Plate thickness estimation device, plate thickness estimation method, and program

Abstract

To provide a plate thickness estimation device and the like that can easily estimate a thickness of a steel plate even if it is a steel plate forming an existing structure.SOLUTION: A plate thickness estimation device 400 includes a thermal image acquisition unit 411, a weather information acquisition unit 412, a heat balance analysis unit 413, and a plate thickness estimation unit 414. The thermal image acquisition unit 411 acquires time-series data of thermal images of a steel plate. The weather information acquisition unit 412 acquires time-series data of weather information at a point related to a location where the steel plate is installed. The heat balance analysis unit 413 calculates time-series data of a net radiation amount and a sensible heat transport amount in a target area for plate thickness estimation based on the time-series data of the thermal images, the time-series data of the weather information, and a heat balance analysis model. The plate thickness estimation unit 414 has a trained model that outputs an estimated value of a plate thickness when time-series data of the net radiation amount and the sensible heat transport amount of the target area are input.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a plate thickness estimation device, plate thickness estimation method, and program for estimating the thickness of a steel plate. [Background technology]

[0002] The thickness of steel plates that make up a structure changes due to corrosion and other factors, so estimating this thickness is important for evaluating the safety of the structure. For example, Patent Document 1 describes a method for estimating the wall thickness of pipes that make up a structure using infrared radiation. This method comprises the steps of heating and irradiating a pipe that has been pre-coated with blackbody paint having a known surface emissivity in a predetermined irradiation pattern, measuring the temperature change distribution pattern using an infrared camera, and estimating the wall thickness distribution of the pipe based on the measured temperature change distribution pattern, etc. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2008-157806 [Overview of the project] [Problems that the invention aims to solve]

[0004] The method disclosed in Patent Document 1 is extremely time-consuming because it requires pre-applying blackbody paint to the portion of the plate to be estimated. In particular, this method is not practical for estimating the plate thickness of steel plates constituting existing structures such as water facilities.

[0005] This invention has been made in view of the above circumstances, and aims to provide a plate thickness estimation device, a plate thickness estimation method, and a program that can easily estimate the plate thickness of steel plates that make up existing structures. [Means for solving the problem]

[0006] To achieve the above objective, the plate thickness estimation device according to the first aspect of the present invention is: A thermal image acquisition means that acquires a thermal image showing the thermal distribution of a steel plate from an infrared camera that measures the thermal distribution of the steel plate, and acquires time-series data of the thermal image on a certain measurement day, A weather information acquisition means for acquiring time-series data of weather information on the measurement date at a location related to the installation location of the steel plate, A heat balance analysis means calculates time-series data of net radiation and sensible heat transport in the target area for thickness estimation of the steel plate, based on the time-series data of the thermal image acquired by the thermal image acquisition means, the time-series data of the meteorological information acquired by the meteorological information acquisition means, and a predetermined heat balance analysis model. The system includes a plate thickness estimation means that calculates an estimated value of the plate thickness of the target region based on time-series data of the net radiation and sensible heat transport of the target region calculated by the heat balance analysis means, The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. The plate thickness estimation means includes a trained model that has undergone machine learning using training data, which outputs an estimated value of the plate thickness of the target region when time-series data of the net radiation and sensible heat transport of the target region are input.

[0007] The heat balance analysis means is Based on the time-series data of the weather information acquired by the weather information acquisition means and the heat balance analysis model, it is possible to calculate time-series data of the theoretical surface temperature of the target region. Time-series data of the net radiation and sensible heat transport of the target region may be calculated using time-series data of the surface temperature of the target region based on the thermal image, and, if necessary, using time-series data of the theoretical surface temperature.

[0008] The plate thickness estimation means may calculate an estimated value of the plate thickness of the target area based on time-series data of the net radiation and sensible heat transport in the target area during the time period in which the temperature rises on the measurement day.

[0009] The heat balance analysis means divides the thermal image into a plurality of cell images each corresponding to the target area, and based on the time-series data of each target area of the thermal image, the time-series data of the meteorological information, and the heat balance analysis model, calculates the time-series data of the net radiation amount and the sensible heat transport amount for each target area. The plate thickness estimation means may calculate an estimated value of the plate thickness for each target area based on the time-series data of the net radiation amount and the sensible heat transport amount for each target area.

[0010] The plate thickness estimation device may further include a diagnosis means for diagnosing the corrosion state of the target area based on the estimated value of the plate thickness of the target area calculated by the plate thickness estimation means.

[0011] The meteorological information acquisition means acquires time-series data of prior meteorological information, which is the meteorological information at the location on a day before the measurement day. The plate thickness estimation device may further include a measurement time zone determination means for determining a measurement time zone on the measurement day suitable for estimating the plate thickness of the target area based on the time-series data of the prior meteorological information acquired by the meteorological information acquisition means.

[0012] The measurement time zone determination means calculates time-series data of the theoretically surface temperature of the target area based on the time-series data of the prior meteorological information acquired by the meteorological information acquisition means and the heat balance analysis model. The measurement time zone may be determined using the calculated time-series data of the theoretically surface temperature.

[0013] To achieve the above object, a plate thickness estimation method according to a second aspect of the present invention includes a thermal image acquisition step of acquiring a thermal image showing the thermal distribution of a steel plate from an infrared camera that measures the thermal distribution of the steel plate, and acquiring time-series data of the thermal image on a certain measurement day; a meteorological information acquisition step of acquiring time-series data of meteorological information on the measurement day at a location related to the installation location of the steel plate; A heat balance analysis step calculates time-series data of the net radiation and sensible heat transport in the target area for thickness estimation of the steel plate, based on the time-series data of the thermal image acquired in the thermal image acquisition step, the time-series data of the meteorological information acquired in the meteorological information acquisition step, and a predetermined heat balance analysis model. The system includes a plate thickness estimation step which calculates an estimated value of the plate thickness of the target region based on time-series data of the net radiation and sensible heat transport of the target region calculated in the heat balance analysis step, The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. In the plate thickness estimation step, a trained model that has undergone machine learning using training data is used to calculate the estimated plate thickness of the target area, so as to output an estimated value of the plate thickness of the target area when time-series data of the net radiation and sensible heat transport of the target area are input.

[0014] To achieve the above objective, the program according to the third aspect of the present invention is: Computers, A thermal image acquisition means that obtains a thermal image showing the thermal distribution of a steel plate from an infrared camera that measures the thermal distribution of the steel plate, and obtains time-series data of the thermal image on a given measurement day. A weather information acquisition means for acquiring time-series data of weather information on the measurement date at a location related to the installation location of the steel plate, A thermal image acquisition means obtains time-series data of thermal images, a thermal image acquisition means obtains time-series data of meteorological information, and a thermal balance analysis means calculates time-series data of net radiation and sensible heat transport in the target area for thickness estimation of the steel plate based on a predetermined thermal balance analysis model. Based on the time-series data of net radiation and sensible heat transport in the target region calculated by the heat balance analysis means, it functions as a plate thickness estimation means that calculates an estimated value of the plate thickness of the target region. The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. The plate thickness estimation means includes a trained model that has undergone machine learning using training data, which outputs an estimated value of the plate thickness of the target region when time-series data of the net radiation and sensible heat transport of the target region are input. [Effects of the Invention]

[0015] According to the present invention, even steel plates that make up existing structures can have their thickness easily estimated. [Brief explanation of the drawing]

[0016] [Figure 1] This figure shows the overall configuration of a plate thickness estimation system according to one embodiment of the present invention. [Figure 2] This is a block diagram showing the configuration of the plate thickness estimation device according to the same embodiment. [Figure 3] This flowchart shows an example of the plate thickness estimation process according to the same embodiment as above. [Figure 4] This flowchart shows an example of the measurement time period determination process according to the above embodiment. [Figure 5] This is a diagram showing a portion of a thermal image, and is intended to explain the multiple cell images obtained by dividing the thermal image. [Figure 6] This is a schematic diagram of the heat balance analysis model according to the same embodiment. [Figure 7] This figure shows a portion of the diagnostic images according to the same embodiment as described above. [Figure 8] (a) and (b) are diagrams showing the experimental setup according to one embodiment. [Figure 9] This diagram shows the evaluation flow of the experiment. [Figure 10] The figure shows time-series data of weather information obtained in the experiment, with (a) temperature, (b) solar radiation, (c) average wind speed, and (d) water vapor pressure. [Figure 11] (a) and (b) are diagrams showing thermocouples placed on steel sheet piles in the same experiment. [Figure 12]Figures (a) to (f) show the relationship between the surface temperature and the back surface temperature measured by thermocouples installed on the flanges of steel sheet piles in each case under consideration. [Figure 13] This figure shows the changes in surface temperature of steel sheet piles, comparing calculated values ​​based on weather information with measured values. [Figure 14] This figure shows the difference in calculated temperatures of steel sheet piles depending on the plate thickness. (a) shows data from May 4, 2021, and (b) shows data from May 6, 2021. [Figure 15] Figures (a) to (d) show the relationship between thermal image data and thermocouple data on the south side of the web before standardization on each measurement day. [Figure 16] Figures (a) to (d) show the relationship between thermal image data and thermocouple data on the north side of the web before standardization on each measurement day. [Figure 17] Figures (a) to (d) show the relationship between thermal image data and thermocouple data on the south side of the web after standardization on each measurement day. [Figure 18] Figures (a) to (d) show the relationship between thermal image data and thermocouple data on the north side of the web after standardization on each measurement day. [Figure 19] This figure shows the machine learning analysis flow for the experiment. [Figure 20] This is a conceptual diagram of a random forest. [Figure 21] This figure compares the importance of explanatory variables with temperature. (a) shows data from May 4, 2021, and (b) shows data from May 5, 2021. [Figure 22] This figure compares the importance of explanatory variables with temperature, with (a) showing data from May 6, 2021, and (b) showing data from May 7, 2021. [Figure 23] This figure compares the importance of explanatory variables with wind speed. (a) shows data from May 4, 2021, and (b) shows data from May 5, 2021. [Figure 24]This figure compares the importance of explanatory variables with wind speed; (a) shows data from May 6, 2021, and (b) shows data from May 7, 2021. [Modes for carrying out the invention]

[0017] One embodiment of the present invention will be described with reference to the drawings.

[0018] Figure 1 shows the overall configuration of the plate thickness estimation system 100 according to this embodiment. The plate thickness estimation system 100 estimates the plate thickness of steel plates 2 that constitute structures used in water facilities 1 such as agricultural irrigation canals. The steel plates 2 are, for example, steel sheet piles. In Figure 1, reference numeral 3 indicates a waterway.

[0019] The plate thickness estimation system 100 comprises an infrared camera 200, a weather observation device 300, and a plate thickness estimation device 400. The infrared camera 200 and the weather observation device 300 are each connected to the plate thickness estimation device 400 via a wired or wireless communication network.

[0020] The infrared camera 200 is an infrared thermography camera that measures the thermal distribution of the steel plate 2 and generates a thermal image (thermography) showing the thermal distribution of the steel plate 2 in the measurement area. In this embodiment, the infrared camera 200 measures the thermal distribution of the steel plate 2 over a specific time period on a given measurement day and transmits the time-series data of the thermal image for that measurement day to the plate thickness estimation device 400.

[0021] The meteorological observation device 300 observes meteorological information, including temperature, solar radiation, average wind speed, and water vapor pressure, at locations related to the installation location of the steel plate 2. Here, "locations related to the installation location of the steel plate 2" is preferably the location where the steel plate 2 is installed, but any location can be arbitrarily selected as long as it is a location where meteorological conditions similar to those of the steel plate 2 can be observed. In the following description, the meteorological observation device 300 will be described as observing meteorological information at the location where the steel plate 2 is installed. The meteorological observation device 300 transmits time-series data of the observed meteorological information to the plate thickness estimation device 400. In this embodiment, the meteorological observation device 300 transmits at least time-series data of meteorological information for the measurement day to the plate thickness estimation device 400.

[0022] The plate thickness estimation device 400 is a terminal device operated by an operator, consisting of a personal computer, tablet terminal, etc. As shown in Figure 2, the plate thickness estimation device 400 includes a control unit 410, a storage unit 420, an operation unit 430, a display unit 440, and a communication unit 450. These units are connected by a bus for transmitting signals. The plate thickness estimation device 400 may also be composed of multiple computers that cooperate with each other.

[0023] The control unit 410 includes a CPU (Central Processing Unit), ROM (Read Only Memory), and RAM (Random Access Memory). In the control unit 410, the CPU reads the control program stored in the ROM, and while using the RAM as work memory, it controls the operation of the entire plate thickness estimation device 400.

[0024] The storage unit 420 is a non-volatile memory such as flash memory or a hard disk. The storage unit 420 stores programs and data used by the control unit 410 to perform various processes, including the OS (Operating System) and application programs, as needed. The storage unit 420 also stores data generated or acquired by the control unit 410 as a result of various processes.

[0025] The control unit 430 is equipped with input devices such as a keyboard, mouse, buttons, touchpad, and touch panel, and accepts operations from the operator. By operating the control unit 430, the operator can input commands to the plate thickness estimation device 400.

[0026] The display unit 440 is a display device consisting of a liquid crystal display, an organic EL (Electro-Luminescence) display, etc. The display unit 440 can display various types of information obtained as a result of processing by the control unit 410. For example, the display unit 440 can display thermal images acquired by the control unit 410 from the infrared camera 200, weather information acquired by the control unit 410 from the weather observation device 300, and so on.

[0027] The communication unit 450 is an interface for wired or wireless communication with external devices, including the infrared camera 200 and the weather observation device 300. Under the control of the control unit 410, the communication unit 450 communicates with the infrared camera 200 and the weather observation device 300, respectively, to acquire time-series data of thermal images and time-series data of weather information. The communication unit 450 can also connect to a wide-area network such as the Internet via wired or wireless communication.

[0028] As shown in Figure 2, the control unit 410 includes, functionally, a thermal image acquisition unit 411, a weather information acquisition unit 412, a heat balance analysis unit 413, a plate thickness estimation unit 414, a diagnostic unit 415, an output unit 416, and a measurement time zone determination unit 417. The control unit 410 functions as each of these units by having the CPU execute a program stored in ROM. These functional units will be described below along with the plate thickness estimation process and measurement time zone determination process executed by the control unit 410. For example, the control unit 410 starts the plate thickness estimation process shown in Figure 3 and the measurement time zone determination process shown in Figure 4 in response to a command input using the operation unit 430.

[0029] (Plate thickness estimation process) When the plate thickness estimation process is started, the thermal image acquisition unit 411 acquires a thermal image showing the heat distribution of the steel plate 2 from the infrared camera 200 which measures the heat distribution of the steel plate 2, and acquires time-series data of the thermal image on a certain measurement day (i.e., time-series data of the heat distribution of the steel plate 2) (step S101). This time-series data of the thermal image is data from multiple thermal images obtained by measuring the steel plate 2 with the infrared camera 200 at predetermined time intervals (for example, every hour) on the measurement day. In addition, the weather information acquisition unit 412 acquires time-series data of weather information including temperature, solar radiation, average wind speed, and water vapor pressure at the location where the steel plate 2 is installed on the measurement day (step S102). This time-series data of weather information is data obtained by the weather observation device 300 at predetermined time intervals (for example, every 10 minutes) on the measurement day.

[0030] Next, the control unit 410, functioning as a heat balance analysis unit 413, divides each of the multiple thermal images acquired by the thermal image acquisition unit 411 into multiple cell images (step S103). Figure 5 shows a portion of the thermal image at a certain time. As shown by the dashed lines in the figure, the heat balance analysis unit 413 obtains multiple cell images 4 by dividing the thermal image into a matrix. Each of the multiple cell images 4 corresponds to the target region for plate thickness estimation (hereinafter also simply referred to as the "target region"). The processing from step S103 onwards is executed after the measurement date (specifically, after the acquisition of time-series data of thermal images and time-series data of weather information is completed). For example, the processing from step S103 onwards may be executed on the day following the measurement date.

[0031] Next, the heat balance analysis unit 413 performs a heat balance analysis based on the time-series data for each cell image 4 of the thermal image (i.e., time-series data for each target area of ​​the thermal image), the time-series data of meteorological information, and a predetermined heat balance analysis model (step S104). Then, for each target area, the heat balance analysis unit 413 calculates time-series data of net radiation and sensible heat transport as explanatory variables to be used in the trained model described later (step S105). The heat balance analysis model and the method for calculating net radiation and sensible heat transport will be explained below.

[0032] Here, a schematic diagram of the heat balance analysis model is shown in Fig. 6. Regarding the plate thickness, the plate thickness of the sound steel sheet pile was set to 5 mm, and considering the reduction in plate thickness due to corrosion, the spatial step was discretized at 0.0005 m (0.5 mm). Also, the time step was discretized at 0.008 s. The heat balance analysis unit 413 calculates the heat balance at the surface layer of the steel sheet pile (Δx = 0.0005 m, cell number: 0 shown in Fig. 6). In the heat balance analysis, the heat balance of the ground surface was referred to (Jun Muto, 2000: Science of the Atmosphere Near the Ground Surface - Understanding and Application. The University of Tokyo Press) for modeling. For the inflow and outflow of fluxes at the surface layer of the steel sheet pile, the net radiation amount (radiative heat transfer), the sensible heat transport amount (convective heat transfer), and the conductive heat inside the steel sheet pile (conductive electrothermal) were assumed. The net radiation amount was set to be positive for the inflow to the surface layer. The sensible heat transport amount and the conductive heat inside the steel sheet pile were set to be positive for the inflow from the surface layer. In this model, the latent heat transport amount is not considered. In this case, the heat balance equation at the surface of the steel sheet pile is expressed by the following equation (Equation 1). The symbols in Equation (1) are as follows. R n : Net radiation amount (W / m 2 ) H: Sensible heat transport amount (W / m 2 ) G: Conductive heat inside the steel sheet pile (W / m 2 )

[0033]

Equation

[0034] (Net radiation amount R n ) Net radiation amount R n is expressed by the following equation (Equation 2). The symbols in Equation (2) are as follows. S ↓ : Solar radiation amount (W / m 2 ) ref: Albedo ε sfc : Emissivity σ: Stefan - Boltzmann constant T s : Surface temperature (°C) L ↓ : Atmospheric radiation amount (W / m2 ) Of these, solar radiation is obtained from meteorological information acquired by the meteorological information acquisition unit 412. An example of surface temperature calculation will be described later. For albedo and emissivity, the values ​​shown in (Table 1) below were used.

[0035]

number

[0036] (Equation 2) Atmospheric radiation L ↓ This can be calculated using the following formula (Equation 3). The symbols in formula (Equation 3) are as follows: L df ↓ : Downward longwave radiation σ: Stefan Boltzmann constant T a Temperature (°C) The temperature is obtained from the weather information acquired by the weather information acquisition unit 412. Note that x is represented by the following equation (Equation 4).

[0037]

number

[0038]

number

[0039] In equation (Mathematics 4), w TOP * This represents the total amount of available water vapor, for example, 0.1 cm³. <w TOP * It is <6cm. DEW is the dew point temperature (°C), which can be calculated using the following equation (Equation 5). In equation (Equation 5), e is the water vapor pressure (hPa), which is obtained from the weather information acquired by the weather information acquisition unit 412.

[0040]

number

[0041] (sensible heat transport amount H) The sensible heat transport rate H is expressed by the following equation (Equation 6). In equation (Equation 6), the surface temperature T s Temperature T a This has already been explained. The heat transfer coefficient h (average) is expressed by the following equation (Equation 7).

[0042]

number

[0043]

number

[0044] In equation (7), Nu is the Nusselt number, which is expressed by the following equation (8). L is the characteristic length (m) of the steel sheet pile, and here it is set to half the length of the shorter side in the web portion of the steel sheet pile (e.g., 0.07m). k is the thermal conductivity of air, and the value shown in (Table 2) below was used.

[0045]

number

[0046] In equation (8), Re is the Reynolds number, which is expressed by the following equation (9). Pr is the Prandtl number, and the values ​​listed in (Table 2) below were used.

[0047]

number

[0048] In equation (9), U is the wind speed (m / s), which is obtained from the weather information acquired by the weather information acquisition unit 412. As previously explained, L is the characteristic length (m) of the steel sheet pile. v is the kinematic viscosity of the air, and the value shown in (Table 2) below was used.

[0049] (Conductive heat G inside steel sheet piles) The conductive heat G inside the steel sheet pile is expressed by the following equation (Equation 10). In equation (Equation 10), c is the specific heat and ρ is the density; the values ​​shown in Table 1 below were used. dt(s) corresponds to the discretized time step as described above. dx corresponds to the discretized spatial step as described above (i.e., it corresponds to Δx shown in Figure 6). D is the position where heat conduction is approximately zero. T is the temperature (K) inside the steel sheet pile.

[0050]

number

[0051] The physical properties of the steel used in the above formulas are shown in (Table 1), and the physical properties of air are shown in (Table 2).

[0052] [Table 1]

[0053] [Table 2]

[0054] The heat balance analysis unit 413 uses the heat balance analysis model defined by the above formula, the time-series data for each cell image 4 of the thermal image (i.e., time-series data for each target area of ​​the thermal image), and the time-series data of meteorological information to calculate time-series data of net radiation and sensible heat transport for each target area.

[0055] The surface temperature T used when calculating net radiation and sensible heat transport. s This can be determined based on the cell images 4 that constitute the thermal image. For example, the heat balance analysis unit 413 may determine the temperature shown at the center of one cell image 4 as the surface temperature Ts of the target region corresponding to that cell image 4. Alternatively, the heat balance analysis unit 413 may determine the average of the heat distribution shown by one cell image 4 as the surface temperature T of the target region corresponding to that cell image 4. s You may also request it as follows:

[0056] Furthermore, in equation (10) which shows the conductive heat G, if we assume that the temperature T inside the steel sheet pile is constant, then the temperature T inside the steel sheet pile is equal to the surface temperature T. s It can be replaced with R in the heat balance equation of (Mathematics 1). n The calculation process for each of H and G involves the surface temperature T s Because it includes, the heat balance analysis unit 413 calculates the surface temperature T based on (Equation 1). s The theoretical surface temperature (i.e., an estimated value of the surface temperature) may be calculated by solving an equation with the unknown variable. In other words, the heat balance analysis unit 413 may calculate the theoretical surface temperature from meteorological information and a heat balance analysis model without using thermal image data from the infrared camera 200.

[0057] The heat balance analysis unit 413 calculates net radiation and sensible heat transport using the surface temperature identified from the thermal image, as well as, if necessary, the theoretical surface temperature calculated independently of the thermal image. For example, the heat balance analysis unit 413 may calculate a weighted average that takes into account the importance of the net radiation calculated based on the surface temperature identified from the thermal image and the net radiation calculated based on the theoretical surface temperature, and determine the calculated weighted average value as the net radiation to be used as an explanatory variable. The same applies to sensible heat transport. Furthermore, if it is difficult to identify the surface temperature in a certain cell image 4, the heat balance analysis unit 413 may calculate the net radiation and sensible heat transport for the target region corresponding to that cell image 4 based on the theoretical surface temperature.

[0058] Returning to Figure 3, following the processing in step S105, the plate thickness estimation unit 414 of the control unit 410 estimates the plate thickness of each target region based on the time-series data of net radiation and sensible heat transport calculated in step S105 and the trained model (step S106). Specifically, the plate thickness estimation unit 414 calculates a class label (output value), which will be described later, for each target region.

[0059] The plate thickness estimation unit 414 has a trained model that has undergone machine learning, and estimates the plate thickness based on this trained model. This trained model has been subjected to machine learning using training data so that when time-series data of net radiation and sensible heat transport of the target area are input, it outputs an estimated value of the plate thickness of that area. The estimated value of the plate thickness of the target area is expressed in several predetermined stages depending on the plate thickness. As an example, the trained model is a plate thickness estimation algorithm constructed using a random forest. Training data can be prepared using the methods described below.

[0060] First, multiple cases of steel sheet piles with different corrosion conditions are prepared, including normal steel sheet piles with no corrosion at all. Then, for each of the multiple cases of steel sheet piles, the plate thickness for each target area is measured in advance, and a class label is assigned to each target area, evaluating the plate thickness in stages. In addition, based on time-series data of thermal images and weather information for a specific time period on a given day (for example, 4am to 7pm), the heat balance analysis unit 413 calculates time-series data of net radiation and sensible heat transport for each target area in advance. This calculation is performed over multiple days. Then, the dataset constructed by corresponding the class labels assigned to each target area for each of the multiple cases of steel sheet piles and the time-series data of net radiation and sensible heat transport for each target area over multiple days is divided into training data and evaluation data in a predetermined ratio. Using the training data and evaluation data obtained in this way, machine learning is performed on the model using a random forest to construct a trained model (plate thickness estimation algorithm). The evaluation of the trained model will be explained in the (Examples) section later.

[0061] Following step S106, the control unit 410 performs a corrosion diagnosis using the functions of the diagnostic unit 415 (step S107). For example, based on the estimated values ​​(class label values) calculated for each target area, the diagnostic unit 415 generates a diagnostic image that visualizes areas with a high probability of corrosion and pitting, as shown in Figure 7, by coloring the target areas (cell images 4) where a decrease in plate thickness is observed in the thermal image. For example, the diagnostic unit 415 can apply colors that change in stages according to the degree of plate thickness reduction to the target areas (cell images 4) that are corroded or have a high probability of corrosion. Then, the control unit 410 displays the diagnostic image on the display unit 440 using the functions of the output unit 416. Note that the corrosion diagnosis by the diagnostic unit 415 is not limited to the diagnostic image shown in Figure 7. The diagnostic unit 415 may also generate text data indicating the location of the target area and the degree of corrosion in that target area. In this case, the output unit 416 displays the text data on the display unit 440. This concludes the explanation of the plate thickness estimation process.

[0062] (Measurement time period determination process) The measurement time period determination process will be explained with reference to Figure 4. The measurement time period determination process is performed on a day prior to the measurement day (preferably the day before). When the measurement time period determination process is started, the weather information acquisition unit 412 acquires time-series data of prior weather information, including temperature, solar radiation, average wind speed, and water vapor pressure, at the location where the steel plate 2 is installed on a day prior to the measurement day (step S201).

[0063] Next, the heat balance analysis unit 413 performs a heat balance analysis based on the time-series data of prior weather information acquired by the weather information acquisition unit 412 and the heat balance analysis model described above (step S202). In the measurement time period determination process, thermal image data is not acquired from the infrared camera 200, and the heat balance analysis unit 413 calculates the time-series data of the theoretical surface temperature of the target area for thickness estimation of the steel plate 2 using the same method as described above. The target area here is a virtual area, and it is sufficient that the conditions for performing the heat balance analysis are predetermined based on the results of prior investigations into the thickness, shape, size, etc. of the steel plate 2 used in the water supply facility 1.

[0064] Next, the control unit 410, functioning as a measurement time zone determination unit 417, determines a measurement time zone on the measurement day that is suitable for estimating the plate thickness of the target area, based on the results of the heat balance analysis in step S202 (step S203). For example, the measurement time zone determination unit 417 determines the measurement time zone to be the time period during which the surface temperature rises, by referring to the time series data of the theoretical surface temperature calculated in step S202. Examples of time zones suitable for estimating the plate thickness of the steel plate 2 will be described later in the (Examples) section.

[0065] The measurement time period determination unit 417 may determine the measurement time period based on the time-series data of prior weather information acquired in step S201. For example, the measurement time period determination unit 417 may determine the measurement time period based on the time-series data of prior weather information, such as (i) a time period in which the temperature is expected to be rising, or (ii) a time period in which the temperature is expected to be rising and the wind speed is below a predetermined threshold. In such cases, the processing in step S202 of the measurement time period determination process may be omitted.

[0066] Furthermore, in step S202, the heat balance analysis unit 413 may calculate time-series data of net radiation and sensible heat transport in the target area for thickness estimation of the steel plate 2 based on the calculated time-series data of theoretical surface temperature. Then, in step S203, the measurement time zone determination unit 417 may determine the measurement time zone based on the time-series data of net radiation and sensible heat transport in the target area. For example, the measurement time zone determination unit 417 may determine the measurement time zone as a time zone in which the net radiation or sensible heat transport is highly important as an explanatory variable in thickness estimation, based on a pre-trained model that has already been constructed. In such a case, for example, a pre-trained model that has been subjected to machine learning using training data may be further constructed so as to output the importance of the target area in thickness estimation when time-series data of net radiation and sensible heat transport in the target area is input. This importance is represented by a predetermined number of levels (labels) according to its level. The measurement time zone determination unit 417 calculates the importance of the explanatory variables based on the trained model constructed in this way, and can determine the time period in which the calculated importance is higher than a predetermined threshold as the measurement time zone.

[0067] The plate thickness estimation unit 414 can calculate an estimated value of the plate thickness of a target area based on the time-series data of the net radiation and sensible heat transport of the target area during the measurement time period determined by the measurement time period determination unit 417 as described above. The measurement time period determined by the measurement time period determination unit 417 should only be used as a reference for the measurement date, and the time-series data of thermal images, weather information, etc., used in the plate thickness estimation process are not bound by this measurement time period. Whether or not to reflect the measurement time period determined by the measurement time period determination unit 417 in the plate thickness estimation process is left to the operator's discretion. This concludes the explanation of the measurement time period determination process.

[0068] Here, a characteristic of the performance degradation of a steel sheet pile waterway, which is an example of a water management facility 1, is that corrosion progresses in the water level fluctuation region. In particular, in agricultural steel sheet pile waterways, many facilities have been confirmed where corrosion becomes apparent about 10 years after installation due to extensive facility management, and the risk of pitting corrosion and buckling failure due to the progression of corrosion has been pointed out. Using the plate thickness estimation method described above, the plate thickness of the steel plate 2 used in the steel sheet pile waterway can be estimated by non-destructive and non-contact inspection, and the corrosion status can be evaluated based on the estimated plate thickness. Below, the results of an experimental study of a plate thickness estimation method using the plate thickness estimation device 400 will be described as one embodiment. As one embodiment, the inventors of this application conducted an experiment under the conditions described below.

[0069] (Examples) 1. Overview of the steel sheet pile sample model experiment Three steel sheet pile samples with different thicknesses—Case A (newly installed), Case B (corroded (with holes)), and Case C (corroded (without holes))—were set up in the experimental environment, and a model demonstration experiment was conducted to evaluate the temperature differences depending on weather conditions and plate thickness. The cases considered in this test are shown below (Table 3).

[0070] [Table 3]

[0071] The measurement items in this experiment were meteorological data, steel sheet pile temperature data (using thermocouples, contact), and thermal images (non-contact). The experiment was conducted from May 4th to May 7th, 2021, on the rooftop of the Faculty of Agriculture building at Niigata University. Figures 8(a) and (b) show the experimental setup. The measurement intervals were 10 minutes for meteorological data and thermocouple data, and 1 hour for thermal images. Meteorological data included temperature, solar radiation, average wind speed, maximum instantaneous wind speed, precipitation, atmospheric pressure, and water vapor pressure. Of this meteorological data, the data used in the analysis model for the plate thickness estimation device 400 (corresponding to the aforementioned meteorological information) were temperature, solar radiation, average wind speed, and water vapor pressure. Thermocouples were installed at six locations per steel sheet pile sample. Since the web was the surface for thermal image analysis, thermocouples were not installed on the surface. The following equipment was used: ATMOS-41 combined weather measurement unit (manufactured by METER Corporation) as weather observation equipment, K-type thermocouple (manufactured by Toyo Thermal Chemical Co., Ltd.) as thermocouple, and InfRec R300SR (manufactured by Nippon Avionics Co., Ltd.) as infrared thermography camera.

[0072] 2. Evaluation Flow Figure 9 shows the evaluation flow chart for this experiment.

[0073] 3. Obtaining weather information Figures 10(a) to 10(d) show the time-series data of meteorological information used in the analysis model. Figure 10(a) shows temperature, Figure 10(b) shows solar radiation, Figure 10(c) shows average wind speed, and Figure 10(d) shows water vapor pressure. This meteorological information was acquired at 10-minute intervals.

[0074] 4. Acquisition of temperature data for steel sheet piles using thermocouples Thermocouples were installed at six locations per steel sheet pile sample (1ch: south surface of flange, 2ch: south back surface of flange, 3ch: south back surface of web, 4ch: north back surface of web, 5ch: north surface of flange, 6ch: north back surface of flange). The arrangement of thermocouples is shown in Figures 11(a) and (b). Temperature data for each channel was acquired at 10-minute intervals. Of these, the web temperature data for each case under consideration was used in later analysis. Since thermocouples were installed on both the front and back surfaces of the steel sheet pile sample at the flange, the presence or absence of a temperature difference between the front and back surfaces was verified using measured values. Figures 12(a) to (f) show the relationship between the surface temperature and the back surface temperature measured by the thermocouples installed at the flange. As shown in Figures 12(a) to (f), due to the relatively high thermal conductivity of steel, the temperature difference between the front and back surfaces was small, indicating that the temperature gradient inside the sample was very small (uniform temperature).

[0075] 5. Non-contact temperature measurement using thermal imaging An infrared thermography camera was used to acquire thermal images at one-hour intervals (a total of 16 time points) during the period from 4:00 AM to 7:00 PM each day. A portion of the thermal images is shown in Figure 5.

[0076] 6. Heat balance analysis In the heat balance analysis, we attempted to reproduce the surface temperature by calculating the heat balance of steel sheet piles based on meteorological information. In the analysis results, we focused on reproducing the differences in surface temperature according to the plate thickness and identifying the time period in which these differences occur. The heat balance analysis model is as described in the above embodiment, with reference to Figure 6, (Equation 1) to (Equation 20). In this experiment, considering the reduction in plate thickness due to corrosion, we considered plate thicknesses of 5 mm (sound), 4 mm, 3 mm, 2 mm, and 1 mm. The physical properties used in the heat balance analysis were the values ​​shown in (Table 1) and (Table 2) above. Here, without using thermal images, the theoretical surface temperature of the steel sheet pile was calculated based on time-series data of meteorological information and the physical properties. The analysis conditions are as follows. <Analysis conditions> • Spatial step: 0.0005m (0.5mm) • Time step: 0.008s Initial conditions: Measured thermocouple values ​​at 0:00 (assuming constant values ​​including the interior of the steel sheet pile). • Boundary conditions: The surface temperature was calculated using the heat balance, and the back surface was treated as an adiabatic boundary. • Weather information: Temperature, solar radiation, average wind speed, and water vapor pressure were provided every 10 minutes. • Calculation time: Approximately 90 minutes (time required to calculate 4 days' worth of data) • Characteristic length for determining the Reynolds number: 0.07 m (7 cm), which is half the length of the web in the short direction of the steel sheet pile.

[0077] 7. Results and discussion of heat balance analysis and infrared measurement. 7.1. Verification of fitting between calculated and measured values We compared the calculated surface temperature (i.e., the theoretical surface temperature) obtained from a heat balance analysis assuming a plate thickness of 5 mm with the thermocouple data (measured temperature) of the web in a newly installed steel sheet pile sample, Case A, with a constant plate thickness of 5 mm. As mentioned above, since the temperature difference between the front and back sides was confirmed to be small in a verification experiment using thermocouples installed on both sides of the flange, we used the data from a thermocouple installed on the back side of the web. Figure 13 shows the results of comparing the calculated and measured values. Referring to the figure, it can be seen that on cloudy and rainy days on May 5, 2021 and May 7, 2021, there was a difference between the calculated and measured values ​​in some parts of the daytime. However, on sunny days on May 4, 2021 and May 6, 2021, it was confirmed that the measured values ​​were generally reproduced by the calculated values. Thus, since it was shown that the temperature of steel sheet piles can be reproduced by heat balance analysis, we then investigated the temperature differences depending on the plate thickness set in the analysis.

[0078] Figures 14(a) and (b) show the temperature differences according to plate thickness on May 4th and May 6th, 2021, when the fit between calculated and measured values ​​was good. In this study, calculated values ​​were used in all cases. In Figures 14(a) and (b), in addition to the temperature on the first vertical axis, relative temperature is shown on the second vertical axis. Here, relative temperature represents the ratio when the temperature at a plate thickness of 5 mm is used as the base (denominator) and the temperature at each plate thickness is used as the numerator. Focusing on the morning, a tendency for the relative temperature to be higher as the plate thickness decreased was observed during the time periods of 5-8 am on May 4th, 2021 (Figure 14(a)) and 5-10 am on May 6th, 2021 (Figure 14(b)). Focusing on the afternoon, a tendency for lower relative temperatures with thinner plates was observed during the 17:00-20:00 time slot on May 4, 2021 (Figure 14(a)) and the 16:00-20:00 time slot on May 6, 2021 (Figure 14(b)). During the 10:00-16:00 time slot, on both days, a tendency for larger relative temperature changes (increase / decrease) at specific time intervals with thinner plates was observed. This indicates that the time when temperature differences depending on plate thickness become apparent can be identified through heat balance analysis. Furthermore, we investigated whether a similar trend could be observed in thermocouple data, and a trend similar to that observed in the heat balance analysis in Figures 14(a) and (b) was confirmed. In other words, it was confirmed that temperature differences depending on plate thickness also occur in actual temperature measurements using contact thermocouples.

[0079] 7.2. Standardization of thermal image data We investigated the temporal changes in temperature data obtained from thermal images acquired at one-hour intervals from 4:00 to 19:00 between May 4th and May 7th, 2021. The analysis range for the thermal images was defined as the web (south and north sides). The analysis range was divided into cell images as described above. The size of each cell image was 11 × 13 pixels (height × width), and each of the south and north sides of the web was divided into 51 cell images of 3 × 17 (height × width). In other words, there were 51 cell images for the south side of the web and 51 cell images for the north side of the web, for a total of 102 cell images for the south and north sides of the web. The plate thickness measurement point was positioned in the center of each cell image. To reduce temperature differences due to variations in sample emissivity, the temperature was standardized in each cell to have an average value of 0°C and a standard deviation of 1°C. Figures 15(a) to (d) show the relationship between the thermal image data and thermocouple data for the south side of the web before standardization on each measurement day. Figures 16(a) to 16(d) show the relationship between thermal image data and thermocouple data on the north side of the web before standardization on each measurement day. Figures 17(a) to 17(d) show the relationship between thermal image data and thermocouple data on the south side of the web after standardization on each measurement day. Figures 18(a) to 18(d) show the relationship between thermal image data and thermocouple data on the north side of the web after standardization on each measurement day. Standardization resulted in the thermocouple data and thermal image data approaching a proportional relationship of y=x. Regarding the results of the uncorrelated test, a statistically significant difference at a significance level of 1% was observed in all cases from Figures 15 to 18.

[0080] 8. Evaluation of plate thickness using machine learning 8.1. Analysis Flow Based on the temperature differences depending on plate thickness revealed by heat balance analysis, we attempted to perform a detailed evaluation of plate thickness by using machine learning. Figure 19 shows the analysis flow using machine learning.

[0081] 8.2.Analysis conditions As mentioned above, a total of 102 cell images are obtained for each study case, including images from the south and north sides of the web. Adding these together for Case A, Case B, and Case C yields a total of 306 cell images. These 306 data points (102 data points each for Case A, Case B, and Case C), obtained by measuring the plate thickness at the corresponding locations in each of these 306 cell images, were split 5:5 into training and evaluation data. 32 data points were used as explanatory variables: net radiation (16 data points) and sensible heat transport (16 data points) for each hour from 4:00 to 19:00 each day. As mentioned above, surface temperature was used in the calculation of net radiation and sensible heat transport, and the thermal image data from each cell image was used as this surface temperature. The plate thickness was classified into three stages, and class labels were assigned as shown in (Table 4).

[0082] [Table 4]

[0083] 8.3. Algorithm Construction An algorithm for estimating plate thickness was constructed using random forests, a type of supervised learning. Random forests can use datasets with duplicate data in the training data. Other advantages include higher accuracy in classification without overfitting compared to decision trees, and faster processing speed. The Gini coefficient, which represents the impurity at a given node t, is expressed by the following equation (Equation 11).

[0084]

number

[0085] Here, p(c i |t) represents the probability that the i-th class data is selected at node t. The tree is branched in such a way that the decrease in impurity ΔL(t) due to the Gini coefficient is maximized. This is shown in equation (Equation 12) below.

[0086]

number

[0087] Here, p L and p R t is the probability of being classified into the left and right branches after each division. L and t R These are the nodes at the ends of the left and right edges, respectively. Figure 20 shows a conceptual diagram of a random forest.

[0088] 8.4. Comparison of Model Accuracy The accuracy of pre-trained models using machine learning based on random forests was compared using the following four metrics. 1) Accuracy Rate: The percentage of correct answers relative to all predictions. 2) Recall: The proportion of those that were actually positive that were predicted to be positive. 3) Precision: The percentage of data predicted to be positive that are actually positive. 4) F-score: Harmonic mean of precision and recall The recall, precision, and F-value were calculated for each class and evaluated as average values. A comparison of accuracy is shown in Table 5 below. The results revealed that plate thickness could be classified with high accuracy of 0.8 or higher for all indicators, regardless of the date. The difference in accuracy between May 4, 2021 and other dates is thought to be due to the frequent occurrence of wind speeds of 3 m / s or higher on May 4, 2021. It was revealed that plate thickness can be estimated and evaluated using net radiation and sensible heat transport.

[0089] [Table 5]

[0090] 8.5. Importance of explanatory variables The importance of explanatory variables is an index that shows the contribution rate of each explanatory variable in the classification. In this experiment, importance was calculated by weighting the decrease in the Gini coefficient by the sample size. The importance of explanatory variables compared with meteorological information temperature is shown in Figures 21(a), (b) and 22(a), (b), and the importance compared with wind speed is shown in Figures 23(a), (b) and 23(a), (b). Among these, the explanatory variables with the highest importance are shown in Table 6 below. Sensible heat transport was found to be of high importance during the morning hours from 4 to 12, and net radiation was found to be of high importance during the hours from 10 to 13. From the results of plate thickness estimation using machine learning, it was shown that plate thickness can be estimated well by using explanatory variables in the temperature rise process (times when the temperature rises) on the measurement day.

[0091] [Table 6]

[0092] The present invention is not limited by the embodiments and drawings described above. Modifications (including the deletion of components) can be made as appropriate without altering the essence of the invention. An example of such modification is described below.

[0093] The above example demonstrates how to construct a pre-trained model using random forests, but the pre-trained model may also be constructed using other well-known machine learning methods such as neural networks.

[0094] The above example illustrates how the plate thickness estimation device 400 acquires weather information from the weather observation device 300, but it is not limited to this. The plate thickness estimation device 400 may communicate with existing weather observation devices via a wide-area network and acquire weather information from those devices at locations that can be considered to have weather conditions similar to those at the installation location of the steel plate 2.

[0095] A component of the plate thickness estimation system 100, such as an infrared camera 200, may be mounted on a UAV (unmanned aerial vehicle) (commonly known as a drone) and capable of measuring the steel plate 2 by remote control. Alternatively, the infrared camera 200 and other components may be mounted on other mobile devices, such as a remotely controlled ground-based robot.

[0096] In the above example, steel plate 2 is described as a steel sheet pile used in a water management facility, but the type of steel plate 2 to be used for thickness estimation is arbitrary and not limited to this. The steel plate 2 whose thickness is estimated by the plate thickness estimation system 100 and plate thickness estimation device 400 may be a steel sheet pile for earth retention, a steel plate for civil engineering, a steel plate for a transmission tower, etc.

[0097] The above describes an example in which the meteorological information acquired by the meteorological information acquisition unit 412 includes temperature, solar radiation, average wind speed, and water vapor pressure, but it is not limited to this. For example, when the heat balance analysis unit 413 calculates the net radiation and sensible heat transport of the target area for plate thickness estimation based on the heat balance analysis model, it may calculate the net radiation and sensible heat transport by approximately treating the information that is less significant than the other information among the various pieces of information constituting the meteorological information. In other words, the meteorological information only needs to include at least one of the temperature, solar radiation, average wind speed, and water vapor pressure.

[0098] The program that performs the plate thickness estimation process and measurement time zone determination process described above is assumed to be pre-stored in the ROM of the control unit 410, but it may also be distributed and provided on a removable recording medium. Furthermore, the program may be downloaded from other devices connected to the plate thickness estimation device 400. The plate thickness estimation device 400 may also perform each process according to the program by exchanging various data with other devices via a telecommunications network or the like.

[0099] The plate thickness estimation device 400 described above comprises a thermal image acquisition means (thermal image acquisition unit 411), a weather information acquisition means (weather information acquisition unit 412), a heat balance analysis means (heat balance analysis unit 413), and a plate thickness estimation means (plate thickness estimation unit 414). Furthermore, the plate thickness estimation method using the plate thickness estimation device 400 comprises the steps performed by each of these means. In addition, the program used in the plate thickness estimation device 400 causes a computer (control unit 410) to function as these means. According to the plate thickness estimation device 400, the plate thickness estimation method, and the program, since it is only necessary to analyze thermal images and weather information, the plate thickness of steel plates 2 that make up existing structures can be easily estimated by non-destructive and non-contact inspection.

[0100] Furthermore, the heat balance analysis means can calculate time-series data of the theoretical surface temperature (estimated surface temperature) of the target region based on the time-series data of meteorological information acquired by the meteorological information acquisition means and the heat balance analysis model. The heat balance analysis means may use time-series data of the surface temperature of the target region based on thermal images, as well as time-series data of the theoretical surface temperature, to calculate time-series data of the net radiation and sensible heat transport of the target region, if necessary. With this configuration, not only the surface temperature based on thermal images, but also the theoretical surface temperature can be used in the analysis as needed.

[0101] Furthermore, the plate thickness estimation means may calculate an estimated value of the plate thickness of the target area based on time-series data of the net radiation and sensible heat transport in the target area during the time period in which the temperature rises on the measurement day. This configuration reduces the amount of data that needs to be measured and observed on the measurement day (time-series data of thermal images and time-series data of meteorological information), and allows for efficient estimation of the thickness of steel plate 2.

[0102] Alternatively, the heat balance analysis means may divide the thermal image into multiple cell images 4, each corresponding to a target region, and calculate time-series data of net radiation and sensible heat transport for each target region based on time-series data of the thermal image for each target region, time-series data of meteorological information, and a heat balance analysis model. In this case, the plate thickness estimation means calculates an estimated value of the plate thickness for each target region based on the time-series data of net radiation and sensible heat transport for each target region. With this configuration, the resolution of the object to be analyzed can be arbitrarily adjusted by setting the size of cell image 4.

[0103] Furthermore, the plate thickness estimation device 400 may also include a diagnostic means for diagnosing the corrosion status of the target area based on the estimated plate thickness of the target area calculated by the plate thickness estimation means. This configuration is useful because it allows for the diagnosis of the corrosion status of steel plate 2 through non-destructive and non-contact inspection.

[0104] Furthermore, the weather information acquisition means may acquire time-series data of prior weather information, including temperature, solar radiation, average wind speed, and water vapor pressure at the location on a day prior to the measurement day. The plate thickness estimation device 400 may further include a measurement time period determination means (measurement time period determination unit 417) that determines a measurement time period on the measurement day that is suitable for estimating the plate thickness of the target area, based on the time-series data of prior weather information acquired by the weather information acquisition means. This configuration is efficient because it reduces the amount of data that needs to be measured and observed on the measurement day (time-series data of thermal images and time-series data of meteorological information).

[0105] Alternatively, the measurement time period determination means may calculate time series data of the theoretical surface temperature of the target area based on the time series data of prior meteorological information acquired by the meteorological information acquisition means and a heat balance analysis model, and then determine the measurement time period using the calculated time series data of the theoretical surface temperature. This configuration allows for the prediction of the surface temperature changes of the steel plate 2 on the measurement day, thus enabling efficient estimation of the steel plate thickness 2.

[0106] In the above explanation, explanations of known technical matters have been omitted as appropriate in order to facilitate understanding of the present invention. [Explanation of Symbols]

[0107] 100... Plate thickness estimation system 200...Infrared camera, 300...Meteorological observation equipment 400…Plate thickness estimation device 410... Control Unit 411... Thermal image acquisition unit 412... Weather Information Acquisition Department 413... Heat Balance Analysis Department 414...Plate thickness estimation section 415…Diagnostic Department 416...Output section 417...Measurement time zone determination unit 420...Storage section, 430...Operation section, 440...Display section, 450...Communication section 1...Water supply facilities, 2...Steel plate, 3...Waterway, 4...Cell image

Claims

1. A thermal image acquisition means that acquires a thermal image showing the thermal distribution of a steel plate from an infrared camera that measures the thermal distribution of the steel plate, and acquires time-series data of the thermal image on a certain measurement day, A weather information acquisition means for acquiring time-series data of weather information on the measurement date at a location related to the installation location of the steel plate, A heat balance analysis means calculates time-series data of net radiation and sensible heat transport in the target area for thickness estimation of the steel plate, based on the time-series data of the thermal image acquired by the thermal image acquisition means, the time-series data of the meteorological information acquired by the meteorological information acquisition means, and a predetermined heat balance analysis model. The system includes a plate thickness estimation means that calculates an estimated value of the plate thickness of the target region based on time-series data of the net radiation and sensible heat transport of the target region calculated by the heat balance analysis means, The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. The plate thickness estimation means has a trained model that has been subjected to machine learning using training data, so as to output an estimated value of the plate thickness of the target region when time-series data of the net radiation and sensible heat transport of the target region are input. Plate thickness estimation device.

2. The heat balance analysis means is Based on the time-series data of the weather information acquired by the weather information acquisition means and the heat balance analysis model, it is possible to calculate time-series data of the theoretical surface temperature of the target region. Using the time-series data of the surface temperature of the target region based on the thermal image, and using the theoretical time-series data of the surface temperature as necessary, time-series data of the net radiation and sensible heat transport of the target region are calculated. The plate thickness estimation device according to claim 1.

3. The plate thickness estimation means calculates an estimated value of the plate thickness of the target area based on time-series data of the net radiation and sensible heat transport in the target area during the time period in which the temperature rises on the measurement day. The plate thickness estimation device according to claim 1 or 2.

4. The heat balance analysis means divides the thermal image into a plurality of cell images, each corresponding to the target region, and calculates time-series data of net radiation and sensible heat transport for each target region based on the time-series data of the thermal image for each target region, the time-series data of the meteorological information, and the heat balance analysis model. The plate thickness estimation means calculates an estimated value of the plate thickness for each target region based on time-series data of net radiation and sensible heat transport for each target region. The plate thickness estimation device according to claim 1 or 2.

5. The system further includes a diagnostic means for diagnosing the corrosion status of the target area based on the estimated plate thickness of the target area calculated by the plate thickness estimation means. The plate thickness estimation device according to claim 1 or 2.

6. The weather information acquisition means acquires time-series data of prior weather information, which is weather information for the location on a day prior to the measurement date. The system further comprises a measurement time period determination means that determines a measurement time period on the measurement day that is suitable for estimating the plate thickness of the target area, based on the time-series data of the prior weather information acquired by the weather information acquisition means. The plate thickness estimation device according to claim 1 or 2.

7. The measurement time period determination means is Based on the time-series data of the prior weather information acquired by the weather information acquisition means and the heat balance analysis model, the theoretical time-series data of the surface temperature of the target area is calculated. The time series data of the calculated theoretical surface temperature is used to determine the measurement time period. The plate thickness estimation device according to claim 6.

8. A thermal image acquisition step involves obtaining a thermal image showing the thermal distribution of a steel plate from an infrared camera used to measure the thermal distribution of the steel plate, and obtaining time-series data of the thermal image on a given measurement day. A weather information acquisition step involves acquiring time-series data of weather information on the measurement date at a location related to the installation location of the steel plate, A heat balance analysis step calculates time-series data of the net radiation and sensible heat transport in the target area for thickness estimation of the steel plate, based on the time-series data of the thermal image acquired in the thermal image acquisition step, the time-series data of the meteorological information acquired in the meteorological information acquisition step, and a predetermined heat balance analysis model. The system includes a plate thickness estimation step which calculates an estimated value of the plate thickness of the target region based on time-series data of the net radiation and sensible heat transport of the target region calculated in the heat balance analysis step, The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. In the plate thickness estimation step, when time-series data of the net radiation and sensible heat transport of the target area are input, a trained model that has undergone machine learning using training data is used to calculate the estimated plate thickness of the target area, thereby outputting an estimated value of the plate thickness of the target area. Plate thickness estimation method.

9. Computers, A thermal image acquisition means that obtains a thermal image showing the thermal distribution of a steel plate from an infrared camera that measures the thermal distribution of the steel plate, and obtains time-series data of the thermal image on a given measurement day. A weather information acquisition means for acquiring time-series data of weather information on the measurement date at a location related to the installation location of the steel plate, A thermal image acquisition means obtains time-series data of thermal images, a thermal image acquisition means obtains time-series data of meteorological information, and a thermal balance analysis means calculates time-series data of net radiation and sensible heat transport in the target area for thickness estimation of the steel plate based on a predetermined thermal balance analysis model. Based on the time-series data of net radiation and sensible heat transport in the target region calculated by the heat balance analysis means, it functions as a plate thickness estimation means that calculates an estimated value of the plate thickness of the target region. The estimated thickness of the target area is expressed in a predetermined number of stages according to the thickness. The plate thickness estimation means has a trained model that has been subjected to machine learning using training data, so as to output an estimated value of the plate thickness of the target region when time-series data of the net radiation and sensible heat transport of the target region are input. program.

Images