System for predicting generation of disaster in work site

Abstract

To suppress generation of disasters by predicting generation of disasters and calling for attention.SOLUTION: A disaster generation prediction system A for predicting generation of disasters in a work site such as a construction site or a factory includes: a derivation unit 10 for deriving disaster prediction information and keyword information by performing machine learning processing including previous disaster information in teacher data and by performing text mining processing on previous disaster information; and a report keyword unit 20 for generating a prediction result report C by using at least the disaster prediction information and the keyword information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a system for predicting the occurrence of various types of accidents, such as industrial accidents, at work sites, such as construction sites and factories. [Background technology]

[0002] The following Patent Document 1 discloses a device that predicts the occurrence of disasters related to construction work by using document information related to the construction work (work procedures, construction schedules, execution plans, reports, etc.). As shown in Figures 12 and 13 of Patent Document 1, by specifying the target building, the machine-learned disaster occurrence prediction model makes a disaster prediction, and outputs a prediction result screen showing a comment indicating the possibility of a disaster occurring at the specified building, as well as the type of disaster, the time of occurrence, etc.

[0003] In addition, the following Non-Patent Document 1 discloses a machine learning model that predicts whether or not a disaster will occur in the following month using primary data such as the total number of workers and supervisors, total working hours, and number of accidents at each construction site in the previous month, and secondary data such as time series trends and change rates as features. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2022-189402 A [Non-patent literature]

[0005] [Non-Patent Document 1] On-site construction safety promotion activities using AI to predict the probability of workplace accidents (https: / / www.mhi.co.jp / technology / review / pdf / 593 / 593150.pdf) Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a disaster prediction system that leads to the prevention of disaster occurrence using a method different from the conventional methods described above. [Means for solving the problem]

[0007] The present invention, made to solve the above-mentioned problems, is a disaster occurrence prediction system capable of predicting the occurrence of disasters at work sites such as construction sites and factories, and is characterized by having at least a derivation unit that derives disaster prediction information and keyword information through machine learning processing that includes past disaster information in training data and text mining processing of the past disaster information, and a report generation unit that generates a prediction result report using at least the disaster prediction information and the keyword information. Furthermore, in the present invention, the derivation unit can use future prediction values ​​obtained from any future prediction model independent of the disaster occurrence prediction system to derive the disaster prediction information. In addition, the present invention may be configured so that the report generation unit has a function of generating a forecast chart column illustrating risk information of disaster occurrence for each work site for each specified period based on the disaster prediction information. The present invention may also be configured such that the report generation unit has a function of generating, based on the disaster prediction information, a summary section including a summary description of the disaster prediction information. The present invention may also be configured such that the report generating section has a function of generating a notice column including an introduction to keywords that require attention, based on the keyword information. Effect of the Invention

[0008] According to the present invention, at least one of the following effects can be obtained. (1) Not only does it derive disaster prediction information (such as the number and probability of occurrence at each work site), but it also derives keyword information related to the occurrence of disasters, which enables more specific warnings and helps prevent the occurrence of disasters. (2) By using future prediction values ​​obtained from future prediction models such as weather forecast models to derive disaster prediction information, it is possible to provide disaster prediction information with higher accuracy. (3) By providing the forecast result report with a forecast chart column showing disaster risk information for each work site for each specified period, workers can be alerted in a manner similar to a familiar weather forecast chart. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a functional block diagram of a disaster occurrence prediction system according to the present invention. [Diagram 2] FIG. 4 is a functional block diagram of a derivation unit according to the first embodiment. [Diagram 3] A figure showing an example of a prediction result report output (1). [Figure 4] A figure showing an example of a prediction result report output (2). [Diagram 5] A figure showing an example of a prediction result report output (3). [Figure 6] A figure showing an example of a prediction result report output (4). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, embodiments of the present invention will be described with reference to the drawings. EXAMPLES

[0011] <1> Overall configuration (Fig. 1) FIG. 1 is a functional block diagram of a disaster occurrence prediction system A according to the present invention. The disaster occurrence prediction system A of the present invention is configured to include at least a derivation unit 10 that derives disaster prediction information and keyword information based on the results of machine learning and text mining using document information B, and a report generation unit 20 that generates a prediction result report C using at least the disaster prediction information and keyword information derived by the derivation unit 10. Each unit can be realized by any combination of hardware and software, and each unit can be a separate device, or multiple units can be integrated into a single device. The details of each part will be explained below.

[0012] <2> Lead-out section (Fig. 1, Fig. 2) The derivation unit 10 has a function for deriving disaster prediction information and keyword information. In this embodiment, as shown in FIG. 2, the derivation unit 10 is divided into two functional blocks (a predicted information derivation unit 11 and a keyword derivation unit 12), but the present invention is not limited to this configuration. The derivation process may be performed manually by a user's command, or automatically on a regular basis by the system.

[0013] <2.1> Prediction information derivation part (Fig. 2) The prediction information derivation unit 11 has a function for deriving disaster prediction information. This "disaster prediction information" is information that indicates the details of the predicted disaster (date of occurrence of the disaster, location of occurrence of the disaster, etc.).

[0014] <2.1.1> Example of derivation method The prediction information derivation unit 11 can be configured, for example, using a neural network that has been machine-learned using the accumulated document information B as teacher data (a learning data set). Document information B used as training data for deriving disaster prediction information can use the information written in each section of a disaster report that summarizes the details of past disasters.

[0015] <2.1.2> Example of training data For example, training data for each disaster occurrence can be generated from the following item information included in the summary section 30 in the disaster report. Disaster date: The date the disaster occurred Location of disaster: Location of the work site (work base) where the disaster occurred

[0016] <2.1.3> Example of disaster forecast information output The prediction information derivation unit 11 is equipped with a neural network that is machine-learned using this training data, and so, for example, when a future date and a predicted location are input data to the neural network, the unit can derive as output data the presence or absence (or number) of a disaster occurring at the predicted location, with that date set as the predicted date of disaster occurrence. Then, information linking the predicted disaster occurrence date and predicted location can be used as disaster prediction information. In addition, the prediction information derivation unit 11 can further compile information linking the predicted disaster occurrence date and predicted location, and derive the predicted number of disasters occurring per month for each base station as disaster prediction information.

[0017] <2.2> Keyword derivation section (Fig. 2) The keyword derivation unit 12 has a function for deriving keyword information related to the occurrence of a disaster. This "keyword information" is information that indicates words that are highly relevant to the predicted disaster.

[0018] <2.2.1> Example of derivation method The keyword derivation unit 12 can, for example, perform text mining processing on sentences and natural language in the accumulated document information B, and output analysis results that extract, for each word, the part-of-speech classification, occurrence frequency, co-occurrence relationships (associating words with similar occurrence patterns that appear in a sentence), and the like.

[0019] <2.2.2> An example of text mining The document information B used by the keyword derivation unit 12 can be text entered in a detail section 40, such as a detailed entry section for the disaster content or a free entry section, in a disaster report that summarizes the contents of past disasters.

[0020] <2.2.3> Example of keyword information output As a result of performing text mining processing on the sentences entered in the details column 40, the keyword derivation unit 12 can derive keyword information such as words that appear frequently when a disaster occurs and combinations of words that are considered to have a co-occurrence relationship. The target of the text mining process may be all the accumulated document information B, or may be the document information B created within a specific period of time. For example, when a prediction period is set to January to March in the future, only disaster reports in which disasters occurred between January and March in the past may be subject to text mining processing.

[0021] <2.2.4> Example of grouping keyword information In the derivation unit 10, each word derived from the keyword information may be further grouped by an arbitrary name. For example, in the case of words derived from keyword information, such as "step," "shelf," "outlet," "air," and "entrance / exit," each word can be grouped under the name "building-related" as a concept common to these words. In addition, in the case of words derived from keyword information, such as "extrusion," "mold," "casting," and "metal plate," each word can be grouped under the name "equipment-related" as a common concept to these words.

[0022] <3> Report generation section (Figure 1) The report generating unit 20 has a function for generating a prediction result report C by using at least the disaster prediction information and keyword information derived by the derivation unit 10.

[0023] <3.1> Example of prediction result report It is preferable that the prediction result report C is generated in a manner that allows safety management personnel and workers at each work site to easily understand the predicted future frequency of accidents, the details of the accidents, and the like.

[0024] In this embodiment, a disaster occurrence prediction system A according to the present invention is configured so that a user can log in from a Web browser installed on a user's terminal and view a prediction result report C on the Web browser. The forecast result report C may include a forecast chart column 50, a forecast summary column 60, a forecast caution column 70, a graph column 80, and the like. The display image of each column will be described with reference to FIGS.

[0025] <3.2> Forecast chart column (Figure 3) The forecast chart field 50 is a field that illustrates disaster risk information for each work site for each predetermined period based on the disaster forecast information derived by the derivation unit 10. As shown in Figure 3, the forecast chart section 50 has leading lines drawn from the locations of the work sites (sites A to E) on the map of Japan, and provides risk information 51 in the form of icons representing weather, showing the probability of disasters occurring each month over the next three months. In this embodiment, an icon with a lightning mark indicates an extremely high probability of a disaster occurring (80-99%), an icon with a rain mark indicates a high probability of a disaster occurring (60-79%), and an icon with a cloudy mark indicates a somewhat high probability of a disaster occurring (40-59%). There are various possible methods for determining the probability of a disaster occurring from the disaster prediction information derived by the derivation unit 10, and the present invention does not particularly limit the method. For example, however, a configuration can be made in which multiple thresholds are set for the total number of disasters occurring in a month for each work site, and which icon is to be displayed is determined based on which threshold is exceeded.

[0026] <3.3> Forecast summary section (Fig. 4(a)) The forecast summary field 60 is a field that displays a summary of disaster prediction information for each predetermined period based on the disaster prediction information derived by the derivation unit 10. The report generation unit 20 accumulates template data for summary sentences, and selects template data based on the disaster prediction information derived by the derivation unit 10, and generates a summary sentence by appropriately filling in blanks in the selected template data (such as base name, probability of disaster occurrence, predicted month of disaster occurrence, etc.). It is preferable to prepare various variations of summary template data depending on the number of disasters occurring per month at each base, the total number of disasters discovered at multiple bases, etc.

[0027] <3.4> Forecast caution column (Fig. 4(b)) The forecast caution column 70 is a column that displays an introduction to keywords that require caution, based on the keyword information derived by the derivation unit 10. The report generation unit 20 accumulates template data for introductions, and can generate introductions by filling in blanks in the template data as appropriate using keyword information derived by the derivation unit 10 (such as frequently occurring words, combinations of words that are considered to have a co-occurrence relationship, etc.). It is preferable to prepare various variations of the template data for the introductory text according to the frequency of appearance, the concept names used for the grouping explained in <2.2.4> above, and the like.

[0028] <3.5> Graph column (Figures 5 and 6) The graph column 80 is a column that uses at least the disaster prediction information derived by the derivation unit 10 to display in a graph the number of disaster occurrences for each predetermined period narrowed down by any condition.

[0029] <3.5.1> Display for each work site A graph section 80 shown in FIG. 5 is a graph in which the horizontal axis indicates the time of disaster occurrence and the vertical axis indicates the number of disaster occurrences for each work site (each base). The graph on the left side of Figure 5 plots the number of past and future disaster occurrences from January 2018 to September 2023, while the graph on the right side plots the number of disaster occurrences by month in 2023.

[0030] <3.5.2> Display by attribute A graph column 80 shown in FIG. 6 displays, for each attribute of disaster content, the time of disaster occurrence on the horizontal axis and the number of disaster occurrences on the vertical axis. The "disaster content attributes" may be the names used for the grouping explained in <2.2.4> above, or other names of classification conditions that have been appropriately defined. The graph on the left side of Figure 6 shows the past and future disaster occurrences from January 2018 to September 2023, while the graph on the right side shows the monthly disaster occurrences in 2023.

[0031] <4> summary In this way, the disaster occurrence prediction system A of this embodiment not only derives disaster prediction information (such as the number of disasters accumulated by work site and time period, and the probability of disaster occurrence), but also derives keyword information related to the occurrence of a disaster, making it possible to provide more specific warnings. In addition, the prediction results are displayed in a report format similar to a familiar weather forecast chart, which is beneficial in that the content of the warnings is easy for safety management personnel and workers to understand.

[0032] <5> supplementary explanation Furthermore, if appropriate warnings are issued by using disaster occurrence prediction system A of the present invention, it is possible that a discrepancy will occur between the results predicted by disaster occurrence prediction system A of the present invention and the actual results, resulting in low prediction accuracy. However, since the purpose of the disaster occurrence prediction system A according to the present invention is to prevent disasters from occurring, this purpose is fully achieved. EXAMPLES

[0033] The disaster occurrence prediction system A according to the present invention can further be combined with any future prediction model. This future prediction model is a model that is independent (established) from the disaster occurrence prediction system A, and can utilize various future prediction models that are expected to have a correlation with the occurrence of disasters (for example, a weather forecast model (a model that predicts future weather), a product production volume model (a model that predicts future production volume), a construction volume model (a model that predicts future construction volume), a worker long-term skill model (a model that predicts the future improvement in worker skills), etc.). According to this embodiment, it is possible to provide disaster prediction information with higher accuracy. Below, an example of a combination will be described in which a weather forecast model is used as the future prediction model.

[0034] <1> Combination example (1) Incorporation into prediction information derivation part (not shown) In this example, the teacher data (learning data set) for machine learning performed by the prediction information derivation unit 11 is configured to include actual weather information values ​​for each location and date obtained from another database, and when deriving disaster prediction information, future weather forecast values ​​obtained from a weather forecast model that is independent of the disaster occurrence prediction system A are also used as input data to derive disaster prediction information.

[0035] <2> Combination example (2) Correction of disaster prediction information (not shown) In this example, in addition to the predicted information derivation unit 11 and the keyword derivation unit 12 that constitute the derivation unit 10 in the first embodiment, a correction coefficient output unit is provided as a new functional block. The correction coefficient output unit can be configured as a neural network that uses past weather information for each location and date, and correction coefficients that are quantified using predetermined rules to determine whether or not a disaster has occurred and the number of such disasters, as training data (learning data set). By inputting future weather forecast values ​​obtained by a weather forecast model independent of the disaster occurrence prediction system A of the present invention as input data into this correction coefficient output unit, a correction coefficient is derived for each base station and each date, and the disaster prediction information (such as the number of disaster occurrences per month) obtained by the prediction information derivation unit 11 is multiplied by an appropriate correction coefficient to derive final disaster prediction information (such as the number of disaster occurrences per month after multiplication). [Explanation of symbols]

[0036] A: Disaster prediction system 10: Derivation part 11: Prediction information derivation part 12: Keyword extraction section 13: Correction coefficient output section 20: Report generation section B: Document information 30:Summary column 40:Details column C: Prediction result report 50: Forecast chart column 51: Risk information 60: Forecast summary 70: Forecast warning column 80:Graph column

Claims

1. A disaster occurrence prediction system capable of predicting the occurrence of disasters at work sites such as construction sites and factories, A derivation unit that derives disaster prediction information and keyword information by machine learning processing using past disaster information as training data and text mining processing of the past disaster information; a report generating unit that generates a prediction result report by using at least the disaster prediction information and the keyword information; The present invention is characterized by comprising at least Disaster occurrence prediction system.

2. The derivation unit derives the disaster prediction information by using a future prediction value obtained from an arbitrary future prediction model independent of the disaster occurrence prediction system. The disaster occurrence prediction system according to claim 1 .

3. The report generation unit has a function of generating a forecast chart column showing risk information of disaster occurrence for each work site for each predetermined period based on the disaster prediction information. The disaster occurrence prediction system according to claim 1 .

4. The report generation unit has a function of generating an overview section including an overview description of the disaster prediction information based on the disaster prediction information. The disaster occurrence prediction system according to claim 1 .

5. The report generating unit has a function of generating a warning column including an introduction to keywords that should be noted based on the keyword information. The disaster occurrence prediction system according to claim 1 .

Images