Fault evaluation device for an automatic transmission, fault evaluation method for an automatic transmission, and non-volatile storage medium storing a fault evaluation program for an automatic transmission

Abstract

A fault evaluation device for an automatic transmission (30), wherein the fault evaluation device evaluates a fault of the automatic transmission (30), and wherein the fault evaluation device is used for a vehicle (100) comprising the automatic transmission (30) and a control device (90) designed to carry out a learning process for correcting a target pressure (Z) for oil to be supplied to the automatic transmission (30) so that fluctuations in the acceleration of the vehicle (100) during shifting of the automatic transmission (30) are small, wherein the fault evaluation device comprises: a processor; and a memory (94), wherein: the memory (94) stores image data defining a first image (M1) and a second image (M2); the processor is designed to output an output variable (y) which is an evaluation value indicating the presence or absence of the fault of the automatic transmission (30) when a plurality of input variables (x) are input; the first map (M1) and the second map (M2) comprise as one of the input variables (x) an acceleration variable, which is a variable indicating the acceleration of the vehicle (100) during the shifting of the automatic transmission (30); the first image (M1) is a trained image trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is located in a first area; the second mapping (M2) is a trained mapping trained by machine learning under the condition that the learning progress variable is in a second range different from the first range; and the processor is designed to execute a determination process, which is a process for determining the input variables (x) and the learning progress variable, a first calculation process for calculating a value of the output variable (y) by inputting the input variable (x) determined by the determination process into the first map (M1) when the learning progress variable determined by the determination process is in the first area, and a second calculation process for calculating a value of the output variable (y) by inputting the input variable (x) determined by the determination process into the second map (M2) when the learning progress variable determined by the determination process is in the second area.

Description

[0001] The present invention relates to a failure evaluation device for an automatic transmission, a failure evaluation method for an automatic transmission, and a non-volatile storage medium storing a failure evaluation program for an automatic transmission.

[0002] WO 2012 / 111192 A1 describes a vehicle with an evaluation device that evaluates a shift shock in an automatic transmission. The evaluation device determines an acceleration from an acceleration sensor. The evaluation device evaluates a shift shock in the automatic transmission based on the transition of the acceleration from the start to the completion of the shifting process of the automatic transmission.

[0003] The magnitude of a shift shock in an automatic transmission can vary depending not only on the presence or absence of a fault in the automatic transmission, but also on various other factors. Therefore, even if the evaluation device described in WO 2012 / 111192 A1 evaluates a shift shock as large, there is a significant probability that the large shift shock is not due to a fault in the automatic transmission. Therefore, the evaluation device described in WO 2012 / 111192 A1 requires improvement in assessing the presence or absence of a malfunction in an automatic transmission.

[0004] For a better understanding of the present invention, reference is also made to US 8,095,284 B2 and DE 195 01 671 A1.

[0005] A first aspect of the present invention provides a fault evaluation device for an automatic transmission, wherein the fault evaluation device evaluates a fault of the automatic transmission, and wherein the fault evaluation device is used for a vehicle including the automatic transmission and a control device configured to perform a learning process for correcting a target pressure for oil to be supplied to the automatic transmission so that fluctuations in the acceleration of the vehicle during shifting of the automatic transmission are small. The fault evaluation device includes a processor and a memory. The memory stores map data defining a first map and a second map. The processor is configured to output an output variable that is an evaluation value indicating the presence or absence of the fault of the automatic transmission.when multiple input variables are input. The first map and the second map include, as one of the input variables, an acceleration variable, which is a variable indicating the acceleration of the vehicle during shifting of the automatic transmission; the first map is a trained map trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is in a first range; the second map is a trained map trained by machine learning under the condition that the learning progress variable is in a second range different from the first range; and the processor is configured to perform a determination process, which is a process for determining the input variables and the learning progress variable, a first calculation process,to calculate a value of the output variable by entering the input variable determined by the determination process into the first image when the learning progress variable determined by the determination process is in the first area, and to execute a second calculation process to calculate a value of the output variable by entering the input variable determined by the determination process into the second image when the learning progress variable determined by the determination process is in the second area,

[0006] With the configuration described above, one of the first map and the second map is selectively used in accordance with the learning progress variable, which is a variable indicating the status of the learning progress in the learning process. Consequently, an appropriate value can be output as an output variable, which is an evaluation value indicating the presence or absence of the fault regardless of the status of the learning process progress. That is, with the configuration described above, the presence or absence of the fault of the automatic transmission can be accurately evaluated regardless of the status of the learning process progress.

[0007] In the aspect described above, the learning progress variable may be a variable that takes a first value when learning has converged in the learning process and that takes a second value different from the first value when learning has not converged in the learning process.

[0008] With the configuration described above, it is possible to obtain an evaluation value determined by considering whether the learning has converged and the acceleration fluctuations during the shifting process are relatively large, or whether the learning has not converged and the acceleration fluctuations during the shifting process are relatively large. Thus, the automatic transmission failure can be accurately evaluated.

[0009] In the aspect described above, the learning progress variable may be a variable indicating the execution frequency (number of executions) since the automatic transmission was installed in the vehicle. In the above configuration, the execution frequency of the learning process, which is a value strongly correlated with the degree of progress of the learning process, is input as an input variable. Therefore, an evaluation value that accurately reflects the degree of progress of the learning process can be obtained.

[0010] In the aspect described above, the learning progress variable may be the vehicle's traveled distance since the automatic transmission was installed in the vehicle. In the above configuration, the vehicle's traveled distance, which is a value strongly correlated with the degree of progress of the learning process, is input as an input variable. Therefore, an evaluation value that accurately reflects the progress of the learning process can be obtained.

[0011] In the aspect described above, the target pressure may be calculated by adding or multiplying a learning correction value and a reference pressure, which is an oil pressure at a time point when the learning process has not yet been executed; the learning process may be a process to calculate the learning correction value so that the fluctuations in the acceleration of the vehicle during shifting of the automatic transmission are small; and the mapping may include the learning correction value as the input variable.

[0012] In the aspect described above, the absolute value of the learning correction value tends to be small when deterioration, etc., of the automatic transmission has not progressed, and large when deterioration, etc., of the automatic transmission has progressed. Thus, the presence or absence of the automatic transmission failure can be accurately evaluated by inputting a value that can reflect deterioration, etc., of the automatic transmission.

[0013] In the aspect described above, the automatic transmission may include a plurality of engagement elements and a plurality of gear stages shifted by the engagement elements; and the map may include, as the input variable, a shift type variable indicating a type of gear stages before and after the shifting of the automatic transmission.

[0014] In the aspect described above, the ease of change in acceleration during the gear shift differs according to the type of gear stages before and after the automatic transmission shift. With the configuration described above, the input variable further includes a gear shift type variable indicating the type of gear stages before and after the automatic transmission shift, so that the error can be accurately evaluated according to the type of gear stages.

[0015] In the aspect described above, the automatic transmission may include the engagement elements and the gear stages shifted by the engagement elements; and the map may include, as the input variable, a variable indicating the number of gear shifts since the automatic transmission was installed in the vehicle, the number of gear shifts to one of the gear stages being performed after the automatic transmission was shifted.

[0016] In the above aspect, the number of gear shifts performed after the automatic transmission shifts into one of the gear stages, which is a value that strongly correlates with the degree of wear of the automatic transmission, is input as an input variable. The presence or absence of the automatic transmission failure can thus be accurately assessed by inputting a value that can reflect the degree of wear of the automatic transmission.

[0017] In the aspect described above, the automatic transmission may include the engagement elements and the gear stages shifted by the engagement elements; and the map may include, as the input variable, a variable indicating the number of engagements performed by the engagement elements since the automatic transmission was installed on the vehicle, the number of engagements performed by an engagement element among the engagement elements that is engaged to establish one of the gear stages after the automatic transmission is shifted.

[0018] In the above configuration, the number of engagements performed by the engagement element engaged to establish one of the gear stages after the automatic transmission shifts, which is a value that strongly correlates with the degree of wear of the automatic transmission, is input as an input variable. The presence or absence of the automatic transmission failure can thus be accurately evaluated by inputting a value that can reflect the degree of wear of the automatic transmission.

[0019] In the above-described aspect, the mapping may include an accelerator variable indicating an operation amount of an accelerator pedal during shifting of the automatic transmission as the input variable. In the above configuration, the fluctuations in acceleration during shifting of the automatic transmission differ depending on the operation amount of the accelerator pedal, even if other conditions are the same. With the above-described configuration, a value indicating the operation amount of the accelerator pedal is input, so that an evaluation value reflecting the operation amount of the accelerator pedal can be obtained.

[0020] A second aspect of the present invention provides a failure evaluation method for an automatic transmission, wherein the failure evaluation method is used to evaluate a failure of the automatic transmission, and wherein the failure evaluation method is used for a vehicle including the automatic transmission and a control device configured to perform a learning process for correcting a target pressure for oil to be supplied to the automatic transmission so that fluctuations in the acceleration of the vehicle during shifting of the automatic transmission are small. The failure evaluation method is executed by a failure evaluation device. The failure evaluation device includes a processor and a memory. The memory stores map data defining a first map and a second map.The processor is configured to output an output variable, which is an evaluation value indicating the presence or absence of the fault of the automatic transmission, when a plurality of input variables are input. The first map and the second map include, as one of the input variables, an acceleration variable, which is a variable indicating the acceleration of the vehicle during shifting of the automatic transmission. The first map is a trained map trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is in a first range. The second map is a trained map trained by machine learning under the condition that the learning progress variable is in a second range different from the first range.The error evaluation method includes: inputting the acceleration variable and the learning progress variable as the input variables to the error evaluation device; and calculating a value of the output variable by inputting the input variable into the first map when the learning progress variable is in the first range, and calculating the value of the output variable by inputting the input variable into the second map when the learning progress variable is in the second range.

[0021] With the configuration described above, one of the first map and the second map is selectively used in accordance with the learning progress variable, which is a variable indicating the status of the learning progress in the learning process. Consequently, an appropriate value can be output as an output variable, which is an evaluation value indicating the presence or absence of the fault regardless of the status of the learning process progress. That is, with the configuration described above, the presence or absence of the fault of the automatic transmission can be accurately evaluated regardless of the status of the learning process progress.

[0022] A third aspect of the present invention provides a non-volatile storage medium storing a failure evaluation program for an automatic transmission, the failure evaluation program being configured to cause a computer to function as a failure evaluation device that evaluates a failure of the automatic transmission. The failure evaluation program is used for a vehicle including the automatic transmission and a control device configured to perform a learning process for correcting a target pressure for oil to be supplied to the automatic transmission so that fluctuations in the acceleration of the vehicle during shifting of the automatic transmission are small. The failure evaluation program has map data defining a first map and a second map.The first map and the second map include, as one of a plurality of input variables, an acceleration variable, which is a variable indicating the acceleration of the vehicle during shifting of the automatic transmission. The first map is a trained map trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is in a first range.

[0023] The second map is a trained map trained by machine learning under the condition that the learning progress variable is in a second range different from the first range. The fault evaluation program is configured to cause the computer to execute a function of acquiring the input variables and the learning progress variables, a function of calculating a value of an output variable, which is an evaluation value indicating the presence or absence of the fault of the automatic transmission, by inputting the acquired input variables to the first map when the acquired learning progress variable is in the first range, and a function of calculating the value of the output variable by inputting the acquired input variables to the second map when the acquired learning progress variable is in the second range.

[0024] With the configuration described above, one of the first map and the second map is selectively used in accordance with the learning progress variable, which is a variable indicating the status of the learning progress in the learning process. Consequently, an appropriate value can be output as an output variable, which is an evaluation value indicating the presence or absence of the fault regardless of the status of the learning process progress. That is, with the configuration described above, the presence or absence of the fault of the automatic transmission can be accurately evaluated regardless of the status of the learning process progress.

[0025] Features, advantages and technical and industrial significance of exemplary embodiments of the invention are described below with reference to the accompanying drawings, in which like reference numerals designate like elements and wherein: Fig. 1 is a schematic diagram of a vehicle according to a first embodiment; Fig. 2 shows the relationship between gear stages and engagement elements in an automatic transmission according to the embodiment; Fig. 3 is a flowchart showing learning control according to the embodiment; Fig. 4 is a flowchart showing evaluation control according to the embodiment; Fig. 5 is a schematic diagram of a vehicle according to a second embodiment; and Fig. 6 is a flowchart showing evaluation control according to the embodiment. First embodiment

[0026] The following is a first embodiment of the present invention with reference to the Fig. 1 to 4. First, a configuration of a vehicle 100 is described.

[0027] As it is in Fig. 1, the vehicle 100 includes an internal combustion engine 10, a power split device 20, an automatic transmission 30, drive wheels 46, a hydraulic device 50, a first motor / generator 61, and a second motor / generator 62.

[0028] The power split device 20 is coupled to a crankshaft 11, which is an output shaft of the internal combustion engine 10. The power split device 20 is a planetary gear mechanism including a sun gear S, a ring gear R, and a carrier C. The crankshaft 11 is coupled to the carrier C of the power split device 20. A rotating shaft 61A of the first motor / generator 61 is coupled to the sun gear S. A rotating shaft 62A of the second motor / generator 62 is coupled to a ring gear shaft RA, which is an output shaft of the ring gear R. An input shaft 41 of the automatic transmission 30 is also coupled to the ring gear shaft RA. The right and left drive gears 46 are coupled to an output shaft 42 of the automatic transmission 30 via a differential gear (not shown).

[0029] When the internal combustion engine 10 is operated and a torque is transmitted from the crankshaft 11 to the carrier C of the power split device 20, the torque is split to the sun gear S side and the ring gear R side.

[0030] When the first motor / generator 61 is operated as a motor and a torque is transmitted to the sun gear S of the power split device 20, the torque is split to the carrier C side and the ring gear R side.

[0031] When the second motor / generator 62 operates as a motor and transmits torque to the ring gear shaft RA, the torque is transmitted to the automatic transmission 30. When torque is supplied from the drive wheel 46 side to the second motor / generator 62 via the ring gear shaft RA, the second motor / generator 62 operates as an electric generator and can generate regenerative braking force for the vehicle 100.

[0032] The automatic transmission 30 includes a first planetary gear mechanism 30A, a second planetary gear mechanism 30B, a first clutch C1, a second clutch C2, a first brake mechanism B1, a second brake mechanism B2, and a one-way clutch F1.

[0033] Further, the first planetary gear mechanism 30A includes a sun gear 31, a ring gear 32, a pinion gear 33, and a carrier 34. The ring gear 32 is coupled to the sun gear 31 via the pinion gear 33. The pinion gear 33 is supported by the carrier 34.

[0034] The sun gear 31 is coupled to the first brake mechanism B1. The first brake mechanism B1 can be switched between an engaged state and a disengaged state in accordance with the oil pressure applied to the first brake mechanism B1. Specifically, the first brake mechanism B1 is switched from the disengaged state to the engaged state when the oil pressure applied to the first brake mechanism B1 becomes high. The rotation of the sun gear 31 is decelerated when the first brake mechanism B1 is in the engaged state.

[0035] The one-way clutch F1 is coupled to the carrier 34. The one-way clutch F1 regulates the rotation of the carrier 34 to one side while preventing its rotation to the other side. That is, the one-way clutch F1 is switched between a controlled state in which the rotation of the carrier 34 is regulated and a permitted state in which the rotation of the carrier 34 is permitted. The carrier 34 is coupled to the second brake mechanism B2. Like the first brake mechanism B1, the second brake mechanism B2 can also be switched between an engaged state and a disengaged state in accordance with the oil pressure applied to the second brake mechanism B2. The rotation of the carrier 34 is decelerated when the second brake mechanism B2 is in the engaged state.

[0036] The second planetary gear mechanism 30B includes a sun gear 36, a ring gear 37, a pinion gear 38, and a carrier 39. The ring gear 37 is coupled to the sun gear 36 via the pinion gear 38. The pinion gear 38 is supported by the carrier 39. The output shaft 42 is coupled to the carrier 39.

[0037] In each of the planetary gear mechanisms configured as described above, the carrier 34 of the first planetary gear mechanism 30A is coupled to the ring gear 37 of the second planetary gear mechanism 30B. The ring gear 32 of the first planetary gear mechanism 30A is coupled to the carrier 39 of the second planetary gear mechanism 30B.

[0038] The sun gear 36 of the second planetary gear mechanism 30B is coupled to the input shaft 41 via the first clutch C1. The first clutch C1 can be switched between an engaged state and a disengaged state in accordance with the oil pressure applied to the first clutch C1. Specifically, the first clutch C1 is switched from the disengaged state to the engaged state when the oil pressure applied to the first clutch C1 becomes high. The sun gear 36 of the second planetary gear mechanism 30B rotates together with the input shaft 41 when the first clutch C1 is in the engaged state.

[0039] The carrier 34 of the first planetary gear mechanism 30A is coupled to the input shaft 41 via the second clutch C2. Like the first clutch C1, the second clutch C2 can be switched between an engaged state and a disengaged state in accordance with the oil pressure applied to the second clutch C2. The carrier 34 of the first planetary gear mechanism 30A rotates together with the input shaft 41 when the second clutch C2 is in the engaged state. In the present embodiment, the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2 are each an engaging element.

[0040] In the automatic transmission 30, as it is in Fig. As shown in Fig. 2, gear stages are switched in accordance with the combination of the engaged state and the disengaged state of the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2, and the combination of the controlled state and the permitted state of the one-way clutch F1. The automatic transmission 30 can establish a total of five gear stages, including four gear stages, i.e., "first gear" to "fourth gear," for forward travel and one gear stage, i.e., "R," for reverse travel.

[0041] In Fig. 2, the symbol "◯" indicates that the engaging element such as the first clutch C1 is in the engaged state or that the one-way clutch F1 is in the controlled state. The symbol "(◯)" indicates that the engaging element such as the first clutch C1 is in the engaged state or the disengaged state. The blank space indicates that the engaging element such as the first clutch C1 is in the disengaged state or that the one-way clutch F1 is in the permitted state. For example, when the gear stage of the automatic transmission 30 is the second gear, the first clutch C1 and the first brake mechanism B1 are in the engaged state, the second clutch C2 and the second brake mechanism B2 are in the disengaged state, and the one-way clutch F1 is in the permitted state.

[0042] The hydraulic device 50 is installed in the vehicle 100. The hydraulic device 50 includes an oil pump 51 and a hydraulic circuit 52 through which oil flows from the oil pump 51. The oil pump 51 is a so-called mechanical oil pump that operates upon receiving torque from the crankshaft 11. The hydraulic circuit 52 includes a plurality of solenoid valves (not shown). The hydraulic circuit 52 regulates the oil pressure applied to the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2 by controlling the solenoid valves. That is, in the first embodiment, the engaged and disengaged states of the engaging elements such as the first clutch C1 are controlled using the oil pressure by controlling the solenoid valves of the hydraulic circuit 52.

[0043] The vehicle 100 includes, installed therein, a crank angle sensor 71, an accelerator position sensor 72, a vehicle speed sensor 73, an acceleration sensor 74, a display unit 76, and an accelerator pedal 77. The crank angle sensor 71 detects a crank angle SC, which is the rotation angle of the crankshaft 11. The accelerator position sensor 72 detects an accelerator operation amount ACC, which is the operation amount of the accelerator pedal 77 operated by a driver. The vehicle speed sensor 73 detects a vehicle speed SP, which is the speed of the vehicle 100.The acceleration sensor 74 is a so-called triaxial sensor that detects a front-to-rear acceleration G1, which is the acceleration of the vehicle 100 in the front-to-rear direction, a vehicle lateral acceleration G2, which is the acceleration of the vehicle 100 in the width direction, and an up-down acceleration G3, which is the acceleration of the vehicle 100 in the up-down direction. The acceleration sensor 74 detects the front-to-rear acceleration G1, the vehicle lateral acceleration G2, and the up-down acceleration G3. The display unit 76 provides visual information to the driver, etc., of the vehicle 100. Examples of the display unit 76 include an indicator lamp.

[0044] The vehicle 100 includes a control device 90. The control device 90 receives signals indicative of the crank angle SC, the accelerator operation amount ACC, and the vehicle speed SP input from the crank angle sensor 71, the accelerator position sensor 72, and the vehicle speed sensor 73, respectively. The control device 90 receives a signal indicative of the front-to-rear acceleration G1, the vehicle lateral acceleration G2, and the up-down acceleration G3 input from the acceleration sensor 74. The control device 90 calculates an engine speed NE, which is the rotational speed of the crankshaft 11 per unit time, based on the crank angle SC.

[0045] The control device 90 includes a central processing unit (CPU) 91, a peripheral circuit 92, a read-only memory (ROM) 93, and a memory 94. The CPU 91, the peripheral circuit 92, the ROM 93, and the memory 94 are communicatively connected to each other by a bus 95. The ROM 93 stores various types of programs in advance to enable the CPU 91 to execute various types of control. The memory 94 stores map data 94A, which will be discussed later, in advance. The memory 94 stores data including the accelerator operation amount ACC, the vehicle speed SP, the front-to-rear acceleration G1, the vehicle lateral acceleration G2, the top-to-bottom acceleration G3, and the engine speed NE, which are input to the control device 90 during a period.Peripheral circuit 92 includes a circuit that generates a clock signal that determines an internal operation, a power source circuit, a reset circuit, etc. In the present embodiment, CPU 91 and ROM 93 serve as the processor. Memory 94 serves as the storage. Control device 90 serves as a fault evaluation device that evaluates a fault of automatic transmission 30.

[0046] The CPU 91 controls the engine 10, the first motor / generator 61, the second motor / generator 62, the automatic transmission 30, etc., by executing various types of programs stored in the ROM 93. Specifically, the CPU 91 calculates the required vehicle power, which is a power output value necessary for driving the vehicle 100, based on the accelerator operation amount ACC and the vehicle speed SP. The CPU 91 determines the torque distribution to the engine 10, the first motor / generator 61, and the second motor / generator 62 based on the required vehicle power. The CPU 91 controls the power of the engine 10 and the power operation and regeneration of the first motor / generator 61 and the second motor / generator 62 based on the distribution of torque to the engine 10, the first motor / generator 61, and the second motor / generator 62.

[0047] The CPU 91 calculates a target gear stage, which is a gear stage as a target for the automatic transmission 30, based on the vehicle speed SP and the required vehicle power. The CPU 91 calculates a target pressure Z, which is a target value for the oil pressure to be applied to the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2, based on the target gear stage. Then, the CPU 91 outputs a control signal S1 to the hydraulic device 50 based on the target pressure Z. The hydraulic device 50 changes the oil pressure to be applied to the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2 based on the control signal S1.For example, when the gear stage of the automatic transmission 30 before a change is the second gear, the first clutch C1 and the first brake mechanism B1 are in the engaged state, the second clutch C2 and the second brake mechanism B2 are in the disengaged state, and the one-way clutch F1 is in the permitted state, as shown in FIG. Fig. 2. When the target gear stage for the automatic transmission 30 is set to the third gear, the second clutch C2 is brought from the disengaged state to the engaged state in accordance with the control signal S1, that is, based on the target pressure Z for the second clutch C2, by gradually increasing the oil pressure applied to the second clutch C2 from the hydraulic device 50. On the other hand, the first brake mechanism B1 is brought from the engaged state to the disengaged state in accordance with the control signal S1, that is, based on the target pressure Z for the first brake mechanism B1, by gradually decreasing the oil pressure applied to the first brake mechanism B1 from the hydraulic device 50. As a result, the gear stage of the automatic transmission 30 is changed from the second gear to the third gear.

[0048] The following describes the learning control executed by the CPU 91 during the shifting of the automatic transmission 30. The CPU 91 repeatedly executes the learning control in a predetermined cycle from the start of the shifting of the automatic transmission 30 to the end of the shifting of the automatic transmission 30. The ROM 93 stores a learning program, which is a program for learning control, in advance. The CPU 91 executes the learning control by executing the learning program stored in the ROM 93.

[0049] As it is in Fig. As shown in Fig. 3, when the learning control is started, the CPU 91 calculates an overshoot amount NEA of the engine speed NE in step S11. The value of the total torque obtained by summing the torque that can be transmitted by an engaging element held in the engaged state, a torque that can be transmitted by an engaging element brought from the disengaged state to the engaged state, and the torque that can be transmitted by an engaging element brought from the engaged state to the disengaged state during shifting of the automatic transmission 30 is used as the torque value. The torque value that can occasionally be temporarily transmitted becomes small compared with the torque value that should be transmitted from the input shaft 41 to the output shaft 42. In this case, the engine speed NE becomes temporarily high.The engine speed NE returns to the original state when the difference between the value of the torque that can be transmitted and the value of the torque to be transmitted becomes small. Thus, based on a phenomenon that the engine speed NE temporarily becomes high, or so-called overspeeding of the internal combustion engine 10, it is possible to determine that the torque value that can be transmitted is small compared to the torque value to be transmitted. Thus, in step S11, the CPU 91 determines the engine speed NE for a certain period determined in advance from the time point at which step S11 is executed. Then, the CPU 91 calculates the overspeed amount NEA obtained by subtracting the smallest value of the determined engine speeds NE from the largest value thereof.The memory 94 stores the engine speed NE for the specific period determined in advance from the time at which step S11 is executed. Thereafter, the CPU 91 proceeds to step S12.

[0050] In step S12, the CPU 91 determines whether the over-revolution amount NEA is equal to or lower than a predetermined over-revolution amount NEB. The predetermined over-revolution amount NEB is set in advance as a value for detecting the over-revolution amount NEA that is greater than a predetermined value. If it is determined in step S12 that the over-revolution amount NEA is higher than the predetermined over-revolution amount NEB (S12: NO), the CPU 91 proceeds to step S21.

[0051] In step S21, the CPU 91 corrects the target pressure Z for an engagement element that is operated from the disengaged state to the engaged state. The target pressure Z is calculated by adding a learning correction value CVL to a predetermined reference pressure ZA. The reference pressure ZA is set in advance for the first clutch C1, the second clutch C2, the first brake mechanism B1, and the second brake mechanism B2, respectively, when the automatic transmission 30 is manufactured. In step S21, the CPU 91 changes the learning correction value CVL for an engagement element that is operated from the disengaged state to the engaged state. Specifically, the CPU 91 calculates a new learning correction value CVL by adding a predetermined value determined in advance to the learning correction value CVL before the process in step S21. As a result, the target pressure Z for an engagement element that is operated from the disengaged state to the engaged state is corrected.The initial value of the learning correction value CVL for the target intervention element at the time point when the process in step S21 has not yet been executed is “0”.

[0052] As it is in Fig. As shown in FIG. 2, for example, when the gear stage of the automatic transmission 30 is shifted from second gear to third gear, the second clutch C2 is shifted from the disengaged state to the engaged state. In this case, the target pressure Z for the second clutch C2 is corrected by changing the learning correction value CVL for the second clutch C2. Thereafter, the CPU 91 terminates the learning control executed during the current shifting operation of the automatic transmission 30.

[0053] In the present embodiment, a temporary increase in the engine speed NE is prevented by repeatedly executing the process in step S21. The occurrence of a shock caused in the vehicle 100 during shifting of the automatic transmission 30, a so-called shift shock, due to a temporary increase in the engine speed NE is prevented. As a result, fluctuations in the acceleration of the vehicle 100 during shifting of the automatic transmission 30 are reduced. That is, the process in step S21 is the learning process for correcting the target pressure Z for oil to be supplied to the automatic transmission 30 so that fluctuations in the acceleration of the vehicle 100 during shifting of the automatic transmission 30 are small.

[0054] On the other hand, if it is determined in step S12 that the over-rotation amount NEA is equal to or lower than the set over-rotation amount NEB (S12: YES), the CPU 91 proceeds to step S15. In step S15, the CPU 91 determines whether learning has converged for an engagement element operated from the disengaged state to the engaged state. In a specific example, the CPU 91 acquires information about changes in the learning correction value CVL in the previous learning control for the target engagement element. The CPU 91 determines that learning for the target engagement element has converged if the learning correction value CVL has never been changed in the learning control from the current learning control to the time before a set, predetermined number of times.The memory 94 stores information about changes in the learning correction value CVL in the learning control from the current learning control to the time before the set predetermined number of times. If it is determined in step S15 that the learning for an engagement element operated from the disengaged state to the engaged state has converged (S15: YES), the CPU 91 proceeds to step S16. Thus, in the first embodiment, it is determined that the learning has converged when there is a low probability that the learning correction value CVL for the target engagement element has changed, even if the process in step S21, that is, the learning in the learning process, is repeatedly executed thereafter, and the process proceeds to step S16 when it is determined that the learning has converged.

[0055] In step S16, the CPU 91 sets a learning convergence flag FL indicating that learning has converged for an engagement element operated from the non-engagement state to the engagement state. That is, when learning has converged in the learning process, the learning convergence flag FL is set. Thereafter, the CPU 91 repeatedly executes the processes in and after step S11. The initial value of the learning convergence flag FL for the target engagement element at the time point when the process in step S21 has not yet been executed is cleared (not set).

[0056] On the other hand, when it is determined in step S15 that the learning for an engagement element operated from the non-engagement state to the engagement state has not converged (S15: NO), the CPU 91 proceeds to the process in step S17.

[0057] In step S17, the CPU 91 sets the learning convergence flag FL for an engagement element that is operated from the non-engagement state to the engagement state to CLEAR. Thereafter, the CPU 91 repeatedly executes the processes in and after step S11.

[0058] The following describes the evaluation control in which the CPU 91 evaluates the automatic transmission 30. The CPU 91 executes the evaluation control each time the automatic transmission 30 completes a shift. The ROM 93 stores an evaluation program in advance, which is a program for the evaluation control. The CPU 91 executes the evaluation control by executing the evaluation program stored in the ROM 93.

[0059] As it is in Fig. 4, in step S31, when the evaluation control is started, the CPU 91 acquires various kinds of values ​​by accessing the memory 94. Specifically, in the shifting of the automatic transmission 30 that was completed immediately before the evaluation control, the CPU 91 acquires the front-rear acceleration G1 for a period from the start of the shifting of the automatic transmission 30 to the end of the shifting of the automatic transmission 30. The CPU 91 calculates the maximum value of the amount of fluctuation in the front-rear acceleration G1 per unit time based on the acquired front-rear acceleration G1. The CPU 91 acquires the maximum value of the amount of fluctuation in the front-rear acceleration G1 per unit time as a maximum change value Gmax.As discussed above, the engine speed NE temporarily becomes high when the amount of torque that can be transmitted is temporarily small compared to the amount of torque to be transmitted during shifting of the automatic transmission 30. A shift shock is occasionally caused due to a temporary increase in the engine speed NE. Thus, the CPU 91 determines, as a value indicating the magnitude of the shift shock, the maximum value of the amount of fluctuation of the front-to-rear acceleration G1 per unit time during the shifting of the automatic transmission 30 that was completed immediately before the evaluation control as the maximum change value Gmax. The memory 94 stores the front-to-rear acceleration G1 detected during the shifting of the automatic transmission 30 that was completed immediately before the evaluation control.

[0060] The CPU 91 determines the learning convergence flag FL for an engagement element that is operated from the disengaged state to the engaged state during the shift of the automatic transmission 30 that was completed immediately before the evaluation control. The memory 94 stores the learning convergence flag FL for each engagement element.

[0061] The CPU 91 determines, as the execution frequency NL, or the number NL of executions of the learning process, the execution frequency of step S21 in the learning control for an engagement element operated from the disengaged state to the engaged state during the shift of the automatic transmission 30 that was completed immediately before the evaluation control. The execution frequency NL is the number from the time the automatic transmission 30 is installed in the vehicle 100 during the manufacture of the vehicle 100 until the time step S31 is executed. The memory 94 stores the execution frequency NL for each engagement element. The execution frequency NL is reset, for example, when the automatic transmission 30 is replaced during maintenance, etc., of the automatic transmission 30.In this case, the automatic transmission 30 is considered to be installed in the vehicle 100 when a new automatic transmission 30 is installed to replace the automatic transmission 30.

[0062] The CPU 91 determines the learning correction value CVL for an engagement element that is operated from the disengaged state to the engaged state during the shift of the automatic transmission 30 that was completed immediately before the evaluation control. The memory 94 stores the learning correction value CVL for each engagement element.

[0063] The CPU 91 determines a shift type TL, which indicates the type of gear stages before and after the shift of the automatic transmission 30 that was completed immediately before the evaluation control. For example, if the gear stage of the automatic transmission 30 is changed from second gear to third gear during the shift of the automatic transmission 30, the shift type TL indicates the shift from second gear to third gear. The memory 94 stores the shift type TL during the shift of the automatic transmission 30 that was completed immediately before the evaluation control.

[0064] The CPU 91 determines the number of engagements EN, which is the number of engagements performed by an engagement element operated from the disengaged state to the engaged state during the shift of the automatic transmission 30 that was completed immediately before the evaluation control. The number of engagements EN is described using the number of engagements EN of the second clutch C2 as an example. For example, when the gear stage of the automatic transmission 30 is changed from second gear to third gear after the shift stage of the automatic transmission 30 is changed from first gear to second gear, the number of shifts performed in the automatic transmission 30 is two. However, the second clutch C2 is operated from the disengaged state to the engaged state only when the gear stage of the automatic transmission 30 is changed from second gear to third gear.Therefore, the number EN of engagements of the second clutch C2 is incremented by one when the gear stage of the automatic transmission 30 is changed from the second gear to the third gear.

[0065] The number of engagements EN is the number from the time the automatic transmission 30 is installed in the vehicle 100 during the manufacture of the vehicle 100 until the time at which step S31 is executed. The memory 94 stores the number of engagements EN for each engagement element. The number of engagements EN is reset, for example, when the automatic transmission 30 is replaced during maintenance, etc. of the automatic transmission 30. In this case, the automatic transmission 30 is considered installed in the vehicle 100 when a new automatic transmission 30 is installed to replace the automatic transmission 30.

[0066] The CPU 91 determines the accelerator operation amount ACC at the time of the maximum change value Gmax during the shift of the automatic transmission 30 that was completed immediately before the evaluation control. The memory 94 stores the accelerator operation amount ACC detected during the shift of the automatic transmission 30 that was completed immediately before the evaluation control in correlation with the front-to-rear acceleration G1. The process in step S31 is the determination process. After that, the CPU 91 proceeds to the process in step S32.

[0067] In step S32, the CPU 91 generates input variables x (1) to x (7) for mapping to evaluate the presence or absence of a failure or malfunction of the automatic transmission 30 and sets the various values ​​obtained in the process in step S31 to the input variables x (1) to x (7).

[0068] Specifically, the CPU 91 sets the maximum change value Gmax to the input variable x(1). The CPU 91 sets the learning convergence flag FL to the input variable x(2). A value of "1" is set to the input variable x(2) when the learning convergence flag FL is SET, while a value of "0" is set to the input variable x(2) when the learning convergence flag FL is CLEAR. That is, a value of "1" is set to the input variable x(2) as a first value when learning has converged in the learning process, while a value of "0" is set to the input variable x(2) as a second value different from the first value when learning has not converged in the learning process.

[0069] The CPU 91 sets the execution frequency NL to the input variable x (3). The CPU 91 sets the learning correction value CVL to the input variable x (4). The CPU 91 sets the shift type TL to the input variable x (5). A numerical value determined in advance in accordance with the shift type TL is set to the input variable x (5). For example, when the gear stage of the automatic transmission 30 is changed from second gear to third gear, a value of "23" is set to the input variable x (5) as a numerical value for a type indicating the shift from second gear to third gear.

[0070] The CPU 91 sets the number EN of interventions to the input variable x (6). The CPU 91 sets the accelerator operation amount ACC to the input variable x (7). After that, the CPU 91 proceeds to step S33.

[0071] In the present embodiment, the input variable x(1) is an acceleration variable, which is a variable indicating the acceleration of the vehicle 100 during the shifting of the automatic transmission 30. The input variable x(2) is a first learning progress variable, which is a variable indicating the status of the learning progress in the learning process. The input variable x(3) is a second learning progress variable, which is a variable indicating the status of the learning progress in the learning process. The input variable x(5) is a shift type variable, which indicates the type of gear stages before and after the shifting of the automatic transmission 30. The input variable x(6) is a variable indicating the number of engagements, that is, the number of engagements by an engagement element engaged to establish a gear stage after the shifting of the automatic transmission 30.The input variable x (7) is an acceleration variable indicating the accelerator operation amount ACC during shifting of the automatic transmission 30.

[0072] In step S33, the CPU 91 calculates the value of an output variable y (i) by inputting the input variables x (1) to x (7) generated in the process in step S32 and an input variable x (0) as a bias parameter into the Fig. which is determined by image data 94A stored in advance in the memory 94. Thereafter, the CPU 91 proceeds to the process in step S34.

[0073] Examples of Fig. , which is defined by the mapping data 94A, comprise a function approximator and a fully connected, feedforward neural network or network with a single intermediate layer. In particular, Fig. , which is specified by the mapping data 94A, the values ​​of nodes in the intermediate layer are determined by substituting into an activation function f all "m" values ​​obtained by converting the input variables x(1) to x(7) and the input variable x(0) as a bias parameter using a linear mapping specified by a coefficient wFjk (j = 1 to m, k = 0 to 7). Further, output variables y(1) to y(2) are determined by substituting into an activation function g all values ​​obtained by converting the values ​​of nodes in the intermediate layer using a linear mapping specified by a coefficient wSij (i = 1 to 2). The output variable y(1) is a variable indicating the probability that the automatic transmission 30 is normal. The output variable y (2) is a variable that indicates the probability that the automatic transmission 30 has a fault or is faulty.In the present embodiment, the output variable y(1) and the output variable y(2) are evaluation values ​​indicating the presence or absence of a fault or malfunction of the automatic transmission 30. The processes in step S32 and step S33 are the calculation process. In the present embodiment, examples of the activation function f include a ReLU (Rectified Linear Unit) function. Examples of the activation function g include a soft max function. Thus, the sum of the output variable y(1) and the output variable y(2) is "1."

[0074] The Fig. , which is specified by the mapping data 94A, is generated, for example, as follows. First, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected by causing a vehicle prototype in which a normal automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100. Further, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected by causing a vehicle prototype in which a faulty automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100. Thereafter, the Fig. which has been trained by learning using the various types of values ​​collected for the normal automatic transmission 30 and the various types of values ​​collected for the faulty automatic transmission 30 as teacher data.

[0075] In step S34, the CPU 91 determines whether the output variable y(1) is equal to or lower than the output variable y(2). If it is determined in step S34 that the output variable y(1) is equal to or lower than the output variable y(2) (S34: YES), the CPU 91 proceeds to step S41.

[0076] In step S41, the CPU 91 determines that the automatic transmission 30 has a fault. The CPU 91 then proceeds to step S42. In step S42, the CPU 91 outputs a signal to the display unit 76 to cause the display unit 76 to display that the automatic transmission 30 has a fault. The CPU 91 then terminates the current evaluation control.

[0077] On the other hand, if it is determined in step S34 that the output variable y(1) is not equal to or lower than the output variable y(2) (S34: NO), the CPU 91 proceeds to step S46. In step S46, the CPU 91 determines that the automatic transmission 30 is normal. Thereafter, the CPU 91 terminates the current evaluation control.

[0078] The functions and effects of the present embodiment are described below. (1) The magnitude of a shift shock in the automatic transmission 30 varies according to not only the presence or absence of a fault of the automatic transmission 30, but also the status of the learning process progress. Specifically, in the vehicle 100, the shift shock in the automatic transmission 30 becomes smaller as the learning process is repeatedly executed. Therefore, the shift shock in the automatic transmission 30 is relatively large when the learning process has not yet fully progressed, such as immediately after the vehicle 100 is delivered. Thus, if the presence or absence of a fault of the automatic transmission 30 is simply judged based on only the acceleration of the vehicle 100, the automatic transmission 30 may be judged to be faulty even if the automatic transmission 30 is not faulty.

[0079] In the present embodiment, the Fig. In this context, the map data 94A determines a judgment value indicating the presence or absence of a fault of the automatic transmission 30 by considering not only the maximum change value Gmax but also a value indicating the progress status of the learning process. Thus, it is possible to prevent the automatic transmission 30 from being judged as faulty even if the maximum change value Gmax has become large when the learning process has not yet fully progressed, such as immediately after delivery of the vehicle 100. That is, in the first embodiment, the presence or absence of a fault of the automatic transmission 30 can be accurately judged regardless of the progress status of the learning process.

[0080] (2) The Fig. , which is specified by the mapping data 94A, outputs an evaluation value indicating the presence or absence of a fault of the automatic transmission 30 by taking into account the learning convergence flag FL, which indicates whether learning has converged. Therefore, it is possible to obtain an evaluation value determined by taking into account whether the maximum change value Gmax during the shifting operation is relatively large even when learning has converged, or the maximum change value Gmax during the shifting operation is relatively large because the learning has not converged, even when the maximum change value Gmax is the same. Thus, the presence or absence of a fault of the automatic transmission 30 can be accurately evaluated compared with a configuration in which the learning convergence flag FL is not taken into account.

[0081] (3) In the vehicle 100, learning converges when the learning process is repeatedly executed. Therefore, the degree of learning progress in the learning process tends to vary according to the execution frequency NL of the learning process, even if the learning convergence flag FL is CLEAR.

[0082] In this regard, the execution frequency NL, which is strongly correlated with the degree of learning progress in the learning process, is used as an input variable in the Fig. which is specified by the mapping data 94A. Therefore, an evaluation value can be obtained that accurately reflects the degree of progress of the learning process.

[0083] (4) The absolute value of the learning correction value CVL tends to become large as the deterioration of the automatic transmission 30 progresses. Thus, the learning correction value CVL, which is a value that can reflect the deterioration of the automatic transmission 30, is used as an input variable in the Fig. inputted, which is specified by the mapping data 94A. Consequently, the presence or absence of a failure of the automatic transmission 30 can be accurately evaluated by reflecting the deterioration of the automatic transmission 30.

[0084] (5) In the vehicle 100, the shift shock during shifting of the automatic transmission 30 differs depending on the shift type TL. Therefore, the magnitude of the shift shock, which is used as a criterion for determining whether the automatic transmission 30 is normal or faulty, also differs for each shift type TL. Here, the shift type TL is used as an input variable in the Fig. inputted, which is specified by the mapping data 94A. Consequently, the presence or absence of a failure of the automatic transmission 30 can be accurately evaluated depending on the shift type TL, compared with a configuration in which the shift type TL is not considered.

[0085] (6) In the vehicle 100, an engagement element is worn or worn each time the engagement element is engaged during shifting of the automatic transmission 30. The timing of engagement of the engagement element may change as the wear of the engagement element increases. When the engagement timing of the engagement element changes in this way, a shift shock caused during shifting of the automatic transmission 30 may become large.

[0086] In the present embodiment, the number EN of interventions is used as an input variable in the Fig. which is specified by the mapping data 94A. That is, the number EN of engagements of an engagement element operated during the shift of the automatic transmission 30 from the disengaged state to the engaged state, which is a value strongly correlated with the degree of wear of the engagement element, is input as an input variable. The presence or absence of a failure of the automatic transmission 30 can be accurately evaluated by inputting a value reflecting the degree of wear of the engagement element in this way, compared with a configuration in which the degree of wear of the engagement element is not reflected.

[0087] (7) A force transmitted from the engine 10, the first motor / generator 61, and the second motor / generator 62 to the automatic transmission 30 tends to be large when the accelerator operation amount ACC is large. A force to be transmitted by an engagement element operated from the disengaged state to the engaged state is large when a force transmitted to the automatic transmission 30 during shifting of the automatic transmission 30 is large. Therefore, the shift shock in the automatic transmission 30, that is, the maximum change value Gmax, varies depending on the accelerator operation amount ACC, even when conditions such as the shift type TL are the same.

[0088] The accelerator actuation amount ACC is used as an input variable in the Fig. which is specified by the map data 94A. Consequently, it is possible to obtain an evaluation value reflecting the accelerator operation amount ACC correlated with the maximum change value Gmax. Second embodiment

[0089] The following is a second embodiment of the present invention with reference to the Fig. 5 and Fig. 6. As described in Fig. 5, the second embodiment differs in that map data 94B is stored in advance in the memory 94 instead of the map data 94A. Two types of maps, that is, a first map M1 and a second map M2, are set in the map data 94B as a map for judging the presence or absence of a failure of the automatic transmission 30. The second embodiment further differs in the judgment control. The CPU 91 always executes the judgment control at the end of the shifting operation of the automatic transmission 30. The ROM 93 stores in advance a judgment program, which is a program for the judgment control. The CPU 91 executes the judgment control by executing the judgment program stored in the ROM 93. The following mainly describes the differences of the second embodiment from the first embodiment.Components according to the second embodiment that are similar to each according to the first embodiment are denoted by the same reference numerals to omit or simplify a description thereof.

[0090] As it is in Fig. As shown in Figure 6, at the start of evaluation control, the CPU 91 acquires various types of values ​​by accessing the memory 94 in step S61. The process in step S61 is the same as the process in step S31. Thereafter, the CPU 91 proceeds to the process in step S62.

[0091] In step S62, the CPU 91 generates input variables x (1) to x (6) for the map for judging the presence or absence of a failure of the automatic transmission 30, and then sets various values ​​obtained in the process in step S61 to the input variables x (1) to x (6).

[0092] Specifically, the CPU 91 sets the maximum change value Gmax to the input variable x (1). The CPU 91 sets the execution frequency NL to the input variable x (2). The CPU 91 sets the learning correction value CVL to the input variable x (3). The CPU 91 sets the shift type TL to the input variable x (4). The CPU 91 sets the number of interventions EN to the input variable x (5). The CPU 91 sets the accelerator operation amount ACC to the input variable x (6). Thereafter, the CPU 91 proceeds to step S63.

[0093] In the present embodiment, the input variable x(1) is an acceleration variable, which is a variable indicating the acceleration of the vehicle 100 during the shifting of the automatic transmission 30. The input variable x(2) is a learning progress variable, which is a variable indicating the status of the learning progress in the learning process. The input variable x(4) is a shift type variable, which indicates the type of gear stages before and after the shifting of the automatic transmission 30. The input variable x(5) is a variable indicating the number of engagements, that is, the number of engagements of an engagement element engaged to establish a gear stage after the shifting of the automatic transmission 30. The input variable x(6) is an acceleration variable, which indicates the accelerator operation amount ACC during the shifting of the automatic transmission 30.

[0094] In step S63, the CPU 91 determines whether the learning convergence flag FL is CLEAR. The learning convergence flag FL is a learning progress variable that indicates the status of the learning progress in the learning process. A case where the learning convergence flag FL is CLEAR corresponds to the learning progress variable located in the first range. A case where the learning convergence flag FL is SET corresponds to the learning progress variable located in the second range. If it is determined in step S63 that the learning convergence flag FL is CLEAR (S63: YES), the CPU 91 proceeds to step S71.

[0095] In step S71, the CPU 91 calculates the value of an output variable y(i) by inputting the input variables x(1) to x(6) generated in the process in step S62 and an input variable x(0) as bias parameters to the first map M1 specified by the map data 94B stored in advance in the memory 94. Thereafter, the CPU 91 proceeds to the process in step S81.

[0096] Examples of the first mapping M1 specified by the mapping data 94B include a function approximator and a fully connected feedforward neural network with a single intermediate layer. Specifically, in the first mapping M1 specified by the mapping data 94B, the values ​​of nodes in the intermediate layer are determined by substituting into an activation function p all "m" values ​​obtained by converting the input variables x(1) to x(6) and the input variable x(0) as a bias parameter using a mapping specified by a coefficient wFjk (j = 1 to m, k = 0 to 6). Further, output variables y(1) to y(2) are determined by substituting into an activation function q all values ​​obtained by converting the values ​​of nodes in the intermediate layer using a linear mapping specified by a coefficient wSij (i = 1 to 2).The output variable y(1) is a variable indicating the probability that the automatic transmission 30 is normal. The output variable y(2) is a variable indicating the probability that the automatic transmission 30 has a fault. In the present embodiment, the output variable y(1) and the output variable y(2) are evaluation values ​​indicating the presence or absence of a fault of the automatic transmission 30. In this case, the processes in step S62 and step S71 are the first calculation process. In the present embodiment, examples of the activation function p include a ReLU function. Examples of the activation function q include a soft max function. Thus, the sum of the output variable y(1) and the output variable y(2) is "1."

[0097] The first map M1, which is specified by the map data 94B, is generated, for example, as follows. First, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected by causing a prototype vehicle in which a normal automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100. Further, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected by causing a prototype vehicle in which a faulty automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100.Thereafter, the first map M1 that has been trained is generated by learning using values ​​for a case where the learning convergence flag FL is CLEAR from the various kinds of values ​​collected for the normal automatic transmission 30 and values ​​for a case where the learning convergence flag FL is CLEAR from the various kinds of values ​​collected for the faulty automatic transmission 30 as teacher data.

[0098] On the other hand, if it is determined in step S63 that the learning convergence flag FL is SET (S63: NO), the CPU 91 proceeds to the process in step S72. In step S72, the CPU 91 calculates the value of an output variable y(i) by inputting the input variables x(1) to x(6) generated in the process in step S62 and an input variable x(0) as a bias parameter to the second map M2 specified by the map data 94B stored in advance in the memory 94. Thereafter, the CPU 91 proceeds to the process in step S81.

[0099] Examples of the second mapping M2 specified by the mapping data 94B include a function approximator and a fully connected feedforward neural network with a single intermediate layer. Specifically, in the second mapping M2 specified by the mapping data 94B, the values ​​of nodes in the intermediate layer are determined by substituting, into an activation function r, all "m" values ​​obtained by converting the input variables x(1) to x(6) and the input variable x(0) as a bias parameter using a linear mapping specified by a coefficient wFjk (j=1 to m, k=0 to 6).Further, output variables y(1) to y(2) are determined by substituting, into an activation function s, all values ​​obtained by converting the values ​​of nodes in the intermediate layer using a linear mapping specified by a coefficient wSij (i = 1 to 2). The output variable y(1) is a variable indicating the probability that the automatic transmission 30 is normal. The output variable y(2) is a variable indicating the probability that the automatic transmission 30 has a failure. In the present embodiment, the output variable y(1) and the output variable y(2) are evaluation values ​​indicating the presence or absence of a failure of the automatic transmission 30. In this case, the processes in step S62 and step S72 are the second calculation process. In the present embodiment, examples of the activation function r include a ReLU function.Examples of the activation function s include a soft-max function. Thus, the sum of the output variable y (1) and the output variable y (2) is "1."

[0100] The second map M2, which is specified by the map data 94B, is generated, for example, as follows. First, by causing a prototype vehicle in which a normal automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected. Further, by causing a prototype vehicle in which a faulty automatic transmission 30 is installed to run in various states, etc., before the delivery of the vehicle 100, various types of values ​​about the automatic transmission 30 at the time a shift shock is caused are collected.Thereafter, the second map M2 that has been trained is generated by learning using values ​​for a case where the learning convergence flag FL is SET from the various kinds of values ​​collected for the normal automatic transmission 30 and values ​​for a case where the learning convergence flag FL is SET from the various kinds of values ​​collected for the faulty automatic transmission 30 as teacher data.

[0101] In step S81, the CPU 91 determines whether the output variable y(1) is equal to or lower than the output variable y(2). If it is determined in step S81 that the output variable y(1) is equal to or lower than the output variable y(2) (S81: YES), the CPU 91 proceeds to step S91.

[0102] In step S91, the CPU 91 determines that the automatic transmission 30 has a fault. The CPU 91 then proceeds to step S92. In step S92, the CPU 91 outputs a signal to the display unit 76 to cause the display unit 76 to display that the automatic transmission 30 has a fault. The CPU 91 then terminates the current evaluation control.

[0103] On the other hand, if it is determined in step S81 that the output variable y(1) is not equal to or lower than the output variable y(2) (S81: NO), the CPU 91 proceeds to step S96. In step S96, the CPU 91 determines that the automatic transmission 30 is normal. Thereafter, the CPU 91 terminates the current evaluation control.

[0104] The following describes the functions and effects of the present embodiment. In the present embodiment, in addition to the effects (3) to (7) described above, the following effect (8) is obtained.

[0105] (8) Characteristics of how the presence or absence of a fault affects a shift shock during shifting of the automatic transmission 30 may differ significantly depending on whether the learning has converged in the learning process. When evaluating the presence or absence of a fault in this way, the accuracy of the entire map cannot be guaranteed when attempting to obtain an evaluation value using the same map, even if the characteristics differ significantly depending on the status of the learning process progress.

[0106] In the present embodiment, the first map M1 is used when the learning convergence flag FL is CLEAR, while the second map M2 is used when the learning convergence flag FL is SET. That is, one of the first map M1 and the second map M2 is selectively used in accordance with the learning convergence flag FL, which serves as a learning progress variable, which is a variable indicating the status of the learning progress in the learning process. Consequently, an appropriate value can be output as an output variable, which is an evaluation value indicating the presence or absence of a fault of the automatic transmission 30, regardless of whether the learning has converged in the learning process.Thus, it is possible to prevent the automatic transmission 30 from being judged as faulty even when the maximum change value Gmax has become large when the learning process has not yet fully progressed, such as immediately after delivery of the vehicle 100. That is, in the first embodiment, the presence or absence of a fault of the automatic transmission 30 can be accurately judged regardless of the status of the progress of the learning process. Further embodiments

[0107] The present embodiment may be modified as follows. The present embodiment and the following modifications may be combined with each other, provided that such embodiment and modifications do not technically contradict each other. Acceleration variable

[0108] In the first embodiment and the second embodiment described above, the acceleration variable to be input into the map is not limited to those according to the above embodiments. For example, the up-down acceleration G3 tends to vary in accordance with the magnitude of a shift shock in the automatic transmission 30. Therefore, the maximum value of the amount of fluctuation in the up-down acceleration G3 per unit time during shifting of the automatic transmission 30 may be used as the acceleration variable to be input into the map, instead of or in addition to the maximum change value Gmax, which is the amount of fluctuation in the front-back acceleration G1 per unit time during shifting of the automatic transmission 30.

[0109] For example, when the number of vibrations generated in the vehicle 100 per unit time varies depending on the magnitude of a shift shock in the automatic transmission 30, the number of vibrations generated in the vehicle 100 per unit time can be used as the acceleration variable to be input into the map. The number of vibrations generated in the vehicle 100 per unit time can be determined based on the front-to-rear acceleration G1 or the top-to-bottom acceleration G3. It is not always necessary to use all of the several acceleration variables described above as the acceleration variable to be input into the map, but only to use at least one of these variables.

[0110] Instead of the amount of fluctuation in the front-to-rear acceleration G1 per unit time, the front-to-rear acceleration G1 itself can be input as an input variable. It can be estimated whether the vehicle is accelerating or decelerating during the gearshift process if, for example, the accelerator operation amount ACC or the brake pedal operation amount is input as an input variable. Thus, the presence or absence of a fault of the automatic transmission 30 can be accurately evaluated even using the front-to-rear acceleration G1 itself as an input variable.There is no problem in that the front-to-rear acceleration G1 itself is input as an input variable as long as the series of processes is executed during steady running of the vehicle 100, even if no parameter that provides an estimate of whether the vehicle is accelerating or decelerating during the gear shifting operation is input as an input variable. Learning progress variable

[0111] In the first embodiment and the second embodiment described above, the learning progress variable input to the map is not limited to those according to the above embodiments. For example, the status of the learning progress in the learning process tends to be greater the longer the travel distance the vehicle 100 has traveled since the automatic transmission 30 was installed in the vehicle 100 during the manufacture of the vehicle 100. Therefore, the travel distance of the vehicle 100 can be used as the learning progress variable to be input to the map. With this configuration, the travel distance of the vehicle 100 is input as an input variable, which is a value that strongly correlates with the degree of progress of the learning process. Consequently, an evaluation value that accurately reflects the progress of the learning process can be obtained.For example, when the automatic transmission 30 is replaced during maintenance, etc. of the automatic transmission 30, the time at which a new automatic transmission 30 is installed to replace the automatic transmission 30 corresponds to the time at which the automatic transmission 30 is installed in the vehicle 100.

[0112] It is not always necessary to use the learning convergence flag FL, the execution frequency NL and the travel distance of the vehicle 100 as the learning progress variable to be input into the map, but it is only necessary to use at least one of these variables, for example.

[0113] In the second embodiment described above, the learning progress variable for determining the map to be used in the calculation process is not limited to that according to the second embodiment. For example, the execution frequency NL may be used as the learning progress variable for determining the map to be used in the calculation process. In this case, in step S63, an affirmative determination may be made when the execution frequency NL is equal to or lower than the set, predetermined execution frequency, and a negative determination may be made when the execution frequency NL is not equal to or lower than the set, predetermined execution frequency.

[0114] Accordingly, the distance traveled by the vehicle 100 since the automatic transmission 30 was installed in the vehicle 100 during the manufacture of the vehicle 100 can be used as the learning progress variable for determining the map used in the calculation process.

[0115] In the second embodiment described above, not only two types of maps but three or more types of maps can be selectively used according to the status of the learning progress when the execution frequency NL or the travel distance of the vehicle 100 is used as the learning progress variable for determining the map to be used in the calculation process.Specifically, in a configuration where the learning progress variable is the execution frequency NL, the first mapping may be used when the execution frequency NL is equal to or lower than the first specified execution frequency, the second mapping may be used when the execution frequency NL is higher than the first specified execution frequency and equal to or lower than the second specified execution frequency, which is higher than the first specified execution frequency, and the third mapping may be used when the execution frequency NL is higher than the second specified execution frequency.

[0116] In a configuration where the learning progress variable is used to determine the map to be used in the calculation process, it is not always necessary to input the learning progress variable as an input variable into the map. With this configuration, the presence or absence of a fault of the automatic transmission 30 can be selectively evaluated according to the learning progress status by using two or more map types. Additional input variables

[0117] In the first embodiment and the second embodiment described above, the input variables to be input to the map are not limited to those according to the above embodiments. For example, the degree of wear of an engaging element of the automatic transmission 30 tends to increase with the increase in the number of shifts since the automatic transmission 30 is installed in the vehicle 100 during the manufacture of the vehicle 100. Therefore, the number of shifts to a gear stage after the automatic transmission 30 is shifted can be used as an input variable to be input to the map. With this configuration, the number of shifts to a gear stage after the automatic transmission 30 is shifted, which is a value that strongly correlates with the degree of wear of an engaging element of the automatic transmission 30, is input as an input variable.Consequently, an evaluation value can be obtained which reflects the degree of wear of an engagement element of the automatic transmission 30.

[0118] It is not always necessary to use the shift type TL, the number of engagements EN, the accelerator operation amount ACC, and the number of shifts of the automatic transmission 30 as the input variables to be input to the map, but some of the variables may be omitted as appropriate. That is, in a configuration where the presence or absence of a failure of the automatic transmission 30 is determined using a map type like the first embodiment, it is only necessary that at least the acceleration variable and the learning progress variable should be included as the input variables to be input to the map.On the other hand, in a configuration in which the presence or absence of a failure of the automatic transmission 30 is selectively determined using two or more types of maps based on the learning progress variable as in the second embodiment, it is only necessary that at least the acceleration variable should be included as the input variable to be inputted to the map. Learning process

[0119] In the first embodiment and the second embodiment described above, the learning process is not limited to that according to the above embodiments. For example, the target pressure Z may be calculated by multiplying the predetermined reference pressure ZA by the learning correction value CVL. In this case, the CPU 91 may calculate a new learning correction value CVL in step S21 by adding a predetermined value determined in advance to the learning correction value CVL before the process in step S21. With this configuration, the initial value of the learning correction value CVL for the target engagement element at the time point when the process in step S21 has not yet been executed is "1." Output variable

[0120] In the first embodiment and the second embodiment described above, the output variable of the map is not limited to those according to the above embodiments. For example, it is not always necessary to calculate two evaluation values, that is, a variable indicating the probability that the automatic transmission 30 is normal and a variable indicating the probability that the automatic transmission 30 has a failure, as the output variable of the map, but it is also possible to calculate only one variable indicating the probability that the automatic transmission 30 has a failure. In this case, the CPU 91 can determine that the automatic transmission 30 has a failure if it is determined that the variable indicating the probability that the automatic transmission 30 has a failure is equal to or greater than a predetermined threshold. illustration

[0121] In the first embodiment and the second embodiment described above, the activation functions of the mapping are exemplary and not limited to those according to the above embodiments. For example, a logistic sigmoid function, etc., can be used as the activation functions of the mapping.

[0122] In the first embodiment and the second embodiment described above, a neural network or a neural network with a single intermediate layer is given as an example of the neural network. However, the neural network may also include two or more intermediate layers.

[0123] In the first embodiment and the second embodiment described above, a fully connected feedforward neural network is given as an example of the neural network. However, the present invention is not limited to this. For example, a recurrent neural network can be used as the neural network.

[0124] In the first and second embodiments described above, the function approximator used as the mapping is not limited to a neural network. For example, the function approximator may be a regression formula that does not include an intermediate layer. Error evaluation device

[0125] In the first embodiment and the second embodiment described above, the failure evaluation device is installed in the vehicle 100. However, the present invention is not limited to this. For example, the failure evaluation device may be installed at a dealer, etc., that maintains the vehicle. In this case, the vehicle stores various types of values ​​including at least the acceleration variable and the learning progress variable in the memory 94. The failure evaluation device installed at a dealer, etc., acquires the various types of values ​​stored in the memory 94 of the vehicle during maintenance, etc. of the vehicle. The failure evaluation device can evaluate the presence or absence of a failure of the automatic transmission 30 by calculating an output variable by inputting the various types of acquired values ​​into the map. processor

[0126] In the first embodiment and the second embodiment described above, the processor is not limited to a processor that includes the CPU 91 and the ROM 93 and that executes software processing. As a specific example, the processor may include a dedicated hardware circuit such as an application-specific integrated circuit (ASIC) that performs hardware processing for at least some processes that are subjected to software processing in the above-described embodiments. That is, the processor may include any of the following configurations (a) to (c). (a) The processor includes a processing device that executes all of the above-described processes in accordance with a program, and a program storage device such as a ROM that stores the program.(b) The processor includes a processing device that executes some of the above-described processes in accordance with a program, a program storage device, and a dedicated hardware circuit that executes the remaining processes. (c) The processor includes a dedicated hardware circuit that executes all of the above-described processes. The processor may include multiple software execution devices, each including a processing device and a program storage device, or dedicated hardware circuits. vehicle

[0127] In the first embodiment and the second embodiment described above, the vehicle is a so-called mixed hybrid vehicle. However, the present invention is not limited to this. For example, the vehicle may be a series hybrid vehicle or a parallel hybrid vehicle.

[0128] In the first embodiment and the second embodiment described above, the vehicle is also not limited to a vehicle that includes an internal combustion engine and a motor / generator. For example, the vehicle may be a vehicle that includes an internal combustion engine but not a motor / generator. Further, the vehicle may be a vehicle that includes a motor / generator but not an internal combustion engine.

Claims

[1] A fault evaluation device for an automatic transmission (30), wherein the fault evaluation device evaluates a fault of the automatic transmission (30), and wherein the fault evaluation device is used for a vehicle (100) comprising the automatic transmission (30) and a control device (90) designed to carry out a learning process for correcting a target pressure (Z) for oil to be supplied to the automatic transmission (30) so that fluctuations in the acceleration of the vehicle (100) during shifting of the automatic transmission (30) are small, the fault evaluation device comprising: a processor; and a memory (94), wherein: the memory (94) stores image data defining a first image (M1) and a second image (M2); the processor is designed to output an output variable (y) which is an evaluation value indicating the presence or absence of the fault of the automatic transmission (30) when a plurality of input variables (x) are input; the first map (M1) and the second map (M2) comprise as one of the input variables (x) an acceleration variable, which is a variable indicating the acceleration of the vehicle (100) during the shifting of the automatic transmission (30); the first image (M1) is a trained image trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is located in a first area; the second mapping (M2) is a trained mapping trained by machine learning under the condition that the learning progress variable is in a second range different from the first range; and the processor is designed to execute a determination process, which is a process for determining the input variables (x) and the learning progress variable, a first calculation process for calculating a value of the output variable (y) by inputting the input variable (x) determined by the determination process into the first map (M1) when the learning progress variable determined by the determination process is in the first area, and a second calculation process for calculating a value of the output variable (y) by inputting the input variable (x) determined by the determination process into the second map (M2) when the learning progress variable determined by the determination process is in the second area. [2] The error evaluation device according to claim 1, wherein the learning progress variable is a variable that takes a first value when the learning in the learning process has converged and that takes a second value different from the first value when the learning in the learning process has not converged. [3] The failure evaluation device according to claim 1, wherein the learning progress variable is a variable indicating the execution frequency of the learning process since the automatic transmission (30) is installed in the vehicle (100). [4] The failure evaluation device according to claim 1, wherein the learning progress variable is a travel distance of the vehicle (100) traveled since the automatic transmission (30) was installed in the vehicle (100). [5] An error evaluation device according to any one of claims 1 to 4, wherein: the target pressure (Z) is calculated by adding or multiplying a learning correction value and a reference pressure, which is an oil pressure at a time when the learning process has not yet been carried out; the learning process is a process for calculating the learning correction value such that the fluctuations in the acceleration of the vehicle (100) during the shifting of the automatic transmission (30) are small; and the first map (M1) and the second map (M2) include the learning correction value as the input variable (x). [6] An error evaluation device according to any one of claims 1 to 5, wherein: the automatic transmission (30) comprises a plurality of engagement elements and a plurality of gear stages which are switched by the engagement elements; and the first map (M1) and the second map (M2) comprise as the input variable (x) a shift type variable indicating a type of gear stages before and after the shifting of the automatic transmission (30). [7] An error evaluation device according to claim 6, wherein: the automatic transmission (30) comprises the engagement elements and the gear stages switched by the engagement elements; and the first map (M1) and the second map (M2) comprise as the input variable (x) a variable which indicates the number of gear shifts since the installation of the automatic transmission (30) in the vehicle (100), wherein the number of gear shifts into one of the gear stages was determined after the automatic transmission (30) was shifted. [8] An error evaluation device according to any one of claims 6 or 7, wherein: the automatic transmission (30) comprises the engagement elements and the gear stages switched by the engagement elements; and the first map (M1) and the second map (M2) comprise as the input variable (x) a variable indicating the number of engagements performed by the engagement elements since the automatic transmission (30) was installed in the vehicle (100), the number of engagements being performed by an engagement element of the engagement elements that is engaged to establish one of the gear stages after the automatic transmission (30) is shifted. [9] The failure evaluation device according to any one of claims 1 to 8, wherein said input variable (x) is an accelerator variable indicating an operation amount of an accelerator pedal during shifting of the automatic transmission (30). [10] A fault evaluation method for an automatic transmission (30), wherein the fault evaluation method is used to evaluate a fault of the automatic transmission (30), and wherein the fault evaluation method is used for a vehicle (100) comprising the automatic transmission (30) and a control device (90) designed to carry out a learning process for correcting a target pressure (Z) for oil to be supplied to the automatic transmission (30) so that fluctuations in the acceleration of the vehicle (100) during shifting of the automatic transmission (30) are small, wherein: the error evaluation method is carried out by an error evaluation device; the error evaluation device comprises a processor and a memory (94); the memory (94) stores image data defining a first image (M1) and a second image (M2); the processor is designed to output an output variable (y) which is an evaluation value indicating the presence or absence of the fault of the automatic transmission (30) when a plurality of input variables (x) are input; the first map (M1) and the second map (M2) include an acceleration variable, which is a variable indicating the acceleration of the vehicle (100) during shifting of the automatic transmission (30), as one of the input variables (x); the first image (M1) is a trained image trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is located in a first area; and the second image (M2) is a trained image trained by machine learning under the condition that the learning progress variable is in a second range different from the first range, where the error assessment procedure includes: Entering the acceleration variables and the learning progress variables as the input variables (x) into the error evaluation device; and Calculating a value of the output variable (y) by inputting the input variable (x) into the first map (M1) when the learning progress variable is in the first range, and calculating the value of the output variable (y) by inputting the input variable (x) into the second map (M2) when the learning progress variable is in the second range. [11] A non-volatile storage medium storing a fault evaluation program for an automatic transmission (30), the fault evaluation program being designed to cause a computer to operate as a fault evaluation device that evaluates a fault of the automatic transmission (30), and the fault evaluation program being used for a vehicle (100) comprising the automatic transmission (30) and a control device (90) designed to carry out a learning process for correcting a target pressure (Z) for oil to be supplied to the automatic transmission (30) so that fluctuations in the acceleration of the vehicle (100) during shifting of the automatic transmission (30) are small, wherein: the fault evaluation program has map data defining a first map (M1) and a second map (M2), wherein the first map (M1) and the second map (M2) include, as one of a plurality of input variables (x), an acceleration variable which is a variable indicating the acceleration of the vehicle (100) during shifting of the automatic transmission (30); the first image (M1) is a trained image trained by machine learning under the condition that a learning progress variable, which is a variable indicating a status of the learning progress in the learning process, is located in a first area; the second image (M2) is a trained image trained by machine learning under the condition that the learning progress variable is in a second range different from the first range; and the fault evaluation program is designed to cause the computer to execute a function for determining the input variable (x) and the learning progress variable, a function for calculating a value of an output variable (y), which is an evaluation value indicating the presence or absence of the fault of the automatic transmission (30), by inputting the determined input variable (x) into the first map (M1) when the determined learning progress variable is in the first range, and a function for calculating the value of the output variable (y) by inputting the determined input variable (x) into the second map (M2) when the determined learning progress variable is in the second range.

Images