valve at the first time point. it can. Therefore, the second device can determine whether or not the control valve is operating normally more accurately than when only one of these abnormality determinations is performed.

  The present invention is not limited to the above embodiments, and various modifications can be adopted within the scope of the present invention.

  For example, in each of the above embodiments, the operating parameters that affect the supercharging state by the supercharger include the low pressure stage turbocharger rotation speed, the high pressure stage turbocharger rotation speed, the throttle valve opening, and the supercharging pressure. Is adopted. However, the operation parameters that can be employed in the control valve abnormality determination device of the present invention are not limited to these. For example, the control valve abnormality determining device according to the present invention includes, as operating parameters, air pressure at a predetermined location in the intake passage, air temperature at a predetermined location in the intake passage, and a compressor of the first supercharger. At least one of the rotational speeds of the two.

  Further, in each of the above embodiments, the dynamic characteristic of “supercharging pressure” is modeled as an ARX model. However, the control valve abnormality determination device of the present invention may be configured to model the dynamic characteristic of “an operation parameter other than the supercharging pressure” as an ARX model.

  In addition, in each of the above-described embodiments, once the modeling of the boost pressure dynamic characteristics is completed, the acquisition of time series data is stopped. However, the control valve abnormality determination device of the present invention continues to acquire time-series data even after the modeling of the boost pressure dynamic characteristics is completed, and every time a predetermined time elapses based on the acquired time-series data. And “re-estimation of separation hyperplane” and “remodeling of dynamic characteristics of supercharging pressure”. That is, the control valve abnormality determination device of the present invention is configured to update the separation hyperplane (first function) and the function (second function) representing the dynamic characteristic of the supercharging pressure every time a predetermined time elapses. Also good. Furthermore, the control valve abnormality determination device of the present invention may be configured to update at least one of the functions representing the separation hyperplane and the dynamic characteristic of the supercharging pressure.

  Further, in each of the above embodiments, only abnormality determination of the intake air switching valve 64 is performed. However, the control valve abnormality determination device of the present invention may be configured to perform abnormality determination of not only the intake switching valve 64 but also one or both of the exhaust switching valve 66 and the exhaust bypass valve 68. More specifically, for example, the control valve abnormality determination device of the present invention models the dynamic characteristics of a predetermined operating parameter as a plurality of ARX models corresponding to the opening degree of the exhaust gas switching valve 66, and the modeling is performed. Further, it is possible to determine whether the exhaust gas switching valve 66 is abnormal by comparing the estimated value of the operating parameter obtained from the dynamic characteristic of the operating parameter with the actual value of the operating parameter. Furthermore, for example, the control valve abnormality determination device of the present invention can be configured to determine abnormality of the exhaust bypass valve 68 as described above.

  Further, in each of the above embodiments, the control valve abnormality determination device of the present invention includes two superchargers (first supercharger 61 and second supercharger 62) arranged in series and three control valves ( The present invention is applied to an internal combustion engine provided with an intake switching valve 64, an exhaust switching valve 66, and an exhaust bypass valve 68). However, the internal combustion engine to which the control valve abnormality determination device of the present invention is applied is not limited to the internal combustion engine. For example, the control valve abnormality determination device of the present invention can be applied to an internal combustion engine including one supercharger and a predetermined number of control valves. Furthermore, for example, the control valve abnormality determination device of the present invention can be applied to an internal combustion engine including three or more superchargers and a predetermined number of control valves. In addition, for example, the control valve of the present invention can be applied to an internal combustion engine in which a plurality of superchargers are arranged in “parallel”.

  Further, in the second embodiment, abnormality determination based on the value of the supercharging pressure at the second time point (first abnormality determination) and abnormality determination based on the open / closed state of the intake air switching valve 64 at the first time point (second operation). Next abnormality determination) is executed. However, the control valve abnormality determination device of the present invention may be configured to perform only the second abnormality determination.

DESCRIPTION OF SYMBOLS 20 ... Engine main body, 22 ... Fuel injection apparatus, 32 ... Intake pipe, 33 ... Throttle valve, 42 ... Exhaust pipe, 61 ... High pressure stage supercharger, 61a ... High pressure stage compressor, 62 ... Low pressure stage supercharger, 62a ... Low pressure stage compressor, 63 ... High pressure stage compressor bypass passage, 64 ... Intake switching valve, 64a ... Intake switching valve actuator, 66 ... Exhaust switching valve, 68 ... Exhaust bypass valve, 73 ... Throttle valve opening sensor, 74 ... Supercharging Pressure sensor, 76 ... high pressure stage turbocharger rotational speed sensor, 77 ... low pressure stage turbocharger rotational speed sensor, 78 ... intake switching valve opening sensor, 80 ... electric control device

Claims (6)

A first turbocharger having a turbine disposed in an exhaust passage of an internal combustion engine and a compressor disposed in an intake passage of the engine, and air that is provided in a passage portion that bypasses the compressor and flows into the compressor A control valve abnormality determining device for an internal combustion engine comprising: a control valve that changes a ratio between the amount of the air and the amount of air passing through the passage portion,
Time series data obtaining means for obtaining a value of a predetermined operation parameter related to a supercharging state by the first supercharger as time series data corresponding to the passage of time;
According to a piecewise affine analysis method, data vectors including the acquired time-series data as elements are classified into a plurality of data vector groups, and the plurality of classified data vectors are classified using the plurality of classified data vector groups. Estimating a first function consisting of a single function or a plurality of functions representing boundaries for dividing a group in a regression space, and using time series data included as elements in each of the plurality of classified data vector groups, Data analysis means for estimating a second function representing the dynamic characteristics of the operation parameter as an ARX model for each of the plurality of classified data vector groups;
A regression vector including time series data acquired by the time series data acquisition unit at a first time point after the time point when the first function and the second function are estimated by the data analysis unit is the first vector. By applying to one function, it is determined which data vector group of the plurality of data vector groups is classified in the regression space, and the same one of the second functions is determined. An operation for estimating the value of the operation parameter at a second time point after the first time point by applying the regression vector to a second function corresponding to a data vector group in which the regression vector is classified in the regression space. State estimation means;
An abnormality for determining whether or not the control valve is abnormal by comparing the estimated value of the operating parameter at the second time point with the actual value of the operating parameter at the second time point A determination means;
A control valve abnormality determination device for an internal combustion engine comprising: The control valve abnormality determination device according to claim 1,
Each of the plurality of classified data vector groups is a control valve abnormality determination device configured to correspond to each of a plurality of opening regions that divide a movable range of the opening of the control valve. In the control valve abnormality determination device according to claim 2,
The operating state estimating means includes
One of the plurality of opening regions corresponding to the data vector group into which the regression vector determined by the regression vector and the first function is classified in a regression space is controlled at the first time point. It is configured to estimate as a control valve estimation affiliation region that is an opening region to which the valve opening belongs,
The abnormality determining means includes
The estimated control valve estimated affiliation region is compared with the control valve actual affiliation region which is one of the plurality of opening regions to which the actual opening of the control valve at the first time point belongs. Accordingly, a control valve abnormality determination device configured to determine whether or not the control valve is abnormal. In the control valve abnormality determination device according to any one of claims 1 to 3,
The plurality of opening regions are:
A control valve abnormality determination device comprising two regions: a region in which the opening of the control valve is a fully closed opening, and a region in which the opening of the control valve is an opening other than the fully closed opening. In the control valve abnormality determination device according to any one of claims 1 to 4,
The operating parameters are:
A control valve abnormality determination device including at least one of an air pressure at a predetermined location in the intake passage, an air temperature at a predetermined location in the intake passage, and a rotation speed of the compressor. A control valve abnormality determination device according to any one of claims 1 to 5,
The agency
A second supercharger different from the first supercharger, wherein the turbine of the second supercharger is disposed downstream of the turbine of the first supercharger in the exhaust passage and the second supercharger. A compressor of the supercharger comprises a second supercharger disposed upstream of the compressor of the first supercharger in the intake passage;
The passage portion is branched from the intake passage at a branch portion between the compressor of the first supercharger and the compressor of the second supercharger, and a merging portion downstream of the compressor of the first supercharger It is configured to join the intake passage at
An intake throttle valve that is disposed downstream of the merging portion of the intake passage and whose opening is changed;
The operating parameters included in the time-series data acquired by the time-series data acquisition means include the supercharging pressure of the engine, the opening of the intake throttle valve, the rotational speed of the compressor of the first supercharger, and Including the rotational speed of the compressor of the second supercharger,
The control valve abnormality determining device, wherein the value of the operating parameter at the second time point estimated by the operating state estimating means is a boost pressure of the engine.

Images