State identification apparatus

Abstract

A state identification apparatus (1) outputs an identification signal for identifying whether an object being identified is in a single prescribed state. A processor (2) of the state identification apparatus executes a program (P) stored in a storage device (3), thereby functioning as a plurality of determination devices (10) that each perform determination. The plurality of determination devices include at least one identification determination device (11) for determining whether the object being identified is in the prescribed state. The plurality of determination devices each acquire determination target data (D) relating to at least one type of physical quantity, calculate the Mahalanobis distance (MD) without depending on each other on the basis of the acquired determination target data and mutually different unit spaces (U) that are set in advance through the Mahalanobis-Taguchi method (MT method), and perform respective determinations on the basis of the calculated MD. At least some of the types of physical quantities in two pieces of determination target data acquired by any two of the plurality of determination devices are different from each other.

Description

Condition Identification Device

[0001] The present invention relates to a state identification device that outputs a signal that identifies whether an object to be identified is in a single predetermined state.

[0002] Conventionally, in various technical fields, various condition identification devices are used that output a signal that identifies whether an object to be identified is in a predetermined state. For example, various condition identification devices are also used in the technical field of engine-mounted devices. For example, Patent Document 1 discloses a condition identification device that outputs a signal that identifies whether a catalyst that purifies exhaust gas emitted from an engine is in a deteriorated state. Furthermore, for example, Patent Document 2 discloses a condition identification device that outputs a signal that identifies whether an engine is in a misfire state. The condition identification device of Patent Document 2 identifies whether an engine is in a misfire state by performing a detailed analysis of a signal from an engine rotation speed sensor.

[0003] Furthermore, various state identification devices have been proposed that use the Mahalanobis-Taguchi method (MT method) to output a signal that identifies whether an object to be identified is in a predetermined state. For example, Patent Document 3 discloses a state identification device that uses the MT method to output a signal that identifies whether knocking is occurring. Also, Patent Document 4 discloses a state identification device that uses the MT method to output a signal that identifies whether an engine is in a misfire state.

[0004] International Publication No. 2021 / 182584 International Publication No. 2016 / 194953 Japanese Patent Application Laid-Open No. 2020-056331 Japanese Patent Application Laid-Open No. 2010-014065

[0005] In state identification devices that do not use the MT method, such as those disclosed in Patent Documents 1 and 2, the load of calculation processing is high in order to ensure high identification accuracy. Therefore, high-performance hardware resources, such as a processor with high processing power and a large-capacity memory, are required. As a result, the degree of freedom in designing the hardware resources of the state identification device is low. In contrast, state identification devices that use the MT method can reduce the load of calculation processing while improving identification accuracy compared to state identification devices that do not use the MT method. By reducing the load of calculation processing, the degree of freedom in designing the hardware resources of the state identification device can be improved. However, there is a demand for state identification devices that can improve the degree of freedom in designing the hardware resources while further improving identification accuracy.

[0006] An object of the present invention is to propose a state identification device that can improve the degree of freedom in designing hardware resources while further increasing the identification accuracy.

[0007] (1) A state identification device according to one embodiment of the present invention has the following configuration: a state identification device having a processor and a storage device, and outputting an identification signal for identifying whether an object to be identified is in a single predetermined state, wherein the processor functions as a plurality of decision devices that each make a decision by executing a program stored in the storage device, and when a decision device that decides whether the object to be identified is in the predetermined state is defined as an identification decision device, and a decision device that decides whether a predetermined condition other than a decision whether the object to be identified is in the predetermined state is established is defined as a condition decision device, the plurality of decision devices are (i) composed of at least one of the condition decision device and one of the identification decision device, or (ii) composed of at least one of the condition decision device and one of the identification decision device, or (iii) the system is composed of a plurality of the discrimination decision devices without including a condition decision device, or (iii) the system is composed of at least one of the condition decision device and a plurality of the discrimination decision devices, wherein the plurality of decision devices each acquires data to be judged relating to at least one type of physical quantity, calculates Mahalanobis distances without mutual dependence on each other based on different unit spaces previously set by the MT method and the acquired data to be judged, and makes a judgment based on the calculated Mahalanobis distances, and at least some of the types of physical quantities in the two data to be judged acquired by any two of the plurality of decision devices are different from each other.

[0008] According to this configuration, the processor of the state identification device functions as multiple decision devices, and the multiple decision devices include at least one decision device that determines whether the object to be identified is in a predetermined state. When the multiple decision devices include only one decision device, the multiple decision devices include at least one condition decision device that determines whether a predetermined condition other than whether the object to be identified is in a predetermined state is satisfied. When the multiple decision devices include multiple decision devices, the multiple decision devices either do not include a condition decision device or include at least one condition decision device. The multiple decision devices each acquire decision object data related to at least one type of physical quantity, calculate Mahalanobis distances based on different unit spaces previously set using the MT method and the acquired decision object data without interdependence with each other, and make a decision based on the calculated Mahalanobis distances. In this way, the state identification device uses multiple decision devices that independently perform decision using the MT method to determine whether the object to be identified is in a single predetermined state. Therefore, the identification accuracy can be improved compared to a state identification device having only one decision device that performs decision using the MT method. Furthermore, compared to a condition identification device having only one determiner that makes a judgment using the MT method, it is possible to simplify the calculation processing performed by the single determiner while improving the identification accuracy. Therefore, the calculation processing load of the entire condition identification device can be reduced. Therefore, it is possible to further improve the identification accuracy while further increasing the design freedom of the hardware resources of the condition identification device. Moreover, the two determiners included in the multiple determiners differ from each other in at least some of the types of physical quantities in the judgment target data they acquire. Therefore, by using the MT method, it is possible to reduce the calculation processing load, while using more types of data to further improve the identification accuracy. As a result, it is possible to further improve the design freedom of the hardware resources while further increasing the identification accuracy.

[0009] (2) In addition to the configuration of (1) above, the state identification device of one embodiment of the present invention may have the following configuration: Any two of the plurality of determiners have the same types of physical quantities in the determination target data that they acquire.

[0010] According to this configuration, any two of the plurality of decision devices calculate the Mahalanobis distance using decision target data that includes the same data, which may reduce the load of the calculation processing performed by the two decision devices.

[0011] (3) A state identification device according to one embodiment of the present invention may have the following configuration in addition to at least one of the configurations (1) and (2) described above: (ii) the plurality of decision devices are composed of the plurality of discrimination decision devices without including the condition decision device, or (iii) the plurality of discrimination decision devices are composed of the at least one condition decision device and the plurality of discrimination decision devices, and when a discrimination decision time is defined as a time required for the discrimination decision device to complete a decision on whether or not the data to be determined is in the predetermined state, the discrimination decision time of the discrimination decision device having the longest discrimination decision time among the plurality of discrimination decision devices is defined as a maximum discrimination decision time, and the discrimination decision time of the discrimination decision device having the shortest discrimination decision time among the plurality of discrimination decision devices is defined as a minimum discrimination decision time, the difference between the maximum discrimination decision time and the minimum discrimination decision time is shorter than half of the maximum discrimination decision time.

[0012] With this configuration, the difference in the time required for the determination by the plurality of determination judges is small, and therefore the variation in the time required for the state recognition device to recognize whether the object to be recognized is in a predetermined state can be reduced.

[0013] (4) A state identification device according to one embodiment of the present invention may have the following configuration in addition to at least one of the configurations (1) to (3) described above: the plurality of decision devices (i) are configured from the at least one condition decision device and the one discrimination decision device, or (iii) are configured from the at least one condition decision device and the plurality of discrimination decision devices, and when a first condition decision device that is any one of the at least one condition decision device determines that the predetermined condition is met, the first discrimination decision device that is any one of the one or more discrimination decision devices determines whether the object to be identified is in the predetermined state, and when the first condition decision device determines that the predetermined condition is not met, the first discrimination decision device does not determine whether the object to be identified is in the predetermined state.

[0014] This configuration can reduce the load on the calculation process of the state identification device by eliminating unnecessary decisions by the first discrimination / determiner, compared to when the first condition decision device makes a decision after the first discrimination / determiner makes a decision, thereby further increasing the degree of freedom in designing hardware resources.

[0015] (5) A condition identification device according to one embodiment of the present invention may have the following configuration in addition to at least one of the configurations described above in (1) to (4): the condition identification device outputs the identification signal for identifying whether an engine mounted on a vehicle traveling on a road surface is in a misfire state, the plurality of determiners are composed of one condition determiner and two determination determiners, the two determination determiners being a continuous misfire determiner that determines whether the engine is in a misfire state by determining whether or not a specific cylinder has continuously misfired, and an intermittent misfire determiner that determines whether or not the engine is in a misfire state by determining whether or not a single cylinder or multiple cylinders have intermittent misfires, and the one condition determiner determines whether or not the road surface is flat.

[0016] For example, engine rotational speed fluctuations are similar when the road surface is uneven and when intermittent misfires occur. Furthermore, engine rotational speed fluctuations also occur when continuous misfires occur. Therefore, with the above-described configuration, a single condition determiner determines whether the road surface is flat using the MT method, thereby reducing the computational load on the condition identification device and improving the accuracy of identifying whether a misfire occurs. Furthermore, because two discriminators determine whether intermittent misfires and continuous misfires based on different unit spaces, the computational load on the condition identification device can be reduced while improving the accuracy of identifying whether a misfire occurs, compared to when a single discriminator determines whether a misfire occurs, including both intermittent and continuous misfires, based on a single unit space. Therefore, the accuracy of identifying whether an engine is misfiring can be improved while increasing the design flexibility of the hardware resources of the condition identification device that identifies whether an engine is misfiring. Note that sporadic misfires in one cylinder refers to intermittent misfires in one cylinder in an engine having at least one cylinder. In addition, sporadic misfires occurring in multiple cylinders means that misfires occur irregularly in multiple cylinders, such as when the misfiring cylinder among the multiple cylinders changes over time.

[0017] (6) In addition to the configuration of (5) above, the condition identification device of one embodiment of the present invention may have the following configuration: When the one condition determiner determines that the road surface is flat, the intermittent misfire identification determiner determines whether the engine is in a misfire state, and when the one condition determiner determines that the road surface is not flat, the intermittent misfire identification determiner does not determine whether the engine is in a misfire state.

[0018] This configuration, compared to when the condition determiner makes a decision after the intermittent misfire detector makes a decision, eliminates unnecessary decisions by the intermittent misfire detector and reduces the processing load on the condition detector, thereby increasing the design flexibility of hardware resources.

[0019] In the present invention and embodiments, the object to be identified is not particularly limited. For example, the object to be identified may be a device or a part of a device. For example, the object to be identified may be a living thing (e.g., a human) or a part of a living thing. The object to be identified is not limited to an object. For example, the object to be identified may be a phenomenon.

[0020] In the present invention and the embodiments, the use of the state identification device is not particularly limited, and the state identification device may be composed of a plurality of devices that can communicate with each other.

[0021] In the present invention and embodiments, the processor includes any circuit, such as a microcontroller, a CPU (Central Processing Unit), a microprocessor, a multiprocessor, an application-specific integrated circuit (ASIC), a programmable logic circuit (PLC), or a field-programmable gate array (FPGA). The storage device stores a program that causes the processor to function as multiple decision devices. The storage device also stores data of a predetermined unit space. The storage device includes a non-transitory storage medium that stores the program and the data of the unit space. The storage device includes semiconductor memory such as a register or cache memory, a main memory (primary storage device / RAM), and storage (external storage device / auxiliary storage device).

[0022] In the present invention and embodiments, a state identification device that outputs an identification signal for identifying whether an object to be identified is in a predetermined state means a state identification device that identifies whether an object to be identified is in a predetermined state and outputs an identification signal indicating whether the object to be identified is in the predetermined state. In the present invention and embodiments, outputting an identification signal for identifying whether an object to be identified is in a predetermined state may mean outputting different signals when the object to be identified is identified as being in the predetermined state and when the object to be identified is not identified as being in the predetermined state. Outputting an identification signal for identifying whether an object to be identified is in a predetermined state may mean outputting a signal only when the object to be identified is identified as being in the predetermined state, and not outputting a signal when the object to be identified is identified as not being in the predetermined state.

[0023] In the present invention and embodiments, outputting an identification signal may mean outputting the identification signal to a device external to the state identification device, or may mean outputting the identification signal to a processor included in the state identification device that is the same as or different from the processors that function as multiple judgers.

[0024] In the present invention and embodiments, when the plurality of determiners includes a plurality of discrimination determiners, the discrimination determiners may determine whether or not the predetermined state exists by determining whether or not a plurality of different states are included in the predetermined state, or may directly determine whether or not the predetermined state exists. Even when the plurality of discrimination determiners perform the former determination, the state identification device ultimately outputs an identification signal that identifies whether or not the state is a single predetermined state.

[0025] In the present invention and the embodiments, the predetermined condition that is not a judgment of whether the object to be identified is in a predetermined state may be, for example, a condition of whether the environment is such that the identification / determining device can make a correct judgment, but is not limited to this. When the multiple determiners include multiple condition determiners, the predetermined conditions in the multiple condition determiners are different from each other.

[0026] In the present invention and embodiments, the judgment target data regarding at least one type of physical quantity is data obtained from at least one detector. In this specification, data obtained from a detector may be the data itself detected by the detector, or may be data generated from the data detected by the detector. The number of types of physical quantities in the judgment target data acquired by one judgment device may be the same as or greater than the number of detectors used to acquire the judgment target data. When different types of detectors are used to acquire two pieces of judgment target data, at least some of the types of physical quantities in the two pieces of judgment target data differ from each other. However, when at least some of the types of physical quantities in the two pieces of judgment target data differ from each other, the types of detectors used to acquire the two pieces of judgment target data do not necessarily differ from each other. The judgment target data may include, for example, time-series data. The judgment target data may include, for example, data of a signal output from a sensor (detector). The judgment target data may include, for example, data generated from a signal output from a sensor (detector). The judgment target data may include, for example, image data captured by an imaging device (detector). The image data is data regarding a physical quantity such as brightness. The judgment target data acquired by the classification decision device is data related to the object to be classified. In the present invention and the embodiments, the judgment target data acquired by the condition decision device is not limited to data related to the object to be classified. The judgment target data acquired by the condition decision device may be, for example, data related to the surrounding environment of the object to be classified.

[0027] In the present invention and its embodiments, the unit spaces set for each of the multiple decision-makers are different from one another. That is, one unit space data is set for each decision-maker. The unit space is set in advance using the MT method based on a reference data group. The Mahalanobis distance indicates the distance from the unit space. The closer the Mahalanobis distance is to 1, the closer the data to be decided is to the reference data group. In the field of quality engineering, where the Mahalanobis distance is used, the square of the Mahalanobis distance is sometimes referred to as the Mahalanobis distance. The reference data group that forms the basis of the unit space used by the discrimination decision-maker may be a data group in a predetermined state or a data group in an unpredictable state. The unit space is set based on at least one type of feature that indicates the characteristics of the reference data group. The decision-maker may extract at least one type of feature from the data to be decided and calculate the Mahalanobis distance based on the extracted at least one type of feature and the unit space. Alternatively, the data to be decided may be data of at least one type of feature. The decision-maker may use multiple types of feature to calculate one Mahalanobis distance.

[0028] In the present invention and the embodiments, the two pieces of judgment target data acquired by any two of the multiple judgment devices may be the same. In other words, the types of physical quantities in the two pieces of judgment target data may be completely the same. However, even in this case, the unit spaces used by the two judgment devices to calculate the Mahalanobis distance are different from each other. The difference between the two unit spaces may be, for example, a difference in the type of feature quantity.

[0029] In the present invention and embodiments, calculating Mahalanobis distances without mutual interdependence means that one decision device does not calculate a Mahalanobis distance based on data to be decided that includes a Mahalanobis distance calculated by another MT decision device.

[0030] In the present invention and embodiments, the processor may execute a program stored in a storage device to function as a plurality of determiners that make a determination using the MT method and at least one non-MT determiner that makes a determination without using the MT method. The at least one non-MT determiner that makes a determination without using the MT method may include a non-MT determiner that determines whether an object to be identified is in a predetermined state. The at least one non-MT determiner that makes a determination without using the MT method may include a non-MT determiner that determines whether a predetermined condition is met that is not a determination of whether an object to be identified is in a predetermined state.

[0031] In the present invention, if the number of a certain component is not clearly specified and the component is expressed in singular when translated into English, the present invention may have a plurality of the component, or the present invention may have only one of the component.

[0032] It should be noted that, in the present invention and embodiments, the words including, having, comprising, and their derivatives are used to encompass additional items in addition to the listed items and equivalents thereof.

[0033] Unless otherwise defined, all terms (including technical and scientific terms) used in this specification and the present invention have the same meaning as commonly understood by those skilled in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted to have a meaning consistent with the meaning in the context of the relevant technology and this disclosure, and should not be interpreted in an idealized or overly formal sense.

[0034] It should be noted that in this specification, the term "preferable" is non-exclusive. "Preferable" means "preferably, but not limited to." In this specification, a configuration described as "preferable" at least achieves the above-mentioned effects obtained by the present invention. In this specification, the term "may (may)" is non-exclusive. "may (may)" means "may (may) but is not limited to." In this specification, "may (may)" implicitly includes the possibility that "does not (is not)." In this specification, a configuration described as "may (may)" at least achieves the above-mentioned effects obtained by the present invention.

[0035] Before describing embodiments of the present invention in detail, it is to be understood that the invention is not limited to the details of construction and the arrangement of components set forth in the following description or illustrated in the drawings. The present invention is capable of embodiments other than those described below. The present invention is also capable of embodiments incorporating various variations of the embodiments described below.

[0036] The state identification device of the present invention aims to propose a state identification device that can further improve the degree of freedom in designing hardware resources while further increasing the identification accuracy.

[0037] FIGS. 1(a) to 1(j) are diagrams illustrating the configuration of a condition identification device according to a first embodiment of the present invention. FIG. 2 is a diagram illustrating the configuration of a condition identification device according to a second embodiment of the present invention. FIG. 3 is a map used by the intermittent misfire identification determinator of the condition identification device according to the second embodiment of the present invention to estimate the misfire rate. FIGS. 4(a) to 4(d) are graphs illustrating temporal changes in the throttle opening, vehicle speed, and engine speed of a vehicle to which the condition identification device according to the second embodiment of the present invention is applied. FIGS. 5(a) to 5(c) are graphs illustrating temporal changes in the throttle opening, vehicle speed, engine speed, and Mahalanobis distance calculated by the condition determinator of a vehicle to which the condition identification device according to the second embodiment of the present invention is applied. FIGS. 6(a) and 6(b) are graphs illustrating temporal changes in the misfire rate, intake air volume, engine speed, and Mahalanobis distance calculated by the intermittent misfire identification determinator of a vehicle to which the condition identification device according to the second embodiment of the present invention is applied. FIG. 7 is a graph showing an enlarged portion of FIG. 6(a). 8(a) and 8(b) are graphs showing temporal changes in the throttle opening, vehicle speed, engine speed, feedback correction coefficient, and Mahalanobis distance calculated by the consecutive misfire discrimination device of the vehicle to which the state discrimination device of the second embodiment is applied. Fig. 9 is a graph showing an enlarged portion of Fig. 8(a).

[0038] First Embodiment A condition identification device 1 according to a first embodiment of the present invention will be described with reference to FIGS. 1(a) to 1(j). FIGS. 1(a), 1(b), and 1(c) show several examples of the condition identification device 1 according to the first embodiment. However, these are merely examples and do not limit the first embodiment. The condition identification device 1 has a processor 2 and a storage device 3, and outputs an identification signal that identifies whether an object to be identified is in a single predetermined state. The processor 2 executes a program P stored in the storage device 3, thereby functioning as several determinators 10 that each make a determination using the MT method.

[0039] For example, as shown in FIG. 1A, the multiple decision devices 10 of the first embodiment may be composed of at least one condition decision device 12 and one discrimination decision device 11. The discrimination decision device 11 is a decision device 10 that decides whether or not the object to be identified is in a predetermined state. The condition decision device 12 is a decision device 10 that decides whether or not a predetermined condition, other than whether or not the object to be identified is in a predetermined state, is satisfied. As shown by the solid line in FIG. 1A, the number of the multiple decision devices 10 may be two. Alternatively, the number of the multiple decision devices 10 may be three, including the condition decision device 12 shown by the two-dot chain line in FIG. 1A. Alternatively, the number of the multiple decision devices 10 may be four or more.

[0040] For example, as shown in Fig. 1(b), the plurality of decision devices 10 of the first embodiment may be configured with a plurality of discrimination decision devices 11 without including a condition decision device 12. As shown by the solid line in Fig. 1(b), the number of the plurality of decision devices 10 may be two. Also, the plurality of decision devices 10 may be three decision devices 10 including the condition decision device 12 shown by the two-dot chain line in Fig. 1(b). Also, the number of the plurality of decision devices 10 may be four or more.

[0041] For example, as shown in Fig. 1(c), the plurality of decision devices 10 of the first embodiment may be configured with at least one condition decision device 12 and a plurality of discrimination decision devices 11. Fig. 1(c) shows an example in which the number of the plurality of decision devices 10 is three, but the number of the plurality of decision devices 10 may be four or more.

[0042] Next, the processing executed by each determinator 10 of the first embodiment will be described using the flowcharts included in FIGS. 1A to 1C. In step S1, each determinator 10 acquires determination target data D related to at least one type of physical quantity. In step S2, each determinator 10 calculates a Mahalanobis distance MD based on a unit space U preset by the MT method and the acquired determination target data D. Data on the unit space U is stored in the storage device 3. The unit spaces U used by the multiple determinators 10 to calculate the Mahalanobis distances in step S2 are different from one another. In the example of FIGS. 1A to 1C, the three determinators 10 each use different unit spaces U1, U2, and U3. In step S2, the multiple determinators 10 calculate the Mahalanobis distances independently of one another. In step S3, each determinator 10 makes a judgment based on the calculated Mahalanobis distance. In step S3, the identification / determinator 11 determines whether the object to be identified is in a predetermined state. In step S3, the condition decision device 12 decides whether or not a predetermined condition is satisfied. In the examples of Figures 1(a) to 1(c), the decision device 10 makes a decision based on one Mahalanobis distance in step S3, but the decision device 10 of the first embodiment may repeat the process of step S1 and the process of step S2, and after the number of calculated Mahalanobis distances reaches a predetermined number, make a decision based on a plurality of Mahalanobis distances in step S3.

[0043] In the first embodiment, at least some of the types of physical quantities in two pieces of judgment target data acquired by any two of the multiple judgment devices 10 are different from each other. Figures 1A to 1C show an example in which three pieces of judgment target data D acquired by three judgment devices 10 each include data on two types of physical quantities. For example, in the example of Figure 1A, the judgment target data D acquired by the discrimination judgment device 11 includes data DA related to physical quantity A and data DB related to physical quantity B, the judgment target data D acquired by the condition judgment device 12 shown by a solid line includes data DA related to physical quantity A and data DC related to physical quantity C, and the judgment target data D acquired by the condition judgment device 12 shown by a two-dot chain line includes data DE related to physical quantity E and data DF related to physical quantity F. 1A, the types of physical quantities in the judgment target data D acquired by the discrimination decision device 11 are different from some of the types of physical quantities in the judgment target data D acquired by the condition decision device 12, which are indicated by solid lines, and are different from the types of physical quantities in the judgment target data D acquired by the condition decision device 12, which are indicated by two-dot chain lines. In the first embodiment, at least some of the types of physical quantities in the two judgment target data acquired by any two of the multiple decision devices 10 may be the same as each other.

[0044] A series of processes executed by the discrimination decision unit 11 is referred to as discrimination decision processing S11, and a series of processes executed by the condition decision unit 12 is referred to as condition decision processing S12. As shown in Figures 1(a) to 1(c), the discrimination decision processing S11 and the condition decision processing S12 each include steps S1 to S3.

[0045] Next, an example of the order of the determination processing by the multiple determiners 10 will be described using the flowcharts of Figures 1(d) to 1(j). However, Figures 1(d) to 1(j) are merely examples of the first embodiment and do not limit the first embodiment. A YES in the identification determination processing S11 in the flowcharts of Figures 1(d) to 1(j) corresponds to a YES in step S3 by the identification determiner 11 in the flowcharts of Figures 1(a) to 1(c). A YES in the condition determination processing S12 in the flowcharts of Figures 1(d), 1(h), and 1(j) corresponds to a YES in step S3 by the condition determiner 12 in the flowcharts of Figures 1(a) and 1(c).

[0046] Fig. 1(d) shows an example where the plurality of decision devices 10 are configured with one discrimination decision device 11 and one condition decision device 12. Figs. 1(e), 1(f), and 1(g) show examples where the plurality of decision devices 10 are configured with two discrimination decision devices 11. Figs. 1(h), 1(i), and 1(j) show examples where the plurality of decision devices 10 are configured with one discrimination decision device 11 and two condition decision devices 12. Figs. 1(g) and 1(j) show cases where the discrimination decision processing S11 by the two discrimination decision devices 11 is executed in parallel. "Executing in parallel" does not only mean executing in physical parallel, but also includes processing in logical parallel.

[0047] In the examples of FIGS. 1(d) and 1(f) to 1(j), if the discrimination judger 11 judges that the object to be classified is in a predetermined state in the discrimination judgment process S11, the processor 2 outputs a discrimination signal indicating that the object to be classified is in a predetermined state in step S4. In the example of FIG. 1(e), if the first discrimination judger 11 of the two discrimination judgers 11 judges that the object to be classified is in a predetermined state in the discrimination judgment process S11, the second discrimination judger 11 executes the discrimination judgment process S11. Then, if the second discrimination judger 11 judges that the object to be classified is in a predetermined state in the discrimination judgment process S11, the processor 2 outputs a discrimination signal indicating that the object to be classified is in a predetermined state in step S4. In the examples of FIGS. 1(d) to 1(j), if the discrimination judger 11 judges that the object to be classified is not in a predetermined state, the process returns to the identification process. In this case, the processor 2 may output a discrimination signal indicating that the object to be classified is not in a predetermined state.

[0048] In the examples of FIGS. 1(d) and 1(h) to 1(j), after the condition decision unit 12 performs the condition decision process S12, one of the discrimination decision units 11 performs the discrimination decision process S11. More specifically, if the condition decision unit 12 determines that a predetermined condition is met in the condition decision process S12, one of the discrimination decision units 11 performs the discrimination decision process S11 to determine whether the object to be identified is in a predetermined state. If the condition decision unit 12 determines that the predetermined condition is not met, the discrimination decision unit 11 does not determine whether the object to be identified is in a predetermined state. This configuration eliminates unnecessary decisions by the discrimination decision unit 11 and reduces the load on the computational processing of the state identification device 1 compared to when the condition decision unit 12 makes a decision after the discrimination decision unit 11 makes a decision. However, when the multiple decision units 10 of the first embodiment have at least one condition decision unit 12, the discrimination decision process S11 and the condition decision process S12 do not have to be performed in this order.

[0049] Here, the time required for the discrimination and judgment device 11 to complete the judgment of whether the data to be judged is in a predetermined state after acquiring the judgment target data is defined as discrimination and judgment time T. The discrimination and judgment time T is, for example, the time from step S1 to step S3. When the multiple judges 10 include multiple discrimination and judgment devices 11, the discrimination and judgment time T of the discrimination and judgment device 11 with the longest discrimination and judgment time among the multiple discrimination and judgment devices 11 is defined as maximum discrimination and judgment time TMAX, and the discrimination and judgment time T of the discrimination and judgment device 11 with the shortest discrimination and judgment time among the multiple discrimination and judgment devices 11 is defined as minimum discrimination and judgment time TMIN. In the first embodiment, the difference between the maximum discrimination and judgment time TMAX and the minimum discrimination and judgment time TMIN may be shorter than half of the maximum discrimination and judgment time TMAX. However, the condition identification device 1 of the first embodiment is not limited to this configuration.

[0050] The MT method is one of the techniques included in the MT system. In addition to the MT method, other major techniques included in the MT system include the MTA method, the RT method, and the two-sided T method. The RT method is a method suitable for setting multiple unit spaces and determining whether the object to be identified is in one of multiple states; it is also possible to set only one unit space. However, the RT method is a method suitable for setting multiple unit spaces based on the same type of feature and identifying multiple states, but is not suitable for changing the type of feature for each unit space. Furthermore, the RT method and the MTA method have difficulty in handling the adjoint Mahalanobis distance used in the calculation process. Furthermore, the two-sided T method and the MTA method require setting the unit space and signal level, making it difficult to ensure identification accuracy. In contrast, the state identification device 1 of this embodiment, which uses multiple classifiers that independently perform judgments using the MT method to determine whether the object to be identified is in a single predetermined state, can increase identification accuracy while reducing the computational load.

[0051] Second Embodiment A condition identification device 1 according to a second embodiment of the present invention will be described with reference to Figures 2 to 9. The condition identification device 1 according to the second embodiment has the same configuration as the condition identification device 1 according to the first embodiment. The condition identification device 1 according to the second embodiment is an example in which the condition identification device according to the present invention is applied to a misfire diagnosis device that outputs an identification signal that identifies whether or not the engine 40 mounted on a vehicle 30 is in a misfire state. However, the application of the condition identification device according to the present invention to a misfire diagnosis device is not limited to the second embodiment.

[0052] The condition identification device 1 is provided in a vehicle 30. The condition identification device 1 also serves as a control device that controls an engine 40. The vehicle 30 is not particularly limited as long as it is a vehicle 30 that travels on a road surface. The vehicle 30 has a plurality of wheels 31 including at least one drive wheel. The drive wheel receives power output from the engine 40 and rotates.

[0053] The vehicle 30 has an engine 40, a fuel supply device (not shown), an ignition device (not shown), a throttle valve 45, a catalyst 46, an engine rotation speed sensor 51, a throttle opening sensor (throttle position sensor) 52, an intake pressure sensor 53, an oxygen sensor 54, and a wheel rotation speed sensor 55. The vehicle 30 also has an alarm lamp (not shown) that lights up when the state identification device 1 outputs an identification signal indicating that the engine 40 is in a misfire state.

[0054] The engine 40 has at least one combustion chamber 41. The engine 40 may be a single-cylinder engine or a multi-cylinder engine. An ignition device (not shown) is provided in the combustion chamber 41. The engine 40 has a crankshaft 42 that rotates when a mixture of fuel and air is burned in the at least one combustion chamber 41. Power of the engine 40 is output from the crankshaft 42. A fuel supply device (not shown) is disposed in the combustion chamber 41 or in an intake passage 43 connected to the combustion chamber 41. The fuel supply device injects fuel into the combustion chamber 41 or the intake passage 43. A throttle valve 45 is provided to adjust the amount of air supplied to the engine 40. The throttle valve 45 may be electronically controlled or mechanically controlled.

[0055] The catalyst 46 is disposed in the exhaust passage 44 connected to the combustion chamber 41 and purifies the exhaust gas emitted from the engine 40. The oxygen sensor 54 is disposed in the exhaust passage 44 and is located upstream of the catalyst 46 in the flow direction of the exhaust gas. The oxygen sensor 54 outputs a signal corresponding to the oxygen concentration in the exhaust gas. The oxygen sensor 54 outputs a signal indicating whether the oxygen concentration is higher or lower than a predetermined oxygen concentration range.

[0056] The engine rotation speed sensor 51 outputs a signal every time the crankshaft 42 rotates a predetermined angle. In other words, it outputs a signal every predetermined crank angle. The predetermined crank angle may be, for example, 15° crank angle (hereinafter abbreviated as °CA). The condition identification device 1 calculates the engine rotation speed ES, which is the rotation speed of the crankshaft 42, based on the signal output from the engine rotation speed sensor 51.

[0057] The throttle opening sensor 52 detects the position of the throttle valve 45 and outputs a signal indicating the opening TH of the throttle valve 45. Hereinafter, the opening TH of the throttle valve 45 will be referred to as the throttle opening TH. The intake pressure sensor 53 is disposed between the throttle valve 45 and the combustion chamber 41 and detects the pressure of the air in the intake passage 43. The condition identification device 1 acquires the intake air amount IA. The condition identification device 1 calculates the intake air amount IA based on, for example, the calculated engine rotation speed ES, a signal from the throttle opening sensor 52, and a signal from the intake pressure sensor 53. The vehicle 30 may have an air flow meter (not shown) that detects the intake air amount IA. In this case, the condition identification device 1 acquires the intake air amount IA detected by the air flow meter.

[0058] The state identification device 1 controls the amount of fuel supplied from the fuel supply device to the combustion chamber 41. The state identification device 1 controls the amount of fuel based on a signal from the oxygen sensor 54 to maintain the air-fuel ratio of the mixture near a target air-fuel ratio. The state identification device 1 sets the amount of fuel based on at least a base fuel amount and a feedback correction coefficient FB set based on the signal from the oxygen sensor 54. The fuel state identification device 1 may set the amount of fuel based on the base fuel amount, the feedback correction coefficient FB, and at least one correction value. The base fuel amount is set based on a calculated or detected intake air amount IA, etc. When the state identification device 1 calculates the amount of fuel, the feedback correction coefficient FB is multiplied by the base fuel amount. If the oxygen sensor 54 outputs a signal indicating that the oxygen concentration is higher than a predetermined oxygen concentration range, the state identification device 1 increases the feedback correction coefficient FB to increase the amount of fuel. If the oxygen sensor 54 outputs a signal indicating that the oxygen concentration is lower than the predetermined oxygen concentration range, the state identification device 1 decreases the feedback correction coefficient FB to decrease the amount of fuel. The feedback correction coefficient FB fluctuates around a value of 1, for example.

[0059] The wheel rotation speed sensor 55 detects the rotation speed of the wheel 31. The condition identification device 1 calculates the rotation speed of the wheel 31 based on the signal output from the wheel rotation speed sensor 55. The condition identification device 1 calculates the vehicle speed VS based on the calculated rotation speed of the wheel 31.

[0060] In the second embodiment, the number of the multiple determiners 10 is three. The three determiners 10 are composed of two discrimination determiners 11 and one condition determiner 12. The two discrimination determiners 11 are composed of an intermittent misfire discrimination determiner 11a and a consecutive misfire discrimination determiner 11b. The intermittent misfire discrimination determiner 11a determines whether the engine 40 is in a misfire state by determining the presence or absence of intermittent misfires based on the calculated Mahalanobis distance MD. Intermittent misfires are a phenomenon in which misfires occur sporadically in one or more cylinders. The consecutive misfire discrimination determiner 11b determines whether the engine 40 is in a misfire state by determining the presence or absence of consecutive misfires based on the calculated Mahalanobis distance MD. Consecutive misfires are a phenomenon in which misfires occur consecutively in a specific cylinder. The condition determiner 12 determines whether the road surface on which the vehicle 30 is traveling is flat.

[0061] The condition identification device 1 of the second embodiment may execute the determination process by the three determiners 10 in the order shown in FIG. 1( h) or 1( j), for example. That is, when the condition determiner 12 determines that the road surface is flat, the two determination determiners 11 determine whether the engine 40 is in a misfire state, and when the condition determiner 12 determines that the road surface is not flat, the two determination determiners 11 do not determine whether the engine 40 is in a misfire state. The condition identification device 1 of the second embodiment may also execute the determination process by the three determiners 10 in the order shown in FIG. 1( i), for example. However, in this case, the consecutive misfire determination determiner 11 b performs the determination process before the determination process by the condition determiner 12. In other words, when the condition determiner 12 determines that the road surface is flat, the intermittence discrimination determiner 11a determines whether the engine 40 is in a misfire state, and when the condition determiner 12 determines that the road surface is not flat, the intermittence discrimination determiner 11a does not determine whether the engine 40 is in a misfire state.

[0062] The condition decision unit 12 acquires, as the decision target data D, a vehicle speed VS obtained from the signal of the wheel rotation speed sensor 55 and an engine rotation speed ES obtained from the signal of the engine rotation speed sensor 51. Instead of the vehicle speed VS, the condition decision unit 12 may acquire the rotation speed of the wheels 31 obtained from the signal of the wheel rotation speed sensor 55. The engine rotation speed ES acquired by the condition decision unit 12 may be a value for each crank angle greater than the crank angle interval at which the engine rotation speed sensor 51 outputs a signal. The engine rotation speed ES acquired by the condition decision unit 12 may be data of values ​​for each 720° CA, for example. The condition decision unit 12 extracts multiple types of feature quantities from the acquired decision target data D and calculates the Mahalanobis distance MD based on the extracted multiple types of feature quantities and a preset unit space U. Specific examples of the feature quantities will be described later. The unit space U for the condition decision unit 12 is set based on a reference data group when the road surface is uneven and has bumps and no misfire occurs. The condition decision unit 12 determines that the road surface is not flat if the calculated Mahalanobis distance MD is smaller than a predetermined judgment threshold, and determines that the road surface is flat if the calculated Mahalanobis distance MD is equal to or larger than the predetermined judgment threshold.

[0063] The condition decision unit 12 extracts at least one type of feature quantity indicating a characteristic of fluctuations in the engine rotation speed ES from data on the engine rotation speed ES during a first determination period. The length of the first determination period may be set independently of the engine rotation speed ES or may be set according to the engine rotation speed ES. The first determination period may be, for example, approximately one second. The at least one type of feature quantity may be, for example, a first feature quantity indicating the magnitude of fluctuations in the engine rotation speed ES and a second feature quantity indicating the frequency of fluctuations in the engine rotation speed ES. The first feature quantity may be, for example, the difference between the maximum and minimum values ​​of the value for each first unit period during the first determination period. The first unit period may be, for example, approximately 10 to 30 msec. The second feature quantity may be, for example, the standard deviation of the differential value for each second unit period during the first determination period. The length of the second unit period may be the same as or different from the length of the first unit period. The second unit period may be, for example, approximately 10 to 30 msec. The second unit period may be set in accordance with the cycle of fluctuations in the engine speed ES so that a large derivative value appears.

[0064] The condition decision unit 12 extracts at least one type of feature quantity indicating the characteristics of fluctuations in the vehicle speed VS (or the rotation speed of the wheels 31) from data on the vehicle speed VS (or the rotation speed of the wheels 31) during the first decision period. The at least one type of feature quantity is, for example, a first feature quantity indicating the magnitude of fluctuations in the vehicle speed VS and a second feature quantity indicating the frequency of fluctuations in the vehicle speed VS. A specific example of the first feature quantity may be the same as the specific example of the first feature quantity indicating the magnitude of fluctuations in the engine rotation speed ES described above. A specific example of the second feature quantity may be the same as the specific example of the second feature quantity indicating the frequency of fluctuations in the engine rotation speed ES described above.

[0065] The intermittent misfire discrimination decision-maker 11a acquires, as the discrimination target data D, the calculated or detected intake air amount IA and the engine speed ES obtained from the signal of the engine speed sensor 51. Like the engine speed ES acquired by the condition decision-maker 12, the engine speed ES acquired by the intermittent misfire discrimination decision-maker 11a may be a value for each crank angle greater than the crank angle interval at which the engine speed sensor 51 outputs a signal. The intermittent misfire discrimination decision-maker 11a extracts multiple types of feature quantities from the acquired discrimination target data D and calculates a Mahalanobis distance MD based on the extracted multiple types of feature quantities and a preset unit space U. Specific examples of the feature quantities will be described later. The unit space U for the intermittent misfire discrimination decision-maker 11a is set based on a reference data group for when the road surface is flat and no misfire occurs. The intermittent misfire discrimination decision-maker 11a estimates the misfire rate MR based at least on the calculated Mahalanobis distance MD. Because the unit space U is set based on a set of reference data for when no misfire occurs, the misfire rate MR can be estimated from the Mahalanobis distance MD. The misfire rate MR is the ratio of the number of misfires to the cumulative number of revolutions of the crankshaft 42. The intermittent misfire discrimination / determinator 11a may estimate the misfire rate MR based on the Mahalanobis distance MD and a value related to the engine load. In this case, the storage device 3 pre-stores a map showing the relationship between the Mahalanobis distance MD and the misfire rate MR for each engine load range. The value related to the engine load is, for example, the intake air amount IA. The intermittent misfire discrimination / determinator 11a may estimate the misfire rate MR based on the Mahalanobis distance MD, the value related to the engine load, and the engine rotational speed ES. In this case, the storage device 3 pre-stores a map showing the relationship between the Mahalanobis distance MD and the misfire rate MR for each engine operating range, as shown in FIG. 3, for example. The engine operating range is a combination of an engine load range and an engine rotational speed range. 3 also shows graphs illustrating the relationship between the Mahalanobis distance MD and the misfire rate MR, which are set for two engine operating regions. In these two graphs, the horizontal axis value X1, which indicates the misfire rate MR, is the same, and the vertical axis value Y1, which indicates the Mahalanobis distance MD, is the same.The intermittent misfire discrimination determiner 11a determines that intermittent misfires are occurring when the estimated misfire rate MR is greater than a predetermined misfire rate determination threshold, and determines that intermittent misfires are not occurring when the estimated misfire rate MR is equal to or less than the misfire rate determination threshold. If the intermittent misfire discrimination determiner 11a determines that intermittent misfires are occurring, it determines that engine 40 is in a misfire state and outputs an identification signal indicating that engine 40 is in a misfire state to an alarm lamp (not shown).

[0066] The intermittent misfire identification determiner 11a extracts at least one feature quantity indicating the characteristics of fluctuations in the engine speed ES from the data on the engine speed ES during the second determination period. The length of the second determination period may be the same as or different from the length of the first determination period. The length of the second determination period may be set independently of the engine speed ES or may be set according to the engine speed ES. The at least one feature quantity is, for example, a feature quantity indicating the frequency and magnitude of fluctuations in the engine speed ES. The feature quantity indicating the frequency of fluctuations in the engine speed ES may be, for example, a standard deviation of a differential value for each unit period during the second determination period. The at least one feature quantity may be a plurality of standard deviations with different unit periods. When the feature quantity is a plurality of standard deviations with different unit periods, the feature quantity indicates the magnitude of fluctuations in the engine speed ES. For example, the at least one type of characteristic amount may be the standard deviation of the differential values ​​obtained every 10 msec during the second determination period, the standard deviation of the differential values ​​obtained every 20 msec during the second determination period, and the standard deviation of the differential values ​​obtained every 30 msec during the second determination period. The intermittent misfire discrimination determiner 11a may acquire, as part of the determination target data D, at least some of the multiple types of characteristic amounts extracted by the condition determiner 12 from the engine speed ES data.

[0067] The intermittent misfire detector 11a extracts at least one characteristic quantity from the data on the intake air amount IA during the second determination period. The at least one characteristic quantity is, for example, a representative value of the intake air amount IA during the second determination period. The representative value may be, for example, an average value.

[0068] The consecutive misfire discrimination determiner 11b acquires, as the discrimination target data D, the throttle opening TH obtained from the signal of the throttle opening sensor 52, the vehicle speed VS obtained from the signal of the wheel rotation speed sensor 55, and a feedback correction coefficient FB set based on the signal of the oxygen sensor 54. The consecutive misfire discrimination determiner 11b may acquire the rotation speed of the wheels 31 obtained from the signal of the wheel rotation speed sensor 55 instead of the vehicle speed VS. The consecutive misfire discrimination determiner 11b may acquire the signal value of the oxygen sensor 54 instead of the feedback correction coefficient FB. The consecutive misfire discrimination determiner 11b extracts multiple types of feature quantities from the acquired discrimination target data D and calculates the Mahalanobis distance MD based on the extracted multiple types of feature quantities and a predetermined unit space U. Specific examples of the feature quantities will be described later. The unit space U for the consecutive misfire discrimination determiner 11b is set based on a reference data group obtained when the road surface is flat and no misfire occurs. The consecutive misfire discrimination determiner 11b determines that consecutive misfires are occurring when the calculated Mahalanobis distance MD is greater than a predetermined threshold value greater than 1, and determines that consecutive misfires are not occurring when the calculated Mahalanobis distance MD is less than the threshold value. The unit space U for the consecutive misfire discrimination determiner 11b may be set based on a reference data set obtained when the road surface is flat and consecutive misfires are occurring. In this case, the consecutive misfire discrimination determiner 11b determines that consecutive misfires are occurring when the calculated Mahalanobis distance MD is less than a predetermined threshold value near 1, and determines that consecutive misfires are not occurring when the calculated Mahalanobis distance MD is greater than the threshold value. Regardless of the setting of the unit space U, the reference data set is preferably, but not limited to, a data set obtained when the vehicle speed VS is maintained constant. If the consecutive misfire discrimination determiner 11b determines that consecutive misfires are occurring, it determines that the engine 40 is in a misfire state and outputs an identification signal to an alarm lamp (not shown) indicating that the engine 40 is in a misfire state. If the consecutive misfire discrimination determiner 11b determines that consecutive misfires are occurring, in addition to outputting the identification signal, it may also output information on the misfire rate MR estimated based on the number of cylinders of the engine 40 to an output destination other than the alarm lamp.For example, if engine 40 is a four-cylinder engine and continuous misfires are occurring, the misfire rate MR can be estimated to be at least 25% because continuous misfires are occurring in at least one of the four cylinders. The misfire rate determination threshold value used by intermittent misfire identification and judgment unit 11a is smaller than the misfire rate when continuous misfires occur in one cylinder in engine 40 with a typical number of cylinders. If continuous misfire identification and judgment unit 11b is configured to output information about the misfire rate MR when it is determined that engine 40 is in a misfire state, intermittent misfire identification and judgment unit 11a is configured to output information about the estimated misfire rate MR when it is determined that engine 40 is in a misfire state.

[0069] The consecutive misfire identification determiner 11b extracts at least one characteristic quantity indicating the magnitude of the throttle opening TH relative to the vehicle speed VS from the throttle opening TH data and the vehicle speed VS (or the rotational speed of the wheels 31) data during the third determination period. The length of the third determination period may be the same as or different from the length of at least one of the first determination period and the second determination period. The length of the third determination period may be set independently of the engine rotational speed ES or may be set according to the engine rotational speed ES. The at least one characteristic quantity may be, for example, a value obtained by dividing the average value of the throttle opening TH during the third diagnosis period by the average value of the vehicle speed VS during the third diagnosis period.

[0070] The consecutive misfire identification determiner 11b extracts a feature quantity indicating the characteristics of fluctuations in the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) from the data on the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) during the third determination period. The feature quantity indicating the characteristics of fluctuations in the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) is, for example, a representative value (feature quantity) of the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) during the third determination period and a feature quantity indicating the frequency of fluctuations in the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) during the third determination period. The representative value may be, for example, an average value. The feature quantity indicating the frequency of fluctuations in the feedback correction coefficient FB (or the signal value of the oxygen sensor 54) during the third determination period may be, for example, the standard deviation of the differential value for each unit period in the third determination period. The unit period is set to a length shorter than the cycle of fluctuations in the feedback correction coefficient FB. This allows the feature quantity indicating the frequency of fluctuations in the feedback correction coefficient FB to be extracted. The unit period may be, for example, about 10 to 30 msec. The period of fluctuation of the feedback correction coefficient FB is usually 100 msec or more.

[0071] 4(a) to 4(d) are graphs showing temporal changes in the throttle opening TH obtained from the signal of the throttle opening sensor 52, the vehicle speed VS obtained from the signal of the wheel rotation speed sensor 55, and the engine rotation speed ES obtained from the signal of the engine rotation speed sensor 51. The range of values ​​on the vertical axis of the throttle opening TH graphs in FIGS. 4(a) to 4(d) is the same. The difference between the upper and lower limit values ​​of the vehicle speed VS graphs in FIGS. 4(a) to 4(d) is also the same. The difference between the upper and lower limit values ​​of the engine rotation speed ES graphs in FIGS. 4(a) to 4(d) is also the same. In FIGS. 4(a) to 4(d), the engine rotation speed ES graphs are graphs of values ​​every 720° CA. The time range of the graphs in FIGS. 4(a) to 4(d) is approximately 10 seconds. Fig. 4(a) is a graph showing the case where the road surface is not flat, Fig. 4(b) is a graph showing the case where intermittent misfires occur, Fig. 4(c) is a graph showing the case where continuous misfires occur, and Fig. 4(d) is a graph showing the case where the engine is normal. In the description of this embodiment, normal refers to the case where no misfires occur and the road surface is flat.

[0072] As shown in FIGS. 4( a ) to 4 ( d ), regardless of whether misfire occurs, the engine rotation speed ES fluctuates in an oscillating manner in response to the timing of a specific stroke (e.g., a combustion stroke) in one cycle of the engine 40. When intermittent misfire occurs, as shown in FIGS. 4( b ) and 4 ( d ), the fluctuation of the engine rotation speed ES becomes larger than under normal conditions. That is, the amplitude of the fluctuation of the engine rotation speed ES becomes larger. This is because the engine rotation speed ES decreases the instant intermittent misfire occurs, and when ignition returns to normal immediately thereafter, the engine rotation speed ES increases compared to under normal conditions. As shown in FIGS. 4( a ) and 4 ( d ), when the road surface is uneven and has bumps, the contact state between the wheels 31 and the road surface is not constant, and the vehicle separates from the road surface or makes strong contact with the road surface, resulting in larger fluctuations in the rotation speed of the wheels 31. As a result, the crankshaft 42, which is mechanically connected to the drive wheels among the wheels 31, is affected by the fluctuations in the rotation speed of the wheels 31, resulting in larger fluctuations in the engine rotation speed ES. As shown in Figures 4(a) and 4(b), the fluctuations in engine speed ES when the road surface is uneven are similar to those when intermittent misfires occur. However, the fluctuations in the rotational speed of the wheels 31 differ between when the road surface is uneven and when intermittent misfires occur. Therefore, it is possible to distinguish between traveling on an uneven road surface and intermittent misfires using the engine speed ES data and the rotational speed of the wheels 31 or the vehicle speed VS calculated from the rotational speed of the wheels 31. However, if the unit space U for the condition decision unit 12 is set based on a reference data group for when the road surface is flat and no misfires occur, the Mahalanobis distance MD will be large not only when the road surface is uneven but also when intermittent misfires occur. Therefore, if the condition decision unit 12 were to determine whether the road surface is flat or whether intermittent misfires occur based on this unit space U, the Mahalanobis distance MD would be large in both cases where the road surface is uneven and where intermittent misfires occur, and it would be impossible to correctly determine whether the road surface is flat. In this embodiment, the unit space U for the condition decision unit 12 is set based on a reference data group when the road surface is not flat and no misfires occur.This prevents the condition determiner 12 from erroneously determining that the road surface is not flat when intermittent misfires are occurring. Furthermore, in this embodiment, the intermittent misfire determination determiner 11a determines whether or not an intermittent misfire is occurring after the condition determiner 12 determines that the road surface is flat. Therefore, even if the unit space U for the intermittent misfire determination determiner 11a is set based on a set of reference data for when the road surface is flat and no misfires are occurring, the intermittent misfire determination determiner 11a can correctly determine whether or not an intermittent misfire is occurring.

[0073] When continuous misfires occur, the fluctuations in engine speed ES differ from normal. However, as shown in FIGS. 4(c) and 4(d), this difference cannot be determined from engine speed ES data taken every 720° CA. For this reason, conventional misfire diagnosis uses engine speed ES data with high time resolution. Conventional misfire diagnosis uses engine speed ES data taken every 15° CA, for example. In contrast, in this embodiment, engine speed ES is not used to determine whether continuous misfires occur.

[0074] 5(a) to 5(c) are graphs showing temporal changes in the throttle opening TH obtained from the signal of the throttle opening sensor 52, the vehicle speed VS obtained from the signal of the wheel rotation speed sensor 55, the engine rotation speed ES obtained from the signal of the engine rotation speed sensor 51, and the Mahalanobis distance MD calculated by the condition decision unit 12. Strictly speaking, the vertical axis of the Mahalanobis distance MD graphs in FIGS. 5(a) to 5(c) represents the square of the Mahalanobis distance MD. The same applies to the vertical axes of the Mahalanobis distance MD graphs in FIGS. 6(a), 6(b), 7, 8(a), 8(b), and 9, which will be described later. The upper limit of the vertical axis of the Mahalanobis distance MD graphs in FIGS. 5(a) to 5(c) is 45. The range of values ​​on the vertical axis of the throttle opening TH graphs in FIGS. 5(a) to 5(c) is the same. Like the throttle opening TH graph, the vehicle speed VS graph and the engine rotation speed ES graph in Figures 5(a) to 5(c) also have the same range of values ​​on the vertical axis. The time range of the graphs in Figures 5(b) and 5(c) is approximately 30 minutes. Figures 5(b) and 5(c) are graphs showing values ​​from when the vehicle 30 starts traveling, and the vehicle 30 is driven so that the changes in vehicle speed VS are the same in Figures 5(b) and 5(c). Figure 5(a) is a graph showing a case where the road surface is uneven and no misfire occurs, Figure 5(b) is a graph showing a case where the road surface is flat and intermittent misfire occurs, and Figure 5(c) is a graph showing a normal state. As shown in Figure 5(a), when the road surface is uneven, the Mahalanobis distance MD calculated by the condition decision unit 12 is distributed near 1. Furthermore, as shown in Figures 5(b) and 5(c), when the road surface is flat, the Mahalanobis distance MD calculated by the condition decision unit 12 remains greater than 1 regardless of whether a misfire occurs.

[0075] 6(a) and 6(b) are graphs showing temporal changes in the misfire rate MR, intake air flow rate IA, engine speed ES, and the Mahalanobis distance MD calculated by the intermittent misfire discrimination / determining device 11a when the engine 40 is operated so that intermittent misfires occur at a predetermined misfire rate MR. The upper limit of the vertical axis of the Mahalanobis distance MD graphs in FIGS. 6(a) and 6(b) is 19,000. The range of values ​​on the vertical axis of the misfire rate MR graphs in FIGS. 6(a) and 6(b) is the same. The range of values ​​on the vertical axis of the intake air flow rate IA and the engine speed ES graphs in FIGS. 6(a) and 6(b) is the same as the misfire rate MR graph. The time range of the graphs in FIGS. 6(a) and 6(b) is approximately one minute. FIG. 6( a ) is a graph showing the case where the engine load is low, and FIG. 6( b ) is a graph showing the case where the engine load is high. Note that in FIGS. 6( a ) and 6( b ), misfires have not yet occurred during period A1. Comparing FIGS. 6( a ) and 6( b ) reveals that the amount of fluctuation in engine speed ES during intermittent misfires is greater under high loads than under low loads. In other words, engine load correlates with the amount of fluctuation in engine speed ES during intermittent misfires. As shown in FIGS. 6( a ) and 6( b ), the magnitude of fluctuation in engine speed ES is not correlated with the misfire rate MR. Because the time interval between misfires decreases as the misfire rate MR increases, the frequency of engine speed ES fluctuations increases as the misfire rate MR increases. When the frequency of engine speed ES fluctuations increases, the lines in the engine speed ES graph become denser.

[0076] The intermittent misfire identification / determination device 11a of this embodiment acquires the intake air amount IA and the engine speed ES as the determination target data D, extracts a feature value representing the frequency of fluctuations in the engine speed ES from the engine speed ES data, and calculates a Mahalanobis distance MD based on the extracted feature value representing the frequency of fluctuations in the engine speed ES, a representative value (feature value) of the intake air amount IA, and the unit space U. As shown in FIGS. 6(a) and 6(b), the higher the misfire rate MR, the larger the Mahalanobis distance MD. Therefore, the calculated Mahalanobis distance MD can be used to estimate the misfire rate MR. Furthermore, although the determination target data D includes the intake air amount IA, as shown in FIGS. 6(a) and 6(b), the Mahalanobis distance MD varies depending on the engine load, even for the same misfire rate MR. Therefore, by estimating the misfire rate MR using the calculated Mahalanobis distance MD and a value related to the engine load (e.g., the intake air amount IA), the misfire rate MR can be accurately estimated. Furthermore, by estimating the misfire rate MR using the calculated Mahalanobis distance MD, a value related to the engine load (for example, the intake air amount IA), and the engine rotation speed ES, the misfire rate MR can be estimated with higher accuracy.

[0077] Figure 7 is a graph showing three regions with different misfire rates selected from the three graphs in Figure 6(a) excluding the intake air volume IA, and expanded along the horizontal axis. The upper limit of the vertical axis of the Mahalanobis distance MD graph in Figure 7 is 9000. The arrows in Figure 7 indicate a time range of, for example, 5 seconds. If the second determination period is, for example, 1 second, the Mahalanobis distance MD may not be stable 1 or 2 seconds after the misfire rate MR changes. However, approximately 5 seconds after the misfire rate MR changes, regardless of the engine load, the intermittent misfire identification / determination unit 11a can determine that intermittent misfires are occurring and accurately estimate the misfire rate MR.

[0078] 8(a) and 8(b) are graphs showing temporal changes in the throttle opening TH, vehicle speed VS, engine rotation speed ES, feedback correction coefficient FB, and Mahalanobis distance MD calculated by the consecutive misfire discrimination determiner 11b after the vehicle 30 starts traveling. The upper limit of the vertical axis of the Mahalanobis distance MD graphs in FIGS. 8(a) and 8(b) is 900. The range of values ​​on the vertical axis of the throttle opening TH graphs in FIGS. 8(a) and 8(b) is the same. The range of values ​​on the vertical axis of the vehicle speed VS graph, the engine rotation speed ES graph, and the feedback correction coefficient FB graphs in FIGS. 8(a) and 8(b) is the same as the throttle opening TH graph. In the examples of FIGS. 8(a) and 8(b), the unit space U for the consecutive misfire discrimination determiner 11b is set based on a reference data set obtained when the road surface is flat and no misfires occur. The time range of the graphs in FIGS. 8(a) and 8(b) is approximately 30 minutes. FIG. 8(a) is a graph showing a case where consecutive misfires are occurring, and FIG. 8(b) is a graph showing a normal state. In FIGS. 8(a) and 8(b), the vehicle 30 is driven so that the changes in vehicle speed VS are the same. When consecutive misfires are occurring, the output of the engine 40 is reduced compared to normal operation. Therefore, as shown in FIGS. 8(a) and 8(b), when consecutive misfires are occurring, the throttle opening TH is larger than normal operation even at the same vehicle speed VS. Furthermore, when consecutive misfires are occurring, the feedback correction coefficient FB is higher compared to normal operation. In FIG. 8(a), the feedback correction coefficient FB is maintained at its upper limit for a long period of time. In other words, the feedback correction coefficient FB fluctuates less frequently. The reason the feedback correction coefficient FB is higher is that when consecutive misfires occur, the mixture is not combusted in a particular cylinder, and even if the fuel amount is increased, the oxygen sensor 54 continues to output a signal indicating that the oxygen concentration is higher than the predetermined range. Focusing on the behavior of physical quantities that occur when such consecutive misfires occur, the consecutive misfire identification and judgment unit 11b of this embodiment determines whether or not consecutive misfires are occurring using the vehicle speed VS (or the rotational speed of the wheels 31), the throttle opening TH, and the feedback correction coefficient FB (or the signal from the oxygen sensor 54).As shown in Figure 8(a), when consecutive misfires occur, the Mahalanobis distance MD is distributed over values ​​significantly greater than 1. As shown in Figure 8(b), when consecutive misfires do not occur, the Mahalanobis distance MD is distributed over a range close to 1. In Figure 8(b), there are points where the Mahalanobis distance MD temporarily becomes significantly greater than 1. This occurs when the throttle opening TH suddenly increases due to starting or shifting gears.

[0079] FIG. 9 is a graph of FIG. 8(a) enlarged in the horizontal direction for the region 25 seconds after the vehicle 30 starts traveling. The upper limit value of the vertical axis of the graph of the Mahalanobis distance MD in FIG. 9 is 900. In FIGS. 8(a), 8(b), and 9, the vehicle 30 is stopped in an idling state before starting to travel, and the throttle opening TH is still minimal even 5 seconds after the vehicle 30 starts traveling. Therefore, if continuous misfires occur, the feedback correction coefficient FB begins to increase approximately 5 seconds after the vehicle 30 starts traveling, and the Mahalanobis distance MD becomes a value slightly higher than 1. Then, approximately 10 seconds after the vehicle 30 starts traveling, the Mahalanobis distance MD stabilizes at a value clearly higher than 1. Therefore, in the examples of Figures 8(a) and 9, if the above-mentioned third judgment period is, for example, 1 second, approximately 10 seconds after the vehicle 30 starts to move, the consecutive misfire identification judger 11b can correctly judge that the engine 40 is in a misfire state and can accurately estimate the misfire rate MR.

[0080] The misfire rate MR threshold, which is the standard for determining whether the engine 40 is misfiring, may be as low as 3%, for example. In a conventional method for calculating the misfire rate MR by counting the number of misfires, the number of misfires is counted during 1,000 revolutions of the crankshaft 42 in order to ensure accuracy in the misfire rate MR. Furthermore, to ensure accuracy, the period during which the engine speed ES changes due to changes in the throttle opening TH is excluded. Therefore, in a conventional misfire diagnosis that does not use the MT method, if the engine speed ES is low, a diagnosis time of approximately several tens of seconds is required to complete 1,000 revolutions.

[0081] The condition identification device 1 of this embodiment focuses on the "visual appearance" of physical quantities related to the vehicle's operating state when they are graphed, and quantifies the "visual appearance" using the MT method, a pattern recognition technology, to perform misfire diagnosis. Therefore, the condition decision unit 12 and the intermittent misfire decision unit 11a can make their decision without using engine speed ES data with high time resolution, as used in conventional misfire diagnosis without the MT method. The consecutive misfire decision unit 11b can also make their decision without using engine speed ES data. Furthermore, the consecutive misfire decision unit 11b can also make their decision without using data with high time resolution, as used in conventional misfire diagnosis without the MT method. Therefore, the computational load can be significantly reduced compared to conventional misfire diagnosis without the MT method. Furthermore, in the condition identification device 1 of the second embodiment, the three decision units 10 independently make decisions using the MT method to determine whether the engine 40 is in a misfire state. Therefore, compared to the condition identification device of Patent Document 4, in which a single determiner uses the MT method to determine whether the engine is misfiring, the computational processing performed by the single determiner 10 can be simplified. This reduces the computational processing load on the entire condition identification device 1. This increases the design flexibility of the hardware resources of the condition identification device 1 while further improving the identification accuracy. Because the three determiners 10 of this embodiment use the MT method to make the determination, the lengths of the first, second, and third determination periods can be significantly shorter than those in conventional misfire diagnosis that do not use the MT method. Therefore, the condition identification device 1 of the second embodiment can determine whether the engine 40 is misfiring in real time and accurately estimate the misfire rate MR.

[0082] In addition, the processor 2 of the condition identification device 1 of the first embodiment may function as at least one non-MT judger (not shown) that makes judgments without using the MT method, in addition to the multiple judgers 10 that make judgments using the MT method.

[0083] In the second embodiment, the MT method is used to determine whether an intermittent misfire occurs, whether a continuous misfire occurs, and whether the road is flat. However, it is also possible to use the MT method to determine two of the intermittent misfire, continuous misfire, and whether the road is flat, and not to use the MT method for the remaining one. For example, it is also possible to use the MT method to determine whether an intermittent misfire occurs and whether the road is flat, but not to use the MT method for the continuous misfire determination. It is also possible to use the MT method to determine whether the road is flat and whether the road is continuous, but not to use the MT method for the intermittent misfire determination.

[0084] In the second embodiment, the condition identification device 1 is provided in the vehicle 30, but when the condition identification device 1 of the present invention is applied to a misfire diagnosis device, the condition identification device 1 may not be provided in the vehicle 30 but may be a device capable of communicating with a device provided in the vehicle 30.

[0085] 1: State identification device, 2: Processor, 3: Storage device, 10: Judgement device, 11: Identification and judgment device, 11a: Intermittent misfire identification and judgment device, 11b: Continuous misfire identification and judgment device, 12: Condition judgment device, 30: Vehicle, 40: Engine, D: Data to be judged, MD: Mahalanobis distance, P: Program, U: Unit space

Claims

1. A state identification device having a processor and a storage device, and outputting an identification signal for identifying whether an object to be identified is in a single predetermined state, wherein the processor functions as a plurality of decision devices each making a decision by executing a program stored in the storage device, wherein a decision device that decides whether the object to be identified is in the predetermined state is defined as a decision decision device, and a decision device that decides whether a predetermined condition other than a decision whether the object to be identified is in the predetermined state is established is defined as a condition decision device, wherein the plurality of decision devices are: (i) composed of at least one of the condition decision device and one of the decision decision devices, (ii) composed of a plurality of decision decision devices without including the condition decision device, or (iii) composed of at least one of the condition decision device and a plurality of decision decision devices, wherein the plurality of decision devices each acquire data to be identified regarding at least one type of physical quantity, calculate Mahalanobis distances based on previously set different unit spaces and the acquired data to be identified by the MT method (Mahalanobis-Taguchi method) without mutual dependence on each other, and make a decision based on the calculated Mahalanobis distances, A state identification device, characterized in that at least some of the types of physical quantities in the two pieces of judgment target data acquired by any two of the plurality of judgers are different from each other.

2. The state identification device according to claim 1, wherein any two of the plurality of decision devices have at least some of the same types of physical quantities in the decision target data they acquire.

3. The state identification device according to claim 1 or 2, wherein the plurality of decision-makers are either (ii) composed of the plurality of discrimination decision-makers without including the condition decision-maker, or (iii) composed of the at least one condition decision-maker and the plurality of discrimination decision-makers, and wherein the time required for the discrimination decision-maker to acquire the data to be judged until it finishes judging whether or not the data is in the predetermined state is defined as a discrimination decision time, the discrimination decision time of the discrimination decision-maker having the longest discrimination decision time among the plurality of discrimination decision-makers is defined as a maximum discrimination decision time, and the discrimination decision time of the discrimination decision-maker having the shortest discrimination decision time among the plurality of discrimination decision-makers is defined as a minimum discrimination decision time, the difference between the maximum discrimination decision time and the minimum discrimination decision time is shorter than half of the maximum discrimination decision time.

4. A state identification device according to any one of claims 1 to 3, characterized in that the plurality of decision-makers are (i) composed of the at least one condition decision-maker and the one discrimination decision-maker, or (iii) composed of the at least one condition decision-maker and the plurality of discrimination decision-makers, and when a first condition decision-maker that is any of the at least one condition decision-maker determines that the predetermined condition is met, the one discrimination decision-maker or a first discrimination decision-maker that is any of the plurality of discrimination decision-makers determines whether the object to be identified is in the predetermined state, and when the first condition decision-maker determines that the predetermined condition is not met, the first discrimination decision-maker does not determine whether the object to be identified is in the predetermined state.

5. The condition identification device outputs the identification signal that identifies whether or not an engine mounted on a vehicle traveling on a road surface is in a misfire state, and the multiple determiners are composed of one condition determiner and two identification determiners, and the two identification determiners are composed of a continuous misfire identification determiner that determines whether or not the engine is in a misfire state by determining whether or not there is a continuous misfire, which is when misfires occur consecutively in a specific cylinder, and an intermittent misfire identification determiner that determines whether or not the engine is in a misfire state by determining whether or not there is an intermittent misfire, which is when misfires occur sporadically in one or more cylinders, and the condition identification device described in any one of claims 1 to 4 is characterized in that the one condition determiner determines whether or not the road surface is flat.

6. A state identification device as described in claim 5, characterized in that when the one condition decision device determines that the road surface is flat, the intermittent misfire identification decision device determines whether or not the engine is in a misfire state, and when the one condition decision device determines that the road surface is not flat, the intermittent misfire identification decision device does not determine whether or not the engine is in a misfire state.

Images