Temperature distribution estimation device

Abstract

To provide a temperature distribution estimation device capable of estimating the temperature distribution of an object.SOLUTION: A temperature distribution estimation device 1 comprises: a multiple regression inference unit 11 which, designed to estimate the temperature distribution of a cylinder block and a cylinder head, estimates the maximum temperature and minimum temperature of the cylinder block and the cylinder head on the basis of input data 1a; a DNN inference unit 12 having been trained using a large number of temperature distribution images of the cylinder block and the cylinder head and capable of outputting the normalized values of temperature distribution of the cylinder block and the cylinder head; and a temperature distribution calculation unit 13 for calculating the temperature distribution of the cylinder block and the cylinder head, on the basis of the output of the multiple regression inference unit 11 and the output of the DNN inference unit 12.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a temperature distribution estimation device. [Background technology]

[0002] 2. Description of the Related Art Conventionally, a control device applied to a vehicle is known (see, for example, Patent Document 1).

[0003] The control device of Patent Document 1 is configured to estimate the amount of NOx in the exhaust gas using a neural network and control urea injection based on the estimated value of the NOx amount. That is, a virtual sensor that estimates the amount of NOx in the exhaust gas is constructed using a neural network. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2012-7593 A Summary of the Invention [Problem to be solved by the invention]

[0005] In this regard, it is desirable to estimate the temperature distribution of an object, i.e., a virtual sensor that can estimate the temperatures of multiple points on an object is desirable.

[0006] The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a temperature distribution estimation device capable of estimating the temperature distribution of an object. [Means for solving the problem]

[0007] The temperature distribution estimation device according to the present invention estimates the temperature distribution of an object, and includes a statistical inference unit that estimates the maximum and minimum temperatures of the object based on input data, a learning inference unit that learns using a large number of temperature distribution images of the object and is capable of outputting normalized values ​​of the temperature distribution of the object, and a temperature distribution calculation unit that calculates the temperature distribution of the object based on the output of the statistical inference unit and the output of the learning inference unit.

[0008] With this configuration, the maximum and minimum temperatures of the object are estimated by the statistical analysis method, and the normalized value of the temperature distribution of the object is output by the machine learning method. The normalized value is then converted back to temperature by the temperature distribution calculation unit using the maximum and minimum temperatures of the object. Effect of the Invention

[0009] According to the temperature distribution estimation device of the present invention, it is possible to estimate the temperature distribution of an object. [Brief description of the drawings]

[0010] [Figure 1] 1 is a functional block diagram for explaining a temperature distribution estimation device of the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, one embodiment of the present invention will be described.

[0012] First, the configuration of a temperature distribution estimation device 1 according to the present embodiment will be described with reference to FIG.

[0013] The temperature distribution estimation device 1 is mounted on a vehicle, for example, and configured to estimate the temperature distribution of a cylinder block and a cylinder head (hereinafter referred to as "cylinder") of an internal combustion engine for a vehicle (not shown). That is, the temperature distribution estimation device 1 is a virtual sensor that estimates the temperature of multiple points of the cylinder. The temperature distribution estimation device 1 is configured to estimate the temperature distribution of the cylinder based on input data 1a, and to display the estimation result on a display 2. The cylinder is an example of the "object" of the present invention.

[0014] 1, the temperature distribution estimation device 1 includes a multiple regression inference unit 11, a DNN inference unit 12, a temperature distribution calculation unit 13, and a basis explanation unit 14. For example, the temperature distribution estimation device 1 is provided with a calculation unit and a storage unit (not shown), and the multiple regression inference unit 11, the DNN inference unit 12, the temperature distribution calculation unit 13, and the basis explanation unit 14 are realized by the calculation unit executing a program stored in the storage unit.

[0015] The multiple regression inference unit 11 is configured to estimate the maximum and minimum cylinder temperatures based on the input data 1a. The input data 1a includes various information during operation of the internal combustion engine for a vehicle, such as the engine speed, engine torque, and air-fuel ratio. The multiple regression inference unit 11 is an example of the "statistical inference unit" of the present invention.

[0016] Specifically, the multiple regression inference unit 11 is configured to estimate the maximum and minimum temperatures of the cylinder by substituting the input data 1a into the multiple regression equation obtained by the multiple regression analysis. For example, the maximum temperature Tmax of the cylinder is calculated using the following multiple regression equation (1), and the minimum temperature Tmin of the cylinder is calculated using the following multiple regression equation (2).

[0017] Tmax=β a0 +β a1 x1+β a2 x2+β a3 x3 (1) Tmin=β b0 +β b1x1+β b2 x2+β b3 x3 (2) In the multiple regression equation (1), Tmax is a response variable, x1 to x3 are explanatory variables, x1 is the engine speed, x2 is the engine torque, and x3 is the air-fuel ratio. a0 ~ a3 is a partial regression coefficient, which is obtained in advance by the least squares method. In the multiple regression equation (2), Tmin is a response variable, x1 to x3 are explanatory variables, x1 is the engine speed, x2 is the engine torque, and x3 is the air-fuel ratio. β b0 ~ b3 is a partial regression coefficient, which is previously obtained by the least squares method.

[0018] In addition, the standard partial regression coefficient, which is an index representing the importance of the explanatory variables in the multiple regression equation (1), is calculated in advance, and the standard partial regression coefficient, which is an index representing the importance of the explanatory variables in the multiple regression equation (2), is calculated in advance.

[0019] The DNN inference unit 12 is configured to learn using a large number of temperature distribution images of the cylinder and to be able to output a normalized value of the temperature distribution of the cylinder. The DNN inference unit 12 is configured to perform learning and inference using a deep neural network. The DNN inference unit 12 is an example of the "learning and inference unit" of the present invention.

[0020] Specifically, the DNN inference unit 12 is configured to output a normalized value of a position when position information (three-dimensional coordinate position, i.e., X-axis value, Y-axis value, and Z-axis value) in the cylinder is input. For example, position information of an area displayed on the display 2 in the cylinder is input to the DNN inference unit 12, and normalized values ​​of each position in the area are output from the DNN inference unit 12. Note that the area displayed on the display 2 in the cylinder can be specified by the user, for example.

[0021] The temperature distribution calculation unit 13 is configured to calculate the temperature distribution of the cylinder based on the output of the multiple regression inference unit 11 and the output of the DNN inference unit 12. That is, the temperature distribution calculation unit 13 is provided to convert the normalized value to a temperature (°C) using the maximum temperature Tmax and minimum temperature Tmin of the cylinder. This process of converting the value back to a temperature is performed for each position input from the DNN inference unit 12, so that the temperature distribution of the cylinder can be obtained.

[0022] For example, when the maximum temperature Tmax is 250°C and the minimum temperature Tmin is 150°C, and the normalized value of the predetermined position is 0.5, the temperature of the predetermined position is calculated to be 200°C. The temperature distribution calculation unit 13 is configured to output the calculated temperature distribution of the cylinder to the display 2 for display.

[0023] The basis explanation unit 14 is provided to improve the explainability of the estimation result of the temperature distribution. The standard partial regression coefficients of the multiple regression equations (1) and (2) are input to the basis explanation unit 14, and it is possible to determine items that contribute highly to the estimation result of the temperature distribution based on the standard partial regression coefficients. Note that explanatory variables with large standard partial regression coefficients are determined to be items that contribute highly.

[0024] The basis explanation unit 14 is configured to display a basis explanation screen (not shown) on the display 2 when the user requests the basis for the temperature distribution estimation result. The basis explanation screen lists items that have a high contribution to the temperature distribution estimation result. Therefore, by displaying the basis explanation screen, it is possible to present the items that have a high contribution to the temperature distribution estimation result to the user.

[0025] The display 2 is capable of displaying the temperature distribution of the cylinder, and is configured to be capable of displaying a screen explaining the reason in response to a request from the user.

[0026] -Operation- Next, the operation of the temperature distribution estimation device 1 according to this embodiment will be described.

[0027] First, input data 1a is output from, for example, an engine ECU (not shown), and the input data 1a is input to a multiple regression inference unit 11. The input data 1a includes engine speed, engine torque, and air-fuel ratio. In the multiple regression inference unit 11, a maximum cylinder temperature Tmax is calculated based on the input data 1a using the above-mentioned multiple regression equation (1), and a minimum cylinder temperature Tmin is calculated based on the input data 1a using the above-mentioned multiple regression equation (2). The calculation results in the multiple regression inference unit 11 are output to a temperature distribution calculation unit 13.

[0028] Moreover, position information of the area displayed on the display 2 is input to the DNN inference unit 12. The DNN inference unit 12 calculates normalized values ​​of each position of the area displayed on the display 2. The calculation result in the DNN inference unit 12 is output to the temperature distribution calculation unit 13.

[0029] Then, in the temperature distribution calculation unit 13, the temperature distribution of the cylinder is calculated based on the input from the multiple regression inference unit 11 and the input from the DNN inference unit 12. That is, the normalized value of each position is converted back to temperature using the maximum temperature Tmax and the minimum temperature Tmin of the cylinder. That is, the normalized value of each position is converted back to the temperature scale (unit). The calculation result (temperature distribution of the cylinder) in the temperature distribution calculation unit 13 is output to the display 2.

[0030] The display 2 displays the temperature distribution of the cylinder estimated by the temperature distribution estimation device 1. This temperature distribution of the cylinder is displayed, for example, in the form of a heat map. In other words, temperature information of multiple points on the cylinder is displayed simultaneously.

[0031] By repeating the above operations, the user can easily grasp the change in temperature distribution of the cylinder over time. Also, by changing the area of ​​the cylinder displayed on the display 2 (position information input to the DNN inference unit 12), the user can grasp the three-dimensional temperature distribution of the cylinder.

[0032] Here, when the user requests the basis for the estimation result of the temperature distribution, the basis explanation unit 14 displays a basis explanation screen on the display 2. For example, when the standard partial regression coefficient of the engine speed is large, the engine speed is displayed on the basis explanation screen as an item that contributes highly to the estimation result of the temperature distribution. Note that any number of items may be listed on the basis explanation screen, and when multiple items are listed, they may be arranged in descending order of contribution.

[0033] -effect- In this embodiment, as described above, there are provided a multiple regression inference unit 11 that estimates the maximum and minimum temperatures of the cylinder, a DNN inference unit 12 that can output a normalized value of the temperature distribution of the cylinder, and a temperature distribution calculation unit 13 that returns the normalized value to the temperature using the maximum and minimum temperatures of the cylinder. By configuring in this way, it is possible to estimate the temperature distribution of the cylinder and to speed up the estimation.

[0034] In this embodiment, a basis explanation unit 14 is provided to display a basis explanation screen on the display 2, and items with high contributions to the estimation result are displayed on the basis explanation screen, thereby improving explainability. Note that items with high contributions can be presented to the user based on the standard partial regression coefficient, and explanations are possible even if explanation information is not stored in advance, making it possible to deal with unknown problems.

[0035] -Other embodiments- It should be noted that the embodiments disclosed herein are illustrative in all respects and are not intended to be limiting. Therefore, the technical scope of the present invention is not interpreted solely by the above-described embodiments, but is defined by the claims. The technical scope of the present invention includes all modifications within the scope and meaning equivalent to the claims.

[0036] For example, in the above embodiment, a cylinder is shown as an example of the target of the present invention, but the present invention is not limited to this, and the target of the present invention may be something other than a cylinder.

[0037] Further, in the above embodiment, an example was shown in which the estimated temperature distribution was displayed on the display 2, but the present invention is not limited to this, and the estimated temperature distribution may be used for controlling the vehicle, for example.

[0038] In addition, in the above embodiment, an example was shown in which the position information of the area displayed on the display 2 is input to the DNN inference unit 12, but this is not limited to this, and all position information of the cylinder may be input to the DNN inference unit.

[0039] In the above embodiment, the engine speed, engine torque, and air-fuel ratio are included in the input data 1a. However, the input data 1a may include other parameters related to the cylinder temperature.

[0040] In addition, in the above embodiment, an example was shown in which the maximum and minimum temperatures of the cylinder were estimated by multiple regression analysis, but this is not limited to the above, and the maximum and minimum temperatures of the cylinder may be estimated by a statistical analysis other than multiple regression analysis.

[0041] In addition, in the above embodiment, an example was shown in which a normalized value of the cylinder temperature distribution was output using a deep neural network, but this is not limited to this, and a normalized value of the cylinder temperature distribution may be output using a machine learning algorithm other than a deep neural network. [Explanation of symbols]

[0042] 1...Temperature distribution estimation device, 1a...Input data, 11...Multiple regression inference unit (statistical inference unit), 12...DNN inference unit (learning inference unit), 13...Temperature distribution calculation unit, 14...Basis explanation unit

Claims

[Claim 1] A temperature distribution estimation device that estimates a temperature distribution of an object, comprising: A statistical inference unit that estimates a maximum temperature and a minimum temperature of the object based on input data; a learning and inference unit that is trained using a large number of temperature distribution images of the object and is capable of outputting a normalized value of the temperature distribution of the object; a temperature distribution calculation unit that calculates a temperature distribution of the object based on an output of the statistical inference unit and an output of the learning inference unit,

Images