Pipeline damage prediction system, pipeline damage prediction method, and pipeline damage prediction program

Abstract

This pipeline damage prediction system uses a computer to predict pipeline damage due to a natural disaster, the pipeline damage predication system comprising: a basic data storage unit that stores pipeline attribute data indicating an attribute of a pipeline, stores ground characteristic data which indicates a characteristic of the ground in which the pipeline is buried and which includes at least liquefaction data, ground boundary data, or terrain modification data in units of polygons in a map information system GIS, and stores damage data indicating the degree of damage due to a natural disaster occurring in the past; and a damage prediction calculation unit that generates damage prediction information in which damage occurring in the pipeline is predicted on the basis of the pipeline attribute data, the ground characteristic data, and the damage data stored in the basic data storage unit.

Description

Pipeline damage prediction system, pipeline damage prediction method, and pipeline damage prediction program

[0001] The present invention relates to a pipeline damage prediction system, a pipeline damage prediction method, and a pipeline damage prediction program that use a computer to predict damage to pipelines caused by natural disasters.

[0002] For the aging of pipelines such as water supply pipe networks, it is desirable to formulate an efficient renewal plan. At that time, by predicting the damage to pipelines caused by natural disasters such as earthquakes, landslides, volcanic eruptions, and floods, and preferentially renewing pipelines with large predicted damage, it becomes possible to reconstruct a pipe network that is resistant to natural disasters.

[0003] For example, as a method for predicting damage caused by an earthquake, the damage estimation formula Rm(v) shown in Fig. 8(a) proposed by the Japan Water Works Association (JWWWA), a public interest incorporated foundation, and the damage estimation formula Rm(v) shown in Fig. 8(b) proposed by the Japan Water Resources Center (JWRC), a public interest foundation, have been used.

[0004] The former is obtained by multiplying the standard damage rate R(v) (=3.11×10 ―３ ×(v - 15) １．３０ ) based on the maximum ground surface velocity v by the pipe type correction coefficient Cp, the diameter correction coefficient Cd, the terrain correction coefficient Cg, and the liquefaction correction coefficient Cl. The latter is classified according to the presence or absence of liquefaction. When there is no liquefaction, it is obtained by multiplying the standard damage rate R(v) (=9.92×10 ―３ ×(v - 15) １．１４ ) based on the maximum ground surface velocity v by the pipe type correction coefficient Cp, the diameter correction coefficient Cd, and the terrain correction coefficient Cg. When there is liquefaction, it is obtained by multiplying the liquefaction standard damage rate R Ｌ by the pipe type correction coefficient Cp and the diameter correction coefficient Cd.

[0005] In both cases, each correction coefficient is a value obtained using methods such as multiple regression analysis based on the energy and damage situation of past earthquakes (number of damages per pipe length (damage rate)). In both cases, each correction coefficient was determined by dividing a standard regional mesh set based on latitude and longitude into four sections, each section being approximately 250m on each side, and classifying the main microtopographic data contained in each mesh as a representative microtopographic feature of that mesh, setting a topographic correction coefficient Cg for each section. The Japan Water Works Association (JWWA) classified all 24 types of microtopographic features into four sections, while the Japan Water Research Center (JWRC) classified all 24 types of microtopographic features into five sections.

[0006] In other words, the values ​​calculated using the damage estimation formula Rm(v) based on the terrain correction coefficient Cg set for each mesh were adopted as damage prediction information for pipelines laid in that mesh.

[0007] Japanese Patent Publication No. 2001-329574 Japanese Patent Publication No. 2004-310307

[0008] However, with the conventional pipeline damage prediction information calculated on a mesh-based basis, even if the microtopography of the actually laid pipeline differs, the same value is obtained for multiple pipelines laid in the same mesh. This makes it difficult to properly evaluate which pipeline should be replaced first. For example, pipeline damage is said to be common in soft ground areas such as residential land developed by embankment or land created by modifying old rivers or water bodies (such as reservoirs), but conventional methods have not been able to adequately consider the effects of such soft ground conditions when evaluating each pipeline.

[0009] The objective of the present invention is to provide a pipeline damage prediction system, a pipeline damage prediction method, and a pipeline damage prediction program that can obtain appropriate damage prediction information for each pipeline by taking into account the influence of soft ground, ground boundaries, and other factors.

[0010] To achieve the above objectives, the first characteristic configuration of the pipeline damage prediction system according to the present invention is a pipeline damage prediction system that uses a computer to predict damage to pipelines caused by natural disasters, comprising: pipeline attribute data indicating the attributes of the pipeline; ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of liquefaction data, ground boundary data, and topographic modification data in polygon units in a geographic information system (GIS); a basic data storage unit that stores damage data indicating the degree of damage caused by natural disasters that have occurred in the past; and a damage prediction calculation unit that generates damage prediction information predicting damage to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit.

[0011] The damage prediction calculation unit generates damage prediction information that predicts damage to the pipeline based on pipeline attribute data indicating the attributes of the pipeline, ground characteristic data, and damage data indicating the extent of damage caused by natural disasters such as earthquakes that have occurred in the past. By adopting ground characteristic data that includes at least one of the following as the unit of polygons in a geographic information system (GIS) that demarcates boundaries such as the ground: liquefaction data, ground boundary data, and topographic modification data, it is possible to obtain damage prediction information based on the ground characteristics of the polygons in which the pipeline is actually laid, and as a result it becomes possible to formulate an appropriate pipeline renewal plan.

[0012] The second characteristic configuration is that, in addition to the first characteristic configuration described above, the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the pipe length, using the polygon as the unit.

[0013] Even when pipelines are laid across the boundaries of adjacent polygons, making it difficult to appropriately evaluate the degree of influence of the ground constituting each polygon, by defining the overlapping area when the boundaries of adjacent ground are expanded by a predetermined width based on the length of the pipeline, using the outer length of each polygon as the unit, it becomes possible to appropriately evaluate the degree of influence of the ground on the pipeline included in each ground boundary data.

[0014] The third characteristic feature is that, in addition to the first characteristic feature described above, the topographic modification data includes either former river channels, former water bodies, or residential land development ground.

[0015] As topographic modification data, former river channels, former reservoirs, and residential land development areas that have been reclaimed with earthworks are areas of land with a high risk of being affected by natural disasters. This data will allow for an appropriate evaluation of the degree of impact on pipelines laid on such land.

[0016] The fourth characteristic configuration is that, in addition to the first characteristic configuration described above, the damage prediction calculation unit includes a machine learning unit that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates the damage prediction information based on the decision tree model generated by the machine learning unit.

[0017] Preferably, the damage prediction calculation unit includes a machine learning unit that generates the damage prediction information. The machine learning unit prepares a dataset for each of the multiple pipelines that includes the pipeline attribute data, ground characteristic data, and past damage data, and generates a decision tree model by repeatedly performing ensemble learning. By inputting the pipeline attribute data and ground characteristic data for each pipeline to be evaluated into the decision tree model generated by the machine learning unit, appropriate damage prediction information can be obtained.

[0018] The first characteristic configuration of the pipeline damage prediction method according to the present invention is a pipeline damage prediction method that uses a computer to predict damage to a pipeline caused by a natural disaster, and comprises: a basic data storage step in which pipeline attribute data indicating the attributes of the pipeline, ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, and ground characteristic data including at least one of liquefaction data, ground boundary data, and topographic modification data in the form of polygons in a geographic information system (GIS), and damage data indicating the degree of damage caused by natural disasters that have occurred in the past are stored in a basic data storage unit; and a damage prediction calculation step in which a damage prediction calculation unit generates damage prediction information predicting the damage to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit in the basic data storage step.

[0019] The second characteristic configuration is that, in addition to the first characteristic configuration described above, the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the pipe length, using the polygon as the unit.

[0020] The third characteristic feature is that, in addition to the first characteristic feature described above, the topographic modification data includes either former river channels, former water bodies, or residential land development ground.

[0021] The fourth characteristic configuration is that, in addition to the first characteristic configuration described above, the damage prediction calculation step performs a machine learning step that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates the damage prediction information based on the decision tree model generated by the machine learning step.

[0022] The first characteristic configuration of the pipeline diagram design support program according to the present invention is that the computer functions as a basic data storage processing unit that stores pipeline attribute data indicating the attributes of the pipeline, ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of liquefaction data, ground boundary data, and topographic modification data in the form of polygons in a geographic information system (GIS), and damage data indicating the degree of damage caused by natural disasters that have occurred in the past, in a basic data storage unit, and a damage prediction calculation unit that generates damage prediction information predicting the damage that will occur to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit.

[0023] The second characteristic configuration is that, in addition to the first characteristic configuration described above, the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the pipe length, using the polygon as the unit.

[0024] The third characteristic feature is that, in addition to the first characteristic feature described above, the topographic modification data includes either former river channels, former water bodies, or residential land development ground.

[0025] The fourth characteristic configuration is that, in addition to the first characteristic configuration described above, the damage prediction calculation unit includes a machine learning unit that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates the damage prediction information based on the decision tree model generated by the machine learning unit.

[0026] As described above, the present invention provides a pipeline damage prediction system, a pipeline damage prediction method, and a pipeline damage prediction program that can obtain appropriate damage prediction information for each pipeline by taking into account the effects of soft ground and ground boundaries.

[0027] Figure 1 is an explanatory diagram of the computer configuration on which the pipeline damage prediction system is built. Figure 2 is an explanatory diagram of the functional blocks of the pipeline damage prediction system. Figure 3 is an explanatory diagram of the configuration of the dataset for learning or evaluation. Figure 4A is an explanatory diagram for formulating ground boundary data. Figure 4B is an explanatory diagram for formulating ground boundary data. Figure 5 is a flowchart of the procedure for the pipeline damage prediction method. Figure 6 is an explanatory diagram of the prediction results by the pipeline damage prediction system. Figure 7A is an explanatory diagram of the damage prediction calculation unit showing another embodiment. Figure 7B is an explanatory diagram of the damage prediction calculation unit showing another embodiment. Figure 8A is an explanatory diagram of the damage estimation formula by JWWA. Figure 8B is an explanatory diagram of the damage estimation formula by JWRC.

[0028] The pipeline damage prediction system, pipeline damage prediction method, and pipeline damage prediction program according to the present invention will be described below with reference to the drawings. [Pipeline Damage Prediction System] As shown in Figure 1, the pipeline damage prediction system 10 consists of a computer C connected by a communication bus, which includes a motherboard equipped with an integrated circuit using semiconductors such as a CPU and chipset, a memory board equipped with semiconductor memory such as ROM and RAM, and an I / O board equipped with an input / output interface.

[0029] Computer C is connected to storage devices such as hard disk drives and flash memory devices, portable memory devices such as USB memory, input devices such as keyboards and mice, and output devices such as LCD displays and printers, and is configured to connect to a cloud computer via a communication interface such as the Wi-Fi standard.

[0030] The pipeline damage prediction system 10 is realized when a pipeline damage prediction program, provided via a recording medium such as a cloud computer, storage device, or portable memory device, is installed in the memory on a memory board, and the pipeline damage prediction program is executed by the CPU.

[0031] Figure 2 shows the functional block configuration of the pipeline damage prediction system 10. The pipeline damage prediction system 10 comprises a basic data storage unit 11 and a damage prediction calculation unit 12. The basic data storage unit 11 is composed of the aforementioned storage devices and cloud computers and stores pipeline attribute data indicating the attributes of the pipeline, ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of the following in the form of polygons in a geographic information system (hereinafter simply referred to as "GIS"): liquefaction data, ground boundary data, and topographic modification data, and damage data indicating the degree of damage caused by natural disasters that have occurred in the past. A polygon in GIS is a type of geometric data used to represent a geographic area or range, and is a closed figure formed by connecting multiple vertices (points). It represents a smaller area than a conventional mesh (for example, a mesh with sides of approximately 250m obtained by dividing a standard regional mesh set based on latitude and longitude into four sections).

[0032] The damage prediction calculation unit 12 is a calculation unit that generates damage prediction information predicting damage to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit 11.

[0033] The damage prediction calculation unit 12 includes a machine learning unit 13 that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates damage prediction information based on the decision tree model generated by the machine learning unit 13.

[0034] A decision tree is a knowledge representation that describes the procedure for separating features using a branched tree structure. It determines the properties and classification category of a pipeline, which is the target of damage estimation, from multiple attributes that characterize it. Attributes refer to the pipeline attribute data, ground characteristic data, and each classification item that constitutes the damage data. The decision tree algorithm is a machine learning method that automatically generates rules for the tree structure based on the characteristics of each attribute data and classifies pipelines. Branching progresses according to the values ​​of each attribute data, and ultimately, damage estimation information for the pipeline is obtained.

[0035] Figure 3 illustrates an example of a dataset of attributes set for each pipeline Pi (i = 1, 2, ...) that individually identifies each pipeline constituting the pipeline network. Pipeline Pi includes pipeline attribute data such as pipe type (data that can identify the material and seismic resistance), diameter (data indicating the size of the pipe), pipe length (data indicating the length of the pipe), and year of installation. In addition, ground characteristic data includes values ​​that are basically in polygon units, such as the microtopographic classification where the pipeline is laid, liquefaction data indicating liquefaction characteristics, topographic modification data such as old river channels, old water bodies and residential land development ground data, ground boundary data indicating whether the pipeline exists at a ground boundary, maximum ground acceleration, and maximum ground velocity. Furthermore, damage data includes the number of damages per unit pipe length (km) from past earthquakes. Microtopographic classifications are set with values ​​that reflect the degree of earthquake impact, such as velocity amplification factor, according to the microtopography. The "microtopographic classification" includes data indicating the category of "topography" and numerical data of "topography" indicating the degree of impact, such as velocity amplification factor. This takes into account the impact of earthquakes on micro-topography.

[0036] The liquefaction data indicates whether or not liquefaction occurred due to past earthquakes as 1 / 0 ("1" if it occurred, "0" if it did not). The topographic modification data, such as the old river channels, old water bodies, and residential land development ground data, indicates whether or not they apply as 1 / 0 ("1" if they apply, "0" if they do not). The ground boundary data indicates whether or not pipelines exist at the ground boundary as 1 / 0 ("1" if they exist, "0" if they do not).

[0037] If the liquefaction data is "1", it can be predicted that liquefaction will likely occur in the event of a future earthquake, increasing the likelihood of damage. The topographic modification data can be predicted that changes in topography will increase the likelihood of damage in the event of a future earthquake. This takes into account areas where ground deformation is likely to occur, which cannot be evaluated using the current topographic classification. Former river channels and former waterways have soft ground, and residential development areas have soft ground due to embankment, so it can be predicted that ground deformation will increase in the event of a future earthquake.

[0038] When a pipeline exists at the ground boundary, that is, at the boundary of an adjacent polygon, it can be predicted that due to future earthquakes, the ground will not change uniformly, resulting in different ground deformations acting on the pipeline and, as a result, pipeline damage will occur.

[0039] In Fig. 4A, the ground where three polygons represented by solid lines are adjacent is shown. The data source is a seamless geological map at a scale of 1:200,000. Assuming the width of the boundary line is 0.05 mm (about the thickness of a single hair), the ground boundary will have a width of about 10 m (= 0.05 × 200,000 × 1 / 1,000).

[0040] As shown in Fig. 4B, when the ground with different characteristics is adjacent, during an earthquake, the movements of the two will be different, affecting the pipeline at the boundary, and this influence is considered to extend to the adjacent joints. Since the length of one pipeline is 4 m to 6 m, the range of influence of the ground change on the pipeline is, on the safe side, ±6 m from the ground boundary. Considering the influence of the position accuracy as ±5 m and the influence of the ground change on the pipeline as 6 m on the safe side, the range of the ground boundary can be defined as ±11 m. Specifically, the ground boundary data can be defined as the overlapping area when the adjacent boundary with polygons as units is expanded by a predetermined width based on the pipeline extension. In this embodiment, the case where the width of the boundary line of the geological map can be estimated as 10 m is illustrated, but it is not limited to this value, and it is needless to say that it depends on the data source.

[0041] Using the data set as shown in Fig. 3 as learning data, a decision tree is generated by performing machine learning on the machine learning unit 13 using the basic learning algorithms exemplified below. First, appropriate attributes are selected from the learning data set and classified into subsets. If there are no attributes available for classification, the classification knowledge cannot be completed and the algorithm ends. For each subset, this algorithm is applied, the difference between the obtained damage prediction information and the actual damage information is calculated, and the process of creating a new decision tree (classification rule) to reduce the difference is repeated. At this time, instead of creating one decision tree, multiple decision trees are created, and a more appropriate decision tree is generated by using ensemble learning that uses all of them as one piece of knowledge.

[0042] Parallel ensemble methods such as the XGBoost method and the Random Forest method can be preferably used as ensemble learning algorithms. For example, in the Random Forest method, a plurality of learning datasets are created by randomly extracting data from the dataset shown in FIG. 3. Decision trees corresponding to each learning dataset are created using each learning dataset, and the overall result, that is, damage estimation information, can be obtained by using the average of the outputs of the plurality of decision trees.

[0043] FIG. 5 shows the above-described series of procedures. That is, a pipeline damage prediction method for predicting damage to a pipeline due to natural disasters using a computer includes pipeline attribute data indicating the attributes of the pipeline, and ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, the ground characteristic data including at least any one of liquefaction data, ground boundary data, and terrain modification data in units of polygons in GIS, and a basic data storage step (SA1) of storing damage data indicating the degree of damage due to natural disasters that occurred in the past in a basic data storage unit, and based on the pipeline attribute data, the ground characteristic data, and the damage data stored in the basic data storage unit in the basic data storage step, a damage prediction calculation step (SA2 to SA7) of generating damage prediction information predicting the damage occurring in the pipeline by a damage prediction calculation unit.

[0044] The damage prediction calculation step includes a machine learning step (SA2 to SA5) of generating a decision tree model by performing ensemble learning on a plurality of datasets composed of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and a step (SA7) of generating damage prediction information based on the decision tree model generated in the machine learning step.

[0045] Figure 6 shows the recall rate for actual earthquake damage caused by past earthquakes in a certain region, when one or more of the following six features from the aforementioned ground attribute data are removed or combined: presence or absence of liquefaction, presence or absence of residential land development ground, presence or absence of ground boundaries, presence or absence of old river channels / old water bodies, categorical topographic data, and numerical topographic data. It also shows the ratio of the recall rate to Model 0 (M0), which does not take any of these features into account. Here, recall rate refers to the proportion of actual damaged pipelines for which the predicted damage was accurate. A doubling of the recall rate means a doubling of the accuracy rate, and a larger number indicates better prediction accuracy.

[0046] A comparison of prediction results for a specific seismic motion using six features ("topography" includes both categorical and numerical data) that are thought to have a high correlation with actual pipeline damage revealed the following: Models that use at least one of each feature (M1-M13) have a higher recall rate than M0, which uses no features at all. Models that use only one feature have a high recall rate of 2.4 times and 1.8 times higher than M0 for the presence or absence of liquefaction (M1) and ground boundary (M3). Models that do not use any one of the features have the lowest recall rate in the model that only does not use the presence or absence of liquefaction (M7), followed by the model that does not use the presence or absence of residential land development ground (M8). Models that do not use only old rivers / old water bodies (M9) have a higher recall rate than models that do not use only ground boundaries (M10), and models that use both (M11) have an even higher recall rate. 5. By using numerical data on "topography" in addition to the presence or absence of liquefaction, the presence or absence of residential land development ground, ground boundaries, and former rivers / water bodies, the accuracy of reproduction is highest, at 3.1 times that of M0.

[0047] The dataset shown in Figure 3 illustrates an example where polygon data is assigned to pipelines as ground attribute data. However, for attribute data other than liquefaction, former river channels / water bodies, and ground boundaries, it is also permissible to use conventional mesh data.

[0048] The application program that embodies the pipeline damage prediction system 10 described above becomes the pipeline damage prediction program of the present invention, and is installed on a computer via the various recording media and communication interfaces described above.

[0049] In other words, the pipeline damage prediction program is a pipeline damage prediction program that causes a computer to function as a basic data storage unit that stores pipeline attribute data indicating the attributes of the pipeline, ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of liquefaction data, ground boundary data, and topographic modification data in the form of polygons in a GIS, and damage data indicating the degree of damage caused by natural disasters that have occurred in the past, and a damage prediction calculation unit that generates damage prediction information predicting damage that will occur to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit.

[0050] Furthermore, the damage prediction calculation unit includes a machine learning unit that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and is a program that generates damage prediction information based on the decision tree model generated by the machine learning unit.

[0051] It is preferable to have a pipeline renewal plan formulation device that formulates a renewal plan for existing pipelines based on the damage prediction information obtained by the pipeline damage prediction system 10 described above. For example, the renewal plan formulation device can be realized by comprising: a data storage unit capable of storing pipeline attribute data linked to pipeline IDs; a deterioration value unit that performs deterioration evaluation due to aging from the pipeline attribute data; a pipeline damage estimation unit that performs pipeline damage estimation due to natural disasters from the pipeline attribute data; a pipeline attribute data update unit that updates the pipeline attribute data by combining the deterioration evaluation result calculated by the deterioration evaluation unit and the pipeline damage estimation result calculated by the pipeline damage estimation unit based on the pipeline ID and storing them in the data storage unit; and a pipeline renewal plan calculation unit that calculates a pipeline renewal plan based on the updated pipeline attribute data. The pipeline damage prediction system of the present invention is applied to the pipeline damage estimation unit.

[0052] In the embodiments described above, the pipeline damage prediction system is equipped with a machine learning unit that generates a decision tree model by ensemble learning of multiple datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and a damage prediction calculation unit that generates damage prediction information based on the decision tree model generated by the machine learning unit. However, the damage prediction calculation unit may be configured to generate damage prediction information by employing the Markov chain Monte Carlo method (MCMC method).

[0053] The Markov chain Monte Carlo method is a method of defining a chain of probability distributions and estimating the parameters of the distribution to match the data. The number of damages within a mesh follows a Poisson distribution, and the distribution parameter λi in mesh i can be obtained from the formula shown in Figure 7A. The first term on the right-hand side represents the influence of each factor, and the second term Li on the right-hand side represents the length of the pipeline in mesh i. As shown in Figure 7B, each coefficient β of the first term on the right-hand side 〇 These are calculated by fitting a normal distribution. Each coefficient β 〇 By calculating this, we can derive the damage prediction formula. Damage prediction formula model: Number of damages = explanatory variable 1 × explanatory variable 2 × ... × pipeline length log(number of damages) = log(explanatory variable 1) + log(explanatory variable 2) + ... + log(pipe length) + error

[0054] Furthermore, instead of including a machine learning unit as the damage prediction calculation unit, the terrain correction coefficient Cg included in the conventional damage estimation formula Rm(v), as shown in Figures 8A and 8B, may be arranged into ground characteristic data that includes either ground boundary data or terrain modification data in polygon units in a GIS, and the liquefaction correction coefficient may be arranged into liquefaction data in polygon units.

[0055] Furthermore, the ground boundary data, topographic modification data, and liquefaction data may not be set to a value of 1 / 0 as in the example above, but rather as weighting coefficients that further correct the correction coefficient set for the micro-topographic division corresponding to the polygon where the pipeline is laid.

[0056] In the embodiments described above, a pipeline damage prediction system due to earthquakes was explained, but the pipeline damage prediction system according to the present invention can also be applied to predicting pipeline damage due to natural disasters such as ground collapse, volcanic eruptions, and floods.

[0057] The embodiments described above represent one aspect of the present invention, and the technical scope of the present invention is not limited based on this description. It goes without saying that the design can be modified as appropriate within the scope in which the effects of the present invention are achieved.

[0058] 10: Pipeline damage prediction system 11: Basic data storage unit 12: Damage prediction calculation unit

Claims

1. A pipeline damage prediction system that uses a computer to predict damage to pipelines caused by natural disasters, comprising: pipeline attribute data indicating the attributes of the pipeline; ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of liquefaction data, ground boundary data, and topographic modification data in polygon units in a geographic information system (GIS); a basic data storage unit that stores damage data indicating the degree of damage caused by natural disasters that have occurred in the past; and a damage prediction calculation unit that generates damage prediction information predicting damage to the pipeline based on the pipeline attribute data stored in the basic data storage unit, the ground characteristic data, and the damage data.

2. The pipeline damage prediction system according to claim 1, wherein the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the length of the pipe, using the polygon as the unit.

3. The pipeline damage prediction system according to claim 1, wherein the terrain modification data includes any of the former river channel, former water body, and residential land development ground.

4. The pipeline damage prediction system according to claim 1, wherein the damage prediction calculation unit comprises a machine learning unit that generates a decision tree model by ensemble learning of a plurality of datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates the damage prediction information based on the decision tree model generated by the machine learning unit.

5. A pipeline damage prediction method for predicting damage to pipelines caused by natural disasters using a computer, comprising: a basic data storage step of storing pipeline attribute data indicating the attributes of the pipeline, ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, ground characteristic data including at least one of liquefaction data, ground boundary data, and topographic modification data in polygon units in a geographic information system (GIS), and damage data indicating the degree of damage caused by natural disasters that have occurred in the past, in a basic data storage unit; and a damage prediction calculation step of generating damage prediction information predicting the damage to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit in the basic data storage step, using a damage prediction calculation unit.

6. The pipeline damage prediction method according to claim 5, wherein the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the length of the pipe, using the polygon as the unit.

7. The pipeline damage prediction method according to claim 5, wherein the topographic modification data includes any of the former river channel, former water body, and residential land development ground.

8. The pipeline damage prediction method according to claim 5, wherein the damage prediction calculation step involves executing a machine learning step to generate a decision tree model by ensemble learning of a plurality of datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generating the damage prediction information based on the decision tree model generated by the machine learning step.

9. A pipeline damage prediction program for a computer to function as follows: pipeline attribute data indicating the attributes of a pipeline; ground characteristic data indicating the characteristics of the ground in which the pipeline is buried, which includes at least one of liquefaction data, ground boundary data, and topographic modification data in polygon units in a geographic information system (GIS); and damage data indicating the extent of damage caused by natural disasters that have occurred in the past; and a damage prediction calculation unit that generates damage prediction information predicting damage to the pipeline based on the pipeline attribute data, ground characteristic data, and damage data stored in the basic data storage unit.

10. The pipeline damage prediction program according to claim 9, wherein the ground boundary data is defined as the overlapping area when adjacent boundaries are expanded by a predetermined width based on the length of the pipe, using the polygon as the unit.

11. The pipeline damage prediction program according to claim 9, wherein the terrain modification data includes any of the former river channel, former water body, and residential land development ground.

12. The pipeline damage prediction program according to claim 9, wherein the damage prediction calculation unit comprises a machine learning unit that generates a decision tree model by ensemble learning of a plurality of datasets consisting of pipeline attribute data, ground characteristic data, and damage data associated with each pipeline, and generates the damage prediction information based on the decision tree model generated by the machine learning unit.

Images