JP7750210B2 - Learning device and learning model generation method - Google Patents

Description

The present invention relates to a learning device and a learning model generation method.

Patent Document 1 discloses a conventional hybrid vehicle control device that is configured to change the engine start determination threshold associated with the driver's driving force request operation in response to fluctuations in the stiffness of a mount member disposed between the internal combustion engine and the vehicle body. According to Patent Document 1, this allows the load applied to the mount member when starting the hybrid vehicle's internal combustion engine to be adjusted in response to fluctuations in the mount's stiffness due to factors such as ambient temperature and deterioration, thereby enabling the mount member to effectively absorb vibrations from the internal combustion engine and prevent them from being transmitted to the vehicle body.

JP 2013-107512 A

However, the technology in Patent Document 1 mentioned above changes the engine start determination threshold associated with the driving force request operation of a hybrid vehicle by taking into account fluctuations in mount hardness due to factors such as the effects of ambient temperature and deterioration, and because it relates to a control device for a hybrid vehicle, it has the problem of being difficult to apply to vehicles other than hybrid vehicles. Furthermore, while it is possible to attach a body vibration measurement sensor capable of measuring body vibrations, such as an acceleration sensor, to the vehicle body and control the excitation force of the internal combustion engine based on the detection value of the body vibration measurement sensor, it is not realistic to attach body vibration measurement sensors to all vehicles available on the market.

The present invention was developed in response to these problems, and aims to enable changes in the vibration isolation characteristics of mount members to be acquired and used to control the internal combustion engine, regardless of vehicle type, even in vehicles that do not have a vehicle body vibration measurement sensor installed.

To solve the above problem, a learning device according to one aspect of the present invention generates a learning model by performing machine learning using a training dataset that includes input parameters that affect the vibration isolation characteristics of a mount member placed between an internal combustion engine and a vehicle body, and output parameters that indicate the vibration isolation characteristics of the mount member. The input parameters include the temperature in the area where the mount member is used, time series data on changes in the ambient temperature of the mount member over a specified period, and the number of years the mount member has been in use, and the output parameters include vehicle body vibration when the internal combustion engine is started.

In addition, a learning model generation method according to one aspect of the present invention generates a learning model by having a learning device perform machine learning using a training dataset that includes input parameters that affect the vibration isolation characteristics of a mount member placed between an internal combustion engine and a vehicle body, and output parameters that indicate the vibration isolation characteristics of the mount member. The input parameters include the temperature in the area where the mount member is used, time-series data on changes in the ambient temperature of the mount member over a predetermined period, and the number of years the mount member has been in use, and the output parameters include vehicle body vibration when the internal combustion engine is started.

According to these aspects of the present invention, a learning model is generated using machine learning to learn about changes in the vibration isolation characteristics of the mount member. Therefore, regardless of vehicle type, even in vehicles that do not have a body vibration measurement sensor installed, the learning model can be used to obtain changes in the vibration isolation characteristics of the mount member and use them in controlling the internal combustion engine.

1 is a schematic configuration diagram of a learning model generation and utilization system according to one embodiment of the present invention. FIG. 10 is a diagram illustrating an example of a learning model. FIG. 2 is a schematic diagram illustrating a hardware configuration of a data providing vehicle. 10A and 10B are diagrams illustrating a method for acquiring the thermal history of a vibration-isolating rubber. FIG. 2 is a schematic diagram showing the hardware configuration of a vehicle using a learning model. 10 is a flowchart illustrating an example of a training data set providing process executed between a data providing vehicle and a server. 10 is a flowchart illustrating an example of a learning process for a learning model executed by a server. 10 is a flowchart illustrating an example of a learning model utilization process executed between a learning model utilization vehicle and a server. FIG. 10 is a diagram illustrating how an increase in vehicle body vibration can be suppressed by changing the operating point of the internal combustion engine.

One embodiment of the present invention will now be described in detail with reference to the drawings. In the following description, similar components will be designated by the same reference numerals.

An internal combustion engine installed in a vehicle is attached to the body (the vehicle's framework) via a mounting member, and vibration-damping rubber, one of the components that make up the mounting member, prevents the vibrations of the internal combustion engine from being transmitted to the body.

The hardness of the vibration-damping rubber of the mount member, i.e., the vibration-damping characteristics of the mount member, change depending on factors such as the temperature of the region in which the vehicle is used, the heat history of the vibration-damping rubber from surrounding heat sources (the thermal history of the vibration-damping rubber), and the number of years the vibration-damping rubber has been used. If the vibration-damping characteristics of the mount member deteriorate from their initial characteristics due to these influences, vehicle body vibration will increase accordingly, and the vehicle's ride quality will deteriorate. Therefore, if it were possible to control the excitation force that excites the internal combustion engine in accordance with changes in the vibration-damping characteristics of the mount member, it would be possible to prevent a deterioration in the vehicle's ride quality, even if the vibration-damping characteristics of the mount member deteriorate from their initial characteristics.

For vehicles equipped with a sensor capable of measuring body vibrations, such as an acceleration sensor (hereinafter referred to as a "body vibration measurement sensor"), changes in the vibration isolation characteristics of the mount member can be identified by periodically measuring body vibrations while the engine is running under the same conditions using the body vibration measurement sensor and comparing the results. However, from the perspective of cost and other factors, it is not realistic to install body vibration measurement sensors on all vehicles available on the market.

In this embodiment, a learning model (artificial intelligence model) is generated using machine learning to learn about changes in the vibration isolation characteristics of mount members, and by using this learning model, it is possible to obtain changes in the vibration isolation characteristics of mount members even in vehicles that do not have body vibration measurement sensors installed. Below, we will explain the learning model generation and utilization system 100 according to one embodiment of the present invention, with reference to Figure 1 and other figures.

Figure 1 is a schematic diagram of a learning model generation and utilization system 100 according to this embodiment.

The learning model generation and utilization system 100 includes a server 1, one or more data providing vehicles 2, and one or more learning model utilization vehicles 3.

The server 1 includes a communication unit 11, a memory unit 12, and a processing unit 13.

The communication unit 11 has a communication interface circuit for connecting the server 1 to the network 4, for example, via a gateway, and is configured to be able to communicate with each of the data providing vehicles 2 and the learning model using vehicles 3 via the network 4.

The memory unit 12 has storage media such as a hard disk drive (HDD), solid state drive (SSD), optical recording medium, and semiconductor memory, and stores various computer programs and data used for processing by the processing unit 13.

The processing unit 13 has one or more CPUs (Central Processing Units) and their peripheral circuits, and is, for example, a processor, which executes various computer programs stored in the memory unit 12 to comprehensively control the overall operation of the server 1. As an example of the processing performed by the processing unit 13, the processing unit 13 uses a training dataset provided by the data-providing vehicle 2 to learn (generate and re-learn) a learning model used to control the internal combustion engine installed in the learning-model-using vehicle 3.

In this embodiment, the learning model is a neural network model that uses a deep neural network (DNN) or a convolutional neural network (CNN), and is subjected to deep learning, a machine learning method. Therefore, the learning model in this embodiment can also be said to be a trained NN model that has undergone deep learning.

Figure 2 shows an example of a learning model.

The circles in Figure 2 represent artificial neurons. Artificial neurons are usually called nodes or units (referred to as "nodes" in this specification). In Figure 2, L=1 indicates the input layer, L=2 and L=3 indicate hidden layers, and L=4 indicates the output layer. Hidden layers are also called intermediate layers. Note that Figure 2 illustrates a neural network model with two hidden layers, but the number of hidden layers is not particularly limited, and the number of nodes in each of the input, hidden, and output layers is also not particularly limited.

In Figure 2, _x1 and _x2 indicate each node in the input layer (L = 1) and its output value, y indicates a node in the output layer (L = 4) and its output value, _z1 ^{(L = 2)} _{, z2} ^{(L = 2)} , and _z3 ^{(L = 2)} indicate each node in the hidden layer (L = 2) and its output value, and _z1 ^{(L = 3)} and _z2 ^{(L = 3)} indicate each node in the hidden layer (L = 3) and its output value.

Each node in the input layer outputs the input as is. Meanwhile, each node in the hidden layer (L=2) receives the output values _x1 and _x2 of each node in the input layer, and each node in the hidden layer (L=2) calculates a total input value u using the corresponding weight w and bias b. For example, the total input value u _k (L=2) calculated at each node indicated by z _k ^(L=2) (k=1, 2, 3) in the hidden layer ( ^L=2 ) in Figure 2 is expressed by the following equation (M is the number of nodes in the input layer):

Next, this total input value u _k ^(L=2) is transformed by an activation function f, and is output as an output value z _k (L=2) (=f(u _k ( ^L =2))) from the node indicated by z _k ^(L=2) in the hidden layer ( ^L=2 ). Meanwhile, the output values z ₁ (L=2) _, z ₂ ^(L=2) , and z ₃ ^{(L=2) of each node in the hidden layer (L=2} ) are input to each node in the hidden layer ( ^L=3) , and each node in the hidden layer (L=3) calculates a total input value u (=Σz·w+b) using the corresponding weight w and bias b. This total input value u is similarly transformed by an activation function, and is output as output values z ₁ ^(L=3) , z ₂ ^{(L=3) from each node in the hidden layer (L=3} ). The activation function is, for example, a sigmoid function σ.

Furthermore, the output values _z1 (L=3) and _z2 (L=3) of each node in the hidden layer ( ^L=3 ) are input to the nodes in the output layer ^(L=4 ), and the nodes in the output layer calculate the total input value u(Σz·w+b) using the corresponding weights w and biases b, or calculate the total input value u(Σz·w) using only the corresponding weights w. For example, the nodes in the output layer use an identity function as their activation function. In this case, the total input value u calculated in the nodes in the output layer is output directly from the nodes in the output layer as the output value y.

As such, the learning model according to this embodiment comprises an input layer, a hidden layer, and an output layer, and is configured so that when one or more input parameters are input from the input layer, one output parameter corresponding to the input parameters can be output from the output layer. Note that it can also be configured so that when one or more input parameters are input from the input layer, multiple output parameters corresponding to the input parameters can be output from the output layer.

To improve the accuracy of a learning model, the learning model must be trained. To train a learning model, a large number of training data sets are used, including actual measured values of input parameters and actual measured values of output parameters (ground truth data) corresponding to these actual measured values of input parameters. By using a large number of training data sets and repeatedly updating the values of the weights w and bias b in the neural network using the well-known backpropagation method, the values of the weights w and bias b are learned, improving the accuracy of the learning model.

Returning to Figure 1, the data providing vehicle 2 is a vehicle that generates the training dataset necessary to conduct learning of the learning model used by the learning model using vehicle 3, and provides the training dataset to the server 1. The detailed hardware configuration of the data providing vehicle 2 will be described later with reference to Figure 3.

The learning model-using vehicle 3 is a vehicle that uses a learning model as necessary to control the internal combustion engine installed in the vehicle. The use of the learning model in the learning model-using vehicle 3 may be performed on the learning model-using vehicle 3 or on the server 1. For example, the learning model-using vehicle 3 can use the learning model on the learning model-using vehicle 3 by acquiring the learning model itself from the server 1. Also, for example, the learning model-using vehicle 3 can use the learning model on the server 1 by sending input parameters acquired on the learning model-using vehicle 3 to the server 1 and receiving output parameters obtained by inputting the input parameters into the learning model on the server 1 from the server 1. The detailed hardware configuration of the learning model-using vehicle 3 will be described later with reference to Figure 5.

Figure 3 is a schematic diagram showing the hardware configuration of the data providing vehicle 2.

The data-providing vehicle 2 includes an electronic control unit 20, an external communication device 24, various control components 25 including an internal combustion engine attached to the vehicle body via a mount member, and various sensors 26 required to control the control components 25 and to generate a training dataset used by the server 1 when performing machine learning on changes in the vibration-damping characteristics of the mount member. The electronic control unit 20, external communication device 24, control components 25, and sensors 26 are connected to each other via an in-vehicle network 27 that complies with standards such as CAN (Controller Area Network).

The electronic control unit 20 includes an in-vehicle communication interface 21, a memory unit 22, and a processing unit 23. The in-vehicle communication interface 21, the memory unit 22, and the processing unit 23 are connected to each other via signal lines.

The in-vehicle communication interface 21 is a communication interface circuit for connecting the electronic control unit 20 to the in-vehicle network 27.

The memory unit 22 has storage media such as an HDD, SSD, optical recording medium, or semiconductor memory, and stores various computer programs and data used for processing by the processing unit 23.

The processing unit 23, which is, for example, a processor, has one or more CPUs and their peripheral circuits, and executes various computer programs stored in the memory unit 22 to provide overall control of the data providing vehicle 2.

The external vehicle communication device 24 is an on-board terminal with wireless communication capabilities. The external vehicle communication device 24 connects to the network 4 via the wireless base station 5 (see Figure 1) by accessing the wireless base station 5, which is connected to the network 4 via a gateway (not shown). This allows mutual communication between the data-providing vehicle 2 and the server 1.

As mentioned above, the vibration-damping characteristics of the mount member (the hardness of the vibration-damping rubber of the mount member) change depending on factors such as the temperature in the area where the vehicle is used, the heat history that the vibration-damping rubber receives from surrounding heat sources (the thermal history of the vibration-damping rubber (the history of temperature changes around the vibration-damping rubber)), and the number of years the vibration-damping rubber has been used.

For example, anti-vibration rubber tends to be harder when the temperature in the area where the vehicle is used is low compared to when it is high. Furthermore, anti-vibration rubber gradually deteriorates and hardens due to the heat it receives from surrounding heat sources such as the internal combustion engine. In other words, the hardness of anti-vibration rubber changes depending on its thermal history. Furthermore, anti-vibration rubber gradually deteriorates and hardens over the years of use.

In this embodiment, the data-providing vehicle 2 acquires the outside air temperature, the thermal history of the vibration-isolating rubber, and the number of years the vibration-isolating rubber has been in use as input parameters for the learning model, and acquires the vehicle body vibration when the internal combustion engine is started while the vehicle is stopped as an output parameter corresponding to these input parameters.

To this end, the data providing vehicle 2 is equipped with a first temperature sensor as one of the sensors 26 (sensors for acquiring input parameters) that is necessary for acquiring the outside air temperature or a parameter correlated with the outside air temperature. Examples of parameters correlated with the outside air temperature include the temperature of the engine coolant during cold start, the temperature of the engine lubricating oil, or the temperature of the transmission lubricating oil. The data providing vehicle 2 according to this embodiment is equipped with an outside air temperature sensor as the first temperature sensor that directly detects the outside air temperature.

The data-providing vehicle 2 also includes a second temperature sensor, which is necessary for acquiring the thermal history of the vibration-isolating rubber, as one of the sensors 26 (sensors for acquiring input parameters). The thermal history of the vibration-isolating rubber can be, for example, as shown in FIG. 4, the integrated temperature value (the amount of heat input to the vibration-isolating rubber during a predetermined period) obtained by subtracting the outside air temperature from the ambient temperature of the vibration-isolating rubber, acquired at a predetermined sampling period (e.g., 10 seconds) during a predetermined period, such as while the vehicle is in operation (e.g., the period from when the vehicle's start switch is turned on to when it is turned off, or the period from when the vehicle's start switch is turned on to when a predetermined time has elapsed since it was turned off). In other words, the thermal history of the vibration-isolating rubber is time-series data on changes in the ambient temperature of the mount member during a predetermined period. Therefore, the second temperature sensor is a sensor capable of acquiring the ambient temperature of the vibration-isolating rubber or a parameter correlated with the ambient temperature of the vibration-isolating rubber. Parameters that are correlated with the ambient temperature of the vibration-damping rubber include the temperature of the engine coolant or engine lubricant in the internal combustion engine near the vibration-damping rubber mounting position, and the temperature of the transmission lubricant in the transmission near the vibration-damping rubber mounting position. In this embodiment, a water temperature sensor that detects the temperature of the internal combustion engine coolant is used as the second temperature sensor.

The data providing vehicle 2 also includes a vehicle body vibration measuring sensor as one of the sensors 26 (sensors for acquiring output parameters). The vehicle body vibration measuring sensor can be attached to the vehicle body, for example, under the seat.

Figure 5 is a schematic diagram showing the hardware configuration of the learning model-using vehicle 3.

The learning model-using vehicle 3 includes an electronic control unit 30 having an in-vehicle communication interface 31, a memory unit 32, and a processing unit 33, an external communication device 34, various control components 35 including an internal combustion engine attached to the vehicle body via mounting members, and various sensors 36 required to control the control components 35 and to acquire input parameters to be input into the learning model. The electronic control unit 30, external communication device 34, control components 35, and sensors 36 are connected to each other via an in-vehicle network 37 that complies with standards such as CAN.

As such, the hardware configuration of the learning model-using vehicle 3 is basically the same as that of the data-providing vehicle 2, except that the learning model-using vehicle 3 does not have a vehicle body vibration measurement sensor, so a detailed explanation will be omitted here.

Below, an example of the processing performed by the server 1, data providing vehicle 2, and learning model using vehicle 3 is described.

Figure 6 is a flowchart showing an example of the training dataset provision process executed between the data providing vehicle 2 and the server 1.

In step S1, the electronic control unit 20 of the data-providing vehicle 2 acquires the actual measured values of the input parameters and output parameters of the learning model used by the learning-model-using vehicle 3, generates a training dataset necessary to train the learning model, and stores the generated training dataset in the memory unit 22.

In this embodiment, the electronic control unit 20 of the data-providing vehicle 2 acquires, as actual measured values of input parameters, the outside air temperature at vehicle start, the thermal history of the vibration-isolating rubber at vehicle start (i.e., the thermal history of the vibration-isolating rubber at the end of the previous vehicle run or a predetermined time after the end of the vehicle run), and the age of the vibration-isolating rubber at vehicle start. The age of the vibration-isolating rubber at vehicle start can be, for example, the elapsed time since the data-providing vehicle 2 was lined off. The electronic control unit 20 of the data-providing vehicle 2 then acquires, as actual measured values of output parameters, the vehicle body vibration when the internal combustion engine is started for the first time while the vehicle is stopped after start-up. By limiting the timing of acquiring vehicle body vibration to when the vehicle is stopped, it is possible to eliminate the effects of external disturbances such as vibrations transmitted from the suspension to the vehicle body. Naturally, the timing of acquiring the actual measured values of input parameters and output parameters is not limited to these timings.

In step S2, the electronic control unit 20 of the data-providing vehicle 2 determines whether the amount of data in the training data set stored in the memory unit 22 is equal to or greater than the specified transmission amount. If the amount of data in the training data set is equal to or greater than the transmission amount, the electronic control unit 20 proceeds to processing in step S12. On the other hand, if the amount of data in the training data set is less than the transmission amount, the electronic control unit 20 ends this processing.

In step S3, the electronic control unit 20 of the data-providing vehicle 2 transmits the training dataset together with the vehicle type information of the data-providing vehicle 2 to the server 1, and after transmission, erases the data of the training dataset stored in the memory unit 22.

Vehicle type information is information about the vehicle type, such as the vehicle's body shape (sedan, minivan, SUV, etc.), drive system (FF, FR, MR, RR, AWD, etc.), and whether it is a hybrid vehicle. The training dataset is sent along with this vehicle type information because the installation position of the mounting member may differ depending on the vehicle type, and as a result, the sensor used as the second temperature sensor, for example, may differ depending on the vehicle type. Therefore, generating a learning model for each vehicle type can improve the accuracy of the learning model.

In step S4, when the server 1 receives the training dataset and vehicle type information from the data-providing vehicle 2, it references the vehicle type information, selects a database that matches the vehicle type of the data-providing vehicle 2 that is the source of the training dataset from among multiple databases prepared for each vehicle type, and stores the received training dataset in the selected database. Each database is formed in the memory unit 12. Examples of multiple datasets prepared for each vehicle type include, but are not limited to, a database for FR sedans, a database for FR SUVs, a database for FF sedans, a database for FF SUVs, a database for FF minivans, and a database for AWD sedans.

Figure 7 is a flowchart showing an example of the learning process for the learning model executed on server 1.

In step S21, server 1 determines whether there is any database in which the increase in the amount of data in the training dataset since a predetermined time is equal to or greater than a predetermined learning start amount. The predetermined time can be, for example, the time when the learning model was last trained using the training dataset. If there is any database in which the increase in the amount of data in the training dataset since the predetermined time is equal to or greater than the learning start amount, server 1 proceeds to processing in step S22. On the other hand, if there is no database in which the increase in the amount of data in the training dataset since the predetermined time is equal to or greater than the learning start amount, server 1 terminates this processing.

In step S22, server 1 uses a training dataset stored in a database where the increase in the data volume of the training dataset since a predetermined time is equal to or greater than the learning start volume to train (re-train) the learning model of the vehicle type corresponding to that database, and updates the learning model of that vehicle type.

In step S23, server 1 deletes the data in the training dataset in the database used to train the learning model.

Figure 8 is a flowchart showing an example of the learning model usage process executed between the learning model usage vehicle 3 and the server 1.

In step S31, the electronic control unit 30 of the learning model-using vehicle 3 acquires input parameters to be input into the learning model. In this embodiment, the electronic control unit 30 of the learning model-using vehicle 3 acquires as input parameters the outside air temperature at vehicle start, the thermal history of the vibration-isolating rubber at vehicle start (i.e., the thermal history of the vibration-isolating rubber at the end of the previous vehicle run), and the number of years of use of the vibration-isolating rubber at vehicle start.

In step S32, the electronic control unit 30 of the learning model-using vehicle 3 transmits the acquired input parameters and vehicle type information of the vehicle to the server 1.

In step S33, when the server 1 receives input parameters and vehicle type information from the learning model-using vehicle 3, it inputs the input parameters into the learning model corresponding to the vehicle type of the learning model-using vehicle 3 and obtains output parameters. In this embodiment, the output parameters obtained in step S33 are predicted vehicle body vibration values when the internal combustion engine of the learning model-using vehicle 3, which is the source of the input parameters, is started.

In step S34, the server 1 transmits the acquired output parameters to the learning model-using vehicle 3, which is the source of the input parameters.

In step S35, the electronic control unit of the learning model-using vehicle 3 receives the output parameters (predicted vehicle body vibration values when the internal combustion engine is started) from the server 1, and controls the internal combustion engine based on the output parameters so that the vehicle body vibration when the internal combustion engine is running falls within the desired vehicle body vibration value range (the allowable range of vehicle body vibration values when the vibration-damping rubber of the mounting member has its initial characteristics).

Here, internal combustion engines are vibrated by the repeated reciprocating motion of the piston due to the inertial force of the piston caused by sudden changes in piston speed near top dead center and bottom dead center. Because the inertial force of the piston is proportional to the square of the engine rotational speed, the vibrating force caused by the inertial force of the piston generally increases as the engine rotational speed increases. Furthermore, when combustion pressure is applied to the piston, a torque is suddenly applied to the crankshaft in the rotational direction, and a reaction torque acts on the internal combustion engine in the opposite direction to the rotational direction as a reaction force. Therefore, internal combustion engines are also vibrated by this reaction torque, and the vibrating force caused by the reaction torque generally increases as the combustion pressure increases.

Therefore, for example, as shown in Figure 9, by changing the operating point of the internal combustion engine from operating point A on the reference operating line to operating point B on the low rotational speed side along the equal power line, it is possible to suppress an increase in vehicle body vibration caused by a deterioration in the vibration-damping characteristics of the vibration-damping rubber. In this case, if the thermal efficiency at operating point B is worse than the thermal efficiency at normal operating point A, this means that the combustion pressure has decreased, and the increase in vehicle body vibration can be further suppressed.

In this embodiment, the operating point of the internal combustion engine is changed to suppress an increase in vehicle body vibration caused by a deterioration in the vibration-damping characteristics of the vibration-damping rubber of the mount member. However, other methods may also be used to reduce the excitation force by retarding the ignition timing to reduce the combustion pressure. Furthermore, if the learning model-using vehicle 3 is a hybrid vehicle, the output of the internal combustion engine may be reduced and the reduced amount compensated for by the output of the electric motor.

Note that the flowchart shown in Figure 8 explains an example in which the learning model in the learning model-using vehicle 3 is used on the server 1, but as mentioned above, the learning model itself may be obtained from the server 1 so that the learning model can be used on the learning model-using vehicle 3.

The server 1 (learning device) according to the present embodiment described above is configured to generate a learning model by performing machine learning using a training dataset that includes input parameters that affect the vibration isolation characteristics of the mount member placed between the internal combustion engine and the vehicle body, and output parameters that indicate the vibration isolation characteristics of the mount member. The input parameters include the temperature in the area where the mount member is used, time-series data on changes in the ambient temperature of the mount member over a specified period, and the number of years the mount member has been in use, and the output parameters include vehicle body vibration when the internal combustion engine is started.

In this way, by generating a learning model that uses machine learning to learn about changes in the vibration-damping characteristics of the mount member, it is possible to obtain changes in the vibration-damping characteristics of the mount member by using this learning model, regardless of vehicle type, even in vehicles that do not have body vibration measurement sensors installed.As a result, the excitation force that excites the internal combustion engine can be controlled in accordance with changes in the vibration-damping characteristics of the mount member, and deterioration in vehicle ride comfort can be suppressed even if the vibration-damping characteristics of the mount member deteriorate from their initial characteristics.

For example, the learning model generation and utilization system 100 according to this embodiment includes a server 1 (learning device) and a learning model utilization vehicle 3 configured to be able to communicate with the server 1. The electronic control unit 30 (control device) of the learning model utilization vehicle 3 acquires input parameters related to the vehicle itself, and controls the internal combustion engine based on output parameters obtained by inputting the input parameters into the learning model, so as to suppress an increase in vehicle body vibration of the vehicle itself caused by a deterioration in the vibration-damping characteristics of the mount member.

More specifically, the electronic control unit 30 of the learning model-using vehicle 3 is configured to suppress an increase in vehicle body vibration caused by a deterioration in the vibration-damping characteristics of the mount member by changing the operating point of the internal combustion engine to an operating point on the side where the engine rotation speed decreases at equal output or an operating point on the side where thermal efficiency decreases.

The learning model generation and utilization system 100 according to this embodiment further includes a data-providing vehicle 2 configured to be able to communicate with a server 1 (learning device), and the electronic control unit 20 (control device) of the data-providing vehicle 2 is configured to acquire input parameters and output parameters related to the vehicle to generate a training dataset and transmit the generated training dataset together with the vehicle type of the vehicle to the server 1, and the server 1 is configured to separate and store the received training dataset by vehicle type, and generate a learning model for each vehicle type based on the training dataset separated by vehicle type.

The installation position of the mounting components may differ depending on the vehicle type, and the types and installation positions of sensors that acquire input and output parameters may also differ depending on the vehicle type. Therefore, by generating a learning model for each vehicle type based on training datasets separated by vehicle type in this way, the accuracy of the learning model can be improved.

When using a learning model to control the internal combustion engine of the vehicle, the electronic control unit 30 (control device) of the learning model-using vehicle 3 can be configured to send input parameters acquired by the vehicle and the vehicle type of the vehicle to the server 1 (learning device), receive output parameters from the server 1 obtained by inputting the input parameters into a learning model corresponding to the vehicle type of the vehicle, and control the internal combustion engine based on the received output parameters.

For example, the electronic control unit 30 (control device) of the learning model-using vehicle 3 can be configured to receive a learning model corresponding to the vehicle type of the vehicle from the server 1 (learning device), and control the internal combustion engine based on the output parameters obtained by inputting the acquired input parameters into the received learning model.

The above describes embodiments of the present invention, but these embodiments merely illustrate some of the application examples of the present invention, and are not intended to limit the technical scope of the present invention to the specific configurations of the above embodiments.

For example, in the above embodiment, the vehicle body vibration when the internal combustion engine is started is obtained as an output parameter, but the output parameter obtained from the learning model is not limited to such direct vehicle body vibration values, and may be, for example, a correction coefficient for the initial vibration value (actual vehicle body vibration value/initial vibration value).

1 Server 2 Data providing vehicle 3 Learning model using vehicle 100 Learning model generation and use system

Claims (5)

A learning model generation and utilization system including a learning device and a learning model utilization vehicle configured to be able to communicate with the learning device,
The learning device
generating a learning model by performing machine learning using a training dataset including input parameters that affect the vibration isolation characteristics of a mount member disposed between an internal combustion engine and a vehicle body, and output parameters that indicate the vibration isolation characteristics of the mount member;
The input parameters include the temperature of the area where the mounting member is used, time series data of changes in the ambient temperature of the mounting member over a predetermined period of time, and the number of years the mounting member has been in use;
the output parameters include vehicle body vibration when the internal combustion engine is started ;
The learning model-using vehicle is
the internal combustion engine attached to a vehicle body via the mount member;
a control device;
Equipped with
The control device of the learning model-using vehicle
acquiring the input parameters relating to the host vehicle;
the internal combustion engine is controlled based on the output parameters obtained by inputting the input parameters into the learning model so as to suppress an increase in body vibration of the vehicle caused by a deterioration in the vibration-damping characteristics of the mount member;
The operating point of the internal combustion engine is changed to an operating point on the side where the engine rotation speed decreases at the same output or an operating point on the side where the thermal efficiency decreases, thereby suppressing an increase in body vibration of the vehicle caused by a deterioration in the vibration isolation characteristics of the mount member.
Learning model generation and utilization system . a data providing vehicle configured to be able to communicate with the learning device;
The data providing vehicle is
the internal combustion engine attached to a vehicle body via the mount member;
a control device;
Equipped with
The control device of the data providing vehicle
obtaining the input parameters and the output parameters related to an ego-vehicle to generate the training data set;
configured to transmit the training data set to the learning device together with a vehicle type of the ego-vehicle;
The learning device
storing the training data set separately for each vehicle type;
and generating the learning model for each vehicle type based on the training data set separated for each vehicle type.
The learning model generation and utilization system according to claim 1 . The control device of the learning model-using vehicle
transmitting the acquired input parameters and the vehicle type of the host vehicle to the learning device; receiving the output parameters from the learning device, which are obtained by inputting the input parameters into the learning model corresponding to the vehicle type of the host vehicle;
The learning model generation and utilization system according to claim 2 , configured to control the internal combustion engine based on the received output parameters. The control device of the learning model-using vehicle
receiving the learning model according to the vehicle type from the learning device;
The internal combustion engine is controlled based on the output parameters obtained by inputting the acquired input parameters into the received learning model.
The learning model generation and utilization system according to claim 2 . A learning model generation and utilization method used in a learning model generation and utilization system including a learning device and a learning model utilization vehicle configured to be able to communicate with the learning device,
The learning model generation and utilization method includes causing the learning device to perform machine learning using a training data set including input parameters that affect the vibration isolation characteristics of a mount member disposed between an internal combustion engine and a vehicle body, and output parameters that indicate the vibration isolation characteristics of the mount member, thereby generating a learning model ;
The input parameters include the temperature of the area where the mounting member is used, time series data of changes in the ambient temperature of the mounting member over a predetermined period of time, and the number of years the mounting member has been in use;
the output parameters include vehicle body vibration when the internal combustion engine is started;
The learning model generation and utilization method includes the step of:
acquiring the input parameters related to the host vehicle;
controlling the internal combustion engine based on the output parameters obtained by inputting the input parameters into the learning model so as to suppress an increase in body vibration of the vehicle caused by a deterioration in the vibration isolation characteristics of the mount member;
A learning model generation and utilization method further including: suppressing an increase in body vibration of the vehicle caused by a deterioration in the vibration-damping characteristics of the mount member by changing the operating point of the internal combustion engine to an operating point on the side where the engine rotation speed decreases at equal output or an operating point on the side where thermal efficiency decreases.