JP2022129145A - Damage degree prediction device, learned model, method for generating learned model, and damage degree prediction method - Google Patents

Abstract

To predict a structural analysis result without using a previously recorded database.SOLUTION: A damage degree prediction device includes: input means for receiving input of numerical data of a deterioration degree or soundness of a girder structure including a plurality of digits to be a target, an identification number indicating the position of the digit in the plurality of digits, and an assumed external force applied to the girder structure to be a target, which are input data to prediction means using a learned model in which, for each of the digits in the girder structure having a plurality of digits, numerical data indicating a degree of deterioration or soundness of the digits, an identification number indicating the position of the digits in the plurality of digits, and an external force applied to the girder structure are used as input data, and teacher data in which information indicating the damage state of the girder structure obtained from the result of a structural analysis when the external force is applied to the girder structure is used as output data are given and machine learning is performed; and output means for outputting information indicating a damaged state of each digit predicted when an assumed external force is applied to the girder structure to be a target, which is obtained from the learned model.SELECTED DRAWING: Figure 1

Description

There is an application for exception to loss of novelty

The present invention relates to a technique for predicting damage when an external force is applied to the girder structure of a pier or bridge. In this specification, a bridge means an aerial structure for vehicles defined in Article 2 of the Road Traffic Law to pass over water or land.

Techniques for construction or maintenance of structures are known. In general, in order to predict or evaluate damage when external force is applied to the girder structure of a pier, the detailed periodic inspection diagnosis described in Non-Patent Document 1 is performed, and structural analysis is performed based on the results. will be However, since it takes time to obtain a residual strength evaluation using the detailed periodic inspection results, there is a high possibility that the structure subject to evaluation will be subjected to a large external force during the evaluation. Once you receive it, the evaluation result itself loses its value. In addition, due to time and cost considerations, the detailed inspection itself may not be carried out, or even if the detailed periodic inspection is carried out, the remaining strength of the target structure may not be evaluated. For this reason, the repair must be limited to the premise of restoring the original shape, and it is not a rational repair. From the above, since detailed periodic inspection diagnosis and structural analysis based on the results are costly and time-consuming, a relatively simple method for evaluating residual strength from the deterioration degree judgment results obtained from general inspection diagnosis has been proposed. (Non-Patent Document 2).

In addition, the present inventors previously disclosed in Patent Document 1 that a structure analysis result of damage to a beam structure of a pier or bridge when an external force acts is stored as a pier or bridge damage evaluation support device, evaluation method, and program. We proposed a device that acquires and outputs the structural analysis results corresponding to the degree of deterioration of the superstructure of the pier calculated from the deterioration degree data of each beam of the actual pier.

Japanese Patent Application Laid-Open No. 2020-153111

Ministry of Land, Infrastructure, Transport and Tourism, Port and Harbor Bureau, "Maintenance and Management Technical Manual for Port Facilities (Revised Edition)", Coastal Technology Research Center, July 2018, p. 296, p. 317 Kunihiko Uno and Mitsuyasu Iwanami, "Application of residual strength evaluation method to actual pier using deterioration degree determination results", JSCE Proceedings B3 (Marine Development), Vol. 74, No. 2, pp. I_55-I_60, 2018.

However, in the technique described in Patent Document 1, in order to predict damage to a pier or bridge while considering the influence of the degree of deterioration, it is necessary to record the structural analysis results of the target pier or bridge in advance in a database. was there.

In contrast, the present invention can determine the degree of damage caused by an external force to an arbitrary girder of a structure to be evaluated without using a database in which structural analysis results are recorded in advance. It provides technology for spatially and quantitatively predicting.

The present invention provides, for each digit in a digit structure having a plurality of digits, numerical data indicating the degree of deterioration or soundness of the digit, an identification number indicating the position of the digit in the plurality of digits, and the digit structure. Machine learning is performed by giving teacher data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. Numerical data on the degree of deterioration or soundness of each of the plurality of girders obtained from the inspection results of the girder structure including the target plurality of girders, which is the input data for the prediction means using the finished model in the prediction means , an identification number indicating the position of the digit in the plurality of girders, an input means for receiving input of an assumed external force applied to the target girder structure, and the target girder structure obtained from the learned model On the other hand, there is provided a damage prediction apparatus having output means for outputting information indicating the damage state of each digit predicted when the above-mentioned unexpected external force is applied.

The teacher data includes, as the input data, numerical data indicating the degree of deterioration or soundness of other digits around the digit and the distance between the digit and the other digit, Input of numerical data indicating the degree of deterioration or soundness of other girders around the girder included in the girder structure and the distance between the girder and the other girder may be accepted.

Any one of Chebyshev distance, Manhattan distance, and Euclidean distance may be used as the distance measurement method.

wherein the teacher data includes, as the input data, a dimensionless distance as information indicating the position of each of the plurality of digits, and the input means specifies the position of each of the plurality of digits included in the target digit structure. An input of a dimensionless distance to indicate may be accepted.

The target girder structure and the girder structure in the training data are the girder structures of a pier, the training data includes, as the input data, the number of rows of girders orthogonal to the normal line of the pier, and the input Means may receive an input of the number of rows of girders in the target girder structure that are orthogonal to the pier normal.

The target girder structure and the girder structure in the teacher data are the girder structures of a pier, the teacher data includes characteristic values of piles of the pier as the input data, and The input of characteristic values of the piles of the pier may be received in the girder structure.

The teacher data includes, as the input data, information indicating whether or not the girder is located at the end of the girder structure, and the input means controls the girder structure to be located at the end of the girder structure in the target girder structure. You may receive the input of the information which shows whether it is located in.

The training data includes, as the input data, information indicating whether the girder is single-jointed or double-jointed at the end of the girder structure, and the input means controls the girder to be single-sided at the end of the girder structure. An input of information indicating whether it is a joint or a double joint may be received.

The teacher data may include information indicating the length of the digit as the input data, and the input means may receive input of information indicating the length of the digit in the target digit structure.

The girder structure and the target girder structure in the training data are a girder structure of a pier, the training data includes the projection length of the pile of the pier as the input data, and An input of information indicating the projection length of the pile in the girder structure may be received.

The target girder structure is a girder structure of a bridge with an n-layer structure, and the teacher data is, as the input data, information indicating what layer the girder structure is and the distance from other adjacent layers. and the input means may receive input of information indicating what layer the girder structure is in the target girder structure and information indicating a distance to another adjacent layer.

The forecasting means may estimate the remaining strength at any future date by using a probabilistic model of future forecasts for the target girder structure.

Further, the present invention provides numerical data indicating the degree of deterioration or soundness of each girder in a girder structure having a plurality of girder, an identification number indicating the position of the girder in the plurality of girder, and the girder structure. An input layer to which the external force applied to the girder structure is input, an output layer to output information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure, and For each girder, the numerical data indicating the degree of deterioration or soundness of the girder, the identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure are input data, and the girder structure is subjected to the above and an intermediate layer whose parameters are learned using teacher data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when an external force is applied, and a plurality of target girders. Numerical data on the degree of deterioration or soundness of each of the multiple girders obtained from the inspection results of the girder structure, the identification number indicating the position of the girder in the multiple girder, and the target girder structure given Receiving the input of the assumed external force, numerical data of the degree of deterioration or soundness of each of the multiple girders obtained from the inspection results of the girder structure including the target multiple girders, and the position of the girder in the multiple girder. The identification number to be indicated and the data indicating the assumed external force given to the girder structure of interest are input to the input layer, and the intermediate layer performs calculations, and the girder structure when the assumed external force is applied to the girder structure. and providing a trained model for causing a computer to output information indicative of the damage state of the body from the output layer.

Furthermore, the present invention is a model generation method for generating a trained model to be applied to a damage degree prediction device that predicts the degree of damage to a pier or bridge, wherein each girder in a girder structure having a plurality of girders , numerical data indicating the degree of deterioration or soundness of the girder, an identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure are input data, and the external force is applied to the girder structure. a step of acquiring training data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the girder structure is damaged; Input the numerical data indicating the degree of deterioration or soundness, the identification number indicating the position of the girder among the girder, and the external force applied to the girder structure, and structural analysis when the external force is applied to the girder structure. and generating a trained model that outputs information indicating the damage state of the girder structure obtained from the results.

Further, the present invention provides, for each digit in a digit structure having a plurality of digits, numerical data indicating the degree of deterioration or soundness of the digit, an identification number indicating the position of the digit in the plurality of digits, and the digit structure The input data is the external force applied to the girder structure, and the output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. In the prediction means using the trained model, the degree of deterioration or soundness of each of the multiple girders obtained from the inspection results of the girder structure including the target multiple girders, which is the input data for the prediction means a step of receiving input of numerical data, an identification number indicating the position of the digit in the plurality of digits, and an assumed external force applied to the target girder structure; and the target girder structure obtained from the learned model. and outputting information indicating the damage state of each girder predicted when the assumed external force is applied to the girder structure.

ADVANTAGE OF THE INVENTION According to this invention, the prediction of the damage condition when an external force is given to the structure used as object can be implemented in a shorter time compared with the case where structural analysis is performed each time.

The figure which illustrates the functional structure of the damage degree prediction apparatus 1 which concerns on one Embodiment. 2 is a diagram illustrating a functional configuration of a learning device 2; FIG. FIG. 2 is a diagram exemplifying the hardware configuration of the damage degree prediction device 1; The figure which illustrates the structure of the machine-learning model 9 which concerns on one Embodiment. 4 is a flowchart illustrating processing for preparing teacher data; The figure which illustrates the criterion of the pier deterioration degree. The figure which illustrates the criterion of bridge soundness. The figure explaining the handling of girder structure. FIG. 9 is a diagram illustrating inspection results of the degree of deterioration in the girder structure of FIG. 8 ; The figure which illustrates the information which shows deterioration degree distribution of a pier. The figure which illustrates the information which shows the planar shape of a girder structure. The figure which illustrates the external force conditions in a pier. The figure which illustrates the result of structural analysis. 4 is a flowchart illustrating learning processing of a machine learning model 9; 4 is a sequence chart illustrating damage degree prediction processing; The figure which illustrates the information which shows a predicted damage state. The figure which illustrates the dimensionless distance which shows the position of a girder. The figure which illustrates the identification information given to each girder in the multi-layered girder structure. A diagram illustrating an irregular girder structure. The figure which illustrates the girder structure to which the pile head of the dummy was added. The figure which illustrates the girder structure to which the dummy girder was added.

1. Configuration FIG. 1 is a diagram illustrating the functional configuration of a damage degree prediction device 1 according to one embodiment. The damage prediction device 1 is a system that predicts damage when an external force is applied to a target structure. The target structure is the girder structure of a pier or bridge. Here, "girder" means a rod-shaped structure that spans between two piles (or columns) in the superstructure of a pier, or one pile (or column) at either end of the structure. A structure supported by a beam, sometimes called a beam. The damage prediction device 1 predicts damage to a target structure using machine learning without structural analysis of the structure.

The damage degree prediction device 1 is a client device or user terminal that receives an instruction or data input from a user and outputs information according to the input instruction or data. The damage degree prediction device 1 has storage means 11 , input means 12 , prediction means 13 and output means 14 . Storage means 11 stores various data and programs. In this example, the data stored in the storage means 11 include the trained model M. FIG. The learned model M is a machine learning model that is machine-learned by giving teacher data. The trained model M is, for example, a model trained and constructed using neural networks, random forests, decision trees, support vector machines, or deep learning. This training data consists of (1) numerical data indicating the degree of deterioration or soundness of each girder in the girder structure of a pier or bridge, and (2) identification indicating the position of the girder in the girder structure as input data (or explanatory variables). information (or identification number), and (3) at least the external force applied to the beam structure. Further, this teaching data includes at least information indicating the damage state of each girder of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure as output data (or objective variable). This training data is data relating to at least one pier or bridge. A trained model M is generated by the learning device 2 .

FIG. 2 is a diagram illustrating the functional configuration of the learning device 2. As shown in FIG. The learning device 2 gives teacher data to the model to perform machine learning, and obtains a trained model M. FIG. The learning device 2 has storage means 21 and learning means 22 . The storage means 21 stores teacher data and learned models (or pre-learned models). The learning means 22 gives teacher data to the model and performs machine learning to obtain a learned model. The learning device 2 outputs the learned model M to the damage degree prediction device 1 .

Refer to FIG. 1 again. The input unit 12 receives input of explanatory variables to be input to the trained model M. FIG. The prediction means 13 gives the explanatory variables input through the input means 12 to the learned model M, and obtains the corresponding objective variables. The explanatory variables input to the learned model M are (a) numerical data indicating the deterioration degree distribution or soundness judgment distribution in the girder structure obtained from the inspection result of the target structure, and identification information indicating the position of the girder. (or deterioration distribution or soundness judgment distribution map), (b) planar shape and dimensions of the girder structure in the target structure, and (c) assumed external force applied to the target structure. The objective variable (that is, prediction result) obtained from the learned model M for these explanatory variables includes at least information indicating the damage state of the target structure. The output means 14 outputs the information indicating the damage state of the structure obtained by the prediction means 13 .

FIG. 3 is a diagram illustrating the hardware configuration of the damage degree prediction device 1. As illustrated in FIG. The damage degree prediction device 1 is a computer device, such as a personal computer, having a CPU (Central Processing Unit) 101 , a memory 102 , a storage 103 , an input device 104 and a display device 105 . The CPU 101 is a control device that executes programs to perform various calculations and controls other hardware elements in the damage degree prediction device 1 . A memory 102 is a main storage device that functions as a work area when the CPU 101 executes a program, and includes, for example, a RAM. The storage 103 is a non-volatile storage device that records various programs and data, and includes, for example, SSD (Solid State Drive) or HDD (Hard Disk Drive). The input device 104 is a device for the user to input instructions or data to the damage degree prediction device 1, and includes, for example, at least one of a touch screen, a keyboard, and a microphone. The display device 105 is a device that presents information to the user, and includes, for example, an OLED (Organic Light Emitting Diode) display or an LCD (Liquid Crystal Display).

In this example, the program stored in the storage 103 stores a program (hereinafter referred to as “client program”) for causing the computer device to function as the damage degree prediction device 1 . The function of the damage degree prediction device 1 is implemented in the computer device by the CPU 101 executing the client program. At least one of the memory 102 and the storage 103 is an example of the storage unit 11 while the CPU 101 is executing the client program. The CPU 101 is an example of the prediction means 13 . The input device 104 is an example of the input means 12 . The display device 105 is an example of the output means 14 .

Although illustration of the hardware configuration of the learning device 2 is omitted, it is a computer device having the same hardware configuration as the damage degree prediction device 1 . The learning device 2 may be implemented in the same computer device as the damage degree prediction device 1, or may be implemented in a computer device different from the damage degree prediction device 1, such as a server device on a network.

2. Operation Prior to explaining the operation of the damage degree prediction device 1, an outline of the machine learning model used in this embodiment will be explained.

FIG. 4 is a diagram illustrating the structure of the machine learning model 9 according to one embodiment, illustrating a neural network. In this example, the machine learning model has an input layer 91 , an intermediate layer 92 and an output layer 93 . In this example, the input layer 91 includes (1) numerical data indicating the degree of deterioration or soundness of each girder in the girder structure of a pier or bridge, (2) identification information indicating the position of the girder in the girder structure (or identification number), (3) external force applied to the girder structure, (4) numerical data indicating the degree of deterioration or soundness of other girder surrounding the girder, and (5) distance between the girder and other girder is entered. The output layer 93 outputs information indicating the damage state of each girder of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. The information indicating the damage state includes information on at least one of the extent of damage, degree of damage, and rate of damage. More specifically, the information indicating the damage state includes, for example, a damage degree distribution map. The intermediate layer 92 is located between the input layer 91 and the output layer 93 and includes parameters for obtaining data to be output via the output layer 93 from data input via the input layer 91 . In the intermediate layer 92, (1) numerical data indicating the degree of deterioration or soundness of each girder in the girder structure of the pier or bridge, (2) identification information (or identification number) indicating the position of the girder in the girder structure, ( 3) external forces applied to the girder structure, (4) numerical data indicating the degree of deterioration or soundness of other girders surrounding the girder, and (5) numerical data digits indicating the distance between the girder and other girders. is input data, and the parameters are learned using teacher data whose output data is information indicating the damage state of each girder of the girder structure obtained from the structural analysis results when the external force is applied to the girder structure. be. In this example, the "damage state" includes at least one of damage degree, damage distribution, and damage area ratio.

In addition, as a method of estimating the damage state of the girder structure by machine learning, it is conceivable to give the deterioration degree of the girder structure as input data, not numerical data but a deterioration distribution map, that is, image data. Specifically, the machine learning model uses as input data a map of the degree of deterioration distribution or soundness judgment distribution in the girder structure of a pier or bridge, the planar shape and dimensions of the girder structure, and the external force given to the girder structure, It is conceivable to provide teacher data whose output data is information indicating the damage state of each girder of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure, and to perform machine learning. However, according to the research of the inventors of the present application, according to such a method, when predicting a pier with an unknown shape (that is, a shape that has not been learned), for example, the pier shape learned in the past There are times when the pier shape is misrecognized, such as outputting the one with the closest shape. This embodiment addresses this problem.

The operations of the damage degree prediction device 1 and the learning device 2 will be described below. The operations of the damage degree prediction device 1 and the learning device 2 are generally divided into three stages: preparation of teacher data, generation of a trained model M (that is, learning), and prediction using the trained model M. Each process of preparation of teacher data, learning, and prediction will be described below. In the following, functional elements such as the learning means 22 may be described as the subject of processing, but this means that hardware elements such as the CPU 101 execute processing in cooperation with other hardware elements according to a program. means. Moreover, below, the pier is demonstrated as an example as a target structure.

2-1. Preparation of Teacher Data FIG. 5 is a flowchart illustrating a process of preparing teacher data. Here, structural analysis is performed by applying a plurality of analysis conditions (for example, deterioration degree distribution and external force conditions) to each of piers or bridges of various sizes. Known structural analysis software (for example, "Engineer's Studio" manufactured by FORUM8 Co., Ltd.) is used for the structural analysis. Damage distribution, degree of damage, and damage area ratio, which is the area ratio of damage to the total area of the pier or bridge superstructure, are obtained as the damage state as a result of the structural analysis. These data are used as teacher data.

In step S101, (learning means 22 of) the learning device 2 acquires information indicating the deterioration degree distribution or soundness determination distribution of the pier or bridge. Here, the data format of the information indicating the degree of deterioration or soundness may be of any type. In one example, the information indicating the degree of deterioration or soundness includes information specifying a girder in the target girder structure (for example, information indicating the identifier of the girder or information indicating the relative position of the girder) and the degree of deterioration or soundness of the girder. It is a combination with numerical data indicating In one example, the information on the deterioration degree distribution or the soundness determination distribution is character string data indicating this combination. The degree of deterioration or soundness is information indicating the degree of deterioration or soundness of the girder. This is information obtained by judgment.

FIG. 6 is a diagram illustrating criteria for judging the degree of pier deterioration (quoted from Non-Patent Document 1). In this example, datums are provided for each of the side and bottom surfaces of the girder. Four categories of a, b, c, and d are defined as the degree of deterioration for both the side surface portion and the bottom surface portion. The degree of deterioration a is the highest degree of deterioration (or the highest degree of deterioration), and the degree of deterioration d is the lightest degree of deterioration or no deterioration (or the lowest degree of deterioration). The girder structure is divided into a plurality of girder, and the degree of deterioration is determined for each girder. The inspector observes the lower surface and side surface of the target girder in the girder structure and compares it with the judgment criteria to determine the degree of deterioration of the girder. In this example, the inspector determines the degree of deterioration of the lower surface portion and the degree of deterioration of the side portions of the girder, respectively, and adopts the degree of deterioration that is greater as the degree of deterioration of the girder. For example, if the deterioration degree of the lower surface is c, and the deterioration degree of one of the two side surfaces is b and the deterioration degree of the other is d, the deterioration degree of the girder is b.

If the target structure is not a pier but a bridge (for example, a road bridge), the index of soundness is used instead of the degree of deterioration.

FIG. 7 is a diagram illustrating criteria for judging the soundness of a bridge structure (cited from Road Bridge Periodic Inspection Guidelines “Ministry of Land, Infrastructure, Transport and Tourism Road Bureau 2019.2”). In this example, four categories of highway bridge health are defined: categories I, II, III, and IV. Category I is the most healthy (or least deteriorated) state, and Category IV is the most severely deteriorated state. The inspector observes the target girder in the girder structure and compares it with the judgment criteria to determine the soundness of the girder. Note that the criteria for determining the degree of deterioration and soundness illustrated in FIGS. 6 and 7 are merely examples, and criteria other than these may be used. The number of categories of deterioration and soundness is not limited to four, and may be five or more or three or less.

FIG. 8 is a diagram explaining how to handle the girder structure. FIG. 8A is a diagram illustrating a girder structure. In this figure, gray rectangles represent girders (or beams), black squares represent stakes, and white rectangles represent floorboards. In the actual girder structure, the girders, piles, and floor plates have different sizes, but here, for the purpose of facilitating handling as data or facilitating illustration, these are all squares of the same size. Standardize. FIG. 8B shows a standardized girder structure. In this girder structure, the white square at the upper left of the figure is set as the origin, the x-axis is set to the right, and the y-axis is set to the bottom. Hereinafter, the x direction and y direction in other drawings will be described in the same way as in this description, and for example, the digit positioned third from the left and second from the top will be represented as digit [3, 2].

FIG. 9 is a diagram illustrating an inspection result of the degree of deterioration in the girder structure of FIG. FIG. 9(A) is a diagram in which deterioration degree inspection results are mapped to the standardized girder structure of FIG. 8(B). In this example, the degradability of digit [3,2] is c. Here, a plurality of digits are observed one by one, and the distance from the digit of interest and the degree of deterioration thereof are tallied for the digits around the digit of interest (hereinafter referred to as "the digit of interest"). FIG. 9(B) is a diagram showing the total result of the surrounding digits for the digit [3, 2]. In this example, there are four digits at a distance of 1 starting from the digit [3, 2], and the degrees of deterioration are d, b, c, and c.

In this example, the Chebyshev distance in the girder structure normalized as the girder-to-girder distance is used. The Chebyshev distance is one of the concepts of distance, along with the Euclidean distance and the Manhattan distance. The Chebyshev distance corresponds to the difference in coordinate values between the vertical and horizontal axes in a two-dimensional space, whichever has the larger difference. For example, the Chebyshev distance between digits [3,2] and digits [2,1] is one, and the Chebyshev distance between digits [3,2] and digits [6,2] is three. In this embodiment, it is thought that it is important to adopt the concept of distance that emphasizes the number of connections between girders, considering that the girders are connected to each other and affect each other. Below, the Chebyshev distance is used. However, the distance between digits is not limited to the Chebyshev distance, and the above-mentioned Euclidean distance, Manhattan distance, or distance according to another definition may be used.

FIG. 10 is a diagram exemplifying information indicating the deterioration degree distribution of the pier, which is acquired in step S101. This figure shows, for a single digit of interest (digit [3, 2]), the degree of deterioration of the digit of interest itself and the degree of deterioration of other surrounding digits. The degree of deterioration is classified according to the distance from the target digit. The zero distance column indicates the degree of deterioration of the target digit itself. Columns after distance 1 indicate the degree of deterioration of other surrounding digits, and the ratio of digits corresponding to each level of deterioration is indicated. For example, there are four digits at a distance of 1 from digit [3, 2] (see FIG. 9(b)), and their degrees of deterioration are d, b, c, and c, respectively. Aggregating these, the deterioration degree d is 25%, the deterioration degree c is 50%, the deterioration degree b is 25%, and the deterioration degree a is 0%.

Although FIG. 10 shows the degree of deterioration for a single digit of interest, information such as that shown in FIG. 10 is prepared for all digits that make up the target digit structure and is acquired in step S101. The degree of deterioration of the target girder, the degree of deterioration of the surrounding girder, and the Chebyshev distance of each surrounding girder are acquired as a set. (For example, Chebyshev distances 10, 20, 25 . . . ) Among them, the distance of the beam that gives a higher accuracy rate than the structural analysis result is adopted as the objective variable.

In step S102, the learning means 22 acquires information indicating the planar shape of the target girder structure. Here, the data format of the information indicating the planar shape of the girder structure may be of any type. In one example, the information indicating the planar shape is a combination of information specifying the girder in the girder structure of the target pier and information indicating the size of the girder.

FIG. 11 is a diagram illustrating information indicating the planar shape of the girder structure. Here, FIG. 11A shows identification numbers of digits. Identification numbers are assigned to the digits in the vertical and horizontal directions, respectively. The table of FIG. 11(B) includes information indicating the size of the target girder structure (the number of girders in each of the vertical and horizontal directions) and information indicating the dimensions of each girder. This girder structure consists of 7 girders in the x direction and 5 girders in the y direction. For example, the digit with digit number 101 has a length of 0.95 m in the x direction and a length of 0.50 m in the y direction.

Refer to FIG. 5 again. In step S103, the learning means 22 acquires external force conditions applied to the target girder structure.

FIG. 12 is a diagram illustrating external force conditions on a pier. Incidentally, the external force condition on the bridge may be defined separately (illustration is omitted). Rows of this table indicate external force conditions, and columns indicate explanatory variables. In this example, five types of explanatory variables are used: static (sea←land), static (sea→land), L1, L2-1, and L2-2. "Static (sea <- land)" represents, for example, the tractive force of a given ship. “Static (sea→land)” indicates, for example, the berthing force of a given ship. "L1" indicates a level 1 seismic motion, that is, an earthquake that has a high probability of occurring during the service period. "L2" indicates a level 2 earthquake motion, that is, an earthquake that has a low probability of occurring during the service period but releases a large amount of energy. earthquake, etc.). L2-2 indicates Type II: Level 2 inland seismic ground motion (for example, Hyogo-ken Nanbu Earthquake). In addition, static (sea ← land) and static (sea → land) described as external force conditions are applied not only as a prescribed ship traction force and berthing force, but also as a design load. It is also used to see if there is any damage that should not occur on the pier. The external force conditions L1 and L2 are also used as design loads depending on the importance of the design target pier.

In this table, the value of each cell is either "0" or "1". A value of "0" indicates that the force is not applied and a value of "1" indicates that the force is applied. For example, the external force condition "static (sea <-> land)" indicates that the symmetrical girder structure is subjected to, for example, only the tractive force of a ship, and no other external force is applied. In this example, only an external force condition in which a single type of external force is applied is shown, but external force conditions in which a plurality of types of external force are applied may be used. Specifically, for example, an external force condition in which a static (sea←land) external force and a level 1 seismic motion are applied at the same time may be used.

Refer to FIG. 5 again. In step S105, the learning means 22 assumes that an external force indicated by the external force condition obtained in step S104 is applied to the girder structure having the planar shape obtained in step S103 and the deterioration degree distribution obtained in step S101. Acquire the structural analysis result of the hypothetical case. Structural analysis may be performed by the learning means 22 itself, or may be performed by a device other than the learning device 2 .

FIG. 13 is a diagram illustrating the results of structural analysis. In this example, as a result of the structural analysis, data on damage degree, damage degree distribution, and damage area ratio are obtained. In one example, the structural analysis results are mapped onto a normalized image of the girder structure. In this map, all girders have a common size and may not contain information indicating the actual size of the girder structure, or each girder may be displayed at a scaled size according to its actual size. good.

In this map, each digit's damage is displayed in a form such as color, hatching, or character information that indicates the type of damage. (Since colors cannot be expressed in this figure, they are distinguished by hatching instead.) For example, crack damage is yellow, yield damage is red, and ultimate damage is black. Since these damages are displayed on the map of the girder structure, it can be said that this figure shows the extent and degree of damage.

This map is converted into information containing a set of digit identification information and degree of damage. Alternatively, as a result of structural analysis, in addition to or instead of the map of FIG. 13, information including a set of girder identification information and degree of damage is output.

The learning device 2 repeats the processing of steps S101 to S104 for each of a plurality of piers having various shapes and sizes, and accumulates data. The data thus accumulated is used as teacher data in the learning process described below. In this training data, the numerical data of the deterioration distribution obtained in step S101, the planar shape obtained in step S102, and the external force condition obtained in step S103 are input data, and the structural analysis result obtained in step S104 is output. Data. The learning device 2 stores these teacher data in the storage means 21 .

2-2. Learning FIG. 14 is a flowchart illustrating learning processing of the machine learning model 9 . The flow of FIG. 14 is started, for example, when a learning instruction is given by the administrator or user of the damage degree prediction device 1 . Alternatively, the flow of FIG. 14 may be started when a predetermined amount of teacher data is accumulated.

In step S201, the learning means 22 acquires teacher data used for learning. In this example, the learning means 22 acquires teacher data prepared in an external device different from the damage degree prediction device 1 .

In step S202, the learning means 22 gives teacher data to the machine learning model 9 to perform machine learning. This teacher data includes input data and output data. In one example, the input data are the numerical data of the deterioration degree distribution acquired in step S101, the planar shape acquired in step S102, and the external force condition acquired in step S103.

In step S<b>203 , the learning means 22 stores the machine learning model 9 in which the parameters of the intermediate layer 92 are learned by machine learning as a learned model M in the storage means 21 . The learning device 2 outputs the learned model M to the damage degree prediction device 1 . The damage degree prediction device 1 stores the trained model M in the storage means 11 . Thus, the damage prediction device 1 obtains the learned model M. FIG.

2-3. Prediction FIG. 15 is a sequence chart illustrating a damage degree prediction process. Prediction of the degree of damage is performed in response to a request from the user.

The user instructs the damage degree prediction device 1 to predict the degree of damage. Specifically, the user inputs explanatory variables for the target structure. The input unit 12 receives input of explanatory variables, specifically, the following data (a) to (c) (step S301). (a) Numerical data of the deterioration degree distribution or soundness judgment distribution in the girder structure obtained from the inspection results of the target structure. (b) Planar shape and dimensions of the girder structure in the target structure. (c) Unexpected force applied to the target structure.

In one example, the damage prediction device 1 itself provides a user interface for inputting inspection results of structures. The user inputs the inspection result to the damage degree prediction device 1 . Alternatively, the damage degree prediction device 1 may receive data on the deterioration degree distribution or the soundness determination distribution from a device that records inspection results of a target structure. The data of the deterioration degree distribution or the soundness determination distribution may have any data format as in the case of the teacher data.

The user inputs the data (for example, image data) of the planar shape and dimensions of the target structure to the damage degree prediction device 1 . Alternatively, the data of the plane shape and dimensions of the structure are recorded in a database in an external device such as a server, and the damage degree prediction device 1 retrieves the plane shape and dimensions of the target structure from this database in accordance with the user's instruction. Shape data may be acquired.

For the unexpected force, for example, the same conditions as those used in the teacher data are used. Alternatively, the damage degree prediction device 1 may use only the conditions designated by the user among the external force conditions in the table of FIG. 12 as explanatory variables.

At step S302, the input means 12 generates a request for predicting the degree of damage. This request includes explanatory variables that were input, obtained, or specified in step S301. In step S303, the input means 12 requests the prediction means 13 to predict the degree of damage.

In this example, it can be said that the input means 12 substantially inputs the following three types of explanatory variables to the learned model M. (a) Numerical data of the deterioration distribution or soundness judgment distribution of each girder in the girder structure obtained from the inspection results of the target structure, (b) Planar shape and dimensions of the girder structure in the target structure, and (c) unexpected forces applied to the structure of interest.

In step S304, the prediction means 13 receives a request to predict the degree of damage. In step S305, the prediction means 13 inputs the girder deterioration map, the planar shape and dimensions of the girder structure, and the assumed external force into the learned model M as explanatory variables. The learned model M outputs an objective variable, that is, information indicating the damage state according to the input explanatory variables. In one example, the information indicating the damage state is data in the same format as the damage distribution map used as teacher data illustrated in FIG. 13 . In step S306, the prediction means 13 acquires the objective variable output from the trained model M. The prediction unit 13 transmits the acquired objective variable to the device that sent the request (step S307). In step S308, the output unit 14 outputs information indicating a predicted damage state according to the objective variable output from the learned model M. The output means 14 may output the information output from the trained model M as it is to the user, or may output the information output from the trained model M after some processing.

FIG. 16 is a diagram exemplifying information indicating a predicted damage state output from the output means 14. As shown in FIG. In this example, the degree of damage (that is, degree of damage) of each digit is divided into a plurality of divisions (that is, damage divisions), and each division is visually represented (for example, with different colors). The divisions used here are common to those illustrated in FIG. 13, eg, crack damage is yellow, yield damage is red, and ultimate damage is black. Moreover, the deterioration degree displayed in FIG. 6 and the soundness displayed in FIG. 7 may be displayed in different colors.

Further, the information output from the output means 14 includes the rate of damage (damage area ratio). Here, the percentage of damage means the ratio of the area of girders having a predetermined damage among the girders of the target girder structure. For example, when the total area of girders in the target girder structure is 5,000 m2, and the total area of girders with crack damage at least in part is 3,000 m2, the crack area ratio is 60%. . Alternatively, the damage ratio may be indicated by the ratio of the number of damaged digits instead of the area ratio.

Note that the format of the information indicating the predicted damage state is not limited to the example of FIG. 16 . The information indicating the damage state should include at least one or more information of the extent of damage, the degree of damage, and the rate of damage. For example, the output means 14 may express the presence or absence of damage in two values (that is, damaged/not damaged) without distinguishing between damage categories. Alternatively, the output means 14 may output only the percentage of damage without outputting the damage state in the form of a map.

Thus, according to the damage degree prediction device 1, information indicating the damage state can be obtained without performing structural analysis of the target structure each time.

3. Modifications The present invention is not limited to the above-described embodiments, and various modifications are possible. Some modifications will be described below. Two or more of the items described in the following modified examples may be used in combination.

ここで、Ｂは杭幅［ｍ］、ＫＣＨは横方向地盤反力係数［ｋＮｍ^－３］、ＥＩは杭の曲げ剛性［ｋＮｍ^２］である。

ここで、Ｄは杭の直径［ｍ］、ｋＨは水平方向地盤反力係数［ｋＮｍ^－３］、ＥＩは杭の曲げ剛性［ｋＮｍ^２］である。
（ウ）その桁が端部に位置するという情報。
（エ）端部の桁の接合（片接合か両接合か）という情報。
（オ）桁の位置を示す無次元距離。
（カ）杭の突出長又は杭の長さ。 3-1. Explanatory Variables Explanatory variables used in teacher data in learning and input data in prediction are not limited to those exemplified in the embodiments. The explanatory variable may include, for example, at least one of the following (a) to (f).
(a) The number of rows of girders perpendicular to the pier normal.
(b) Pile characteristic value β [m−1].
For example, the characteristic value β is quoted from the following formula (1) (“(Public Corporation) Japan Port and Harbor Association: Technical Standards and Commentary on Port Facilities (second volume), pp.704-705, 2018” ), or (2) (quoted from “(Public Corporation) Japan Road Association: Road Bridge Specifications and Commentary IV Substructure Edition, pp.259-260, 2017”).

Here, B is the pile width [m], KCH is the lateral ground reaction force coefficient [kNm ⁻³ ], and EI is the bending rigidity of the pile [kNm ² ].

Here, D is the pile diameter [m], kH is the horizontal ground reaction force coefficient [kNm ⁻³ ], and EI is the bending rigidity of the pile [kNm ² ].
(c) Information that the girder is located at the end.
(d) Information on the joint of the girder at the end (one joint or both joints).
(e) A dimensionless distance indicating the position of the girder.
(f) Pile projection length or pile length.

Regarding (a), even if the total number of girders is the same, it is thought that the magnitude of the effect of external force will change depending on the number of rows of girders perpendicular to the normal line of the pier, so it is meaningless to use this as an explanatory variable. be. Regarding (a), it is significant to use this as an explanatory variable because it is thought that the magnitude of the influence received from the external force changes according to the characteristic value of the pile (for example, 1/β). Regarding (c), the girder located at the end of the girder structure may behave differently against external force than the girder located at the end, so explain whether or not the girder is located at the end. Adding variables may improve prediction accuracy. Regarding (d), there are two ways to join the girders located at the ends: single joints and double joints. Prediction accuracy may be improved by adding information as to whether the joint is single-sided or double-sided when positioned at the edge. A double joint means a joint structure in which both ends of the girder are supported by piles, and a single joint means a joint structure in which only one end of the girder is supported by piles.

Regarding (e), FIG. 17 is a diagram illustrating dimensionless distances indicating positions of girders. This figure shows the distance in the x-direction. The distance in this example indicates a numerical value obtained by dividing the distance from the reference position (upper left corner) by the maximum value (13 in this example) in the standardized digit structure, multiplying the result by 10, and rounding up the decimal part. By using dimensionless distances in this way, it is possible to use uniform distances for girder structures of various sizes. It should be noted that the specific method of calculating the dimensionless distance is not limited to this, and whether or not it is an integer and how many digits the numerical value is (two digits in the example of the figure) are arbitrary.

Regarding (f), the behavior against external force may differ depending on not only the girder structure but also the protrusion length of the pile or the length of the pile. Potentially improved prediction accuracy.

3-2. Length of digit The damage degree prediction device 1 may use the length of the digit as an explanatory variable, but the length used as an explanatory variable may be a real value. A value normalized by a reference value of (for example, the maximum value of the digit length in the digit structure) may be used. Alternatively, the damage prediction device 1 may treat the digit length as a classified class instead of treating it as a numerical value. As an example, when classifying into two classes, digits with a length less than or equal to the average length of all digits included in the digit structure are set to "zero" as the length, and those with a length exceeding the average value. For digits, use "1" as the length. Also, the beams may be classified in the x direction and the y direction respectively. The same is true for piles as well as girders.

3-3. Multi-Layer Structure The target girder structure is a girder structure of a bridge, and a multiple-layer (n-layer) deck structure can also be subject to prediction by the damage prediction device 1 . An example of a multi-level deck structure is a double-deck structure of a road bridge and a railway bridge, or a double-deck structure of two road bridges (eg, an inbound line and a down line). A multi-layer deck structure has a plurality of superstructures (eg, road faces) in the vertical direction. In this example, distances to other girders are taken into account in each layer as in the case of a single layer, but vertical distances are not considered. However, if the correlation of multiple layers is not considered, there is a problem that the girder structure located in the upper layer is more likely to shake during an earthquake, but there is a problem that learning or prediction is performed without distinction. Therefore, in this example, the deck structure of multiple layers is taken into account by including the information of what layer the digit is in the deterioration degree information. In addition, if the overall length of the bridge piers, the length of the piers between each story, or the foundation structure type of the target bridge is a pile foundation, it is also included in the information as the explanatory variable (f) the protrusion length of the pile or the length of the pile. You may do so.

FIG. 18 is a diagram exemplifying identification information given to each girder in a multi-layered girder structure. In the figure, each layer is indicated as "the k-th layer", and this information is also used as an explanatory variable. It is preferable that the multi-layer structure machine learning model is learned using teacher data (of a multi-layer structure bridge) different from that of the single-layer structure machine learning model.

3-4. Irregular Girder Structure FIG. 19 is a diagram illustrating an irregular girder structure. The target girder structure is not limited to a regular grid (square) structure, but may be irregular as shown in this figure. Here, we describe how to deal with such an irregular digit structure. Specifically, a dummy pile head is installed between long girders to divide the girders.

20 is a diagram illustrating a girder structure to which a dummy pile head is added in the example of FIG. 19. FIG. In this example, a dummy pile head divides the long girder into two halves. If it is a regular grid, there should be girders between the dummy pile heads, but the non-existing girders are not shown here. Although this example is not a perfect grid (that is, a square), there is no processing problem because Chebyshev distances can be given to all digits of interest.

When learning or predicting in this state, even if there is originally one digit, it will be divided into two, and there will be two digits, so damage results will be output separately. Since it is originally one girder, the same damage result should be obtained, but there is a possibility that different damage results are obtained for these two girder. In this case, it is preferable from the viewpoint of safety to adopt the result with the highest degree of damage, for example. However, a representative value such as an average value or an intermediate value of these two damage results may be used as the degree of damage to the target girder.

Also, in the example of FIG. 20, a dummy girder may be set in addition to the dummy pile head.

FIG. 21 is a diagram illustrating a girder structure with dummy girder added in the example of FIG. In this example, a dummy girder is added in addition to the dummy pile head, and the girder structure becomes a complete grid (square). The Chebyshev distance can be given to the dummy digits, but since the dummy digits do not exist in reality, the degree of degradation observed in the real digits cannot be given. Therefore, a new degree of deterioration (for example, degree of deterioration Z) is defined for dummy girders and used as an explanatory variable. Corresponding to this, a damage degree Z, which indicates that the damage degree cannot be determined or is unnecessary, is defined and used as an objective variable. If learning is performed using the degree of deterioration and the degree of damage defined in this way, there is no problem even if dummy digits are added. In the result of the degree of damage output from the trained model M, processing such as not displaying the digit of the degree of damage Z may be performed.

3-5. Probabilistic Model The prediction means 13 may predict the future degree of deterioration or soundness using a predetermined probabilistic model. This probabilistic model is a model for estimating future state transitions based on probabilities based on past state transitions, and an example is a Markov chain model. In this example, the prediction means 13 makes a prediction by combining the learned model M and the probabilistic model. More specifically, the prediction means 13 inputs the output from the trained model M to the probability model. The prediction means 13 further inputs the current time and the future time to be predicted into the probability model. These times are input by the user via the input means 12, for example.

3-6. Others When the target girder structure has multiple layers of girder structures, the girder structure to be predicted is not limited to any single layer. Learning and prediction processes may be applied to multiple layers.

The order of processing in teacher data preparation, learning, and prediction is not limited to those exemplified in FIGS. For example, explanatory variables may be input in any order.

The hardware configuration of the damage degree prediction device 1 is not limited to that illustrated in the embodiment. The damage degree prediction device 1 may have any hardware configuration as long as it can implement the required functions. For example, part of the functions of the damage degree prediction device 1, such as the prediction means 13 and the learned model M, may be implemented in an external device such as a server device. In this case, a computer device having input means 12 and output means 14 cooperates with this external device and functions as a whole as the damage degree prediction device 1 according to the embodiment. This server device may be a real server or a virtual server.

Software elements such as programs and trained models may be provided in a state recorded on a computer-readable recording medium such as a CD-ROM (Compact Disc Read only memory), or via a communication network such as the Internet. may be downloaded.

DESCRIPTION OF SYMBOLS 1... Damage degree prediction apparatus 2... Learning apparatus 9... Machine learning model 11... Storage means 12... Input means 13... Prediction means 14... Output means 21... Storage means 22... Learning means 91... Input layer 92 Intermediate layer 93 Output layer 101 CPU 102 Memory 103 Storage 104 Input device 105 Display device

Claims (15)

For each girder in a girder structure having a plurality of girder, input numerical data indicating the degree of deterioration or soundness of the girder, an identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure. Using a trained model that has been machine-learned by giving teacher data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. numerical data of the degree of deterioration or soundness of each of the plurality of girders obtained from the inspection result of the girder structure including the target plurality of girders, which is the input data for the prediction means, an input means for receiving input of an identification number indicating the position of the girder in the girder and an input of an unexpected external force applied to the target girder structure;
and output means for outputting information indicating the damage state of each girder predicted when the assumed external force is applied to the target girder structure obtained from the learned model. The teacher data includes, as the input data, numerical data indicating the degree of deterioration or soundness of other digits around the digit and the distance between the digit and the other digit,
The input means receives input of numerical data indicating the degree of deterioration or soundness of other girders surrounding the girder included in the target girder structure and the distance between the girder and the other girder. Item 2. A damage degree prediction device according to item 1. 3. The damage degree prediction device according to claim 2, wherein any one of Chebyshev distance, Manhattan distance, and Euclidean distance is used as the distance measurement method. wherein the teacher data includes, as the input data, a dimensionless distance as information indicating the position of each of the plurality of digits;
3. The damage degree prediction device according to claim 1, wherein said input means receives input of dimensionless distances indicating respective positions of a plurality of girders included in said target girder structure. The target girder structure and the girder structure in the training data are girder structures of a pier,
wherein the training data includes, as the input data, the number of columns of girders orthogonal to the normal line of the pier;
The damage degree prediction device according to any one of claims 1 to 4, wherein the input means receives an input of the number of columns of girders orthogonal to a normal line of the pier in the target girder structure. The target girder structure and the girder structure in the training data are girder structures of a pier,
wherein the training data includes characteristic values of piles of the pier as the input data;
The damage degree prediction device according to any one of claims 1 to 4, wherein the input means receives input of characteristic values of piles of the pier in the target girder structure. wherein the teacher data includes, as the input data, information indicating whether or not the girder is positioned at the end of the girder structure;
7. The damage degree prediction device according to any one of claims 1 to 6, wherein the input means receives input of information indicating whether or not the girder in the target girder structure is positioned at an end of the girder structure. . The teacher data includes, as the input data, information indicating whether the girder is one-sided joint or two-sided jointed at the end of the girder structure;
8. The damage degree prediction device according to claim 7, wherein the input means receives input of information indicating whether the girder is single jointed or double jointed at the end of the girder structure. wherein the teacher data includes, as the input data, information indicating the length of the digit;
The damage degree prediction device according to any one of claims 1 to 8, wherein the input means receives input of information indicating the length of the girder in the target girder structure. The girder structure and the target girder structure in the training data are girder structures of a pier,
wherein the training data includes, as the input data, the projection length of the piles of the pier;
10. The damage degree prediction device according to any one of claims 1 to 9, wherein the input means receives input of information indicating a protrusion length of the pile in the target girder structure. The target girder structure is a girder structure of an n-layer bridge,
The teacher data includes, as the input data, information indicating what layer the girder structure is and information indicating the distance from other adjacent layers,
5. The input means receives input of information indicating what layer the girder structure is in the target girder structure and information indicating a distance to another adjacent layer. The damage degree prediction device according to . 12. The damage degree prediction device according to any one of claims 1 to 11, wherein the prediction means evaluates the remaining strength on any future day by using a future prediction probability model for the target girder structure. Numerical data indicating the degree of deterioration or soundness of each girder in a girder structure having a plurality of girder, an identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure are input. and an input layer that
an output layer for outputting information indicating a damage state of the girder structure obtained from structural analysis results when the external force is applied to the girder structure;
For each girder in the girder structure, numerical data indicating the degree of deterioration or soundness of the girder, an identification number indicating the position of the girder in the plurality of girder, and an external force applied to the girder structure are input data, an intermediate layer whose parameters are learned using teacher data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure,
Numerical data on the degree of deterioration or soundness of each of the multiple girders obtained from the inspection results of the girder structure including the target multiple girders, the identification number indicating the position of the girders in the multiple girders, and the target Accepts the input of the assumed external force given by the girder structure,
Numerical data on the degree of deterioration or soundness of each of the multiple girders obtained from the inspection results of the girder structure including the target multiple girders, the identification number indicating the position of the girders in the multiple girders, and the target Data indicating the assumed external force given to the girder structure is input to the input layer, the intermediate layer performs calculations, and information indicating the damage state of the girder structure when the assumed external force is applied to the girder structure is obtained. A trained model for operating a computer to output from the output layer. A model generation method for generating a learned model to be applied to a damage prediction device that predicts the degree of damage to a pier or bridge,
For each girder in a girder structure having a plurality of girder, input numerical data indicating the degree of deterioration or soundness of the girder, an identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure. a step of acquiring training data, the output data being information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure;
Using the training data, for each digit in the digit structure, numerical data indicating the degree of deterioration or soundness of the digit, an identification number indicating the position of the digit in the plurality of digits, and given to the digit structure inputting an external force and generating a trained model outputting information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. Method. For each girder in a girder structure having a plurality of girder, input numerical data indicating the degree of deterioration or soundness of the girder, an identification number indicating the position of the girder in the plurality of girder, and the external force applied to the girder structure. Using a trained model that has been machine-learned by giving teacher data whose output data is information indicating the damage state of the girder structure obtained from the structural analysis result when the external force is applied to the girder structure. numerical data of the degree of deterioration or soundness of each of the plurality of girders obtained from the inspection result of the girder structure including the target plurality of girders, which is the input data for the prediction means, receiving input of an identification number indicating the position of the girder in the girder and an assumed external force applied to the target girder structure;
and outputting information indicating the damage state of each girder predicted when the assumed external force is applied to the target girder structure, which is obtained from the trained model. Method.

Images