Abnormal detection device for power transmission device

Abstract

This disclosure provides an anomaly determination device for a power transmission device. A storage device stores mapping data, which includes data obtained through machine learning and a defined mapping. When oil temperature correlation data, which corresponds to time-series data of an oil temperature detection value, is input as an input variable, the mapping output determines whether an anomaly has occurred in the friction engagement element. The oil temperature detection value is the value detected by an oil temperature sensor. A processing circuit performs: an acquisition process to acquire the input variable; and an anomaly determination process, determining whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable acquired in the acquisition process into the mapping.

Description

Technical Field

[0001] This disclosure relates to an anomaly determination device for a power transmission device that determines whether an anomaly has occurred in the friction engagement element of the power transmission device. Background Technology

[0002] Japanese Patent Application Publication No. 2011-58510 describes an example of a power transmission device equipped with friction engagement elements such as clutches and brakes. In such a power transmission device, when an abnormality occurs in the friction engagement elements and the oil circulating within the power transmission device deteriorates, the oil becomes foul-smelling. When an abnormality occurs in the friction engagement elements and the oil deteriorates, the composition of the foul odor emitted by the oil changes to a different composition than when no abnormality occurs in the friction engagement elements and the oil does not deteriorate.

[0003] Therefore, in the aforementioned patent document, an odor sensor is installed in the oil tray where the oil is stored to detect the odor components emitted by the oil, and the abnormality of the power transmission device is predicted based on the detection value of the sensor.

[0004] To predict malfunctions in power transmission devices using the aforementioned methods, odor sensor readings are required. Therefore, a technique is desired to detect malfunctions in friction-bonded components without using odor sensor readings. Summary of the Invention

[0005] This document describes several methods disclosed herein and their effects.

[0006] Method 1. An anomaly determination device for a power transmission device, applied to a vehicle, the vehicle comprising: a power transmission device having a friction engagement element configured to transmit power output from a power source of the vehicle to drive wheels; and an oil temperature sensor configured to detect oil temperature as the temperature of oil circulating within the power transmission device, the anomaly determination device comprising a processing circuit and a storage device, the storage device storing mapping data including data obtained by learning a predetermined mapping through machine learning, wherein when oil temperature-related data corresponding to time-series data of oil temperature detection values ​​is input as an input variable, the mapping outputs an output variable determining whether an anomaly has occurred in the friction engagement element, the oil temperature detection value being the detection value of the oil temperature sensor, the processing circuit being configured to perform: an acquisition process acquiring the input variable; and an anomaly determination process determining whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable acquired in the acquisition process into the mapping.

[0007] When the power transmission device operates, heat is generated in the friction engagement element, and this heat is transferred to the oil circulating in the power transmission device. The oil temperature, which is the temperature of the oil, sometimes varies.

[0008] Here, the heat generated by the friction coupling element when an anomaly occurs differs from the heat generated when no anomaly occurs. The oil temperature changes as the heat generated changes. Therefore, by analyzing the oil temperature correlation data corresponding to the time-series data of the oil temperature detection values, it is possible to predict whether an anomaly has occurred in the friction coupling element.

[0009] In the above structure, oil temperature-related data is used as an input variable, and a mapping data that determines whether an anomaly has occurred in the friction engagement element is stored in a storage device. Furthermore, when the power transmission device operates, the output variable output from the mapping, obtained by inputting the input variable into the mapping, is used to determine whether an anomaly has occurred in the friction engagement element. Therefore, according to the above structure, it is possible to determine whether an anomaly has occurred in the friction engagement element without using the detection value of an odor sensor.

[0010] Method 2. The abnormality determination device for the power transmission device according to Method 1, wherein the acquisition process includes: a detection value acquisition process, acquiring time-series data of the oil temperature detection values ​​including a plurality of oil temperature detection values, wherein the plurality of oil temperature detection values ​​are detected for each detection cycle within a predetermined measurement period; and an association data generation process, generating the oil temperature association data by standardizing the plurality of oil temperature detection values ​​included in the time-series data of the oil temperature detection values.

[0011] In the time-series data of oil temperature readings under abnormal conditions in the friction coupling element, there are points of difference compared to the time-series data of oil temperature readings under normal conditions. However, the degree of difference may differ depending on whether the oil temperature reading is high or low. When using the time-series data of oil temperature readings as the input variable for mapping, the accuracy of the above determination tends to decrease when the degree of difference is small compared to when the degree of difference is large. In other words, there is a possibility that the accuracy of the determination may be deviated based on the magnitude of the oil temperature reading at that time.

[0012] Regarding this point, based on the above structure, oil temperature correlation data is input as an input variable to the mapping. Oil temperature correlation data is obtained by standardizing the time-series data of oil temperature detection values. Therefore, the degree of difference between oil temperature correlation data under conditions of anomalies in the friction coupling element and oil temperature correlation data under conditions of no anomalies does not change significantly whether the oil temperature detection value is large or small. Therefore, by using oil temperature correlation data as an input variable to the mapping, the deviation in the accuracy of the aforementioned determination caused by the magnitude of the oil temperature detection value can be suppressed.

[0013] Method 3. The abnormality determination device of the power transmission device according to Method 2, wherein the value obtained by standardizing the oil temperature detection value is a standardized oil temperature detection value, and the processing circuit is configured to, in the associated data generation process, standardize multiple oil temperature detection values ​​included in the time series data of the oil temperature detection value, derive time series data of the standardized oil temperature detection values ​​including multiple standardized oil temperature detection values, and generate data representing the distribution of the magnitudes of the multiple standardized oil temperature detection values ​​included in the time series data of the standardized oil temperature detection value, as the oil temperature associated data.

[0014] Based on the above structure, the oil temperature correlation data represents the distribution of the magnitudes of the standardized oil temperature detection values ​​within the time-series data of standardized oil temperature detection values. In cases where no anomalies occur in the friction engagement element, and in cases where anomalies occur, there is a possibility that the magnitudes of the standardized oil temperature detection values ​​will deviate in different ways. Therefore, by using this oil temperature correlation data as the input variable for mapping, the aforementioned determination can be performed with high precision.

[0015] Method 4. An anomaly determination device for a power transmission device according to any one of methods 1 to 3, wherein the input variable includes vehicle speed.

[0016] The amount of action of the power transmission device at high vehicle speeds differs from that at low vehicle speeds. Furthermore, if the amount of action of the power transmission device changes, the temperature of the oil circulating within the power transmission device also changes. Therefore, based on the above structure, vehicle speed is used as the input variable for the mapping. That is, the output variable from this mapping is a variable that takes into account the vehicle speed at that moment. Therefore, by using such an output variable, the accuracy of the above determination can be improved.

[0017] Method 5. An abnormality determination device for a power transmission device according to any one of methods 1 to 4, wherein the power transmission device includes a clutch as the friction engagement element, and the input variable includes at least one of the following: the rotational speed of the input-side element of the clutch, the rotational speed of the output-side element of the clutch, and the difference in rotational speed between the input-side element and the output-side element.

[0018] For example, when the clutch is engaged, the heat generated by the clutch may vary depending on the rotational speed of the input-side components, the rotational speed of the output-side components, and the speed difference between the input-side and output-side components. Therefore, according to the above structure, at least one of the rotational speed of the input-side components, the rotational speed of the output-side components, and the speed difference is used as the input variable for the mapping. That is, the output variable from this mapping is a variable that takes into account at least one of the rotational speed of the input-side components, the rotational speed of the output-side components, and the speed difference. Therefore, by using such an output variable, the accuracy of the above determination can be improved.

[0019] Method 6. An abnormality determination device for a power transmission device according to any one of methods 1 to 5, wherein the speed difference between the input part that inputs torque to the friction engagement element and the output part that outputs torque from the friction engagement element is an input-output speed difference, and the input variable includes a calculated value of the heat generated by the friction engagement element calculated based on the product of the torque input to the friction engagement element and the input-output speed difference.

[0020] The calculated heat generation of the friction coupling element is based on the assumption that no abnormalities occur within the friction coupling element. Therefore, the relationship between the calculated heat generation and the oil temperature reading when an abnormality occurs in the friction coupling element may differ from the relationship when no abnormality occurs. Therefore, according to the above structure, the calculated heat generation of the friction coupling element is used as the input variable for the mapping. That is, the output variable from this mapping becomes the variable that considers the calculated heat generation. Therefore, by using such an output variable, the accuracy of the above determination can be improved.

[0021] Method 7. An anomaly determination device for a power transmission device according to any one of methods 1 to 6, wherein the input variable includes the engagement force of the friction engagement element.

[0022] For example, the heat generated by the friction coupling element can differ depending on whether the coupling force is large or small. For instance, when the coupling force is small and the friction coupling element slides together, more heat is generated compared to when the coupling force is large and the friction coupling element is fully engaged. Therefore, according to the above structure, the coupling force of the friction coupling element is used as the input variable for the mapping. That is, the output variable from this mapping becomes a variable that takes into account the coupling force of the friction coupling element. Therefore, by using such an output variable, the accuracy of the above determination can be improved.

[0023] Method 8. An anomaly determination device for a power transmission device according to any one of methods 1 to 7, wherein the input variable includes the component gap of the friction engagement element.

[0024] Generally speaking, the larger the component gap of a friction-bonded element, the easier it is for the bonding force to increase when the friction-bonded elements are joined. That is, the heat generated during the joining of the friction-bonded elements sometimes varies depending on the size of the component gap. Therefore, according to the above structure, the component gap of the friction-bonded element is used as an input variable for the mapping. That is, the output variable from this mapping becomes a variable that takes into account the component gap of the friction-bonded element. Therefore, by using such an output variable, the accuracy of the above determination can be improved.

[0025] Method 9. An anomaly determination device for a power transmission device according to any one of methods 1 to 8, wherein the input variable includes the detection value of an acceleration sensor mounted on the vehicle.

[0026] Vibrations occur in the power transmission device caused by the movement of the friction engagement elements when they are engaged or disengaged. These vibrations are detected by an onboard accelerometer. Furthermore, discrepancies sometimes arise in the accelerometer's readings depending on whether an anomaly occurs in the friction engagement elements or not. Therefore, according to the above structure, the accelerometer's readings are used as the input variable for the mapping. That is, the output variable from this mapping becomes a variable that takes into account the accelerometer's readings. Therefore, by using such an output variable, the accuracy of the aforementioned determination can be improved.

[0027] Method 10. An anomaly determination device for a power transmission device according to any one of methods 1 to 9, wherein the mapping output determines whether burn-out has occurred in the friction engagement element, and the processing circuit is configured to determine whether burn-out has occurred in the friction engagement element in the anomaly determination process based on the output variable output from the mapping obtained by inputting the input variable obtained in the acquisition process to the mapping.

[0028] In the case of burn-out within the friction coupling element when it is engaged, the change in heat generation of the friction coupling element compared to the case where no burn-out has occurred is observed. That is, the change in the detected oil temperature value. Therefore, by using the output variable derived from the above mapping, it is possible to determine whether burn-out has occurred within the friction coupling element.

[0029] Method 11. An anomaly determination device for a power transmission device according to any one of methods 1 to 10, wherein the mapping output determines whether a poor engagement has occurred in the friction engagement element, and the processing circuit is configured to determine, in the anomaly determination process, whether a poor engagement has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process to the mapping.

[0030] The change in heat generation of the friction coupling element during engagement, both when poor engagement occurs and when it does not, is reflected in the shift in the detected oil temperature. Therefore, by using the output variable derived from the above mapping, it is possible to determine whether poor engagement has occurred in the friction coupling element.

[0031] Method 12. An anomaly determination device for a power transmission device according to any one of methods 1 to 11, wherein the mapping output determines whether adhesion has occurred in the friction bonding element, and the processing circuit is configured to determine, in the anomaly determination process, whether adhesion has occurred in the friction bonding element based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process to the mapping.

[0032] If adhesion occurs within the friction coupling elements, they cannot be separated. Therefore, the heat generated by the friction coupling elements changes depending on whether adhesion has occurred or not. That is, the oil temperature reading changes over time. Therefore, by using the output variable derived from the above mapping, it is possible to determine whether adhesion has occurred within the friction coupling elements.

[0033] Method 13. An abnormality determination device for a power transmission device according to any one of methods 1 to 9, wherein the storage device stores a plurality of mapping data, the plurality of mapping data including a first mapping data, a second mapping data, and a third mapping data, wherein the first mapping data specifies that when the input variable is input, it outputs a mapping to determine whether burn-out has occurred in the friction joint element; the second mapping data specifies that when the input variable is input, it outputs a mapping to determine whether poor bonding has occurred in the friction joint element; and the third mapping data specifies that when the input variable is input, it outputs a mapping to determine whether adhesion has occurred in the friction joint element.

[0034] According to the above structure, the first mapping data is obtained through machine learning specifically designed to determine whether burn-off has occurred in the friction-bonded element. The second mapping data is obtained through machine learning specifically designed to determine whether poor bonding has occurred in the friction-bonded element. The third mapping data is obtained through machine learning specifically designed to determine whether adhesion has occurred in the friction-bonded element. In this case, by inputting the mapping input variable defined by the first mapping data and outputting the output variable from the mapping, it is possible to determine whether an anomaly caused by burn-off has occurred in the friction-bonded element. Furthermore, by inputting the mapping input variable defined by the second mapping data and outputting the output variable from the mapping, it is possible to determine whether an anomaly caused by poor bonding has occurred in the friction-bonded element. Furthermore, by inputting the mapping input variable defined by the third mapping data and outputting the output variable from the mapping, it is possible to determine whether an anomaly caused by adhesion has occurred in the friction-bonded element.

[0035] Method 14. An anomaly determination device for a power transmission device according to any one of Methods 1 to 12, wherein the storage device stores a plurality of mapping data individually corresponding to the operating states of each of the friction engagement elements, the plurality of mapping data including first mapping data and second mapping data, the first mapping data specifying a mapping that outputs an output variable determining whether an anomaly has occurred in the friction engagement element when the oil temperature-related data when the operating state of the friction engagement element is a first operating state is input as an input variable, and the second mapping data specifying a mapping that outputs an output variable determining whether an anomaly has occurred in the friction engagement element when the oil temperature-related data when the operating state of the friction engagement element is a second operating state different from the first operating state is input as an input variable, the processing circuit being configured to: perform a data selection process that selects the mapping data corresponding to the operating state of the friction engagement element from the plurality of mapping data stored in the storage device, and in the anomaly determination process, determine whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process to the mapping specified by the mapping data selected in the data selection process.

[0036] Even if an anomaly occurs in the friction engagement element, the oil temperature detection value may shift differently depending on the operating state of the friction engagement element. Therefore, according to the above structure, the first mapping data is obtained by machine learning specifically for the case where the operating state of the friction engagement element is the first operating state. The second mapping data is obtained by machine learning specifically for the case where the operating state of the friction engagement element is the second operating state. The mapping data corresponding to the current operating state is selected from multiple mapping data sets, and input variables are input for the mapping defined by that mapping data. Furthermore, based on the output variables from that mapping, it is determined whether an anomaly has occurred in the friction engagement element. By using the mapping data separately according to the operating state in this way, the accuracy of the above determination can be improved.

[0037] Method 15. An anomaly determination device for a power transmission device according to any one of methods 1 to 12, wherein the storage device stores a plurality of mapping data corresponding to the degree of annual variation of the characteristics of the power transmission device, and the processing circuit is configured to: perform a data selection process that selects the mapping data corresponding to the degree of annual variation of the characteristics of the power transmission device from the plurality of mapping data stored in the storage device; and in the anomaly determination process, determine whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping determined by inputting the input variable obtained in the acquisition process to the mapping specified by the mapping data selected in the data selection process.

[0038] For example, even if no abnormality occurs in the power transmission device, the heat generation of the power transmission device may vary depending on the degree of change in its characteristics over the years. Therefore, according to the above structure, mapping data is used separately according to the degree of change in the characteristics of the power transmission device over the years. That is, mapping data corresponding to the degree of change in the characteristics of the power transmission device over the years is selected, and input variables are input for the mapping specified by the mapping data. Moreover, based on the output variables output from the mapping, it is determined whether an abnormality has occurred in the friction engagement element. By using mapping data separately according to the degree of change in characteristics over the years, the accuracy of the above determination can be improved. Attached Figure Description

[0039] Figure 1 This is a diagram showing the control device and the drive system of the vehicle controlled by the control device in the first embodiment.

[0040] Figure 2 This is a schematic cross-sectional view of a portion of the torque converter with the lock-up clutch disengaged.

[0041] Figure 3This is a schematic cross-sectional view of a portion of the torque converter with the lock-up clutch engaged.

[0042] Figure 4 This is the first half of a flowchart illustrating a series of processes performed by the control device.

[0043] Figure 5 This is the latter half of a flowchart illustrating a series of processes performed by the control device.

[0044] Figure 6 It is a graph showing the shift in oil temperature readings.

[0045] Figure 7 It is a chart obtained by histogramizing the oil temperature correlation data.

[0046] Figure 8 It is a chart obtained by histogramizing the oil temperature correlation data.

[0047] Figure 9 This is a block diagram showing the storage device of the control device in the second embodiment.

[0048] Figure 10 This is the first half of a flowchart illustrating a series of processes performed by the control device.

[0049] Figure 11 This is the latter half of a flowchart illustrating a series of processes performed by the control device.

[0050] Figure 12 This is a block diagram showing the storage device of the control device in the third embodiment.

[0051] Figure 13 This is the first half of a flowchart illustrating a series of processes performed by the control device.

[0052] Figure 14 This is the latter half of a flowchart illustrating a series of processes performed by the control device. Detailed Implementation

[0053] (First Embodiment)

[0054] The following is in accordance with Figures 1-8 This describes the first embodiment of the abnormality detection device for the power transmission device.

[0055] First, the general structure of a vehicle equipped with an anomaly detection device will be described.

[0056] like Figure 1As shown, the vehicle VC includes an internal combustion engine 10, a transmission 40, and drive wheels 60. A torque converter 70 of the transmission 40 is connected to the crankshaft 11 of the internal combustion engine 10. An input shaft 81 of a transmission mechanism 80 is connected to the torque converter 70. A plurality of drive wheels 60 are connected to the output shaft 82 of the transmission mechanism 80 via a differential (not shown).

[0057] The torque converter 70 includes a front cover 71, a pump impeller 72, a turbine 73, and a stator 74. The crankshaft 11 of the internal combustion engine 10 is connected to the front cover 71. The pump impeller 72 rotates integrally with the front cover 71. The turbine 73 rotates integrally with the input shaft 81 of the transmission mechanism 40. The stator 74 amplifies the torque between the pump impeller 72 and the turbine 73. When the pump impeller 72 rotates according to the operation of the internal combustion engine, the rotation of the pump impeller 72 is transmitted to the turbine 73 via oil within the torque converter 70, and the output torque of the internal combustion engine 10 is input to the transmission mechanism 80.

[0058] The torque converter 70 has a lock-up clutch 75. When the lock-up clutch 75 is engaged, the pump wheel 72 and the turbine 73 are mechanically connected via the lock-up clutch 75. Therefore, when the lock-up clutch 75 is engaged, the output torque of the internal combustion engine 10 is transmitted from the front cover 71 to the turbine 73 via the lock-up clutch 75 and input to the transmission mechanism 80.

[0059] like Figure 2 as well as Figure 3 As shown, the lock-up clutch 75 includes a support 76 that rotates integrally with the turbine 73, multiple output-side elements 77 supported in a state where they can rotate integrally with the support 76, and multiple input-side elements 78 supported by the front cover 71. When the left-right direction, i.e., the extension direction of the input shaft 81 of the transmission mechanism 80, is defined as the axial direction AD, each input-side element 78 can slide relative to the front cover 71 in the axial direction AD. However, for the input-side element 78A closest to the turbine 73, the sliding movement towards the turbine 73 side (i.e., the right side in the figure) in the axial direction AD is restricted by the blocking member 79.

[0060] Each output-side element 77 is supported by the support 76 in a state that allows it to slide in the axial direction AD. Furthermore, friction material 77a is attached to both sides of each output-side element 77. Moreover, each output-side element 77 is arranged such that one output-side element 77 is interposed between adjacent input-side elements 78 in the axial direction AD.

[0061] By adjusting the hydraulic pressure in the pressure regulating region 70a within the torque converter 70, the lock-up clutch 75 can be engaged or disengaged. That is, by... Figure 2In the state shown, the hydraulic pressure in the pressure regulating region 70a is increased, and the input-side elements 78 (excluding input-side element 78A) and the output-side elements 77 slide towards the turbine 73 side (right side in the figure) in the axial direction AD. Thus, as... Figure 3 The lock-up clutch 75 shown is in an engaged state, with the input-side element 78 and output-side element 77, which are adjacent to each other in the axial direction AD, pressed against each other. On the other hand, in Figure 3 When the hydraulic pressure in the pressure regulating region 70a decreases in the state shown, the adjacent input-side element 78 and output-side element 77 separate. Thus, as... Figure 2 The lock-up clutch 75 shown is in the disengaged state.

[0062] Furthermore, the input-side element 78 that is furthest from the turbine 73 in the axial direction AD among all input-side elements 78 is designated as "input-side element 78B". This is done when the lock-up clutch 75 is in an operating state from... Figure 2 The separation state shown transitions to Figure 3 In the engaged state shown, the sliding movement of the input-side element 78B is greater than that of the other input-side elements 78 and each output-side element 77. In this embodiment, the sliding movement of the input-side element 78B when the operating state of the lock-up clutch 75 is changed from the disengaged state to the engaged state is referred to as the "assembly clearance PCtc of the lock-up clutch 75".

[0063] like Figure 1 As shown, the transmission mechanism 80 includes a first clutch C1, a second clutch C2, a brake mechanism B1, and a one-way clutch F1. Furthermore, the transmission stages of the transmission device 40 are switched by combining the engagement and disengagement states of the first clutch C1, the second clutch C2, and the brake mechanism B1 with the limiting and permissive states of the one-way clutch F1.

[0064] The vehicle VC includes an oil supply unit 50 that supplies oil to the transmission 40. The oil supply unit 50 includes an oil reservoir 51 for storing oil and a mechanically driven oil pump 52. The driven shaft 52a of the oil pump 52 is connected to the crankshaft 11 of the internal combustion engine 10. The oil pump 52 draws in oil from the oil reservoir 51 and discharges it to the transmission 40. The pressure of the oil discharged from the oil pump 52 is adjusted via a hydraulic control circuit 41 of the transmission 40. The hydraulic control circuit 41 includes multiple solenoid valves 41a. The hydraulic control circuit 41 controls the flow state and pressure of the oil by energizing each solenoid valve 41a.

[0065] The control device 90 controls the internal combustion engine 10 and operates various operating parts of the internal combustion engine 10 to control the torque and exhaust component ratio, which are the control quantities. Additionally, the control device 90 controls the transmission device 40 and operates the solenoid valves 41a of the hydraulic control circuit 41.

[0066] When controlling the aforementioned control quantity, the control device 90 refers to the output signal Scr of the crank angle sensor 101 and the output signal Sin of the input shaft rotation angle sensor 102, which detects the rotation angle of the input shaft 81 of the transmission 40. Additionally, the control device 90 refers to the oil temperature detection value Toil, which is the oil temperature detected by the oil temperature sensor 103; the vehicle speed SPD, which is the vehicle speed VC, detected by the vehicle speed sensor 104; and the vehicle acceleration G, which is the acceleration of the vehicle VC, detected by the acceleration sensor 105.

[0067] The control device 90 includes a CPU 91, a ROM 92, a storage device 93 as an electrically rewritable non-volatile memory, and peripheral circuitry 94, which can communicate via a local network 95. The peripheral circuitry 94 includes circuitry for generating clock signals that define internal operations, a power supply circuit, and a reset circuit. The control device 90 controls various control quantities by executing a program stored in the ROM 92 through the CPU 91.

[0068] The storage device 93 stores multiple mapping data DM1, DM2, and DM3. Each mapping data DM1, DM2, and DM3 includes data that specifies the mapping of the output variable corresponding to the input variable when various input variables are input (described later) and is obtained through machine learning.

[0069] However, when the transmission 40 is in operation, an abnormality sometimes occurs in the lock-up clutch 75 of the transmission 40. Examples of possible abnormalities in the lock-up clutch 75 include the following.

[0070] • Burnout of the lock-up clutch 75.

[0071] • The lock-up clutch 75 is not engaging properly.

[0072] ･The adhesion of the lock-up clutch 75.

[0073] When the lock-up clutch 75 is engaged, and in its normal state, the friction material 77a of the output element 77 is pressed against the input element 78. Both the input element 78 and the output element 77 are made of metal, so the friction material 77a is pressed against the metal (input element 78). However, as the friction material 77a wears down, the surface of the output element 77 becomes exposed. When the lock-up clutch 75 is engaged in this state, the output element 77 directly contacts the input element 78. That is, it is not the friction material that is pressed against the metal (input element 78), but rather the metal (output element 77) that is pressed against the metal (input element 78). As a result, there is a possibility of burn-out occurring in the lock-up clutch 75. The heat generated when the lock-up clutch 75 is pressed against metal is different from the heat generated when the friction material is pressed against metal.

[0074] When the lock-up clutch 75 is engaged or disengaged, the hydraulic pressure in the pressure regulating region 70a within the torque converter 70 is adjusted. For example, if an anomaly occurs in the hydraulic control circuit 41, there is a possibility that the hydraulic pressure in the pressure regulating region 70a cannot be properly adjusted, and the operation of the lock-up clutch 75 cannot be properly controlled. That is, when the lock-up clutch 75 is engaged, if the hydraulic pressure in the pressure regulating region 70a cannot be sufficiently increased, there is a possibility of poor engagement due to insufficient force applied to the input-side element 78 pressing the output-side element 77. When poor engagement occurs, the output-side element 77 slides relative to the input-side element 78, reducing the torque transmission efficiency via the lock-up clutch 75. On the other hand, when the engagement force is large, the sliding of the output-side element 77 relative to the input-side element 78 is suppressed. Furthermore, the heat generated by the lock-up clutch 75 when the output-side element 77 slides relative to the input-side element 78 is greater than the heat generated when the output-side element 77 does not slide relative to the input-side element 78.

[0075] When the locking clutch 75 is switched from the engaged state to the disengaged state, if the hydraulic pressure in the pressure regulating region 70a cannot be reduced due to an abnormality in the hydraulic control circuit 41, there is a possibility that the engaged state may continue. This situation, where the locking clutch 75 remains engaged even though it is desired to disengage, is called locking clutch 75 sticking. When locking clutch 75 sticks, it generates more heat than when the locking clutch 75 can normally disengage.

[0076] In the event of an abnormality as described above occurring in the lock-up clutch 75, the shift in oil temperature—the temperature of the oil circulating within the transmission 40—differs from the situation when no abnormality occurs in the lock-up clutch 75. Therefore, in this embodiment, the control device 90 determines whether an abnormality has occurred in the lock-up clutch 75 based on the shift in the oil temperature detection value Toil. At this time, the control device 90 uses the mapping data DM1, DM2, and DM3 stored in the storage device 93.

[0077] In this embodiment, mapping data DM1 is the data that defines the output variable Y(1) for determining whether burn-out has occurred in the lock-up clutch 75. Mapping data DM2 is the data that defines the output variable Y(2) for determining whether poor engagement has occurred in the lock-up clutch 75. Mapping data DM3 is the data that defines the output variable Y(3) for determining whether sticking has occurred in the lock-up clutch 75.

[0078] Reference Figure 4 as well as Figure 5 This describes a series of processes performed by the control device 90 to determine whether an abnormality has occurred in the lock-up clutch 75. Figure 4 as well as Figure 5 The series of processes shown is implemented by CPU 91 executing a program stored in ROM 92. This series of processes is executed repeatedly at a predetermined cycle. That is, when the elapsed time since the point when the series of processes was temporarily terminated reaches the time corresponding to the predetermined cycle, CPU 91 restarts the execution of the series of processes.

[0079] First, in step S11, CPU 91 sets the coefficient z to "1". In the next step S13, CPU 91 obtains the current oil temperature detection value Toil as the oil temperature detection value Toil(z). In the next step S15, CPU 91 increments the coefficient z by "1". Next, in step S17, CPU 91 determines whether the coefficient z is greater than the coefficient determination value zTh. In this embodiment, to determine whether an abnormality has occurred in the lock-up clutch 75, timing data of the oil temperature detection value Toil is used. The timing data of the oil temperature detection value Toil refers to data including multiple oil temperature detection values ​​Toil that are sequentially consecutive. The coefficient determination value zTh is set as the criterion for determining whether the acquisition of the required number of oil temperature detection values ​​Toil has been completed. If the coefficient z is less than or equal to the coefficient determination value zTh (S17: "No"), CPU 91 transfers the process to step S13. That is, the acquisition of the oil temperature detection value Toil continues. On the other hand, if the coefficient z is greater than the coefficient determination value zTh (S17: "Yes"), the timing data of the oil temperature detection value Toil, which consists of "z" oil temperature detection values ​​Toil, is obtained, so the CPU91 transfers the process to the next step S19.

[0080] In step S19, CPU91 standardizes the timing data of the oil temperature detection value Toil. For example, CPU91 takes the largest value among the multiple oil temperature detection values ​​Toil(1), Toil(2), ..., Toil(z) contained in the timing data of the oil temperature detection value Toil as the reference oil temperature detection value ToilB. Next, CPU91 standardizes each oil temperature detection value Toil(1), Toil(2), ..., Toil(z) by dividing each oil temperature detection value Toil(1), Toil(2), ..., Toil(z) by the reference oil temperature detection value ToilB. The standardized oil temperature detection values ​​Toil(1), Toil(2), ..., Toil(z) are called standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z). For example, the value obtained by dividing the oil temperature detection value Toil(1) by the reference oil temperature detection value ToilB is called the standardized oil temperature detection value ToilN(1). The data including the standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z) are also referred to as "time series data of standardized oil temperature detection values ​​ToilN".

[0081] Then, in the next step S21, the CPU91 generates oil temperature associated data RDToil based on the time-series data of the standardized oil temperature detection values ​​ToilN. In this embodiment, each standardized oil temperature detection value ToilN(1), ToilN(2), ..., ToilN(z) is greater than "0" and less than or equal to "1". Therefore, the region of values ​​from "0" to "1" is divided into multiple segments. For example, the region of values ​​from "0" to "1" is divided for every "0.2". Then, for each segmented region, the CPU91 counts the number of standardized oil temperature detection values ​​ToilN contained in the segmented region. For example, if there are "4" standardized oil temperature detection values ​​ToilN greater than "0.4" and less than or equal to "0.6" among the standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z), the CPU91 sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0.4" to "0.6" to "4". CPU91 calculates the oil temperature correlation data RDToil based on the count results for each segmented region. That is, the data representing the distribution of the magnitudes of multiple standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z) contained in the time series data of the standardized oil temperature detection value ToilN is the oil temperature correlation data RDToil.

[0082] For example, CPU91 sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0" to "0.2" as the count value Cnt (1), and sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0.2" to "0.4" as the count value Cnt (2). In addition, CPU91 sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0.4" to "0.6" as the count value Cnt (3), and sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0.6" to "0.8" as the count value Cnt (4). In addition, CPU91 sets the number of standardized oil temperature detection values ​​ToilN contained in the segmented region from "0.8" to "1" as the count value Cnt (5). That is, the oil temperature associated data RDToil is data composed of count values ​​Cnt (1), Cnt (2), Cnt (3), Cnt (4), and Cnt (5).

[0083] Here, Figure 6 The solid and dashed lines shown represent the time-series data of the oil temperature detection value Toil. Figure 7 According to the basis Figure 6 The graph is obtained by histogramting the oil temperature correlation data RDToil generated from the time-series data of the oil temperature detection value Toil (represented by dashed lines). Figure 8 According to the basis Figure 6 The graph is obtained by histogramting the oil temperature correlation data RDToil, which is generated from the time-series data of the oil temperature detection values ​​(represented by solid lines). Figure 6 The time-series data for the oil temperature reading Toil, represented by the dashed line, indicates that the oil temperature reading Toil rises slowly at a roughly constant rate. Therefore, in Figure 7 When the oil temperature correlation data RDToil is shown, the deviations of the count values ​​Cnt(1) to Cnt(5) are small. On the other hand, in Figure 6 The time-series data for the oil temperature reading Toil, represented by the solid line, shows how the rate of increase of the Toil reading varies along the way. Therefore, regarding... Figure 8 The oil temperature correlation data RDToil shown has a large deviation for each of the count values ​​Cnt(1) to Cnt(5).

[0084] Return to Figure 4 And in step S23, CPU91 obtains the vehicle speed SPD, internal combustion engine speed NE, input shaft speed Nat, speed difference ΔNtc, calculated heat generation value CVtc of lock-up clutch 75, engagement force EFtc of lock-up clutch 75, component clearance PCtc of lock-up clutch 75, and vehicle acceleration G. The internal combustion engine speed NE is either the crankshaft speed 11 calculated based on the output signal Scr of crank angle sensor 101, or the speed of input-side element 78 of lock-up clutch 75. The input shaft speed Nat is either the speed of input shaft 81 of transmission mechanism 80 calculated based on the output signal Sin of input shaft rotation angle sensor 102, or the speed of output-side element 77 of lock-up clutch 75. The speed difference ΔNtc is either the difference between internal combustion engine speed NE and input shaft speed Nat, or the speed difference between input-side element 78 and output-side element 77 in lock-up clutch 75. Furthermore, the speed difference ΔNtc can also be referred to as the speed difference between the front cover 71 that inputs torque to the input-side element 78 and the input shaft 81 that outputs torque from the output-side element 77. When the lock-up clutch 75 is engaged, the heat generation value CVtc of the lock-up clutch 75 is calculated based on the product of the input torque to the lock-up clutch 75 and the speed difference ΔNtc. The input torque to the lock-up clutch 75 is the torque input from the internal combustion engine 10 to the torque converter 70. On the other hand, when the lock-up clutch 75 is disengaged, the heat generation value CVtc is "0". The engagement force EFtc of the lock-up clutch 75 is the force that presses the output-side element 77 against the input-side element 78, which can be derived from the hydraulic pressure of the pressure regulating region 70a. The component clearance PCtc is a value measured during the factory inspection of the transmission 40 and is pre-stored in the storage device 93.

[0085] In the next step S25, the CPU91 substitutes the oil temperature correlation data RDToil generated in step S21 and various data obtained in step S23 into the input variables x(1)~x(13) used to determine whether an abnormality has occurred in the lock-up clutch 75. That is, the CPU91 substitutes the count value Cnt(1) of the oil temperature correlation data RDToil into the input variable x(1), the count value Cnt(2) into the input variable x(2), and the count value Cnt(3) into the input variable x(3). In addition, the CPU91 substitutes the count value Cnt(4) into the input variable x(4) and the count value Cnt(5) into the input variable x(5). In addition, the CPU91 substitutes the vehicle speed SPD into the input variable x(6), the internal combustion engine speed NE into the input variable x(7), and the input shaft speed Nat into the input variable x(8). In addition, CPU91 substitutes the rotational speed difference ΔNtc into the input variable x (9), the calculated heat value CVtc into the input variable x (10), and the engagement force EFtc into the input variable x (11). Then, CPU91 substitutes the component gap PCtc into the input variable x (12) and the vehicle acceleration G into the input variable x (13).

[0086] Next, in step S27, CPU 91 sets the decision coefficient mP to "1". In the following step S29, CPU 91 selects the mapping data corresponding to the decision coefficient mP from the mapping data DM1, DM2, and DM3 stored in the storage device 93. For example, CPU 91 selects mapping data DM1 when the decision coefficient mP is "1", selects mapping data DM2 when the decision coefficient mP is "2", and selects mapping data DM3 when the decision coefficient mP is "3".

[0087] Then, in step S31, CPU91 calculates the output variable Y(MP) by using the mapping input variables x(1)~x(13) specified by the selected mapping data.

[0088] In this embodiment, the mapping is configured as a fully connected feedforward neural network with one intermediate layer. The neural network includes an activation function h(x) that performs a nonlinear transformation on the input-side coefficients wFjk (j=0~n, k=0~13) and the output of the input-side linear mapping defined by the input-side coefficients wFjk. In this embodiment, the hyperbolic tangent "tanh(x)" is exemplified as the activation function h(x). Furthermore, the neural network includes an activation function f(x) that performs a nonlinear transformation on the output-side nonlinear mapping that performs a nonlinear transformation on the output-side coefficients wSj (j=0~n) and the output of the output-side linear mapping defined by the output-side coefficients wSj. In this embodiment, the hyperbolic tangent "tanh(x)" is exemplified as the activation function f(x). Additionally, the value n represents the dimension of the intermediate layer. In this embodiment, the value n is less than "13", which is the dimension of the input variable x. The input-side coefficients wFj0 ​​are bias parameters, becoming the coefficients of the input variable x(0). The input variable x(0) is defined as "1". Additionally, the output coefficient wS0 is a bias parameter.

[0089] The mapping data DM1 is a learned model obtained by learning from a vehicle with the same specifications as the vehicle VC before installation onto the vehicle VC. Here, during the learning of the mapping data DM1, training data including teacher data and input data is obtained beforehand. That is, time-series data of the oil temperature detection value Toil is obtained while the vehicle is actually being driven. Then, oil temperature correlation data RDToil is obtained as input data by performing the same processing as steps S19 and S21 described above on the time-series data of the oil temperature detection value Toil. Additionally, vehicle speed SPD, internal combustion engine speed NE, input shaft speed Nat, speed difference ΔNtc, calculated heat generation value CVtc of the lock-up clutch 75, engagement force EFtc of the lock-up clutch 75, and vehicle acceleration G are also obtained as input data. Furthermore, burn-out occurrence information, which indicates whether burn-out has occurred in the lock-up clutch 75, is obtained as teacher data. For example, burn-out occurrence information in cases of burn-out is set to "0", and burn-out occurrence information in cases of no burn-out is set to "1". In addition, the component clearance PCtc of the lock-up clutch 75 of the vehicle is also obtained as input data before the vehicle is driven.

[0090] Then, multiple training data are generated by driving the vehicle under various conditions. For example, a lock-up clutch, which has the potential to burn out during engagement, is attached to the vehicle, and the vehicle is driven. Then, if no burn-out occurs while the vehicle is driving, various input data for that scenario are obtained, and burn-out occurrence information (indicating no burn-out) is obtained as teacher data. Conversely, if burn-out occurs while the vehicle is driving, various input data for that scenario are obtained, and burn-out occurrence information (indicating burn-out) is obtained as teacher data.

[0091] Using multiple training datasets, the mapping data DM1 is learned. That is, the input-side variables and output-side variables are adjusted respectively so that the error between the output variable, which maps the input data to the output, and the actual burn-off information converges to below a predetermined value.

[0092] Similarly, the mapping data DM2 is a learned model obtained by learning from a vehicle with the same specifications as the vehicle VC before it is installed on the vehicle VC. Here, during the learning of the mapping data DM2, training data, including teacher data and input data, is also obtained beforehand. That is, various input data are obtained as described above by actually driving the vehicle. Additionally, at this time, engagement failure occurrence information, which serves as information on whether engagement failure occurs in the lock-up clutch 75, is obtained as teacher data. For example, engagement failure occurrence information in the case of engagement failure is set to "0", and engagement failure occurrence information in the case of no engagement failure is set to "1".

[0093] Then, by driving the vehicle under various conditions, multiple training data sets, including teacher data and input data, are generated. For example, a lock-up clutch, which has the potential to cause mis-engagement, is fitted to the vehicle, and the vehicle is driven. Then, when the lock-up clutch is engaged while the vehicle is moving and no mis-engagement occurs, various input data for the case where no mis-engagement occurs can be obtained, and mis-engagement occurrence information indicating that no mis-engagement occurred can be obtained as teacher data. Furthermore, when the lock-up clutch is engaged while the vehicle is moving and mis-engagement occurs, various input data for the case where mis-engagement occurs can be obtained, and mis-engagement occurrence information indicating that mis-engagement occurred can be obtained as teacher data.

[0094] Using multiple training datasets, the mapping data DM2 is learned. That is, the input-side variables and output-side variables are adjusted separately in a way that the error between the output variable, which maps the input data to the output, and the actual information on the occurrence of misalignment converges to below a predetermined value.

[0095] Similarly, the mapping data DM3 is a learned model obtained by learning from a vehicle with the same specifications as the vehicle VC before it is installed on the vehicle VC. Here, during the learning of the mapping data DM3, training data, including teacher data and input data, is also obtained beforehand. That is, various input data are obtained by actually driving the vehicle. Additionally, at this time, adhesion occurrence information, which serves as information about whether adhesion has occurred in the lock-up clutch 75, is obtained as teacher data. For example, the adhesion occurrence information when adhesion has occurred is set to "0", and the adhesion occurrence information when adhesion has not occurred is set to "1".

[0096] Then, by driving the vehicle under various conditions, multiple training data sets, including teacher data and input data, are generated. For example, a lock-up clutch, which has the potential to stick, is mounted on the vehicle, and the vehicle is driven. Then, while the vehicle is driving, if sticking does not occur when the lock-up clutch is disengaged, various input data for the case where sticking does not occur can be obtained, and sticking occurrence information indicating that sticking has not occurred can be obtained as teacher data. Furthermore, if sticking occurs while the vehicle is driving and the lock-up clutch is disengaged, various input data for the case where sticking occurs can be obtained, and sticking occurrence information indicating that sticking has occurred can be obtained as teacher data.

[0097] Using multiple training datasets, the mapping data DM3 is learned. That is, the input-side variables and output-side variables are adjusted separately in a way that the error between the output variable, which maps the input data to the output, and the actual bonding information converges to below a predetermined value.

[0098] After calculating the output variable Y(MP) in step S31, in the next step S33, the CPU91 evaluates the output variable Y(MP) calculated in step S31. That is, the CPU91 determines whether an abnormality has occurred in the lock-up clutch 75 based on the output variable Y(MP). For example, when the determination coefficient mP is "1", the CPU91 determines that burn-out has occurred in the lock-up clutch 75 when the output variable Y(1) is below the abnormality determination value. On the other hand, the CPU91 does not determine that burn-out has occurred when the output variable Y(1) is greater than the abnormality determination value. In addition, for example, when the determination coefficient mP is "2", the CPU91 determines that poor engagement has occurred in the lock-up clutch 75 when the output variable Y(2) is below the abnormality determination value. On the other hand, the CPU91 does not determine that poor engagement has occurred when the output variable Y(2) is greater than the abnormality determination value. In addition, for example, when the determination coefficient mP is "3", the CPU91 determines that sticking has occurred in the lock-up clutch 75 when the output variable Y(3) is below the abnormality determination value. On the other hand, when the output variable Y(3) is greater than the abnormality judgment value, the CPU91 does not determine that adhesion has occurred. That is, when the output variable Y is below the abnormality judgment value, it can be determined that there is a high probability that an abnormality has occurred in the lock-up clutch 75, so it can be determined that an abnormality has occurred.

[0099] In the next step S35, if the evaluation result determines that an abnormality has occurred in the lock-up clutch 75 ("Yes"), the CPU 91 transfers the process to the next step S37. In step S37, the CPU 91 stores the meaning of the abnormality in the storage device 93. For example, if the output variable Y(1) is below the abnormality determination value, the CPU 91 stores the meaning of burn-out in the storage device 93. In addition, if, for example, the output variable Y(2) is below the abnormality determination value, the CPU 91 stores the meaning of poor engagement in the storage device 93. In addition, if, for example, the output variable Y(3) is below the abnormality determination value, the CPU 91 stores the meaning of sticking in the storage device 93. Then, the CPU 91 transfers the process to step S39.

[0100] On the other hand, in step S35, if the evaluation result does not determine that an abnormality has occurred in the lock-up clutch 75 ("No"), the CPU91 transfers the process to the next step S39.

[0101] In step S39, CPU 91 determines whether the determination coefficient mP is greater than or equal to "3". If the determination coefficient mP is greater than or equal to "3", all three abnormal determinations related to the lock-up clutch 75 are completed. On the other hand, if the determination coefficient mP is less than "3", there are incomplete abnormal determinations among the three abnormal determinations related to the lock-up clutch 75. Therefore, if the determination coefficient mP is less than "3" (S39: "No"), CPU 91 transfers the process to step S41. In step S41, CPU 91 increments the determination coefficient mP by "1", transferring the process to step S29. On the other hand, if the determination coefficient mP is greater than or equal to "3" ("Yes") in step S39, CPU 91 temporarily terminates the series of processes.

[0102] Explain the function of this implementation method.

[0103] exist Figure 6 The time-series data for the oil temperature reading Toil, represented by the dashed line, indicates the shift of the oil temperature reading Toil even when the lock-up clutch 75 is engaged, without any abnormalities occurring. Figure 6 The time-series data of the oil temperature detection value Toil, represented by a solid line, indicates the shift of the oil temperature detection value Toil when burnout occurs while the lock-up clutch 75 is engaged. In this embodiment, when the time-series data of the oil temperature detection value Toil is obtained, a process is generated based on the time-series data of the oil temperature detection value Toil as follows: Figure 7 as well as Figure 8 The oil temperature correlation data RDToil is shown. That is, the generated oil temperature correlation data RDToil is different depending on whether burn-out occurred during the acquisition of the time series data. That is, the magnitude of the values ​​in each count value Cnt(1)~Cnt(5) of the oil temperature correlation data RDToil is different.

[0104] Furthermore, even in the oil temperature correlation data RDToil based on the timing data of the oil temperature detection value Toil obtained when poor engagement occurs while the lock-up clutch 75 is engaged, and the oil temperature correlation data RDToil based on the timing data of the oil temperature detection value Toil obtained when no abnormality occurs, there are differences in the deviation of the magnitude of each count value Cnt(1) to Cnt(5).

[0105] Furthermore, even in the oil temperature correlation data RDToil based on the timing data of the oil temperature detection value Toil obtained when the locked-up clutch 75 that has become stuck is disengaged, and the oil temperature correlation data RDToil based on the timing data of the oil temperature detection value Toil obtained when no abnormality occurs, there are differences in the deviation of the magnitude of each count value Cnt(1) to Cnt(5).

[0106] When the count values ​​Cnt(1)~Cnt(5) of the oil temperature associated data RDToil are input as input variables to the mapping specified by the mapping data DM1, DM2, and DM3, the mapping outputs the output variable Y(MP) corresponding to the oil temperature associated data RDToil. Then, based on the output variable Y(MP), it is determined whether an anomaly has occurred.

[0107] According to this embodiment, the following effects can be obtained.

[0108] (1-1) In this embodiment, the mapping data of the oil temperature correlation data RDToil as input variable and the output variable Y that determines whether an abnormality has occurred in the lock-up clutch 75 is stored in the storage device 93. Then, when the transmission device 40 is activated, it is determined whether an abnormality has occurred in the lock-up clutch 75 based on the output variable Y output from the mapping by inputting the acquired input variable into the mapping. Therefore, it is possible to determine whether an abnormality has occurred in the lock-up clutch 75 without using the detection value of the odor sensor.

[0109] In addition, in vehicles where the odor sensor is not installed in the vehicle VC, it is also possible to determine whether an abnormality has occurred in the lock-up clutch 75.

[0110] When using odor sensors to determine anomalies, the anomaly cannot be detected until the composition of the odor emitted by the oil deteriorates. In contrast, in this embodiment, the anomaly can be detected by timing the change in the oil temperature reading (Toil). Therefore, early detection of anomalies is possible.

[0111] (1-2) In this embodiment, by standardizing the time-series data of the oil temperature detection value Toil, time-series data of standardized oil temperature detection values ​​ToilN, including multiple standardized oil temperature detection values ​​ToilN, is obtained. Then, oil temperature correlation data RDToil is generated based on the time-series data of the standardized oil temperature detection values ​​ToilN. Therefore, the degree of difference between the oil temperature correlation data RDToil when an anomaly occurs in the lock-up clutch 75 and the oil temperature correlation data RDToil when no anomaly occurs does not change much when the oil temperature detection value Toil is large or small. Therefore, by using the oil temperature correlation data RDToil as the input variable for mapping, the deviation in the accuracy of the determination caused by the magnitude of the oil temperature detection value Toil can be suppressed.

[0112] (1-3) The more data points in the time-series data of the standardized oil temperature detection value ToilN, the higher the accuracy of the determination. In this embodiment, the oil temperature correlation data RDToil, which serves as the input variable for mapping, is obtained by histogramting the time-series data of the standardized oil temperature detection value ToilN. Therefore, even with a large number of time-series data points, the data capacity of the oil temperature correlation data RDToil is not very large. Thus, high-accuracy determination can be achieved even with a small amount of data.

[0113] (1-4) The amount of operation of the transmission 40 when the vehicle speed SPD is high differs from the amount of operation of the transmission 40 when the vehicle speed SPD is low. Furthermore, if the amount of operation of the transmission 40 changes, the temperature of the oil circulating within the transmission 40 also changes. Therefore, in this embodiment, the vehicle speed SPD is used as the input variable for mapping. That is, the output variable Y output from this mapping is a variable that takes into account the vehicle speed SPD at that time. Therefore, by using such an output variable Y, the accuracy of determining whether an abnormality has occurred in the lock-up clutch 75 can be improved.

[0114] (1-5) If no abnormality occurs in the lock-up clutch 75, the temperature shift when the lock-up clutch 75 is engaged can be estimated to some extent. Similarly, the temperature shift when the lock-up clutch 75 is disengaged, the temperature shift when transitioning from disengaged to engaged, and the temperature shift when transitioning from engaged to disengaged can also be estimated to some extent.

[0115] Here, the timing for transitioning the lock-up clutch 75 from the disengaged state to the engaged state, and the timing for transitioning it from the engaged state to the disengaged state, are determined by the vehicle speed SPD. That is, the operating state of the lock-up clutch 75 is controlled according to the vehicle speed SPD. Therefore, by comparing the shift in oil temperature estimated based on the operating state of the lock-up clutch 75 determined according to the vehicle speed SPD, with the shift in the oil temperature detection value Toil, it is possible to infer whether an abnormality has occurred in the lock-up clutch 75.

[0116] Therefore, as described in this embodiment, by adding the vehicle speed SPD to the mapped input variable, the accuracy of determining whether an abnormality has occurred in the lock-up clutch 75 can be improved.

[0117] (1-6) For example, when the lock-up clutch 75 is engaged, the heat generated by the lock-up clutch 75 may vary depending on the rotational speed of the input-side element 78, the rotational speed of the output-side element 77, and the speed difference between the input-side element 78 and the output-side element 77. Therefore, in this embodiment, the internal combustion engine speed NE, which corresponds to the rotational speed of the input-side element 78, the input shaft speed Nat, which corresponds to the rotational speed of the output-side element 77, and the speed difference ΔNtc are used as input variables for the mapping. That is, the output variable Y output from this mapping is a variable that takes into account the internal combustion engine speed NE, the input shaft speed Nat, and the speed difference ΔNtc. Therefore, by using such an output variable Y, the accuracy of the determination of whether an abnormality has occurred in the lock-up clutch 75 can be improved.

[0118] (1-7) The calculated heat generation value CVtc of the lock-up clutch 75 is calculated under the premise that no abnormality has occurred in the lock-up clutch 75. Therefore, the relationship between the calculated heat generation value CVtc and the oil temperature detection value Toil under abnormal conditions may differ from the relationship between the calculated heat generation value CVtc and the oil temperature detection value Toil under normal conditions. Therefore, in this embodiment, the calculated heat generation value CVtc is used as the input variable for the mapping. That is, the output variable Y from this mapping becomes a variable that takes into account the calculated heat generation value CVtc. Therefore, by using such an output variable Y, the accuracy of the determination of whether an abnormality has occurred can be improved.

[0119] (1-8) When the lock-up clutch 75 is engaged, the heat generated by the lock-up clutch 75 sometimes differs depending on whether the engagement force EFtc is large or small. For example, when the engagement force EFtc is small and the lock-up clutch 75 is slipping, the lock-up clutch 75 generates more heat than when the engagement force EFtc is large and the lock-up clutch 75 is fully engaged. Therefore, in this embodiment, the engagement force EFtc is used as the input variable for the mapping. That is, the output variable Y from this mapping becomes a variable that takes into account the engagement force EFtc. Therefore, by using such an output variable Y, the accuracy of the determination of whether an abnormality has occurred in the lock-up clutch 75 can be improved.

[0120] (1-9) The larger the component clearance PCtc of the lock-up clutch 75, the easier it is for the engagement force to increase when the lock-up clutch 75 is engaged. That is, the heat generated by the lock-up clutch 75 when it is engaged varies depending on the size of the component clearance PCtc. Therefore, in this embodiment, the component clearance PCtc is used as an input variable for mapping. That is, the output variable Y from this mapping becomes a variable that takes into account the component clearance PCtc. Therefore, by using such an output variable Y to determine whether an abnormality has occurred in the lock-up clutch 75, the accuracy of the determination can be improved.

[0121] (1-10) When the lock-up clutch 75 is moved from the disengaged state to the engaged state or from the engaged state to the disengaged state, vibrations occur in the transmission device 40 due to changes in the operating state of the lock-up clutch 75. Such vibrations can be detected by the acceleration sensor 105. Furthermore, in cases where an abnormality occurs in the lock-up clutch 75 and in cases where no abnormality occurs, there may be a difference in the vehicle acceleration G detected by the acceleration sensor 105. Therefore, in this embodiment, the vehicle acceleration G is used as the input variable for mapping. That is, the output variable Y from this mapping becomes a variable that takes into account the vehicle acceleration G. Therefore, by using such an output variable Y to determine whether an abnormality has occurred, the accuracy of the determination can be improved.

[0122] (1-11) The change in heat generation of the lock-up clutch 75 in the case of burn-out and the case of no burn-out. That is, the change in the oil temperature detection value Toil. Therefore, by using the output variable Y, which is output from the mapping data DM1 and is input to the mapping input variable specified by the mapping data DM1, it is possible to determine whether burn-out has occurred.

[0123] (1-12) When the lock-up clutch 75 is engaged, the heat generated by the lock-up clutch 75 changes when poor engagement occurs and when poor engagement does not occur. That is, the change in the oil temperature detection value Toil. Therefore, by using the output variable Y, which is output from the mapping data DM2, it is possible to determine whether poor engagement has occurred.

[0124] (1-13) If sticking occurs in the lock-up clutch 75, the lock-up clutch 75 cannot be disengaged. Therefore, the heat generated by the lock-up clutch 75 changes depending on whether sticking occurs. That is, the oil temperature detection value Toil changes over time. Therefore, whether sticking has occurred can be determined by using the output variable from the mapping data DM3, which is the input variable specified by the mapping data DM3.

[0125] (1-14) In this embodiment, mapping data DM1, DM2, and DM3 are prepared for each type of abnormality in the lock-up clutch 75. Then, mapping data DM1 is used when determining whether burning has occurred. Mapping data DM2 is used when determining whether poor engagement has occurred. Mapping data DM3 is used when determining whether sticking has occurred. By using mapping data DM1, DM2, and DM3 separately according to the type of determination, the accuracy of each determination can be improved.

[0126] (Second Implementation)

[0127] Hereinafter, focusing on the differences from the first embodiment, the second embodiment will be described with reference to the accompanying drawings.

[0128] like Figure 9 As shown, in this embodiment, the storage device 93 stores multiple mapping data DM11, DM12, DM13, and DM14, each corresponding individually to the operating state of the lock-up clutch 75. Mapping data DM11 is obtained through machine learning specifically for the case where the operating state of the lock-up clutch 75 is disengaged. Mapping data DM12 is obtained through machine learning specifically for the case where the operating state of the lock-up clutch 75 is engaged. Mapping data DM13 is obtained through machine learning specifically for the case where the operating state of the lock-up clutch 75 is engaged. Mapping data DM14 is obtained through machine learning specifically for the case where the operating state of the lock-up clutch 75 is disengaged. The engagement transition state refers to the operating state of the lock-up clutch 75 when transitioning from the disengaged state to the engaged state. The disengagement transition state refers to the operating state of the lock-up clutch 75 when transitioning from the engaged state to the disengaged state.

[0129] Reference Figure 10 as well as Figure 11 This describes a series of processes performed by the control device 90 to determine whether an abnormality has occurred in the lock-up clutch 75. Figure 10 as well as Figure 11 The series of processes shown is implemented by CPU 91 executing a program stored in ROM 92. This series of processes is executed repeatedly at a predetermined cycle. That is, when the elapsed time since the point when the series of processes was temporarily terminated reaches the time corresponding to the predetermined cycle, CPU 91 restarts the execution of the series of processes.

[0130] Initially, in step S51, CPU 91 sets the coefficient z to "1". In the next step S53, CPU 91 obtains the current oil temperature detection value Toil as the oil temperature detection value Toil(z). In the next step S55, CPU 91 increments the coefficient z by "1". Next, in step S57, CPU 91 determines whether the coefficient z is greater than the aforementioned coefficient determination value zTh. If the coefficient z is less than or equal to the coefficient determination value zTh (S57: "No"), CPU 91 transfers the process to step S53. On the other hand, if the coefficient z is greater than the coefficient determination value zTh (S57: "Yes"), CPU 91 transfers the process to the next step S59.

[0131] In step S59, CPU91, similarly to step S19 above, acquires time-series data of standardized oil temperature detection values ​​ToilN, including multiple standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z). In the next step S61, similarly to step S21 above, CPU91 generates oil temperature correlation data RDToil based on the time-series data of standardized oil temperature detection values ​​ToilN. Next, in step S63, similarly to step S23 above, CPU91 acquires vehicle speed SPD, internal combustion engine speed NE, input shaft speed Nat, speed difference ΔNtc, calculated heat generation value CVtc of lock-up clutch 75, engagement force EFtc of lock-up clutch 75, component clearance PCtc of lock-up clutch 75, and vehicle acceleration G. Then, in the next step S65, CPU91, in the same way as in step S25 above, substitutes the oil temperature correlation data RDToil calculated in step S61 and various data obtained in step S63 into the input variables x(1)~x(13) used to determine whether an abnormality has occurred in the lock-up clutch 75.

[0132] In the next step S67, CPU91 acquires the operating state of the lock-up clutch 75. That is, CPU91 selects the current operating state of the lock-up clutch 75 from the disengagement state, engagement transition state, engagement state, and disengagement transition state. The "current operating state of the lock-up clutch 75" referred to here means the operating state that CPU91 grasps in terms of control. Therefore, if an anomaly occurs in the lock-up clutch 75, the operating state acquired here may differ from the actual operating state.

[0133] Next, in step S69, CPU 91 sets the value corresponding to the operation state obtained in step S67 as the state coefficient SV. For example, when the operation state is the separation state, CPU 91 sets "1" as the state coefficient SV, and when the operation state is the engagement / transfer state, it sets "2" as the state coefficient SV. Alternatively, for example, when the operation state is the engagement state, CPU 91 sets "3" as the state coefficient SV, and when the operation state is the separation / transfer state, it sets "4" as the state coefficient SV.

[0134] In the next step S71, CPU 91 selects the mapping data corresponding to the state coefficient SV from the mapping data DM11, DM12, DM13, and DM14 stored in storage device 93. For example, CPU 91 selects mapping data DM11 when the state coefficient SV is "1", selects mapping data DM12 when the state coefficient SV is "2", selects mapping data DM13 when the state coefficient SV is "3", and selects mapping data DM14 when the state coefficient SV is "4".

[0135] Then, in step S73, CPU91 calculates the output variable Y(SV) by using the mapping input variables x(1)~x(13) specified by the selected mapping data.

[0136] The mapping data DM11 is a learned model obtained by learning from a vehicle with the same specifications as the vehicle VC before installation onto the vehicle VC. Here, during the learning of the mapping data DM11, training data including teacher data and input data is obtained beforehand. That is, various input data are obtained by driving the vehicle with the lock-up clutch disengaged. Furthermore, anomaly occurrence information, which indicates whether an anomaly has occurred in the lock-up clutch, is obtained as teacher data. In this embodiment, the type of anomaly, i.e., burn-out, poor engagement, or adhesion, is not distinguished. For example, the anomaly occurrence information in the case of an anomaly in the lock-up clutch is set to "0", and the anomaly determination information in the case of no anomaly is set to "1".

[0137] Then, multiple training data sets are generated by driving the vehicle under various conditions. For example, a lock-up clutch, which has the potential to cause an anomaly, is attached to the vehicle, and the vehicle is driven. Then, when no anomaly occurs during vehicle operation, various input data for the absence of an anomaly can be obtained, and anomaly occurrence information indicating that no anomaly occurred can be obtained as teacher data. Furthermore, when an anomaly occurs during vehicle operation, various input data for the occurrence of the anomaly can be obtained, and anomaly occurrence information indicating that an anomaly occurred can be obtained as teacher data.

[0138] Using multiple training datasets, the mapping data DM11 is learned. That is, the input-side variables and output-side variables are adjusted respectively so that the error between the output variable, which maps the input data to the output, and the actual anomaly detection information converges to below a predetermined value.

[0139] Similarly, the mapping data DM12, DM13, and DM14 are also learned models obtained using vehicle learning with the same specifications as the vehicle VC before being installed on the vehicle VC. When learning the mapping data DM12, input data is obtained based on various data acquired during vehicle operation when the lock-up clutch is engaged, and anomaly occurrence information at this time is obtained as teacher data. This generates multiple training datasets. Using these multiple training datasets, the mapping data DM12 is learned. That is, the input-side variables and output-side variables are adjusted respectively so that the error between the output variable mapped from the input data and the actual anomaly determination information converges to below a predetermined value.

[0140] When learning the mapping data DM13, input data is obtained based on various data acquired during vehicle operation when the lock-up clutch is engaged, and anomaly occurrence information at this time is obtained as teacher data. This generates multiple training datasets. Using these multiple training datasets, the mapping data DM13 is learned. That is, the input-side variables and output-side variables are adjusted respectively in a way that the error between the output variable mapped from the input data and the actual anomaly determination information converges to a predetermined value.

[0141] When learning the mapping data DM14, input data is obtained based on various data acquired during vehicle operation when the lock-up clutch is in a disengaged state, and anomaly information at this time is obtained as teacher data. This generates multiple training datasets. Using these multiple training datasets, the mapping data DM14 is learned. That is, the input-side variables and output-side variables are adjusted respectively in a way that the error between the output variable mapped from the input data and the actual anomaly determination information converges to below a predetermined value.

[0142] Next, in step S75, CPU 91 evaluates the output variable Y(SV) calculated in step S73. That is, CPU 91 determines whether the output variable Y(SV) is less than or equal to the above-mentioned anomaly determination value. If the output variable Y(SV) is less than or equal to the anomaly determination value, it is considered that an anomaly has occurred in the lock-up clutch 75. On the other hand, if the output variable Y(SV) is greater than the anomaly determination value, it is not considered that an anomaly has occurred in the lock-up clutch 75. In the next step S77, if the evaluation result determines that an anomaly has occurred in the lock-up clutch 75 ("Yes"), CPU 91 transfers the process to the next step S79. In step S79, CPU 91 stores the meaning of the anomaly in the storage device 93. After that, CPU 91 temporarily terminates the series of processes.

[0143] On the other hand, in step S77, if the evaluation result does not determine that an abnormality has occurred in the lock-up clutch 75 ("No"), the CPU91 temporarily terminates the series of processes. That is, for example, if the output variable Y (SV) is greater than the abnormality determination value, the CPU91 does not execute the process of step S79 and temporarily terminates the series of processes.

[0144] According to this embodiment, in addition to the effects described in (1-1) to (1-10) above, the following effects can also be obtained.

[0145] (2-1) Even if an abnormality occurs in the lock-up clutch 75, the shift in the oil temperature detection value Toil may differ depending on the operating state of the lock-up clutch 75. Therefore, in this embodiment, multiple mapping data DM11, DM12, DM13, and DM14 are prepared in advance, each corresponding individually to the operating state of the lock-up clutch 75. From these multiple mapping data DM11, DM12, DM13, and DM14, the mapping data corresponding to the current operating state is selected, and the input variable is input for the mapping specified by this mapping data. Then, based on the output variable Y(SV) output from this mapping, it is determined whether an abnormality has occurred in the lock-up clutch 75. By using the mapping data separately according to the operating state in this way, the accuracy of the determination can be improved.

[0146] (Third implementation)

[0147] Hereinafter, focusing on the differences from the first and second embodiments, the third embodiment will be described with reference to the accompanying drawings.

[0148] like Figure 12As shown, in this embodiment, the storage device 93 stores multiple mapping data DM21, DM22, DM23, ... corresponding to the degree of annual variation in the characteristics of the transmission device 40. Mapping data DM11 is the mapping data for the case where the degree of annual variation in the characteristics is minimal. Mapping data DM12 is the mapping data for the case where the degree of annual variation in the characteristics is second minimal. Mapping data DM13 is the mapping data for the case where the degree of annual variation in the characteristics is third minimal.

[0149] Reference Figure 13 as well as Figure 14 This describes a series of processes performed by the control device 90 to determine whether an abnormality has occurred in the lock-up clutch 75. Figure 13 as well as Figure 14 The series of processes shown is implemented by CPU 91 executing a program stored in ROM 92. This series of processes is executed repeatedly at a predetermined cycle. That is, when the elapsed time since the point when the series of processes was temporarily terminated reaches the time corresponding to the predetermined cycle, CPU 91 restarts the execution of the series of processes.

[0150] Initially, in step S91, CPU 91 sets the coefficient z to "1". In the next step S93, CPU 91 obtains the current oil temperature detection value Toil as the oil temperature detection value Toil(z). In the next step S95, CPU 91 increments the coefficient z by "1". Next, in step S97, CPU 91 determines whether the coefficient z is greater than the aforementioned coefficient determination value zTh. If the coefficient z is less than or equal to the coefficient determination value zTh (S97: "No"), CPU 91 transfers the process to step S93. On the other hand, if the coefficient z is greater than the coefficient determination value zTh (S97: "Yes"), CPU 91 transfers the process to the next step S99.

[0151] In step S99, CPU91, similarly to step S19 above, acquires time-series data of standardized oil temperature detection values ​​ToilN, including multiple standardized oil temperature detection values ​​ToilN(1), ToilN(2), ..., ToilN(z). In the next step S101, similarly to step S21 above, CPU91 generates oil temperature correlation data RDToil based on the time-series data of standardized oil temperature detection values ​​ToilN. Next, in step S103, similarly to step S23 above, CPU91 acquires vehicle speed SPD, internal combustion engine speed NE, input shaft speed Nat, speed difference ΔNtc, calculated heat generation value CVtc of lock-up clutch 75, engagement force EFtc of lock-up clutch 75, component clearance PCtc of lock-up clutch 75, and vehicle acceleration G. Then, in the next step S105, CPU91, in the same way as in step S25 above, substitutes the oil temperature correlation data RDToil calculated in step S101 and various data obtained in step S103 into the input variables x(1)~x(13) used to determine whether an abnormality has occurred in the lock-up clutch 75.

[0152] In the next step S107, CPU91 obtains the annual variation coefficient AGI, which is a coefficient corresponding to the annual variation of the characteristics of the transmission device 40. For example, CPU91 obtains the annual variation coefficient AGI based on the travel distance of the vehicle VC. In this case, CPU91 sets a larger value as the annual variation coefficient AGI if the travel distance of the vehicle VC is longer.

[0153] In the next step S109, the CPU 91 selects the mapping data corresponding to the annual variation coefficient AGI from the mapping data DM21, DM22, DM23, ... stored in the storage device 93. For example, when the annual variation coefficient AGI is "1", the CPU 91 selects the mapping data DM21, and when the annual variation coefficient AGI is "2", it selects the mapping data DM22.

[0154] Then, in step S111, CPU91 calculates the output variable Y(AGI) by using the mapping input variables x(1)~x(13) specified by the selected mapping data.

[0155] The mapping data DM21 is a learned model obtained by learning from a vehicle with the same specifications as the vehicle VC before it is installed on the vehicle VC. Here, during the learning of the mapping data DM21, training data including teacher data and input data is obtained beforehand. That is, various input data are obtained by actually driving the vehicle at a distance such that the annual variation coefficient AGI is set to "1". Furthermore, anomaly occurrence information, which indicates whether an anomaly has occurred in the lock-up clutch, is obtained as teacher data. In this embodiment, the type of anomaly, i.e., burn-out, poor engagement, or adhesion, is not distinguished. For example, the anomaly occurrence information in the case of an anomaly in the lock-up clutch is set to "0", and the anomaly determination information in the case of no anomaly is set to "1".

[0156] Then, multiple training data sets are generated by driving the vehicle under various conditions. For example, a lock-up clutch, which has the potential to cause an anomaly, is attached to the vehicle, and the vehicle is driven. Then, when no anomaly occurs during vehicle operation, various input data for the absence of an anomaly can be obtained, and anomaly occurrence information indicating that no anomaly occurred can be obtained as teacher data. Furthermore, when an anomaly occurs during vehicle operation, various input data for the occurrence of the anomaly can be obtained, and anomaly occurrence information indicating that an anomaly occurred can be obtained as teacher data.

[0157] Using multiple training datasets, the mapping data DM21 is learned. That is, the input-side variables and output-side variables are adjusted respectively in a way that the error between the output variable, which maps the input data to the output, and the actual anomaly detection information converges to below a predetermined value.

[0158] Similarly, the mapping data DM22, DM23, ... are also learned models obtained by learning from vehicles with the same specifications as the vehicle VC before being installed on the vehicle VC. When learning the mapping data DM22, various input data are obtained by actually driving the vehicle at distances where the annual variation coefficient AGI is set to "2," and the anomaly occurrence information at this time is obtained as teacher data. Then, multiple training data sets including the input data and teacher data are generated. Using these multiple training data sets, the mapping data DM22 is learned. That is, the input-side variables and output-side variables are adjusted respectively so that the error between the output variable mapped from the input data and the actual anomaly determination information converges to below a predetermined value.

[0159] In the case of learning the mapping data DM23, various input data are obtained by actually driving vehicles with a driving distance such as an annual variation coefficient (AGI) of "3". Anomaly occurrence information at this time is obtained as teacher data. Then, multiple training data sets including the input data and teacher data are generated. Using these multiple training data sets, the mapping data DM23 is learned. That is, the input-side variables and output-side variables are adjusted respectively so that the error between the output variable mapped from the input data and the actual anomaly determination information converges to below a predetermined value.

[0160] In step S113, CPU 91 evaluates the output variable Y(AGI) calculated in step S111. That is, CPU 91 determines whether the output variable Y(AGI) is less than or equal to the above-mentioned anomaly determination value. If the output variable Y(AGI) is less than or equal to the anomaly determination value, an anomaly is considered to have occurred in the lock-up clutch 75. On the other hand, if the output variable Y(AGI) is greater than the anomaly determination value, an anomaly is not considered to have occurred in the lock-up clutch 75. In the next step S115, if the evaluation result determines that an anomaly has occurred in the lock-up clutch 75 ("Yes"), CPU 91 transfers the process to the next step S117. In step S117, CPU 91 stores the meaning of the anomaly in the storage device 93. After that, CPU 91 temporarily terminates the series of processes.

[0161] On the other hand, in step S115, if the evaluation result does not determine that an abnormality has occurred in the lock-up clutch 75 ("No"), the CPU91 temporarily terminates the series of processes. That is, for example, if the output variable Y (AGI) is greater than the abnormality determination value, the CPU91 does not execute the process of step S117 and temporarily terminates the series of processes.

[0162] According to this embodiment, in addition to the effects described in (1-1) to (1-10) above, the following effects can also be obtained.

[0163] (3-1) Even if no abnormality occurs in the transmission 40, the heat generation of the transmission 40 may vary depending on the degree of change in its characteristics over the years. Therefore, in this embodiment, multiple mapping data DM21, DM22, DM23, ... corresponding to the degree of change in the characteristics of the transmission 40 over the years are prepared in advance. From the multiple mapping data DM21, DM22, DM23, ..., the mapping data corresponding to the degree of change in the characteristics over the years at this time is selected, and the input variable is input for the mapping specified by the mapping data. Then, based on the output variable Y (AGI) output from the mapping, it is determined whether an abnormality has occurred in the lock-up clutch 75. By using the mapping data separately according to the degree of change in the characteristics over the years, the accuracy of the determination can be improved.

[0164] (Correspondence)

[0165] The correspondence between the items in the above embodiments and the items recorded in the "Summary of the Invention" column is as follows. Hereinafter, the correspondence is shown for each number of the method recorded in the "Summary of the Invention" column.

[0166] [1] The anomaly detection device corresponds to the control device 90. The vehicle's power source corresponds to the internal combustion engine 10. The drive wheels correspond to the drive wheels 60. The power transmission device corresponds to the transmission device 40. The friction engagement element corresponds to the lock-up clutch 75. The oil temperature sensor corresponds to the oil temperature sensor 103. The vehicle corresponds to the vehicle VC. The execution device, i.e., the processing circuit, corresponds to the CPU 91 and ROM 92. The storage device corresponds to the storage device 93. The oil temperature detection value corresponds to the oil temperature detection value Toil. The oil temperature associated data corresponds to the oil temperature associated data RDToil. The mapping data corresponds to... Figure 1 The mapping data shown are DM1, DM2, DM3, Figure 9 The mapping data shown are DM11, DM12, DM13, DM14 and Figure 12 The mapping data shown corresponds to DM21, DM22, DM23, ... . The processing and acquisition are then performed. Figure 4 as well as Figure 5 Each process in steps S13 to S23, Figure 10 as well as Figure 11 The processing steps S53~S63 and Figure 13 as well as Figure 14 The corresponding processes in steps S93~S103. Anomaly detection and handling. Figure 4 as well as Figure 5 Each process in steps S27~S41, Figure 10 as well as Figure 11 The processing steps S73~S79 and Figure 13 as well as Figure 14 The corresponding processes in steps S111 to S117 are as follows.

[0167] [2] Detection value acquisition and processing Figure 4 Each process in steps S13 to S17, Figure 10 as well as Figure 11 The processing steps S53~S57 and Figure 13 as well as Figure 14 The corresponding processes in steps S93-S97. Related data generation process and... Figure 4 as well as Figure 5 The processing steps in S19 and S21, Figure 10 as well as Figure 11 The processing steps S59 and S61 in the process and Figure 13 as well as Figure 14 The corresponding processes in steps S99 and S101.

[0168] [3] The standardized oil temperature detection value corresponds to the standardized oil temperature detection value ToilN.

[0169] [4] Vehicle speed corresponds to vehicle speed SPD.

[0170] [5] The clutch corresponds to the lock-up clutch 75. The input-side component corresponds to the input-side component 78. The output-side component corresponds to the output-side component 77. The speed of the input-side component corresponds to the internal combustion engine speed NE. The speed of the output-side component corresponds to the input shaft speed Nat. The speed difference between the input-side and output-side components corresponds to the speed difference ΔNtc.

[0171] [6] The input section corresponds to the front cover 71. The output section corresponds to the input shaft 81. The input-output speed difference corresponds to the speed difference ΔNtc. The calculated value of heat generation corresponds to the calculated value of heat generation CVtc.

[0172] [7] The joint force corresponds to the joint force EFtc.

[0173] [8] The component gap corresponds to the component gap PCtc.

[0174] [9] The acceleration sensor corresponds to acceleration sensor 105. The detection value of the acceleration sensor corresponds to the vehicle acceleration G.

[0175]

[10] Anomaly detection and handling and in Figure 4 as well as Figure 5 The corresponding processing steps S31 to S37 when the determination coefficient mP is "1".

[0176]

[11] Anomaly detection and handling and in Figure 4 as well as Figure 5 The corresponding processing steps S31 to S37 when the determination coefficient mP is "2".

[0177]

[12] Anomaly detection and handling and in Figure 4 as well as Figure 5 The corresponding processing steps S31 to S37 when the determination coefficient mP is "3".

[0178]

[13] The first mapping data corresponds to mapping data DM1. The second mapping data corresponds to mapping data DM2. The third mapping data corresponds to mapping data DM3.

[0179]

[14] The first mapping data corresponds to one mapping data among DM11, DM12, DM13, and DM14. The second mapping data corresponds to the mapping data other than the first mapping data among DM11, DM12, DM13, and DM14. Data selection processing and Figure 10 as well as Figure 11 This corresponds to step S71 in the previous section. Exception handling and... Figure 10 as well as Figure 11 The corresponding processes in steps S73 to S79 ​​are as follows.

[0180]

[15] The mapping data corresponds to mapping data DM21, DM22, DM23, ... . Data selection processing and Figure 13 as well as Figure 14 This corresponds to step S109 in the previous section. Exception handling and... Figure 13 as well as Figure 14 The corresponding processes in steps S111 to S117 are as follows.

[0181] (Example of the change)

[0182] The above embodiments can be implemented with modifications as described below. The above embodiments and the following modifications can be combined with each other to the extent that they are not technically contradictory.

[0183] "About Mapping"

[0184] In the above embodiments, the activation function of the mapping is illustrative and is not limited to the examples in the above embodiments. For example, the Logistic Sigmoid function can also be used as the activation function of the mapping.

[0185] In the above embodiments, a neural network with one intermediate layer is exemplified, but the number of intermediate layers may also be two or more.

[0186] In the embodiments described above, a fully connected feedforward neural network is exemplified as the neural network, but it is not limited to this. For example, a recursively connected neural network can also be used as the neural network. As will be described later, when the time-series data of the standardized oil temperature detection value ToilN and the time-series data of the oil temperature detection value Toil are used as the input variables for mapping instead of the oil temperature correlation data RDToil described above, a recursively connected neural network can be used.

[0187] Regarding the mapping data

[0188] In the second embodiment described above, the mapping specified by the mapping data DM11, DM12, DM13, and DM14 can also output an output variable that can identify the type of abnormality of the interlocking clutch 75. In this case, the magnitude of the output variable corresponds to the type of abnormality, namely, the occurrence of burn-out, the occurrence of poor engagement, or the occurrence of adhesion.

[0189] In the third embodiment described above, the mapping specified by the mapping data DM21, DM22, DM23, ... can also output an output variable that can identify the type of abnormality of the interlocking clutch 75. In this case, the magnitude of the output variable becomes the magnitude corresponding to the type of abnormality, namely, the occurrence of burn-out, the occurrence of poor engagement, and the occurrence of adhesion.

[0190] In this case, information related to the probability of burn-out, the probability of poor bonding, and the probability of adhesion are input to the output layer of the mapping. Then, the output layer outputs an output variable Y corresponding to each of the input information.

[0191] In the first embodiment described above, a mapping data obtained by learning through machine learning can also be stored in the storage device 93 in a manner that enables the determination of whether burn-out has occurred in the lock-up clutch 75, the determination of whether poor engagement has occurred, and the determination of whether sticking has occurred. In this case, the mapping data may not be switched based on the content of the determination.

[0192] Furthermore, in the case where only one mapping data is stored in storage device 93, the mapping data may not be obtained through learning that can distinguish even the types of anomalies.

[0193] In the first embodiment described above, if mapping data DM1 is stored in storage device 93, mapping data DM2 and DM3 may not need to be stored in storage device 93. Even in this case, it is still possible to determine whether burn-out has occurred in the lock-up clutch 75.

[0194] In the first embodiment described above, if mapping data DM2 is stored in storage device 93, mapping data DM1 and DM3 may not need to be stored in storage device 93. Even in this case, it is still possible to determine whether a poor engagement has occurred in the lock-up clutch 75.

[0195] In the first embodiment described above, if mapping data DM3 is stored in storage device 93, mapping data DM1 and DM2 may not need to be stored in storage device 93. Even in this case, it is still possible to determine whether sticking has occurred in the lock-up clutch 75.

[0196] Regarding input variables

[0197] • The input variable may also exclude the vehicle acceleration G.

[0198] Input variables may also exclude component gaps (PCtc).

[0199] The input variables may also exclude the engagement force EFtc.

[0200] If the transmission 40 is equipped with a sensor capable of detecting the engagement force EFtc or a related value of the engagement force EFtc, the detection value of the sensor can also be used as the engagement force EFtc.

[0201] • Input variables may also exclude the calorific value CVtc.

[0202] • If the transmission 40 has a sensor that can detect the heat generated by the lock-up clutch 75 or a related value of the heat generated, the detection value of the sensor can also be used as an input variable.

[0203] If the internal combustion engine speed NE, which is the speed of the input element 78, is used as the input variable, then the input shaft speed Nat, which is the speed of the output element 77, may not be included in the input variable. Additionally, the speed difference ΔNtc may not be included in the input variable.

[0204] If the input shaft speed Nat, which is the speed of the output element 77, is used as the input variable, then the internal combustion engine speed NE, which is the speed of the input element 78, may not be included in the input variable. Additionally, the speed difference ΔNtc may not be included in the input variable.

[0205] • If the speed difference ΔNtc is used as the input variable, the internal combustion engine speed NE, which is the speed of the input-side element 78, may not be included in the input variable. Additionally, the input shaft speed Nat, which is the speed of the output-side element 77, may not be included in the input variable.

[0206] Alternatively, the speed ratio, which is the ratio of the input shaft speed Nat to the internal combustion engine speed NE, can be used instead of the speed difference ΔNtc as the input variable.

[0207] • The input variables may also exclude any one of the following: the internal combustion engine speed NE (which is the speed of the input-side element 78), the input shaft speed Nat (which is the speed of the output-side element 77), and the speed difference ΔNtc.

[0208] • Input variables may also exclude vehicle speed SPD.

[0209] Regarding anomaly detection and handling

[0210] In the first embodiment described above, the meaning of an abnormality occurring in the lock-up clutch 75 can be stored in the storage device 93 in any case where it is determined that burning has occurred, poor engagement has occurred, or adhesion has occurred. That is, if it is possible to store the meaning of an abnormality in the storage device 93, then it is not necessary to store its contents in the storage device 93.

[0211] Regarding oil temperature correlation data

[0212] In the above embodiments, the oil temperature associated data RDToil consists of "5" count values ​​Cnt(1) to Cnt(5), but the number of count values ​​is not limited to "5". For example, if the range of values ​​from "0" to "1" is divided into "0.1" increments, the oil temperature associated data RDToil can be data including "10" count values ​​Cnt(1) to Cnt(10).

[0213] • The oil temperature correlation data used as the input variable for the mapping does not necessarily have to be the oil temperature correlation data RDToil mentioned above. That is, the oil temperature correlation data can also be time-series data of the standardized oil temperature detection value ToilN.

[0214] Oil temperature correlation data can also be time-series data of oil temperature detection values ​​(Toil).

[0215] Regarding the degree of change over the years

[0216] In the third embodiment described above, the degree of change in the characteristics of the transmission device 40 over the years is estimated based on the driving distance of the vehicle VC, but it is not limited to this. For example, the degree of change in the characteristics of the transmission device 40 over the years can also be estimated based on the number of times the lock-up clutch 75 is operated and the total time when the lock-up clutch 75 is in the engaged state.

[0217] Regarding the actuator

[0218] • The execution device is not limited to examples that include a CPU 91 and a ROM 92 to perform software processing. For example, it may also include dedicated hardware circuitry that processes at least a portion of the software processing executed in the above embodiments. Examples of dedicated hardware circuitry include ASICs. ASIC stands for "Application Specific Integrated Circuit". That is, the execution device can be any of the following structures (a) to (c): (a) A processing device that performs all of the above processing according to a program and a program storage device such as a ROM that stores the program. (b) A processing device that performs a portion of the above processing according to a program, a program storage device, and a dedicated hardware circuitry that performs the remaining processing. (c) A dedicated hardware circuitry that performs all of the above processing. Here, there may be multiple software execution devices and dedicated hardware circuitry that include processing devices and program storage devices. That is, the above processing may be performed by a processing circuitry that includes at least one of one or more software execution devices and one or more dedicated hardware circuitry. The program storage device, i.e., computer-readable medium, includes all usable media that can be accessed by a general-purpose or special-purpose computer.

[0219] Regarding power transmission devices

[0220] • A power transmission device is not limited to a stepped transmission as long as it has a friction engagement element. For example, a power transmission device can also be a continuously variable transmission (CVT).

[0221] Regarding friction-bonded elements

[0222] • The friction engagement element is not limited to the lock-up clutch 75. For example, the friction engagement element can be the first clutch C1 of the transmission mechanism 80, the second clutch C2, or the brake mechanism B1.

[0223] Friction engagement elements are not limited to being engaged or disengaged through hydraulic adjustment. For example, friction engagement elements can be engaged or disengaged by electric motor drive or by electromagnetic force adjustment.

[0224] Regarding the vehicle

[0225] The vehicle can also be a hybrid electric vehicle. Alternatively, the vehicle can be one that has an electric generator but no internal combustion engine. In this case, the electric generator is the vehicle's power source.

Claims

1. An anomaly detection device for a power transmission device, applied to a vehicle, The vehicle includes: a power transmission device having friction engagement elements and configured to transmit power from a power source of the vehicle to the drive wheels; and an oil temperature sensor configured to detect the oil temperature, which is the temperature of the oil circulating within the power transmission device. The fault detection device of the power transmission device includes a processing circuit and a storage device. The storage device stores mapping data, which includes data obtained through machine learning and a defined mapping. When oil temperature correlation data, which corresponds to time-series data of oil temperature detection values, is input as an input variable, the mapping output determines whether an anomaly has occurred in the friction engagement element. The oil temperature detection value is the value detected by the oil temperature sensor. The processing circuit is configured to execute: Obtain the processing, obtain the input variables; and The anomaly detection process determines whether an anomaly has occurred in the friction engagement element based on the output variable obtained by inputting the input variable acquired in the acquisition process into the mapping and outputting the output variable from the mapping. The acquisition process includes: The detection value acquisition process acquires time-series data of the oil temperature detection values, which include multiple oil temperature detection values ​​detected for each detection cycle within a predetermined measurement period. as well as The associated data generation process standardizes multiple oil temperature detection values ​​included in the time-series data of the oil temperature detection values ​​to generate the oil temperature associated data. Assuming the value obtained by standardizing the oil temperature detection value is the standardized oil temperature detection value. The processing circuit is configured to, in the associated data generation process, standardize multiple oil temperature detection values ​​included in the time series data of the oil temperature detection values, derive time series data of standardized oil temperature detection values ​​including multiple standardized oil temperature detection values, and generate data representing the distribution of the magnitudes of the multiple standardized oil temperature detection values ​​included in the time series data of the standardized oil temperature detection values, as the oil temperature associated data.

2. The abnormality determination device for the power transmission device according to claim 1, wherein, The input variables include vehicle speed.

3. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The power transmission device includes a clutch as the friction engagement element. The input variables include at least one of the following: the rotational speed of the input-side component of the clutch, the rotational speed of the output-side component of the clutch, and the difference in rotational speed between the input-side component and the output-side component.

4. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The speed difference between the input portion of the torque input to the friction engagement element and the output portion of the torque output from the friction engagement element is the input-output speed difference. The input variables include a calculated value of the heat generated by the friction engagement element, which is calculated based on the product of the torque input to the friction engagement element and the input-output speed difference.

5. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The input variables include the engagement force of the friction engagement element.

6. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The input variable includes the component gap of the friction engagement element.

7. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The input variables include the detection values ​​of the acceleration sensors mounted on the vehicle.

8. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The mapping output determines whether burn-out has occurred in the friction-bonded element. The processing circuit is configured to determine, in the anomaly determination process, whether burn-out has occurred in the friction engagement element based on the output variable output from the mapping obtained by inputting the input variable obtained in the acquisition process into the mapping.

9. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The mapping output is an output variable that determines whether a poor engagement has occurred in the friction-bonded element. The processing circuit is configured to determine, in the anomaly determination process, whether a poor engagement of the friction engagement element has occurred based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process to the mapping.

10. The fault detection device for the power transmission device according to claim 1 or 2, wherein, The mapping output is an output variable that determines whether adhesion has occurred in the friction-bonded element. The processing circuit is configured to determine, in the anomaly determination process, whether adhesion has occurred in the friction bonding element based on the output variable output from the mapping obtained by inputting the input variable obtained in the acquisition process into the mapping.

11. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The storage device stores multiple mapping data. The plurality of mapping data includes first mapping data, second mapping data, and third mapping data. The first mapping data specifies a mapping that, when the input variable is input, outputs a variable to determine whether burn-out has occurred in the friction-bonded element. The second mapping data specifies a mapping that, when the input variable is input, outputs a variable to determine whether a misfit has occurred in the friction engagement element. The third mapping data specifies a mapping that outputs a variable to determine whether adhesion has occurred in the friction-bonding element when the input variable is input.

12. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The storage device stores multiple mapping data that individually correspond to the operating states of each of the friction engagement elements. The plurality of mapping data includes first mapping data and second mapping data. The first mapping data is data that defines the output variable mapping to determine whether an abnormality has occurred in the friction engagement element when the oil temperature correlation data when the operating state of the friction engagement element is the first operating state is used as the input variable. The second mapping data is defined as the data that determines whether an anomaly has occurred in the friction engagement element when the oil temperature correlation data is used as an input variable, and the operating state of the friction engagement element is a second operating state different from the first operating state. The processing circuit is configured as follows: A data selection process is performed to select the mapping data corresponding to the operating state of the friction engagement element from a plurality of mapping data stored in the storage device. In the anomaly determination process, it is determined whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process into the mapping specified by the mapping data selected in the data selection process.

13. The abnormality determination device for the power transmission device according to claim 1 or 2, wherein, The storage device stores multiple mapping data corresponding to the degree of change in the characteristics of the power transmission device over the years. The processing circuit is configured as follows: A data selection process is performed to select, from among the multiple mapping data stored in the storage device, mapping data corresponding to the degree of annual variation in the characteristics of the power transmission device. In the anomaly determination process, it is determined whether an anomaly has occurred in the friction engagement element based on the output variable output from the mapping by inputting the input variable obtained in the acquisition process into the mapping specified by the mapping data selected in the data selection process.

Images