Wear state prediction method, wear state prediction device and wear state prediction program

Abstract

The wear state prediction device (10) comprises: a model generation unit (124) which generates a model for predicting the wear state as a target variable using a prescribed algorithm with temperature data related to an aircraft (31) acquired in advance as an explanatory variable; and a wear state prediction unit (133) which inputs the temperature data into the model generated by the model generation unit (124) to predict the wear state.

Description

Technical Field

[0001] The present invention relates to a wear state prediction method, a wear state prediction device and a wear state prediction program. Background Art

[0002] Conventionally, a technology for predicting the wear state of aircraft tires is known (Patent Document 1). The method described in Patent Document 1 acquires multiple wear energies corresponding to multiple driving states distinguished by usage conditions (e.g., touchdown driving state, post-landing deceleration driving state, taxiing driving state, etc.), and predicts the wear state of the aircraft tire based on the acquired wear energies.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2013-113724 Summary of the Invention

[0006] Problems to be solved by the invention

[0007] In recent years, the analysis of big data has become increasingly active due to the development of AI (e.g., machine learning). However, there has been little research on prediction methods that utilize big data to predict the wear state of aircraft tires.

[0008] Therefore, the present invention has been made in view of such circumstances, and an object thereof is to provide a wear state prediction method, a wear state prediction device, and a wear state prediction program for predicting the wear state of aircraft tires using a model.

[0009] Solutions for solving problems

[0010] In a wear state prediction method involved in one embodiment of the present invention, a prescribed algorithm is used that uses pre-acquired temperature data related to an aircraft (aircraft 31) as an explanatory variable to generate a model for predicting the wear state as a target variable; and the temperature data is input into the generated model to predict the wear state.

[0011] Effects of the Invention

[0012] According to the present invention, the wear state of aircraft tires can be predicted using a model. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a schematic diagram showing the relationship between the wear state prediction device, the network, and the airline company.

[0014] Figure 2It is a schematic configuration diagram of the wear state prediction device according to this embodiment of the present invention.

[0015] Figure 3 This is a diagram illustrating an example of machine learning according to this embodiment of the present invention.

[0016] Figure 4 This is a diagram illustrating an example of the relationship between learning data and teacher data according to this embodiment of the present invention.

[0017] Figure 5 This is a sequence diagram illustrating an operation example of the wear state prediction system according to the present embodiment of the present invention.

[0018] Figure 6 This is a sequence diagram illustrating an operation example of the wear state prediction system according to the present embodiment of the present invention.

[0019] Figure 7 This is a diagram illustrating an example of a method that takes into account the air temperature after the date on which the wear rate is predicted. DETAILED DESCRIPTION

[0020] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the description of the drawings, the same parts are denoted by the same reference numerals and their description will be omitted.

[0021] (Configuration Example of Wear State Prediction System)

[0022] Reference Figure 1 An example of the structure of the wear state prediction system 1 will be described. Figure 1 As shown, the wear state prediction system 1 includes a wear state prediction device 10 , a network 20 , and an airline 30 .

[0023] The wear state prediction device 10 performs two-way communication with the airline 30 via the network 20. Specifically, the wear state prediction device 10 obtains data related to the aircraft 31 from the airline 30 via the network 20. The wear state prediction device 10 uses the obtained data to generate a machine learning model, and uses the generated machine learning model to predict the wear state of the aircraft tire 32 installed on the aircraft 31. The details of the data obtained by the wear state prediction device 10 from the airline 30 will be described later. The network 20 is a communication network that can send and receive data. For example, the network 20 is composed of various communication lines such as dedicated lines set up by telecommunications operators, public switched telephone networks, satellite communication lines, mobile communication lines, etc.

[0024] In this embodiment, the aircraft tires 32 include tires mounted on the main landing gear and tires mounted on the front landing gear. The main landing gear is mounted with multiple (e.g., six) tires. The front landing gear is also mounted with multiple (e.g., two) tires.

[0025] The wear state prediction device 10 is, for example, a general-purpose computer equipped with a CPU, ROM (Read Only Memory), and RAM (Random Access Memory). The CPU reads programs stored in the ROM and other memory devices into the RAM and executes the programs. Furthermore, the wear state prediction device 10 can be a stationary terminal device or a portable terminal device (such as a smartphone) that is easy to carry. Furthermore, the wear state prediction device 10 can also be configured as a server at a management center.

[0026] Next, refer to Figure 2 The structure example of the wear state prediction device 10 is described below. Figure 2 As shown, the wear state prediction device 10 includes a controller 11 (e.g., a CPU) and a storage device 14. The controller 11 includes a first processing function 12 and a second processing function 13 as multiple information processing functions. The first processing function 12 is categorized into a data acquisition unit 121, a data processing unit 122, an algorithm selection unit 123, and a model generation unit 124. The second processing function 13 is categorized into a new data acquisition unit 131, an application model selection unit 132, and a wear state prediction unit 133.

[0027] The data acquisition unit 121 acquires data from the airline 30 via a communication unit (not shown). The communication unit is provided in the wear state prediction device 10 and serves as an interface for connecting to the network 20 to transmit and receive data with the airline 30. The data acquired by the data acquisition unit 121 from the airline 30 includes temperature data of the aircraft 31, acceleration data of the aircraft 31, weight data of the aircraft 31, and data indicating the position at which the aircraft tire 32 is mounted (hereinafter sometimes referred to as mounting position data).

[0028] The air temperature data of the aircraft 31 refers to temperature data measured by a sensor installed in the aircraft 31 .

[0029] The acceleration data of the aircraft 31 includes acceleration data in each of the X, Y, and Z directions of the aircraft 31. The X direction refers to the front-back direction of the aircraft 31. The Y direction refers to the left-right direction of the aircraft 31. The Z direction refers to the up-down direction of the aircraft 31.

[0030] The data processing unit 122 processes the data acquired by the data acquisition unit 121. Details of the processing will be described later.

[0031] The algorithm selection unit 123 selects an algorithm for generating a machine learning model. Examples of the algorithm include multiple linear regression, Lasso regression, nonlinear SVM, random forest, and XGBoost.

[0032] The model generation unit 124 performs machine learning on the data generated by the data processing unit 122 using the algorithm selected by the algorithm selection unit 123 to generate a machine learning model. The model generation unit 124 generates a machine learning model for each algorithm. Thus, multiple machine learning models are generated. The generated machine learning models are stored in the storage device 14.

[0033] The new data acquisition unit 131 acquires new data from the airline 30 to be input into the machine learning model generated by the model generation unit 124. The new data acquired from the airline 30 by the new data acquisition unit 131 has the same attributes (such as temperature and acceleration) and structure (every second, as described later) as the data acquired from the airline 30 by the data acquisition unit 121, but the data itself is different.

[0034] The applied model selection unit 132 reads the machine learning model generated by the model generation unit 124 from the storage device 14 and evaluates the prediction performance of the read machine learning model. Based on the evaluation results, the applied model selection unit 132 selects the machine learning model with the highest prediction performance.

[0035] The wear state prediction unit 133 predicts the wear state by inputting the new data acquired by the new data acquisition unit 131 into the machine learning model selected by the application model selection unit 132. The wear state includes the wear amount and the wear rate. Details of the wear amount and the wear rate will be described later. The predicted data related to the wear state is stored in the storage device 14.

[0036] Next, refer to Figures 3-4 The following describes a method for generating a machine learning model and a method for predicting a wear state using the generated machine learning model.

[0037] like Figure 3 As shown in the figure, machine learning is divided into a learning phase and a prediction phase. In the learning phase, features are extracted from a large amount of data (big data) and a machine learning model is generated. In the prediction phase, new data is input to the generated machine learning model and the results are output. Figure 3 The acquired data is data acquired from the airline 30 as described above, and includes temperature data of the aircraft 31 , acceleration data of the aircraft 31 , weight data of the aircraft 31 , installation position data, and the like.

[0038] In the field of statistics, data input for machine learning is sometimes referred to as explanatory variables. In this embodiment, the explanatory variables include the temperature data of aircraft 31, the acceleration data of aircraft 31, the weight data of aircraft 31, and the installation position data. Furthermore, the data output from the learned model is sometimes referred to as the target variable. In this embodiment, the wear state of aircraft tire 32 is the target variable.

[0039] Machine learning includes supervised learning using teacher data and learning without teacher data. In this embodiment, supervised learning is used. Teacher data means the so-called "correct answer data". Supervised learning is a learning data set (in the Figure 3 The method is represented by the reference numeral 40 in FIG. Figure 3 As shown, in this embodiment, before machine learning is performed, big data is processed to generate a learning data set 40. Hereinafter, such data processing before machine learning is referred to as data preprocessing.

[0040] Next, data preprocessing will be described. In this embodiment, data preprocessing means the aggregation of data and the association of the aggregated data with the teacher data. Such association is sometimes called labeling. The data acquisition unit 121 acquires data from the airline 30 for each flight. As the data structure of a flight, when data is acquired from sensors installed on the aircraft, the data is stored, for example, every second. That is, temperature data, acceleration data, etc. are stored every second. Therefore, even for a single flight, the amount of data is huge.

[0041] Typically, the number of flights from the time the aircraft tire 32 is mounted on the fuselage to the time it is removed is hundreds of times. In the following, the number of flights from the time the aircraft tire 32 is mounted on the fuselage to the time it is removed is sometimes expressed as all flights. In this embodiment, the data for machine learning is the data of all flights (hundreds of flights) from the time the aircraft tire 32 is mounted on the fuselage to the time it is removed. With respect to such big data, the teacher data (wear rate based on the amount of wear) that becomes the associated target is a single value. The amount of wear means the amount of wear of the aircraft tire 32 from the time the aircraft tire 32 is mounted on the fuselage to the time it is removed, and the unit is mm. If the groove depth of the aircraft tire 32 when it is new is X (mm) and the groove depth of the aircraft tire 32 when it is removed is Y (mm), then X-Y (mm) is the amount of wear. Typically, the amount of wear is measured after the aircraft tire 32 is removed, so there is only one value.

[0042] In this embodiment, the wear rate is used as the teacher data. Here, the relationship between the amount of wear and the wear rate is described. In this embodiment, the wear rate is defined as the amount of wear on the aircraft tire 32 during one landing. The unit of the wear rate is expressed as mm / LD. If the amount of wear measured after the aircraft tire 32 is disassembled is divided by the number of all landings, the wear rate is calculated. The wear rate calculated in this way is used as the teacher data. In this embodiment, the number of landings and the number of flights have the same meaning.

[0043] However, the definition of wear rate is not limited to the above. Wear rate can also be defined as the amount of wear relative to the taxiing distance. Taxiing refers to the aircraft 31 using the power of the aircraft 31 to travel on the ground (mainly the runway). If the amount of wear measured after the aircraft tire 32 is disassembled is divided by the taxiing distance of all flights, the wear rate can be calculated. In addition, the taxiing distance in a flight is the total value of the taxiing distance at the departure airport and the taxiing distance at the arrival airport. This total value is usually several thousand meters (km).

[0044] Alternatively, the wear rate can be defined as a parameter obtained by dividing the wear amount by a cumulative value associated with each flight. For example, the cumulative value associated with each flight includes the number of landings and taxiing distance described above. Alternatively, the cumulative value associated with each flight can include the sum of representative acceleration values ​​(e.g., average values) for each flight and the sum of weight values ​​for each flight.

[0045] As an example, the wear rate is described below as the amount of wear on an aircraft tire 32 during a single landing (mm / LD). Of course, as described above, the wear rate can also be a value obtained by dividing the wear amount by the taxiing distance, or a parameter obtained by dividing the wear amount by a cumulative value related to flight. Furthermore, as described above, the wear amount is a single value, so the wear rate calculated using the wear amount is also a single value.

[0046] Since the training data (wear rate) is a single value, the various data items associated with the training data must also be aggregated into a single value. Therefore, the data processing unit 122 aggregates the temperature data, acceleration data, weight data, and so on, into a single value. Furthermore, the installation position data is a fixed value and therefore does not need to be aggregated.

[0047] First, an example of a method for aggregating temperature data will be described. The data processing unit 122 averages the temperature data in one flight to calculate the average temperature data in one flight. The data processing unit 122 repeats this process to calculate the average temperature data for each flight in all flights (the 1st flight, the 2nd flight...the 100th flight...). Then, the data processing unit 122 averages the average temperature data for each flight in all flights to calculate the average temperature data (one value) for all flights (the 1st to the hundreds of flights). In addition, such averaging processing is an example of an aggregation method and is not a necessary processing. As described later, the representative value of the explanatory variable is not limited to the average value, but can also be the median value or the mode value.

[0048] As described above, when acquiring data from sensors installed on an aircraft, temperature data is stored every second. That is, the stored temperature data includes temperature data when the aircraft 31 is stopped, temperature data when the aircraft 31 is traveling, and temperature data when the aircraft 31 is flying. When aggregating the temperature data, the data processing unit 122 may selectively use part of the data based on the state of the aircraft 31. For example, the temperature data when the aircraft 31 is flying may not be used. In other words, the data processing unit 122 may average the temperature data of a portion of each flight to calculate the average temperature data (a single value).

[0049] Next, an example of a method for aggregating acceleration data will be described. As mentioned above, acceleration data includes acceleration data for each of the X, Y, and Z directions. Therefore, data processing unit 122 aggregates acceleration data for the X, Y, and Z directions separately. Here, the acceleration data for the X direction will be described as a representative example.

[0050] The data processing unit 122 calculates the average acceleration data (X direction) for a single flight by averaging the squared sum of the acceleration data (X direction) for that flight. The data processing unit 122 repeats this process to calculate the average acceleration data (X direction) for each of all flights (the first flight, the second flight, the hundredth flight, and so on). The data processing unit 122 then averages the average acceleration data (X direction) for each of all flights to calculate the average acceleration data (a single value in the X direction) for all flights (the first to the hundredth flight). The same process is repeated for the Y and Z directions, so their explanation is omitted.

[0051] Next, an example of a weight data aggregation method will be described. Data processing unit 122 extracts the maximum value from the weight data for a single flight. Data processing unit 122 repeats this process to extract the maximum value for each flight (the first flight, the second flight, the hundredth flight, and so on). Data processing unit 122 then averages the maximum values ​​for each flight to calculate the average weight data (a single value) for all flights (from the first to the hundredth flight).

[0052] In this way, the average temperature data, average acceleration data (X direction, Y direction, Z direction), and average weight data of all flights are calculated as one value by the data processing unit 122. Figure 4 As shown, these one values ​​(learning data 41) are associated with teacher data 42 (wear speed) to generate a learning data set 40. Since the learning data set 40 is generated for each aircraft tire, a plurality of learning data sets 40 are usually generated.

[0053] The algorithm selection unit 123 selects an algorithm for generating a machine learning model. As described above, the algorithms include multiple linear regression, Lasso regression, nonlinear SVM, random forest, XGBoost, and the like. However, these algorithms are not limited to these algorithms; any algorithm capable of supervised learning may be used. These algorithms are well known, so detailed descriptions are omitted.

[0054] like Figure 4 As shown, the model generation unit 124 generates a machine learning model by performing machine learning on the learning data set 40 generated by the data processing unit 122 using the algorithm selected by the algorithm selection unit 123. In this embodiment, the algorithms used include the aforementioned multiple linear regression, Lasso regression, nonlinear SVM, random forest, and XGBoost.

[0055] Generally, the prediction performance of machine learning models is evaluated based on the coefficient of determination (R 2 The algorithm is evaluated using any one or a combination of metrics indicating algorithm performance, such as ΔE, ΔP, ΔS, ΔP ...

[0056] like Figure 3 As shown, the wear state prediction unit 133 inputs new data into the machine learning model (learned model) selected by the application model selection unit 132 to predict the wear rate. Furthermore, the new data acquisition unit 131 acquires new data. The new data is data from the time the aircraft tire 32 is mounted on the fuselage until the wear rate is predicted.

[0057] In this embodiment, average temperature data, average acceleration data, average weight data, and installation location data are used as explanatory variables for machine learning. These explanatory variables have different contribution rates to the target variable. The contribution rate indicates the influence of each explanatory variable on the target variable. The greater the influence of an explanatory variable on the target variable, the more important it is.

[0058] The explanatory variables in this embodiment (average temperature data, average acceleration data, average weight data, installation location data, etc.) are non-cumulative explanatory variables. In other words, in this embodiment, only non-cumulative explanatory variables are used, not cumulative explanatory variables. An example of a cumulative explanatory variable is the coasting distance described above. The reason for not using cumulative explanatory variables is that using cumulative explanatory variables would reduce the accuracy of the machine learning model.

[0059] The inventors analyzed the contribution of each explanatory variable to the target variable. They found that average temperature data had the highest contribution among all the explanatory variables. Furthermore, the inventors found that the only explanatory variable with a lower contribution than average temperature data was average acceleration data. This indicates that using average temperature data as an explanatory variable when generating a machine learning model for predicting wear rate can yield a highly accurate model. Furthermore, using average acceleration data as an explanatory variable in addition to average temperature data can yield an even more accurate model.

[0060] Then refer to Figures 5-6 An operation example of the wear state prediction system 1 is described with reference to a sequence diagram of FIG.

[0061] In step S101 , the air temperature data of the aircraft 31 , the acceleration data of the aircraft 31 , the weight data of the aircraft 31 , the installation position data, and the like are transmitted from the airline company 30 .

[0062] In step S103, the data acquisition unit 121 acquires the data sent in step S101. The process proceeds to step S105, where the data processing unit 122 processes the data acquired in step S103. Specifically, the data processing unit 122 aggregates the temperature data, acceleration data, weight data, and other data into a single value. The data processing unit 122 associates the aggregated data (learning data 41) with the teacher data 42 to generate the learning data set 40.

[0063] The process proceeds to step S107 , where the algorithm selection unit 123 selects an algorithm (multiple linear regression, Lasso regression, nonlinear SVM, random forest, XGBoost, etc.) for generating a machine learning model.

[0064] The process proceeds to step S109 , where the model generation unit 124 performs machine learning on the learning data set 40 generated in step S105 using the algorithm selected in step S107 to generate a machine learning model.

[0065] In step S201, temperature data of the aircraft 31, acceleration data of the aircraft 31, weight data of the aircraft 31, installation position data, etc. are transmitted from the airline 30. The data transmitted in step S201 is new data used for prediction of the wear rate.

[0066] In step S203, the new data acquisition unit 131 acquires the new data sent in step S201. The process proceeds to step S205, where the application model selection unit 132 evaluates the prediction performance of the machine learning models generated in step S109 and selects the machine learning model with the highest prediction performance.

[0067] The process proceeds to step S207 , where the wear state prediction unit 133 inputs the new data acquired in step S203 to the machine learning model selected in step S207 to predict the wear rate.

[0068] (Effect)

[0069] As described above, according to the wear state prediction method according to this embodiment, the following effects can be obtained.

[0070] The wear state prediction device 10 uses a prescribed algorithm with the temperature data of the aircraft 31 obtained from the airline 30 as an explanatory variable to generate a model (machine learning model) that predicts the wear state as a target variable. The wear state prediction device 10 inputs the temperature data into the generated model to predict the wear state. As described above, the inventors found that the contribution rate of the temperature data (average temperature data) among the explanatory variables is the highest compared with other data. By using the temperature data with the highest contribution rate, a machine learning model with high prediction performance can be generated. Moreover, by using a machine learning model with high prediction performance, the wear state can be predicted with high accuracy. In addition, as an example, the wear state represents the wear rate obtained by dividing the wear amount of the aircraft tire 32 by the number of flights.

[0071] Furthermore, as mentioned above, the inventors discovered that the only explanatory variable with a lower contribution rate than air temperature data (average air temperature data) is acceleration data (average acceleration data). By using air temperature data and acceleration data, which have higher contribution rates than other data, a machine learning model with higher predictive performance can be generated, enabling more accurate prediction of wear conditions.

[0072] While the embodiments of the present invention have been described above, the description and drawings constituting part of this disclosure should not be construed as limiting the present invention. Various alternative embodiments, examples, and operational techniques will be apparent to those skilled in the art based on this disclosure.

[0073] In the above-mentioned embodiment, it is described that the new data (data at the time of prediction) input to the machine learning model is data from the time the aircraft tire 32 is installed on the fuselage to the time the wear rate is predicted. On the other hand, it is described that the data used to generate the machine learning model (data at the time of learning) is data from the time the aircraft tire 32 is installed on the fuselage to the time it is removed. The date on which the wear rate is predicted may not coincide with the date on which the aircraft tire 32 is removed. Therefore, the number of data from the date the aircraft tire 32 is installed on the fuselage to the date on which the wear rate is predicted may not coincide with the number of data from the date the aircraft tire 32 is installed on the fuselage to the date on which it is removed. As a result, it is sometimes impossible to obtain a highly accurate result.

[0074] Therefore, the average temperature data with the highest contribution rate to the target variable may be considered, and the temperature after the date when the wear rate is predicted. Figure 7 An example of a method that takes into account the temperature after the date on which the wear rate is predicted will be described.

[0075] Figure 7 January 8 is the date when the latest data is available, and January 11 is the date when the wear rate prediction was performed (hereinafter sometimes referred to as the prediction implementation date). The latest data refers to the latest data sent from the airline 30.

[0076] When predicting the wear rate, that is, on January 11, there is no data for January 9 to 11. In this case, first, the data acquisition unit 121 aggregates the data from the date the aircraft tire 32 was mounted on the fuselage to January 8. Among the aggregated data, the average temperature data (first temperature data) is as follows: Figure 7 The wear state prediction unit 133 inputs the aggregated data into the machine learning model to predict the wear speed. The predicted wear speed is as follows Figure 7 Shown is 0.0130.

[0077] The wear state prediction unit 133 calculates the wear amount by multiplying the predicted wear rate (0.0130) by the number of flights (625). The wear state prediction unit 133 subtracts the calculated wear amount from the groove depth of the aircraft tire 32 when it is new to calculate the groove margin representing the remaining groove depth. The calculated groove margin is as follows: Figure 7Alternatively, the wear state prediction unit 133 may calculate the wear amount by multiplying the predicted wear rate by the cumulative value related to flight.

[0078] If the groove margin is greater than 0, the wear state prediction unit 133 generates data for the next day (January 9). Figure 7 As shown in FIG. 1 , the groove margin (2.38 mm) on January 8 is greater than 0, so the wear state prediction unit 133 generates data for the next day (January 9). The wear state prediction unit 133 obtains the average temperature (second temperature data) of January at the AAA airport where the aircraft 31 arrives and saves it in the database. Therefore, the average temperature of the AAA airport on January 9 is input as 8.9 degrees (refer to Figure 7 ). Next, the wear state prediction unit 133 uses the temperature data of 8.9 degrees to update the average temperature data from the date when the aircraft tire 32 is installed on the fuselage to January 9. The updated average temperature data is as follows Figure 7 Shown is 12.2 degrees.

[0079] The wear state prediction unit 133 inputs the data including the updated average temperature data into the machine learning model and predicts the wear rate again. The predicted wear rate is as follows: Figure 7 The wear state prediction unit 133 calculates the wear amount by multiplying the predicted wear rate (0.0130) by the number of flights (632 times). The wear state prediction unit 133 calculates the groove margin by subtracting the calculated wear amount from the groove depth of the aircraft tire 32 when it is new. The calculated groove margin is as follows: Figure 7 Shown is 2.28mm.

[0080] Since the groove margin (2.28 mm) on January 9th was greater than 0, the wear state prediction unit 133 generates data for the next day (January 10th). The same process is then repeated until the groove margin becomes less than 0. Furthermore, the average temperature in February at AAA Airport is 10.2 degrees Celsius.

[0081] like Figure 7 As shown, the date when the groove margin became less than 0 is February 13 (groove margin is -0.02). After calculating the date when the groove margin became less than 0 (February 13), the wear state prediction unit 133 calculates the date when the aircraft tire 32 was removed. Next, an example of a method for calculating the date when the aircraft tire 32 was removed will be described.

[0082] First, the wear state prediction unit 133 calculates the number of flights ( ) immediately before the groove margin becomes 0, using the groove depth of the new aircraft tire 32 and the wear rate (0.0120) on February 12. Figure 7The wear state prediction unit 133 subtracts the cumulative number of flights from the time the aircraft tire 32 was mounted on the prediction implementation date (January 11) from the number of flights until the groove margin becomes 0 ( Figure 7 ※1)(875-662) shown, calculate the remaining number of landings ( Figure 7 The remaining number of possible landings means the number of times the aircraft tire 32 can be used.

[0083] The wear state prediction unit 133 divides the remaining number of possible landings by the number of flights per day (7 times) to calculate the number of days (days) during which the aircraft tire 32 can be used. Figure 7 Then, the wear state prediction unit 133 calculates February 10, 30 days after January 11, as the date for removing the aircraft tire 32.

[0084] In this way, by taking into account the temperature after the date when the wear rate is predicted (after the prediction time), the wear rate can be predicted with high accuracy.

[0085] Furthermore, by utilizing the highly accurate wear rate, the date on which the aircraft tire 32 will be removed can be calculated with high precision. Conventionally, when replacing an aircraft tire 32, the operator manually measures the wear and replaces it when the wear reaches a specified value. However, if a certain number of aircraft tires 32 are replaced at the same time, replacement delays and inventory shortages may occur. In contrast, the wear state prediction device 10 accurately calculates the date on which the aircraft tire 32 will be removed, enabling efficient replacement operations.

[0086] In the above embodiment, the temperature data of aircraft 31 is used as the explanatory variable for generating a machine learning model. Furthermore, the temperature data of aircraft 31 is described as temperature data measured by a sensor installed on aircraft 31. However, the explanatory variable used to generate a machine learning model is not limited to temperature data measured by a sensor installed on aircraft 31. The explanatory variable used to generate a machine learning model may also be the temperature data of an airport.

[0087] The method for obtaining airport temperature data is not particularly limited. Temperature data observed at fixed points can be obtained, or temperature data publicly available from relevant organizations (for example, the Meteorological Agency in Japan) can be obtained. The method for aggregating airport temperature data is the same as the method for aggregating aircraft 31 temperature data described above. Specifically, airport temperature data is averaged and aggregated into a single value, similar to the temperature data for aircraft 31.

[0088] However, the aggregation method of the temperature data of the aircraft 31 and the temperature data of the airport is not limited to taking the average. The temperature data of the aircraft 31 and the temperature data of the airport used as explanatory variables may be the median or the mode.

[0089] The temperature data related to the aircraft 31 includes both temperature data measured by sensors installed in the aircraft 31 and temperature data of the airport.

[0090] Furthermore, when aggregating the temperature data for the aircraft 31 and the airport, the data processing unit 122 may also divide the period during which the temperature data was stored for aggregation. For example, the data processing unit 122 may divide the period during which the temperature data was stored into a first half and a second half for aggregation. As described above, the number of flights from the time the aircraft tire 32 was mounted on the fuselage to the time it was removed is several hundred. Here, it is assumed that the number of flights from the time the aircraft tire 32 was mounted on the fuselage to the time it was removed is 500.

[0091] In this case, the data processing unit 122 may also divide the period in which the temperature data is stored into the first half (the 1st to the 250th time) and the second half (the 251st to the 500th time) for aggregation. In this way, the average temperature data (one value) of the first half and the average temperature data (one value) of the second half can be obtained as aggregated data. The data associated with the teacher data may be either the average temperature data of the first half or the average temperature data of the second half. Dividing the period in which the temperature data is stored and aggregating it in this way may also obtain the same effect as the above-mentioned effect. In addition, the data to be divided is not limited to the temperature data, and the period in which the acceleration data is stored may also be divided. In addition, the division method is not limited to two divisions of the first half and the second half, but may also be three divisions.

[0092] Furthermore, when aggregating the temperature data for the aircraft 31 and the temperature data for the airport, the data processing unit 122 may also classify the period in which the temperature data is stored into two or more different periods and use the resulting multiple periods for aggregation. As an example of a method for classifying into two or more periods, the data processing unit 122 may classify the period in which the temperature data is stored into the entire period (the 1st to 500th occurrence) and the first half period (the 1st to 250th occurrence). The data processing unit 122 averages the average temperature data for each of the multiple classified periods (the entire period and the first half) to obtain the average temperature data (a single value) for the multiple classified periods (the entire period and the first half). Classifying the period in which the temperature data is stored into multiple periods and aggregating them in this manner can also achieve the same effects as described above. Furthermore, the data to be classified is not limited to temperature data; a period in which acceleration data is stored may also be classified into multiple periods. That is, the data processing unit 122 may classify the period in which explanatory variables with the same attributes are stored into two or more different periods and use the resulting multiple periods for aggregation.

[0093] The entire contents of Japanese Patent Application No. 2020-033208 (filing date: February 28, 2020) are hereby incorporated by reference.

[0094] Description of Reference Numerals

[0095] 1: Wear state prediction system; 10: Wear state prediction device; 11: Controller; 12: First processing function; 13: Second processing function; 14: Storage device; 20: Network; 30: Airlines; 31: Aircraft; 32: Aircraft tire; 121: Data acquisition unit; 122: Data processing unit; 123: Algorithm selection unit; 124: Model generation unit; 131: New data acquisition unit; 132: Application model selection unit; 133: Wear state prediction unit.

Claims

1. A wear state prediction method, wherein: generating a model for predicting the wear state of aircraft tires as a target variable using a predetermined algorithm with previously acquired aircraft-related temperature data as an explanatory variable; inputting the air temperature data into the generated model; and predicting the wear state of the aircraft tire, The wear state prediction method is characterized in that: When predicting the wear state, data including first temperature data and second temperature data are input into the generated model to predict the wear state, wherein the first temperature data is average temperature data related to the aircraft from the date the aircraft tire is mounted on an airframe to the time of prediction, and the second temperature data is average temperature data related to the aircraft after the time of prediction; and A removal date of the aircraft tire is predicted based on the predicted wear state.

2. The wear state prediction method according to claim 1, characterized in that: The removal date is a date on which the groove allowance of the aircraft tire becomes equal to or less than a predetermined value, wherein the wear amount of the aircraft tire is calculated by multiplying the predicted wear state by a cumulative value related to flight, and the groove allowance of the aircraft tire is calculated by subtracting the wear amount from the groove depth of the aircraft tire when it was new.

3. The wear state prediction method according to claim 1, characterized in that: The wear state is a wear rate, which is defined as a parameter obtained by dividing the wear amount of the aircraft tire by a cumulative value related to flight.

4. The wear state prediction method according to claim 3, characterized in that: The number of times the aircraft tire can be used is predicted using the previously acquired groove depth of the aircraft tire when it is new and the predicted wear rate.

5. The wear state prediction method according to any one of claims 1 to 4, characterized in that: The temperature data related to the aircraft is obtained by a sensor provided on the aircraft.

6. The wear state prediction method according to any one of claims 1 to 4, characterized in that: The model is generated using the prescribed algorithm with air temperature data related to the aircraft and previously acquired acceleration data of the aircraft as the explanatory variables.

7. A wear state prediction device for predicting the wear state of an aircraft tire, the wear state prediction device comprising: a model generating unit that generates a model for predicting the wear state of aircraft tires as a target variable using a predetermined algorithm using previously acquired aircraft-related temperature data as an explanatory variable; and a prediction unit that inputs the air temperature data into the model generated by the model generation unit to predict the wear state, The wear state prediction device is characterized in that: When predicting the wear state, the prediction unit inputs data including the first temperature data and the second temperature data into the generated model to predict the wear state, wherein: The first temperature data is average temperature data related to the aircraft from the date the aircraft tire is mounted on the fuselage to the prediction time, and the second temperature data is average temperature data related to the aircraft after the prediction time. The prediction unit predicts a removal date of the aircraft tire based on the predicted wear state.

8. A computer program product comprising a wear state prediction program, wherein the wear state prediction program causes a computer of a terminal device to execute the following steps: generating a model for predicting the wear state of aircraft tires as a target variable using a predetermined algorithm with previously acquired aircraft-related temperature data as an explanatory variable; inputting the air temperature data into the model; and predicting the wear state of the aircraft tire, in, When predicting the wear state, data including first temperature data and second temperature data are input into the generated model to predict the wear state, wherein the first temperature data is average temperature data related to the aircraft from the date the aircraft tire is mounted on an airframe to the time of prediction, and the second temperature data is average temperature data related to the aircraft after the time of prediction; and A removal date of the aircraft tire is predicted based on the predicted wear state.

Images