Control device and control method

Abstract

The purpose of the present invention is to improve the accuracy of control parameters for controlling a control target. This control device is provided with an explanatory variable setting unit, a Mahalanobis distance calculation unit, and a control parameter setting unit. The explanatory variable setting unit sets explanatory variables on the basis of time-series data indicating the status or state of a control target. The Mahalanobis distance calculation unit calculates a Mahalanobis distance on the basis of the explanatory variables. The control parameter setting unit sets control parameters for causing the control target to perform an operation according to the Mahalanobis distance.

Description

Control Device and Control Method

[0001] The present invention relates to a control device and a control method for a vehicle.

[0002] In recent years, in vehicle control, various technologies for performing control using AI (Artificial Intelligence) have been proposed. For example, Patent Document 1 describes a plant state determination device. This plant state determination device inputs the state quantity of the control target into a machine learning model to calculate the Mahalanobis distance. Then, based on the calculated Mahalanobis distance, the plant state determination device performs diagnoses such as sensor failure, sensor degradation determination, and driving region determination.

[0003] Japanese Unexamined Patent Application Publication No. 2019 - 065713

[0004] However, the plant state determination device described in Patent Document 1 performs sensor abnormality determination, sensor degradation determination, and driving state abnormality determination using the Mahalanobis distance, but does not consider the accuracy of the calculated value itself. For example, in an electric motor with a strict temperature upper limit, it is required to expand the driving region to improve the performance per cost. And to expand the driving region, it is necessary to improve the accuracy of the calculated temperature of the electric motor and improve the accuracy of the control parameters.

[0005] In view of the above problems, an object of the present invention is to provide a control device and a control method capable of improving the accuracy of control parameters for controlling a control target.

[0006] To solve the above problems and achieve the above object, a control device according to an aspect of the present invention includes an explanatory variable setting unit, a Mahalanobis distance calculation unit, and a control parameter setting unit. The explanatory variable setting unit sets explanatory variables based on time - series data indicating the situation or state of the control target. The Mahalanobis distance calculation unit calculates the Mahalanobis distance based on the explanatory variables. The control parameter setting unit sets control parameters for causing the control target to execute an operation according to the Mahalanobis distance.

[0007] In one aspect of the present invention, a control method involves an explanatory variable setting unit setting explanatory variables based on time-series data indicating the situation or state of the controlled object. Next, a Mahalanobis distance calculation unit calculates the Mahalanobis distance based on the explanatory variables. Then, a control parameter setting unit sets control parameters to cause the controlled object to perform an action according to the Mahalanobis distance.

[0008] According to one aspect of the present invention, the accuracy of control parameters for controlling a controlled object can be improved. Other problems, configurations, and effects not described above will be clarified by the following description of embodiments.

[0009] This is an overall configuration diagram showing the electric motor system of an electric vehicle according to the first embodiment. This is a block diagram of the control device in the electric motor system according to the first embodiment. This is a block diagram explaining the function of each part in the control device according to the first embodiment. This is a correspondence table showing the relationship between the target variable and explanatory variables of the machine learning unit according to the first embodiment. This is a flowchart showing the parameter calculation process performed by the control parameter setting unit according to the first embodiment. This is a diagram explaining the target torque map, upper limit torque map, and target rotational speed map according to the first embodiment. This is a diagram explaining the coolant pump rotational speed multiplier map, radiator fan PWM table, and radiator fan PWM multiplier map according to the first embodiment. This is a block diagram of the control device in the electric motor system according to the second embodiment. This is a block diagram showing the functional configuration of the control device according to the second embodiment. This is a block diagram of the control device in the electric motor system according to the third embodiment. This is a block diagram showing the functional configuration of the control device according to the third embodiment. This is an overall configuration diagram showing a basic configuration example of an internal combustion engine according to the fourth embodiment. This is a block diagram showing the functional configuration of the control device according to the fourth embodiment. This is a correspondence table showing the relationship between the target variable and explanatory variables of the machine learning unit according to the fourth embodiment. This is a flowchart showing the parameter calculation process performed by the control parameter setting unit according to the fourth embodiment. This graph illustrates the required energization time torque map, upper limit energization time torque map, fuel injection amount map, and fuel enrichment rate map according to the fourth embodiment.

[0010] <First Embodiment> The vehicle control device according to the first embodiment will be described below with reference to Figures 1 to 7. In each figure, common parts are denoted by the same reference numerals.

[0011] [Electric Motor System] First, the configuration of the electric motor system of the electric vehicle according to the first embodiment will be described. Figure 1 is an overall configuration diagram showing the electric motor system of the electric vehicle according to the first embodiment.

[0012] As shown in Figure 1, the electric motor system 1 of the electric vehicle includes a control device 2, a coolant pump controller 3, an electric motor controller 4, a coolant pump 5, an inverter 6, an electric motor 7, a radiator 8, a radiator fan 9, and a coolant path 10. The electric motor system 1 also includes an accelerator opening sensor 11, a current sensor 12, a rotation sensor 13, and a coolant water temperature sensor 14.

[0013] The control device 2, the coolant pump controller 3, the coolant pump 5, the radiator 8, the radiator fan 9, the coolant path 10, and the coolant temperature sensor 14 constitute the cooling system according to the present invention.

[0014] The electric vehicle is equipped with an accelerator position sensor 11 that detects the displacement of the accelerator pedal (accelerator opening). The accelerator opening detected by the accelerator position sensor 11 is output to the control device 2.

[0015] The control device 2 calculates the drive command torque based on the accelerator opening. The control device 2 sends the calculated drive command torque to the motor controller 4. The motor controller 4 generates a drive PWM (Pulse Width Modulation) signal based on the drive command torque sent from the control device 2. The motor controller 4 sends the generated drive PWM signal to the inverter 6.

[0016] The inverter 6 supplies a drive current to the motor 7 based on the drive PWM signal. The current sensor 12 is located on the inverter 6. The current sensor 12 detects the value of the drive current supplied by the inverter 6. The current sensor 12 notifies the motor controller 4 of the detected current value as the inverter current value.

[0017] The electric motor 7 rotates when a drive current flows through it. The electric motor 7 corresponds to the rotating electric machine according to the present invention. The rotation sensor 13 is located on the electric motor 7. The rotation sensor 13 detects the rotation of the electric motor 7. The rotation sensor 13 notifies the electric motor controller 4 of a rotation signal indicating the rotational speed of the electric motor 7.

[0018] The motor controller 4 calculates the rotational speed of the motor 7 based on the rotational signal notified from the rotation sensor 13. The motor controller 4 sends the calculated rotational speed of the motor 7 (hereinafter referred to as "motor rotational speed") to the control device 2. The motor controller 4 also sends the inverter current value notified from the current sensor 12 to the control device 2.

[0019] The inverter 6 and the electric motor 7 are located in the cooling water path 10. A cooling water pump 5 is also located in the cooling water path 10. The cooling water pump 5 circulates the cooling water within the cooling water path 10.

[0020] The cooling water path 10 corresponds to the refrigerant circulation path according to the present invention. The cooling water corresponds to the cooling medium according to the present invention. The cooling water pump 5 corresponds to the cooling medium circulation pump according to the present invention. The inverter 6 and the electric motor 7 correspond to the equipment installed in the refrigerant circulation path according to the present invention.

[0021] The inverter 6 and motor 7 generate heat when a drive current flows through them. Additionally, the motor 7 generates heat through rotation. The heat generated in the inverter 6 and motor 7 is transferred to the cooling water circulating in the cooling water path 10. This cools the inverter 6 and motor 7.

[0022] The cooling water passage 10 downstream of the electric motor 7 passes through the radiator 8. The radiator 8 cools the cooling water circulating in the cooling water passage 10. The radiator fan 9 draws in air from the outside and passes it through the radiator 8. This improves the cooling effect of the radiator 8.

[0023] The coolant temperature sensor 14 is installed, for example, on the outlet side of the radiator 8. The coolant temperature sensor 14 detects the temperature of the coolant in the coolant path 10 on the outlet side of the radiator 8. The coolant temperature sensor 14 notifies the control device 2 of the detected coolant temperature (hereinafter referred to as "radiator outlet coolant temperature").

[0024] The control device 2 calculates the temperature of the motor 7 (hereinafter referred to as "motor temperature") based on the accelerator opening, motor speed, inverter current value, and radiator outlet coolant temperature. Based on the calculated motor temperature, the control device 2 limits the drive command torque. Based on the calculated motor temperature, the control device 2 generates a target rotation speed PWM signal for the radiator fan 9. The control device 2 sends the calculated target rotation speed PWM signal to the radiator fan 9. The radiator fan 9 rotates its blades in accordance with the target rotation speed PWM signal.

[0025] The control device 2 calculates the target rotational speed of the cooling water pump 5 (hereinafter referred to as "target pump rotational speed") based on the calculated motor temperature. The control device 2 sends the calculated target pump rotational speed to the cooling water pump controller 3.

[0026] The cooling water pump controller 3 generates a drive PWM signal based on the target pump rotation speed sent from the control device 2. The cooling water pump controller 3 sends the generated drive PWM signal to the cooling water pump 5. The cooling water pump 5 controls the amount of cooling water discharged according to the drive PWM signal.

[0027] [Control Device] Next, the configuration of the control device 2 in the motor system 1 will be explained with reference to Figure 2. Figure 2 is a block diagram of the control device 2.

[0028] As shown in Figure 2, the control device 2 receives the detection results from various sensors mounted on the vehicle, such as the accelerator pedal position sensor 11 (see Figure 1). The detection results from each sensor are time-series data indicating the status or condition of the controlled object, such as the coolant pump 5 (see Figure 1). Based on the detection results from each sensor, the control device 2 calculates control parameters for controlling the controlled object.

[0029] The control device 2 includes an input processing unit 21, an explanatory variable extraction unit 22, a future data prediction unit 23, a future predicted value calculation unit 24, a future Mahalanobis distance calculation unit 25, and a control parameter setting unit 26. The explanatory variable extraction unit 22 and the future data prediction unit 23 correspond to the explanatory variable setting unit according to the present invention.

[0030] The input processing unit 21 receives time-series data, which are the detection results of each sensor. The input processing unit 21 sends the time-series data necessary for setting the explanatory variables to the explanatory variable extraction unit 22. The explanatory variable extraction unit 22 extracts the data to be defined as explanatory variables from the time-series data sent from the input processing unit 21 and sends it to the future data prediction unit 23.

[0031] The future data prediction unit 23 calculates the target variable (future data) from the explanatory variables sent from the explanatory variable extraction unit 22. This target variable is a predicted value that predicts future data regarding the situation or state of the controlled object. The future data prediction unit 23 sends the calculated target variable (future data) to the future predicted value calculation unit 24 and the future Mahalanobis distance calculation unit 25.

[0032] The future prediction value calculation unit 24 takes the target variable (future data) sent from the future data prediction unit 23 as explanatory variables (future data). The future prediction value calculation unit 24 calculates future prediction values ​​from the explanatory variables (future data). The future prediction values ​​are predicted values ​​that indicate the situation or state of the controlled object necessary for calculating the control parameters. The future prediction value calculation unit 24 sends the calculated future prediction values ​​to the control parameter setting unit 26.

[0033] The future Mahalanobis distance calculation unit 25 takes the target variable (future data) sent from the future data prediction unit 23 as an explanatory variable (future data). The future Mahalanobis distance calculation unit 25 calculates the future Mahalanobis distance from the explanatory variable (future data). The future Mahalanobis distance calculation unit 25 sends the calculated future Mahalanobis distance to the control parameter setting unit 26.

[0034] The control parameter setting unit 26 calculates control parameters based on the status or condition of the controlled object, future predicted values, and future Mahalanobis distance. The control device 2 controls the controlled object using the calculated control parameters. For example, the control device 2 sends the pump rotation speed (see Figure 1), which is one of the control parameters, to the cooling water pump controller 3. The cooling water pump controller 3 controls the operation of the cooling water pump 5 by energizing it with a drive PWM signal corresponding to the pump rotation speed.

[0035] [Functions of each part in the control device] Next, the functions of each part in the control device 2 will be explained with reference to Figure 3. Figure 3 is a block diagram illustrating the functions of each part in the control device 2.

[0036] As shown in Figure 3, the input processing unit 21 sends time-series data to the explanatory variable extraction unit 22. The time-series data includes accelerator opening, radiator outlet coolant temperature, inverter current value, and motor speed (electric motor rotation speed).

[0037] The explanatory variable extraction unit 22 extracts the accelerator opening and motor rotation speed from the time-series data sent from the input processing unit 21 and sends them to the future torque prediction unit 231 of the future data prediction unit 23, which will be described later. The accelerator opening and motor rotation speed are explanatory variables for calculating the motor torque (drive command torque).

[0038] The explanatory variable extraction unit 22 extracts the inverter current value and the radiator outlet coolant temperature from the time-series data sent from the input processing unit 21 and sends them to the future water temperature prediction unit 232 of the future data prediction unit 23, which will be described later. The inverter current value and the radiator outlet coolant temperature are explanatory variables used by the machine learning unit to calculate the temperature of the coolant at the motor (electric motor) inlet.

[0039] The future data prediction unit 23 includes a future torque prediction unit 231 and a future water temperature prediction unit 232.

[0040] The future torque prediction unit 231 calculates the future motor torque (hereinafter referred to as "future motor torque") using a machine learning unit that performs machine learning on time-series data. The machine learning unit is, for example, a neural network model.

[0041] A neural network model is a mathematical model that mimics the structure of the human brain's neural circuits. In this embodiment, it is composed of a multilayer neural network model with an input layer into which explanatory variables are input, an output layer that outputs an objective variable, and an intermediate layer connecting the input and output layers. Neural network models are often used as a means of performing deep learning in so-called machine learning. For example, backpropagation can be applied to the machine learning algorithm. Note that the machine learning unit according to the present invention is not limited to a neural network model, as long as it is possible to calculate motor torque and the like in the future through machine learning.

[0042] The future torque prediction unit 231 calculates the target variable, the future motor torque, using the accelerator opening, motor rotation speed, and the previous value of the future motor torque as explanatory variables. The future torque prediction unit 231 sends the calculated future motor torque to the future water temperature prediction unit 232, the future predicted value calculation unit 24, and the future Mahalanobis distance calculation unit 25.

[0043] The future water temperature prediction unit 232 uses a machine learning unit that performs machine learning on time-series data to calculate the future temperature of the cooling water at the motor inlet (hereinafter referred to as "future motor inlet cooling water temperature"). The future water temperature prediction unit 232 uses the inverter current value, radiator outlet cooling water temperature, future motor torque, and the previous value of the future motor inlet cooling water temperature as explanatory variables to calculate the target variable, the future motor inlet cooling water temperature. The future water temperature prediction unit 232 sends the calculated future motor inlet cooling water temperature to the future prediction value calculation unit 24 and the future Mahalanobis distance calculation unit 25.

[0044] The future predicted value calculation unit 24 calculates the future temperature of the electric motor 7 (hereinafter referred to as "future electric motor temperature") using a machine learning unit that machine-learns time series data. The future predicted value calculation unit 24 calculates the future electric motor temperature, which is the target variable, using the future motor torque and the future motor inlet cooling water temperature as explanatory variables. The future predicted value calculation unit 24 sends the calculated future electric motor temperature to the control parameter setting unit 26.

[0045] The future Mahalanobis distance calculation unit 25 calculates the future Mahalanobis distance (hereinafter referred to as "future Mahalanobis distance") using the future motor torque and the future motor inlet cooling water temperature as parameters. The future Mahalanobis distance calculation unit 25 sends the calculated future Mahalanobis distance to the control parameter setting unit 26.

[0046] The control parameter setting unit 26 calculates various control parameters based on the future electric motor temperature and the future Mahalanobis distance. The calculation of the control parameters will be described later with reference to FIG. 5.

[0047] [Explanatory variables and target variables] Next, the relationship between the target variable and the explanatory variables of the machine learning unit will be described with reference to FIG. 4. FIG. 4 is a correspondence table showing the relationship between the target variable and the explanatory variables of the machine learning unit in the control device 2.

[0048] As shown in FIG. 4, when calculating the motor torque, which is the target variable, using the machine learning unit, the accelerator opening degree and the motor speed are input to the machine learning unit as explanatory variables. Further, the previous value of the motor torque is added to the explanatory variables.

[0049] When calculating the motor inlet cooling water temperature, which is the target variable, using the machine learning unit, the motor torque, the inverter current value, and the radiator outlet cooling water temperature are input to the machine learning unit as explanatory variables. Further, the previous value of the motor inlet cooling water temperature is added to the explanatory variables. When calculating the electric motor temperature, which is the target variable, using the machine learning unit, the motor torque and the motor inlet cooling water temperature are input to the machine learning unit as explanatory variables.

[0050] [Parameter Calculation Process] Next, the parameter calculation process performed by the control parameter setting unit 26 will be explained with reference to Figures 5 to 7. Figure 5 is a flowchart of the parameter calculation process. Figure 6 is a diagram illustrating the target torque map, upper limit torque map, and target rotational speed map. Figure 7 is a diagram illustrating the pump rotational speed multiplier map, radiator fan PWM table, and radiator fan PWM multiplier map.

[0051] First, the control parameter setting unit 26 refers to the target torque map (see Figure 6A) and calculates the target torque based on the accelerator opening and motor rotation speed (S1).

[0052] As shown in Figure 6A, the target torque map is map data that defines multiple target torques by assigning a correspondence between multiple index values ​​related to accelerator opening and multiple index values ​​related to motor rotation speed. The horizontal axis in Figure 6A represents accelerator opening, and the vertical axis in Figure 6A represents motor rotation speed.

[0053] The control parameter setting unit 26 calculates a target torque based on the accelerator opening value and motor rotation speed, according to the values ​​defined in the target torque map. If the point where the accelerator opening value and motor rotation speed intersect lies between multiple index values, the control parameter setting unit 26 calculates the target torque by performing linear interpolation calculations that connect points of multiple close index values.

[0054] Next, the control parameter setting unit 26 refers to the upper limit torque map (see Figure 6B) and calculates the upper limit torque based on the future motor temperature and future Mahalanobis distance (S2).

[0055] As shown in Figure 6B, the upper limit torque map is map data that defines multiple upper limit torques by assigning a correspondence between multiple index values ​​related to future motor temperature and multiple index values ​​related to future Mahalanobis distance. The horizontal axis of Figure 6B represents future motor temperature, and the vertical axis of Figure 6B represents future Mahalanobis distance.

[0056] The control parameter setting unit 26 calculates the upper limit torque based on the values ​​specified in the upper limit torque map, corresponding to the future motor temperature and future Mahalanobis distance. If the point where the future motor temperature and future Mahalanobis distance intersect lies between multiple index values, the control parameter setting unit 26 calculates the upper limit torque by performing a linear interpolation calculation that connects points of multiple close index values.

[0057] Next, the control parameter setting unit 26 refers to the target rotational speed map (see Figure 6C) and calculates the target pump rotational speed based on the target torque and the radiator outlet coolant temperature (S3).

[0058] As shown in Figure 6C, the target rotational speed map is map data that defines multiple target pump rotational speeds by assigning a correspondence between multiple index values ​​related to target torque and multiple index values ​​related to radiator outlet coolant temperature. The horizontal axis of Figure 6C represents the target torque, and the vertical axis of Figure 6C represents the radiator outlet coolant temperature.

[0059] The control parameter setting unit 26 calculates the target pump rotation speed based on the target torque and the radiator outlet coolant temperature, according to the values ​​defined in the target rotation speed map. If the point where the target torque and the radiator outlet coolant temperature intersect lies between multiple index values, the control parameter setting unit 26 calculates the target pump rotation speed by performing a linear interpolation calculation that connects points of multiple close index values.

[0060] Next, the control parameter setting unit 26 refers to the pump rotation speed multiplier map (see Figure 7A) and calculates the pump rotation speed multiplier based on the future motor temperature and future Mahalanobis distance (S4).

[0061] As shown in Figure 7A, the pump speed multiplier map is map data that defines multiple pump speed multipliers by assigning a correspondence between multiple index values ​​related to future motor temperature and multiple index values ​​related to future Mahalanobis distance. The horizontal axis of Figure 7A represents future motor temperature, and the vertical axis of Figure 7A represents future Mahalanobis distance.

[0062] The control parameter setting unit 26 calculates the pump speed multiplier based on the values ​​defined in the pump speed multiplier map, corresponding to the future motor temperature and future Mahalanobis distance. If the point where the future motor temperature and future Mahalanobis distance intersect lies between multiple index values, the control parameter setting unit 26 calculates the pump speed multiplier by performing a linear interpolation calculation that connects points of multiple close index values.

[0063] Next, the control parameter setting unit 26 refers to the radiator fan target PWM table (see Figure 7B) and calculates the radiator fan target PWM based on the radiator outlet coolant temperature (S5).

[0064] As shown in Figure 7B, the radiator fan target PWM table is table data that defines the radiator fan target PWM by associating multiple index values ​​related to the radiator outlet coolant temperature. The horizontal axis of Figure 7B represents the radiator outlet coolant temperature, and the vertical axis of Figure 7B represents the radiator fan target PWM.

[0065] As shown in Figure 7B, the radiator fan target PWM is 0% until the radiator outlet coolant temperature reaches a predetermined lower limit temperature. On the other hand, when the radiator outlet coolant temperature exceeds a predetermined upper limit temperature, the radiator fan target PWM is 100%. The control parameter setting unit 26 calculates the radiator fan target PWM according to the radiator outlet coolant temperature based on the values ​​specified in the radiator fan target PWM table.

[0066] Next, the control parameter setting unit 26 refers to the radiator fan PWM multiplier map (see Figure 7C) and calculates the radiator fan PWM multiplier based on the future motor temperature and future Mahalanobis distance (S6).

[0067] As shown in Figure 7C, the radiator fan PWM multiplier map is map data that defines multiple radiator fan PWM multipliers by assigning a correspondence between multiple index values ​​related to future motor temperature and multiple index values ​​related to future Mahalanobis distance. The horizontal axis of Figure 7C represents future motor temperature, and the vertical axis of Figure 7C represents future Mahalanobis distance.

[0068] The control parameter setting unit 26 calculates the radiator fan PWM multiplier based on the values ​​defined in the radiator fan PWM multiplier map, corresponding to the future motor temperature and future Mahalanobis distance. If the point where the future motor temperature and future Mahalanobis distance intersect lies between multiple index values, the control parameter setting unit 26 calculates the radiator fan PWM multiplier by performing a linear interpolation calculation connecting points of multiple close index values.

[0069] Next, the control parameter setting unit 26 selects the smaller of the target torque calculated in step S1 and the upper limit torque calculated in step S2 to determine the drive command torque (S7). This reduces the margin required to ensure a safety factor and allows for a relaxation of the output limit.

[0070] Next, the control parameter setting unit 26 sends the drive command torque determined in step S7 to the motor controller 4 (see Figure 1) and controls the motor 7, which is the target of control (S8).

[0071] Next, the control parameter setting unit 26 calculates the pump rotation speed (instructed pump rotation speed) by multiplying the target pump rotation speed calculated in step S3 by the pump rotation speed multiplier calculated in step S4 (S9). This reduces the margin required to ensure a safety factor and allows for a relaxation of the output limit.

[0072] Next, the control parameter setting unit 26 sends the pump rotation speed calculated in step S9 to the cooling water pump controller 3 (see Figure 1) and controls the cooling water pump 5, which is the target of control (S10).

[0073] Next, the control parameter setting unit 26 calculates the radiator fan rotation instruction PWM by multiplying the radiator fan target PWM calculated in step S5 by the radiator fan PWM multiplier calculated in step S6 (S11). Next, the control parameter setting unit 26 sends the radiator fan PWM (radiator fan rotation instruction PWM) calculated in step S11 to the radiator fan 9 and controls the drive of the radiator fan 9, which is the target of control (S12). After the processing in step S12, the control parameter setting unit 26 terminates the parameter calculation process.

[0074] The control device 2 of this embodiment calculates future parameters (future motor temperature) and the future Mahalanobis distance of the future parameters for executing predictive operations based on measured time-series data. Then, the control parameter setting unit 26 sets control parameters for controlling the controlled object according to the future parameters and the future Mahalanobis distance. As a result, the control device 2 can improve the accuracy of the control parameters and execute the predictive operations of the controlled object more appropriately compared to the case where control parameters are set without using the future Mahalanobis distance.

[0075] <Second Embodiment> The vehicle control device according to the second embodiment will be described below with reference to Figures 8 and 9. In each figure, common parts are denoted by the same reference numerals.

[0076] The difference between the motor system according to the second embodiment and the motor system 1 according to the first embodiment is the control device 2B. Therefore, the control device 2B will be described here, and the description of the configuration that overlaps with the first embodiment will be omitted.

[0077] [Control Device] First, the configuration of the control device 2B according to the second embodiment will be described with reference to Figure 8. Figure 8 is a block diagram of the control device 2B according to the second embodiment.

[0078] As shown in Figure 8, the control device 2B receives time-series data from each sensor mounted on the vehicle. Based on the time-series data, the control device 2B calculates control parameters for controlling the controlled object.

[0079] The control device 2B includes an input processing unit 21, an explanatory variable extraction unit 22, a future data prediction unit 33, a future predicted value calculation unit 24, a current Mahalanobis distance calculation unit 34, a current Mahalanobis distance storage unit 35, a future Mahalanobis distance calculation unit 36, and a control parameter setting unit 26. The input processing unit 21, the explanatory variable extraction unit 22, the future predicted value calculation unit 24, and the control parameter setting unit 26 are the same as in the first embodiment. The explanatory variable extraction unit 22 and the future data prediction unit 33 correspond to the explanatory variable setting unit according to the present invention.

[0080] The future data prediction unit 33 calculates the future target variable (future data) and the current target variable (current data) from the explanatory variables sent from the explanatory variable extraction unit 22. The future target variable is a predicted value that forecasts future data regarding the situation or state of the controlled object. The current target variable is an estimated value that estimates current data regarding the situation or state of the controlled object.

[0081] The future data prediction unit 33 sends the calculated future target variable (future data) to the future predicted value calculation unit 24. The future data prediction unit 33 also sends the calculated current target variable (current data) to the current Mahalanobis distance calculation unit 34.

[0082] Currently, the Mahalanobis distance calculation unit 34 takes in the current dependent variable (current data) sent from the future data prediction unit 33 as an explanatory variable (current data). The Mahalanobis distance calculation unit 34 calculates the current Mahalanobis distance from the explanatory variable (current data). The Mahalanobis distance calculation unit 34 sends the calculated current Mahalanobis distance to the current Mahalanobis distance storage unit 35.

[0083] The current Mahalanobis distance storage unit 35 stores the current Mahalanobis distance sent from the current Mahalanobis distance calculation unit 34 as time-series data.

[0084] The future Mahalanobis distance calculation unit 36 ​​performs a time-series analysis based on the time-series data currently stored in the Mahalanobis distance storage unit to calculate the future Mahalanobis distance. The future Mahalanobis distance calculation unit 36 ​​sends the calculated future Mahalanobis distance to the control parameter setting unit 26.

[0085] [Functions of each part in the control device] Next, the functions of each part in the control device 2B will be explained with reference to Figure 9. Figure 9 is a block diagram illustrating the functions of each part in the control device 2B.

[0086] As shown in Figure 9, the input processing unit 21 sends time-series data to the explanatory variable extraction unit 22. The time-series data includes accelerator opening, radiator outlet coolant temperature, inverter current value, and motor speed (electric motor rotation speed).

[0087] The explanatory variable extraction unit 22 extracts the accelerator opening, inverter current value, and motor rotation speed from the time-series data sent from the input processing unit 21 and sends them to the current torque estimation unit 331 and future torque prediction unit 332, which will be described later in the future data prediction unit 33. The accelerator opening, inverter current value, and motor rotation speed are explanatory variables used to calculate the motor torque (drive command torque) using the machine learning unit.

[0088] The explanatory variable extraction unit 22 extracts the inverter current value and the radiator outlet coolant temperature from the time-series data sent from the input processing unit 21 and sends them to the current water temperature estimation unit 333 and the future water temperature prediction unit 334, which will be described later in the future data prediction unit 33. The inverter current value and the radiator outlet coolant temperature are explanatory variables for calculating the temperature of the coolant at the inlet of the motor (electric motor).

[0089] The future data prediction unit 33 includes a current torque estimation unit 331, a future torque prediction unit 332, a current water temperature estimation unit 333, and a future water temperature prediction unit 334.

[0090] The current torque estimation unit 331 calculates the current motor torque (hereinafter referred to as "current motor torque") using a machine learning unit that learns time-series data. The current torque estimation unit 331 calculates the current motor torque, which is the target variable, using the inverter current value, accelerator opening, and motor rotation speed as explanatory variables. The current torque estimation unit 331 sends the calculated current motor torque to the future torque prediction unit 332, the current water temperature estimation unit 333, and the current Mahalanobis distance calculation unit 34.

[0091] The future torque prediction unit 332 calculates the future motor torque using a machine learning unit that performs machine learning on time-series data. The future torque prediction unit 332 calculates the target variable, the future motor torque, using the accelerator opening, motor rotation speed, current motor torque, and the previous value of the future motor torque as explanatory variables. The future torque prediction unit 332 sends the calculated future motor torque to the future water temperature prediction unit 334 and the future prediction value calculation unit 24.

[0092] The current water temperature estimation unit 333 uses a machine learning unit that learns time-series data to calculate the current temperature of the coolant at the motor inlet (hereinafter referred to as "current motor inlet coolant temperature"). The current water temperature estimation unit 333 uses the inverter current value, the radiator outlet coolant temperature, and the current motor torque as explanatory variables to calculate the target variable, the current motor inlet coolant temperature. The current water temperature estimation unit 333 sends the calculated current motor inlet coolant temperature to the future water temperature prediction unit 334 and the current Mahalanobis distance calculation unit 34.

[0093] The future water temperature prediction unit 334 calculates the future motor inlet coolant temperature using a machine learning unit that performs machine learning on time-series data. The future water temperature prediction unit 334 calculates the target variable, the future motor inlet coolant temperature, using the inverter current value, radiator outlet coolant temperature, current motor inlet coolant temperature, future motor torque, and the previous value of the future motor inlet coolant temperature as explanatory variables. The future water temperature prediction unit 334 sends the calculated future motor inlet coolant temperature to the future prediction value calculation unit 24.

[0094] The future prediction value calculation unit 24 calculates the future motor temperature using a machine learning unit that performs machine learning on time-series data. The future prediction value calculation unit 24 calculates the future motor temperature, which is the target variable, using the future motor torque and the future motor inlet cooling water temperature as explanatory variables. The future prediction value calculation unit 24 sends the calculated future motor temperature to the control parameter setting unit 26.

[0095] The current Mahalanobis distance calculation unit 34 calculates the current Mahalanobis distance using the current motor torque and the current motor inlet cooling water temperature as parameters. The current Mahalanobis distance calculation unit 34 sends the calculated current Mahalanobis distance to the current Mahalanobis distance storage unit 35. The current Mahalanobis distance storage unit 35 stores the current Mahalanobis distance sent from the current Mahalanobis distance calculation unit 34 as time-series data.

[0096] The future Mahalanobis distance calculation unit 36 ​​calculates the future Mahalanobis distance by performing a time-series analysis based on the time-series data of the current Mahalanobis distance stored in the current Mahalanobis distance storage unit 35. In other words, the future Mahalanobis distance calculation unit 36 ​​analyzes the past change and fluctuation patterns of the time-series data stored in the current Mahalanobis distance storage unit 35 and calculates the future Mahalanobis distance based on that analysis. The future Mahalanobis distance calculation unit 36 ​​passes the calculated future Mahalanobis distance to the control parameter setting unit 26.

[0097] The control parameter setting unit 26 calculates various control parameters based on the future motor temperature and future Mahalanobis distance. The calculation of the control parameters is the same as described with reference to Figure 5.

[0098] The control device 2B of this embodiment calculates future parameters (future motor temperature) and the future Mahalanobis distance of the future parameters for executing predictive operations based on measured time-series data. The control parameter setting unit 26 then sets control parameters for controlling the controlled object according to the future parameters and the future Mahalanobis distance. As a result, the control device 2B can improve the accuracy of the control parameters and execute the predictive operations of the controlled object more appropriately compared to the case where control parameters are set without using the future Mahalanobis distance.

[0099] <Third Embodiment> The vehicle control device according to the third embodiment will be described below with reference to Figures 10 and 11. In each figure, common parts are denoted by the same reference numerals.

[0100] The difference between the motor system according to the third embodiment and the motor system 1 according to the second embodiment is the control device 2C. Therefore, the control device 2C will be described here, and the description of the configuration that overlaps with the first embodiment will be omitted.

[0101] [Control Device] First, the configuration of the control device 2C according to the third embodiment will be described with reference to Figure 10. Figure 10 is a block diagram of the control device 2C according to the third embodiment.

[0102] As shown in Figure 10, the control device 2C receives time-series data from each sensor mounted on the vehicle. Based on the time-series data, the control device 2C calculates control parameters for controlling the controlled object.

[0103] The control device 2C includes an input processing unit 21, a time-series data storage unit 42, a future data prediction unit 43, a future predicted value calculation unit 24, a future Mahalanobis distance calculation unit 25, and a control parameter setting unit 26. The input processing unit 21, the future predicted value calculation unit 24, the future Mahalanobis distance calculation unit 25, and the control parameter setting unit 26 are the same as in the first embodiment. The time-series data storage unit 42 and the future data prediction unit 43 correspond to the explanatory variable setting unit according to the present invention.

[0104] The input processing unit 21 receives time-series data, which is the detection result of each sensor. The input processing unit 21 sends the time-series data to the time-series data storage unit 42. The time-series data storage unit 42 stores the time-series data sent from the input processing unit 21. The time-series data storage unit 42 sends the stored time-series data (hereinafter referred to as "stored time-series data") to the future data prediction unit 43.

[0105] The future data prediction unit 43 calculates explanatory variables (future data) to be used in the future prediction value calculation unit 24 and the future Mahalanobis distance calculation unit 25 based on the stored time series data sent from the time series data storage unit 42. These explanatory variables are predicted values ​​that predict future data regarding the situation or state of the controlled object. The future data prediction unit 43 sends the calculated explanatory variables (future data) to the future prediction value calculation unit 24 and the future Mahalanobis distance calculation unit 25.

[0106] [Functions of each part in the control device] Next, the functions of each part in the control device 2C will be explained with reference to Figure 11. Figure 11 is a block diagram illustrating the functions of each part in the control device 2C.

[0107] As shown in Figure 11, the input processing unit 21 sends time-series data to the time-series data storage unit 42. The time-series data includes accelerator opening, radiator outlet coolant temperature, inverter current value, and motor speed (electric motor rotation speed).

[0108] The time-series data storage unit 42 stores the time-series data sent from the input processing unit 21 and stores the stored time-series data. The time-series data storage unit 42 extracts the stored time-series data of accelerator opening from the stored time-series data and sends it to the future accelerator opening prediction unit 431 of the future data prediction unit 43, which will be described later. The time-series data storage unit 42 extracts the stored time-series data of motor rotation speed from the stored time-series data and sends it to the future motor rotation speed prediction unit 432 of the future data prediction unit 43, which will be described later.

[0109] The time-series data storage unit 42 extracts the stored time-series data of inverter current values ​​from the stored time-series data and sends it to the future inverter current prediction unit 433, which will be described later in the future data prediction unit 43. The time-series data storage unit 42 extracts the stored time-series data of radiator outlet coolant temperature from the stored time-series data and sends it to the future radiator outlet coolant temperature prediction unit 434, which will be described later in the future data prediction unit 43.

[0110] The future data prediction unit 43 includes a future accelerator opening prediction unit 431, a future motor rotation speed prediction unit 432, a future inverter current prediction unit 433, a future radiator outlet coolant temperature prediction unit 434, a future torque prediction unit 435, and a future water temperature prediction unit 436.

[0111] The future accelerator opening prediction unit 431 performs time-series analysis based on the accumulated time-series data of accelerator openings to predict the future accelerator opening (hereinafter referred to as "future accelerator opening"). The future accelerator opening prediction unit 431 sends the calculated future accelerator opening to the future torque prediction unit 435.

[0112] The future motor speed prediction unit 432 performs time-series analysis based on accumulated time-series data of motor speeds to predict future motor speeds (hereinafter referred to as "future motor speeds"). The future motor speed prediction unit 432 sends the calculated future motor speeds to the future torque prediction unit 435.

[0113] The future inverter current prediction unit 433 predicts the future inverter current value (hereinafter referred to as "future inverter current value") based on the future accelerator opening described above. The future inverter current prediction unit 433 sends the calculated future inverter current value to the future torque prediction unit 435 and the future water temperature prediction unit 436.

[0114] The future radiator outlet coolant temperature prediction unit 434 predicts the future radiator outlet coolant temperature (hereinafter referred to as "future radiator outlet coolant temperature") based on the accumulated time-series data of radiator outlet coolant temperature. The future radiator outlet coolant temperature prediction unit 434 sends the calculated future radiator outlet coolant temperature to the future water temperature prediction unit 436.

[0115] The future torque prediction unit 435 calculates the future motor torque using a machine learning unit that performs machine learning on time-series data. The future torque prediction unit 435 calculates the future motor torque, which is the target variable, using the future inverter current value, future accelerator opening, and future motor rotation speed as explanatory variables. The future torque prediction unit 435 sends the calculated future motor torque to the future prediction value calculation unit 24, the future water temperature prediction unit 436, and the future Mahalanobis distance calculation unit 25.

[0116] The future water temperature prediction unit 436 calculates the future motor inlet coolant temperature using a machine learning unit that performs machine learning on time-series data. The future water temperature prediction unit 436 calculates the target variable, the future motor inlet coolant temperature, using the future motor torque, future inverter current value, and future radiator outlet coolant temperature as explanatory variables. The future water temperature prediction unit 436 sends the calculated future motor inlet coolant temperature to the future prediction value calculation unit 24 and the future Mahalanobis distance calculation unit 25.

[0117] The control device 2C of this embodiment calculates future parameters (future motor temperature) and the future Mahalanobis distance of the future parameters for executing predictive operations based on the measured time-series data. Then, the control parameter setting unit 26 sets control parameters for controlling the controlled object according to the future parameters and the future Mahalanobis distance. As a result, the control device 2C can improve the accuracy of the control parameters and execute the predictive operations of the controlled object more appropriately compared to the case where control parameters are set without using the future Mahalanobis distance.

[0118] <Fourth Embodiment> The vehicle control device according to the fourth embodiment will be described below with reference to Figures 12 to 16. In each figure, common parts are denoted by the same reference numerals.

[0119] [Internal Combustion Engine System] First, the configuration of the internal combustion engine system according to the fourth embodiment will be described. Figure 12 is an overall configuration diagram showing an example of the basic configuration of an internal combustion engine according to the fourth embodiment of the present invention.

[0120] The internal combustion engine 100 shown in Figure 12 may have a single cylinder or multiple cylinders, but in the fourth embodiment, a four-cylinder internal combustion engine 100 mounted on a vehicle (mobile body) will be used as an example for explanation.

[0121] As shown in Figure 12, in the internal combustion engine 100, air drawn in from the outside (intake) flows through the air cleaner 110, intake pipe 111, and intake manifold 112. The air that has passed through the intake manifold 112 flows into each cylinder 150 when the intake valve 151 opens. The amount of air flowing into each cylinder 150 is adjusted by the throttle valve 113. The amount of air adjusted by the throttle valve 113 is measured by the flow sensor 114.

[0122] The throttle valve 113 is equipped with a throttle opening sensor 113a for detecting the throttle opening degree. The throttle opening degree information of the throttle valve 113 detected by the throttle opening sensor 113a is output to the control unit (Electronic Control Unit: ECU) 2D.

[0123] In this embodiment, an electronic throttle valve driven by an electric motor is used as the throttle valve 113. However, other types of throttle valves may be used as the throttle valve according to the present invention, as long as they can appropriately adjust the airflow rate.

[0124] The temperature of the air flowing into each cylinder 150 is detected by the intake air temperature sensor 115.

[0125] A crank angle sensor 121 is provided on the radially outer side of the ring gear 120 attached to the crankshaft 123. The crank angle sensor 121 detects the rotation angle of the crankshaft 123. In this embodiment, the crank angle sensor 121 detects the rotation angle of the crankshaft 123 every 10° and for each combustion cycle.

[0126] A water temperature sensor 122 is provided in the water jacket (not shown) of the cylinder head. The water temperature sensor 122 detects the temperature of the coolant in the internal combustion engine 100.

[0127] The vehicle is also equipped with an accelerator position sensor 126 that detects the displacement (pressure amount; accelerator opening) of the accelerator pedal 125. The accelerator opening detected by the accelerator position sensor 126 is output to a control device 2D, which will be described later. The control device 2D calculates the driver's requested torque based on the accelerator opening and controls the throttle valve 113 based on this requested torque.

[0128] Fuel stored in the fuel tank 130 is drawn in and pressurized by the fuel pump 131. The fuel drawn in and pressurized by the fuel pump 131 is adjusted to a predetermined pressure by a pressure regulator 132 installed in the fuel piping 133. The fuel adjusted to the predetermined pressure is then injected into each cylinder 150 from the fuel injector 134. Any excess fuel after pressure adjustment by the pressure regulator 132 is returned to the fuel tank 130 via a return pipe (not shown).

[0129] The fuel injection device 134 is controlled based on the fuel injection pulse (control signal) of the control parameter setting unit 26 (see Figure 13) of the control device 2D, which will be described later.

[0130] An in-cylinder pressure sensor (also called a combustion pressure sensor) 140 is provided in the cylinder head (not shown) of the internal combustion engine 100. The in-cylinder pressure sensor 140 is installed in each cylinder 150 and detects the pressure (combustion pressure) inside the cylinder 150. For example, a piezoelectric or gauge-type pressure sensor is used for the in-cylinder pressure sensor 140. This makes it possible to detect the in-cylinder pressure inside the cylinder 150 over a wide temperature range.

[0131] Each cylinder 150 is fitted with an exhaust valve 152 and an exhaust manifold 160. When the exhaust valve 152 opens, the combustion gases, i.e., exhaust gases, are discharged from the cylinder 150 into the exhaust manifold 160. The exhaust manifold 160 discharges the exhaust gases to the outside of the cylinder 150. A three-way catalytic converter 161 is positioned on the exhaust side of the exhaust manifold 160. The three-way catalytic converter 161 purifies the exhaust gases. The exhaust gases purified by the three-way catalytic converter 161 are discharged into the atmosphere.

[0132] An upstream air-fuel ratio sensor 162 is provided upstream of the three-way catalytic converter 161. The upstream air-fuel ratio sensor 162 outputs a signal corresponding to the oxygen concentration related to the air-fuel ratio of the exhaust gas discharged from each cylinder 150. The upstream air-fuel ratio sensor 162 in this embodiment is a so-called linear air-fuel ratio sensor that detects the air-fuel ratio (oxygen concentration) of the exhaust gas discharged from each cylinder 150 as a voltage that changes proportionally (linearly) to the air-fuel ratio.

[0133] Furthermore, a downstream air-fuel ratio sensor 163 is provided downstream of the three-way catalytic converter 161. The downstream air-fuel ratio sensor 163 outputs a signal corresponding to the oxygen concentration related to the air-fuel ratio of the exhaust gas purified by the three-way catalytic converter 161. The downstream air-fuel ratio sensor 163 in this embodiment is a so-called O2 sensor that outputs a binary detection signal that changes depending on whether the air-fuel ratio is richer or leaner than the stoichiometric air-fuel ratio.

[0134] Each cylinder 150 has a spark plug 200 located in a position facing each combustion chamber. The spark plug 200 generates a spark through discharge (ignition), and this spark ignites the air-fuel mixture inside the cylinder 150. This causes explosive combustion inside the cylinder 150, pushing down the piston 170. The pushing down of the piston 170 causes the crankshaft 123 to rotate. An ignition coil 300 is connected to the spark plug 200, which generates (boosts) the discharge voltage supplied to the spark plug 200.

[0135] The output signals from various sensors, such as the throttle opening sensor 113a, flow sensor 114, crank angle sensor 121, accelerator position sensor 126, water temperature sensor 122, and cylinder pressure sensor 140, are input to the control device 2D. Based on these signals from various sensors, the control device 2D controls the amount of air passing through the throttle valve 113, the amount of exhaust gas returning to the intake side through the EGR valve (not shown), the fuel injection amount of the fuel pump 131 and fuel injector 134, and the ignition timing of the spark plug 200 by the ignition coil 300.

[0136] The control device 2D includes an analog input unit 51a, a digital input unit 51b, an A / D (Analog / Digital) conversion unit 61, a RAM (Random Access Memory) 62, an MPU (Micro-Processing Unit) 63, a ROM (Read Only Memory) 64, an I / O (Input / Output) port 65, and an output circuit 66.

[0137] The analog input section 51a receives analog output signals from various sensors, including a throttle opening sensor 113a, a flow sensor 114, an accelerator position sensor 126, an upstream air-fuel ratio sensor 162, a downstream air-fuel ratio sensor 163, an in-cylinder pressure sensor 140, and a water temperature sensor 122.

[0138] An A / D converter 61 is connected to the analog input unit 51a. Analog output signals from various sensors input to the analog input unit 51a are processed, such as noise reduction, and then converted into digital signals by the A / D converter 61. The digital signals converted by the A / D converter 61 are then stored in the RAM 62.

[0139] The digital input unit 51b receives the digital output signal sent from the crank angle sensor 121. An I / O port 65 is connected to the digital input unit 51b. The digital output signal input to the digital input unit 51b is stored in the RAM 62 via the I / O port 65.

[0140] Each output signal stored in RAM 62 is processed by MPU 63.

[0141] The MPU 63 processes the output signals stored in the RAM 62 according to the control program by executing a control program (not shown) stored in the ROM 64. The MPU 63 calculates control values ​​that define the operating amounts of each actuator (e.g., throttle valve 113, pressure regulator 132, spark plug 200, etc.) that drives the internal combustion engine 100 according to the control program, and temporarily stores these control values ​​in the RAM 62.

[0142] The control value that defines the amount of actuator operation stored in RAM 62 is output to output circuit 66 via I / O port 65. Output circuit 66 outputs signals to control the controlled components such as the throttle valve 113, pressure regulator 132, and spark plug 200.

[0143] [Functional Configuration of the Control Device] Next, the functional configuration of the control device 2D will be explained with reference to Figure 13. Figure 13 is a block diagram illustrating the functions of each part of the control device 2D.

[0144] As shown in Figure 13, the control device 2D includes an input processing unit 51, a time-series data storage unit 52, a future data prediction unit 53, a future predicted value calculation unit 54, a future Mahalanobis distance calculation unit 55, and a control parameter setting unit 56. The input processing unit 51 includes an analog input unit 51a and a digital input unit 51b as shown in Figure 12. The time-series data storage unit 52 and the future data prediction unit 53 correspond to the explanatory variable setting unit according to the present invention.

[0145] The input processing unit 51 receives time-series data, which is information derived from the detection results of each sensor or control parameters inside the control device 2D. The time-series data includes crankshaft rotation speed (engine speed), power supply voltage, intake air flow rate, intake air pressure, intake air temperature, intake air humidity, rainfall detection means, coolant temperature, cooling air velocity, vehicle speed, ignition cylinder number, ignition timing, number of ignition discharges / cycle, time since engine start, and time since engine stop. The input processing unit 51 sends the time-series data to the time-series data storage unit 52.

[0146] The time-series data storage unit 52 stores the time-series data sent from the input processing unit 51. The time-series data storage unit 52 extracts the crankshaft rotation speed from the stored time-series data and sends it to the future crankshaft rotation speed prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts the power supply voltage value from the stored time-series data and sends it to the future power supply voltage prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts the intake air flow rate from the stored time-series data and sends it to the future intake air flow rate prediction unit of the future data prediction unit 53, which will be described later.

[0147] The time-series data storage unit 52 extracts intake pressure from the stored time-series data and sends it to the future intake pressure prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts intake temperature from the stored time-series data and sends it to the future intake temperature prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts intake humidity from the stored time-series data and sends it to the future intake humidity prediction unit of the future data prediction unit 53, which will be described later.

[0148] The time-series data storage unit 52 extracts rainfall detection results from the stored time-series data and sends them to the future rainfall prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts cooling water temperature from the stored time-series data and sends it to the future cooling water temperature prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts cooling wind speed from the stored time-series data and sends it to the future cooling wind speed prediction unit of the future data prediction unit 53, which will be described later.

[0149] The time-series data storage unit 52 extracts the vehicle speed from the stored time-series data and sends it to the future vehicle speed prediction unit of the future data prediction unit 53, which will be described later. The time-series data storage unit 52 extracts the required torque from the stored time-series data and sends it to the future ignition cylinder number prediction unit of the future data prediction unit 53, which will be described later.

[0150] The future data prediction unit 53 includes a future crankshaft rotation speed prediction unit, a future power supply voltage prediction unit, a future intake air flow rate prediction unit, a future intake air pressure prediction unit, a future intake air temperature prediction unit, and a future intake air humidity prediction unit. Furthermore, the future data prediction unit 53 includes a future rainfall prediction unit, a future coolant temperature prediction unit, a future cooling air velocity prediction unit, a future vehicle travel speed prediction unit, a future ignition cylinder number prediction unit, and a future ignition timing prediction unit.

[0151] The future crankshaft rotation speed prediction unit performs time-series analysis based on accumulated time-series data of crankshaft rotation speed to predict the future crankshaft rotation speed (hereinafter referred to as "future crankshaft rotation speed"). The future crankshaft rotation speed prediction unit sends the calculated future crankshaft rotation speed to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0152] The future power supply voltage prediction unit performs time-series analysis based on accumulated time-series data of power supply voltage values ​​to predict future power supply voltage values ​​(hereinafter referred to as "future power supply voltage values"). The future power supply voltage prediction unit sends the calculated future power supply voltage values ​​to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0153] The future intake flow rate prediction unit performs time-series analysis based on accumulated time-series data of intake flow rate to predict future intake flow rate (hereinafter referred to as "future intake flow rate"). The future intake flow rate prediction unit sends the calculated future intake flow rate to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0154] The future intake pressure prediction unit performs a time-series analysis based on accumulated time-series data of intake pressure to predict future intake flow rate (hereinafter referred to as "future intake pressure"). The future intake pressure prediction unit sends the calculated future intake pressure to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0155] The future intake air temperature prediction unit performs time-series analysis based on accumulated time-series data of intake air temperature to predict the future intake air temperature (hereinafter referred to as "future intake air temperature"). The future intake air temperature prediction unit sends the calculated future intake air temperature to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0156] The future intake humidity prediction unit performs a time-series analysis based on accumulated time-series data of intake humidity to predict future intake humidity (hereinafter referred to as "future intake humidity"). The future intake humidity prediction unit sends the calculated future intake humidity to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0157] The future rainfall prediction unit performs time-series analysis based on the accumulated time-series data of rainfall detection results to predict future rainfall (hereinafter referred to as "future rainfall"). The future rainfall prediction unit sends the calculated future rainfall to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0158] The future cooling water temperature prediction unit performs time-series analysis based on accumulated time-series data of cooling water temperature to predict future cooling water temperature (hereinafter referred to as "future cooling water temperature"). The future cooling water temperature prediction unit sends the calculated future cooling water temperature to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0159] The future cooling wind speed prediction unit performs time-series analysis based on accumulated time-series data of cooling wind speed to predict future cooling wind speed (hereinafter referred to as "future cooling wind speed"). The future cooling wind speed prediction unit sends the calculated future cooling wind speed to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0160] The future vehicle speed prediction unit performs time-series analysis based on accumulated time-series data of vehicle speeds to predict future vehicle speeds (hereinafter referred to as "future vehicle speeds"). The future vehicle speed prediction unit sends the calculated future vehicle speeds to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0161] The future ignition cylinder number prediction unit performs time-series analysis based on accumulated time-series data of required torque to predict future required torque (hereinafter referred to as "future required torque"). Furthermore, based on the predicted future required torque, the future ignition cylinder number prediction unit predicts the cylinder number that will perform future ignition (hereinafter referred to as "future ignition cylinder number"). The future ignition cylinder number prediction unit sends the calculated future ignition cylinder number to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0162] The future ignition timing prediction unit predicts the future ignition timing (hereinafter referred to as "future ignition timing") based on multiple future data, including the future intake airflow rate and future crankshaft rotation speed mentioned above. The future ignition timing prediction unit sends the calculated future ignition timing to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0163] The future ignition discharge count / cycle prediction unit predicts the future discharge count / cycle (hereinafter referred to as "future discharge count / cycle") based on multiple future data, including the future crankshaft rotation speed and future power supply voltage values ​​mentioned above. The future ignition discharge count / cycle prediction unit sends the calculated future discharge count / cycle to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55.

[0164] The future engine start time prediction unit predicts the future engine start time (hereinafter referred to as "future engine start time") by adding a predetermined prediction period value to the current engine start time measurement data measured in the control device 2D when the current crankshaft rotation speed is above a predetermined value. The future engine start time prediction unit sends the calculated future engine start time to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55. In this embodiment, the future engine start time is set to zero when the current crankshaft rotation speed is below a value at which it is determined that the crankshaft rotation has substantially stopped.

[0165] The future engine shutdown time prediction unit predicts the future engine shutdown time (hereinafter referred to as "future engine shutdown time") by adding a predetermined prediction period value to the engine shutdown time measurement data measured in the control device 2D when the current crankshaft rotation speed is less than or equal to the value at which the crankshaft rotation is judged to have substantially stopped. The future engine shutdown time prediction unit sends the calculated future engine shutdown time to the future prediction value calculation unit 54 and the future Mahalanobis distance calculation unit 55. In this embodiment, if the current crankshaft rotation speed is greater than or equal to a predetermined value, the future engine shutdown time is set to zero.

[0166] The future prediction value calculation unit 54 calculates the future temperature of the ignition coil 300 (hereinafter referred to as "future ignition coil temperature") using a machine learning unit that performs machine learning on time-series data. The future prediction value calculation unit 54 calculates the future ignition coil temperature using the following explanatory variables: future crankshaft rotation speed, future power supply voltage value, future intake air flow rate, future intake air pressure, future intake air temperature, future intake air humidity, future rainfall, future coolant temperature, future cooling air velocity, future vehicle travel speed, future ignition cylinder number, future ignition timing, future ignition discharge count / cycle, future time after engine start, and future time after engine stop. The future prediction value calculation unit 54 sends the calculated future ignition coil temperature to the control parameter setting unit 56.

[0167] The future Mahalanobis distance calculation unit 55 calculates the future Mahalanobis distance using the following parameters: future crankshaft rotation speed, future power supply voltage, future intake air flow rate, future intake pressure, future intake air temperature, future intake air humidity, future rainfall, future coolant temperature, future cooling air velocity, future vehicle travel speed, future ignition cylinder number, future ignition timing, future ignition discharge count / cycle, future time after engine start, and future time after engine stop. The future Mahalanobis distance calculation unit 55 sends the calculated future Mahalanobis distance to the control parameter setting unit 56.

[0168] The control parameter setting unit 56 calculates various control parameters based on the future ignition coil temperature and the future Mahalanobis distance. The calculation of the control parameters will be explained later with reference to Figure 15.

[0169] [Explanatory Variables and Dependent Variable] Next, the relationship between the dependent and independent variables of the machine learning unit will be explained with reference to Figure 14. Figure 14 is a correspondence table showing the relationship between the dependent and independent variables of the machine learning unit in the control device 2D.

[0170] As shown in Figure 14, when using the machine learning unit to calculate the target variable, ignition coil temperature, the following are input to the machine learning unit as explanatory variables: crankshaft rotation speed, power supply voltage, intake air flow rate, intake pressure, intake air temperature, intake air humidity, rainfall, coolant temperature, cooling air velocity, vehicle speed, ignition cylinder number, ignition timing, number of ignition discharges / cycle, time since engine start, and time since engine stop.

[0171] [Parameter Calculation Process] Next, the parameter calculation process performed by the control parameter setting unit 56 will be explained with reference to Figures 15 and 16. Figure 15 is a flowchart illustrating the parameter calculation process according to the fourth embodiment. Figure 16 is a diagram illustrating the requested energization time map, the upper limit energization time map, the fuel injection amount map, and the fuel enrichment rate map.

[0172] First, the control parameter setting unit 56 refers to the requested energizing time map (see Figure 16A) and calculates the requested energizing time for the ignition coil 300 based on the power supply voltage and engine speed (crankshaft rotation speed) (S21).

[0173] As shown in Figure 16A, the requested energizing time map is map data that defines multiple requested energizing times by assigning a correspondence between multiple index values ​​related to power supply voltage and multiple index values ​​related to engine speed. The horizontal axis of Figure 16A represents power supply voltage, and the vertical axis of Figure 16A represents engine speed.

[0174] The control parameter setting unit 56 calculates the required energizing time based on the power supply voltage value and engine speed, according to the values ​​specified in the required energizing time map. If the point where the power supply voltage value and engine speed intersect lies between multiple index values, the control parameter setting unit 56 calculates the required energizing time by performing a linear interpolation calculation that connects points of multiple close index values.

[0175] Next, the control parameter setting unit 56 refers to the upper limit energizing time map (see Figure 16B) and calculates the upper limit energizing time based on the future ignition coil temperature and the future Mahalanobis distance (S22).

[0176] As shown in Figure 16B, the upper limit energizing time map is map data that defines multiple upper limit energizing times by relating multiple index values ​​related to future ignition coil temperature and multiple index values ​​related to future Mahalanobis distance. The horizontal axis of Figure 16B represents future ignition coil temperature, and the vertical axis of Figure 16B represents future Mahalanobis distance.

[0177] The control parameter setting unit 56 calculates the upper limit energizing time based on the values ​​specified in the upper limit energizing time map, according to the future ignition coil temperature and future Mahalanobis distance. If the point where the future ignition coil temperature and future Mahalanobis distance intersect lies between multiple index values, the control parameter setting unit 56 calculates the upper limit energizing time by performing a linear interpolation calculation that connects points of multiple close index values.

[0178] Next, the control parameter setting unit 56 refers to the target air-fuel ratio map (see Figure 16C) and calculates the target air-fuel ratio based on the throttle opening and engine speed (S23).

[0179] As shown in Figure 16C, the target air-fuel ratio map is map data that defines multiple target air-fuel ratios by associating multiple index values ​​related to throttle opening with multiple index values ​​related to engine speed. The horizontal axis in Figure 16C represents throttle opening, and the vertical axis in Figure 16C represents engine speed.

[0180] The control parameter setting unit 56 calculates a target air-fuel ratio based on the throttle opening and engine speed, using values ​​defined in the target air-fuel ratio map. If the point where the throttle opening and engine speed intersect lies between multiple index values, the control parameter setting unit 56 calculates the target air-fuel ratio by performing linear interpolation calculations that connect points of multiple close index values.

[0181] Next, the control parameter setting unit 56 refers to the fuel enrichment rate map (see Figure 16D) and calculates the fuel enrichment rate based on the future ignition coil temperature and the future Mahalanobis distance (S24).

[0182] As shown in Figure 16D, the fuel enrichment rate map is map data that defines multiple fuel enrichment rates by relating multiple index values ​​related to future ignition coil temperature and multiple index values ​​related to future Mahalanobis distance. The horizontal axis of Figure 16D represents future ignition coil temperature, and the vertical axis of Figure 16D represents future Mahalanobis distance.

[0183] The control parameter setting unit 56 calculates a fuel enrichment rate based on the values ​​specified in the fuel enrichment rate map, corresponding to the future ignition coil temperature and the future Mahalanobis distance. If the point where the future ignition coil temperature and the future Mahalanobis distance intersect lies between multiple index values, the control parameter setting unit 56 calculates the fuel enrichment rate by performing a linear interpolation calculation that connects points of multiple close index values.

[0184] Next, the control parameter setting unit 56 selects the smaller of the requested energizing time calculated in step S21 and the upper limit energizing time calculated in step S22 to determine the energizing time (S25). Next, the control parameter setting unit 56 sends the energizing time determined in step S25 to the ignition control unit (not shown) of the control device 2D and controls the spark plug 200 which is the object of control (S26).

[0185] Next, the control parameter setting unit 56 multiplies the target air-fuel ratio calculated in step S23 by the fuel enrichment rate calculated in step S24 to calculate the fuel injection amount based on the final target air-fuel ratio and air flow rate (S27). Next, the control parameter setting unit 56 sends the fuel injection amount calculated in step S27 to the fuel injection control unit (not shown) of the control device 2D to control the fuel injection device (injector) 134, which is the target of control (S28). After the processing in step S28, the control parameter setting unit 56 terminates the parameter calculation process.

[0186] The control device 2D of this embodiment calculates future parameters (future ignition coil temperature) and the future Mahalanobis distance of the future parameters for executing predictive operations based on measured time-series data. Then, the control parameter setting unit 56 sets control parameters for controlling the controlled object according to the future parameters and the future Mahalanobis distance. As a result, the control device 2D can improve the accuracy of the control parameters and execute the predictive operations of the controlled object more appropriately compared to the case where control parameters are set without using the future Mahalanobis distance.

[0187] The present invention is not limited to the embodiments described above and shown in the drawings, and various modifications can be made without departing from the spirit of the invention as described in the claims.

[0188] Furthermore, the embodiments described above are explained in detail for the purpose of clearly illustrating the present invention, and are not necessarily limited to those comprising all the described configurations. It is also possible to replace parts of the configuration of one embodiment with those of another embodiment, and to add configurations from other embodiments to the configuration of one embodiment. Additionally, it is possible to add, delete, or replace parts of the configuration of each embodiment with those of other embodiments.

[0189] 1...Electric motor system, 2, 2B, 2C, 2D...Control device, 3...Cooling water pump controller, 4...Electric motor controller, 5...Cooling water pump, 6...Inverter, 7...Electric motor, 8...Radiator, 9...Radiator fan, 10...Cooling water path, 11...Accelerator opening sensor, 12...Current sensor, 13...Rotation sensor, 14...Cooling water temperature sensor, 21, 51...Input processing unit, 22...Explanatory variable extraction unit, 23, 33, 43, 53...Future data prediction unit, 24, 54...Future predicted value calculation unit, 25, 36, 55...Future Mahalanobis distance calculation unit, 26, 56...Control parameter setting unit, 34...Current Mahalanobis distance calculation unit, 35...Current Mahalanobis distance storage unit, 42, 52...Time series data storage unit, 61...A / D conversion unit, 62...RAM, 63...MPU, 64...ROM, 65...I / O port, 66...Output circuit, 100...Internal combustion engine, 110...Air cleaner, 111...Intake pipe, 112...Intake manifold, 113...Throttle valve, 113a...Throttle position sensor, 114...Flow sensor, 115...Intake air temperature sensor, 120...Ring gear, 121...Crank angle sensor, 122...Water temperature sensor, 123...Crankshaft, 125...Accelerator pedal, 126...Accelerator position sensor, 130...Fuel tank, 131...Fuel pump, 132...Pressure regulator, 133...Fuel piping, 134...Fuel injector, 140...In-cylinder pressure sensor, 150...Cylinder, 151...Intake valve, 152...Exhaust valve, 160... Exhaust manifold, 161... Three-way catalytic converter, 162... Upstream air-fuel ratio sensor, 163... Downstream air-fuel ratio sensor, 170... Piston, 200... Spark plug, 231, 332, 435... Future torque prediction unit, 232, 334, 436... Future water temperature prediction unit, 300... Ignition coil, 331... Current torque estimation unit, 333... Current water temperature estimation unit, 431... Future accelerator opening prediction unit, 432... Future motor rotation speed prediction unit, 433... Future inverter current prediction unit, 434... Future radiator outlet coolant temperature prediction unit

Claims

1. A control device comprising: an explanatory variable setting unit that sets explanatory variables based on time-series data indicating the status or state of a controlled object; a Mahalanobis distance calculation unit that calculates the Mahalanobis distance based on the explanatory variables; and a control parameter setting unit that sets control parameters for causing the controlled object to perform an action according to the Mahalanobis distance calculated by the Mahalanobis distance calculation unit.

2. The control device according to claim 1, wherein the explanatory variable setting unit comprises a time series data storage unit for storing the time series data, and a future data prediction unit for predicting future data based on the time series data stored in the time series data storage unit, and the Mahalanobis distance calculation unit calculates the future Mahalanobis distance using the future data as parameters.

3. The control device according to claim 1, wherein the explanatory variable setting unit has a future data prediction unit that inputs the explanatory variables to a machine learning unit that performs machine learning on the time series data and calculates a future target variable, and the Mahalanobis distance calculation unit calculates the future Mahalanobis distance using the future target variable as a parameter.

4. The control device according to claim 1, wherein the Mahalanobis distance calculation unit comprises: a current Mahalanobis distance calculation unit that calculates the current Mahalanobis distance based on the explanatory variables; a current Mahalanobis distance storage unit that stores the current Mahalanobis distance calculated by the current Mahalanobis distance calculation unit; and a future Mahalanobis distance calculation unit that calculates the future Mahalanobis distance based on a plurality of current Mahalanobis distances stored in the current Mahalanobis distance storage unit, and the control parameter setting unit sets control parameters for causing the controlled object to perform an operation according to the future Mahalanobis distance.

5. The control device according to claim 1, wherein the controlled object is a rotating electric machine of a vehicle, the time-series data includes parameters related to the field magnet temperature of the rotating electric machine, and the control parameters are the amount of current supplied to the rotating electric machine, the target regeneration amount, or the output torque.

6. The control device according to claim 1, wherein the controlled object is the ignition device of an internal combustion engine of a vehicle, the time-series data includes parameters related to the temperature of the ignition device, and the control parameters are the energization time to the ignition device, the ignition timing, or the number of ignitions per predetermined unit time.

7. The control device according to claim 1, wherein the controlled object is a vehicle cooling system having a cooling medium circulation pump that circulates a cooling medium in a refrigerant circulation path in which one or more devices are installed, the time-series data includes parameters relating to the temperature of the cooling medium, the temperature of the devices, or the ambient temperature, and the control parameters are parameters relating to the discharge amount of the cooling medium circulation pump or parameters relating to instructions given to the refrigerant control valve of the cooling medium circulation pump.

8. The control device according to any one of claims 5 to 7, wherein the time-series data includes parameters relating to the vehicle's operating state, the vehicle's operating environment, or the driving characteristics of the driver operating the vehicle.

9. A control method comprising: an explanatory variable setting unit setting explanatory variables based on time-series data indicating the situation or state of a controlled object; a Mahalanobis distance calculation unit calculating the Mahalanobis distance based on the explanatory variables; and a control parameter setting unit setting control parameters for causing the controlled object to perform an action according to the Mahalanobis distance.

Images