RU2707085C1 - Use of machine learning while ensuring safety of works, method of ensuring safety of works and system for implementation thereof - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: invention relates to preprocessing and analysis of an array of specialized data using machine learning for prediction and prevention of occurrence of hazardous situations in industrial enterprises. Group of inventions essence consists in application of machine training for investigation of incidents and prevention of recurrence of incidents at safety of works, wherein the work safety system comprises a database of objects, a user database, a user access module for objects, a module for monitoring the occurrence of incidents, a module for investigating incidents which is equipped with machine learning means, a module for preventing recurrence of incidents.

EFFECT: technical result is faster and more accurate development and implementation of a plan to prevent repeated incidents.

11 cl, 1 dwg

Description

The group of inventions relates to the field of preliminary processing and analysis of an array of specialized data using machine learning and can be used to predict and prevent the occurrence of dangerous situations in industrial enterprises.

A well-known system of work safety, consisting of a module for loading production data, configured to enter data about an incident, an incident database, a database of causes of incidents, a comparative analysis module for an incident database, made with the possibility of semantic analysis of incident databases and causes incidents, a module for obtaining the results of a comparative analysis, a module for the elimination of incidents and an expert group of enterprises yatiya, take measures to prevent the recurrence of an incident [CN106600126, Publication date: 12/06/2016, IPC: G06Q 10/06, G06Q 50/26].

A known method and system for ensuring the safety of work, including creating a database with information about the facilities of the enterprise, admitting employees to the facilities of the enterprise, monitoring the level of security and the occurrence of incidents at the facilities of the enterprise, identifying and classifying information obtained during monitoring, sending information about incidents to the expert group , analysis of the information received about the incident by the expert group, investigation of the incident by the expert group, identification of the causes of the incident, p zrabotku action plan to address the causes of the incident, the implementation of an action plan to address the causes of the incident at the enterprise objects and tracking of the action plan to address the causes of the incident [CN106485433, Publication date: 12/08/2016, IPC: G06Q 10/06].

As a prototype, a method and system for ensuring work safety was selected, including creating a database with information about the enterprise, creating a database of enterprise workers, monitoring incidents at the enterprise by collecting incident reports, investigating incidents that occurred at the enterprise, and preventing the possibility of re-occurrence of incidents Moreover, the investigation of incidents that have occurred at the enterprise is carried out by means of a statistical analysis of incident reports by an expert a group that is developing an action plan aimed at preventing the possibility of a re-occurrence of incidents at the enterprise [CN102496087, publication date: 12/13/2011, IPC: G06Q 10/06].

The disadvantage of the prototype and well-known technical solutions is the low efficiency of incident investigation and preventing the recurrence of incidents that have occurred at the enterprise, due to the influence of the human factor, which can cause erroneous actions in the statistical analysis of incident reports by an expert group. Erroneous actions can be caused by fatigue, emotional instability or insufficient competence of an individual employee who is involved in the investigation of the incident, while the human factor also significantly reduces the processing speed and increases the time it takes to conduct a comparative analysis of a large unstructured amount of data, as a result of which a developed action plan aimed at warning of the possibility of a re-occurrence of incidents, may be incorrect which can lead to the re-occurrence of incidents and significantly reduces the quality of work safety.

The technical problem to which the group of inventions aims is to improve the quality of work safety.

The technical result, which the group of inventions aims to achieve, is to increase the efficiency of incident investigation and preventing the recurrence of incidents.

The essence of the group of inventions is as follows.

The use of machine learning to investigate incidents and prevent recurrence of incidents while ensuring work safety.

A way to ensure work safety includes forming user permissions, monitoring incidents, investigating incidents and preventing the recurrence of incidents. Unlike the prototype, incident investigation is carried out using machine learning, and incident recurrence is prevented based on the data obtained during the incident investigation.

The work safety system consists of a database of objects, a database of users, a module for user access to objects, a module for monitoring the occurrence of incidents, a module for investigating incidents, and a module for preventing the recurrence of incidents. Unlike the prototype, the incident investigation module is equipped with a machine learning tool.

The database of objects contains information about the types and number of objects, about the qualifications of users that can be admitted to objects, etc. Moreover, objects are understood as workshops, sections, platforms, etc. The user database contains information about the employees of the enterprise, their personal data, qualification data, etc.

The formation of user permissions to objects provides the possibility of distributing users to objects depending on their specialty, experience, personal data, etc. The formation of user permissions to objects can be carried out by the module of user permissions to objects, which can be represented by the system server and access control to objects represented by checkpoints, automatic turnstiles, emergency communications and emergency warning and other means, providing ivayuschimi opportunity to restrict user access to the objects.

Monitoring the occurrence of incidents provides the opportunity to obtain information about the interaction of users allowed on objects. Observation of the occurrence of incidents can be carried out through the module for monitoring the occurrence of incidents, which can be represented by video monitoring, video recording and other means that provide the ability to record user actions and the conditions of occurrence of incidents.

Investigation of incidents is carried out using machine learning, which provides an increase in the speed and accuracy of incident investigation. Investigation of incidents involves the processing of data about the incident and the subsequent identification of the cause of the incident and can be carried out by the incident investigation module, which can be represented by a server, electronic devices spaced in space, etc.

Investigating incidents using machine learning may include compiling an incident database, analyzing an incident database, creating metadata, verifying and adjusting the accuracy of metadata, entering incident data, and receiving the results of an incident investigation.

The incident database may contain photo, video, audio data, text documents and other types of data. The incident database may contain a description of the incidents and the cause of the incidents.

The description of the incident can include any information about the incident, for example, the time and place of the incident, the number and qualifications of the participants in the incident, the location of the participants in the incident when it occurred, the category of danger of the incident, the category of losses when the incident occurred, etc. For example, May 25, 2018 at 12 : 30 in the workshop No. 1, the lathe engine failed, while the lathe operator was at the workplace, no one was injured.

The cause of the incident may include any information about what could have been the source of the incident. The cause of the incident may include a physical factor, for example, a failure of the lathe engine due to a breakdown of the control unit, or a human factor, for example, a failure of the lathe engine due to the fact that the operator did not perform timely maintenance of the lathe, or organizational a factor, for example, a lathe engine failure that occurred because no one was assigned to service the lathe.

In addition, the root cause of the incident can be indicated, including any information about what could have been the initial source of the incident, for example, a lathe engine failure occurred due to a defect on the production line of the processors that the control units of the lathes are equipped with.

Additionally, an action to eliminate the cause of the incident may be indicated, which may include any information on what steps were taken to eliminate the cause of the incident, for example, to eliminate the failure of the lathe engine, the interval between the maintenance of the lathe was reduced. Additionally, measures can be indicated to eliminate the root cause of the incident, for example, to eliminate the failure of the lathe engine, the supplier of the lathe control unit was changed.

Additionally, to increase the speed of database analysis, after compiling the database, database preprocessing can be performed, which may include, for example, removing punctuation marks, numbers, endings, conjunctions of prepositions from text documents, highlighting the necessary fragment in the photo, trimming the video file until when the incident occurred, etc.

Analysis of the incidents database is carried out by correlating any data (photos, videos, audio recordings, text documents, etc.) with each other to determine the frequency of mentioning the data in the database and determine the degree of influence of the data in the database on each other. For example, a database analysis can be performed by comparing incident description data with incident cause data and the event cause elimination action data.

Creating metadata can be done by comparing data that has a certain degree of influence on each other. Moreover, metadata refers to the relationship between the degree of influence of some data on the change in other data. Thus, for example, the relationship between the description of the incident, the cause of the incident and the event to eliminate the cause of the incident and the degree of influence of the description of the incident on the cause of the incident and the event to eliminate the cause of the incident can be monitored.

Verification and adjustment of the reliability of metadata provides the ability to track the reliability of the degree of influence of data on each other and the reliability of the formed relationships between these data. Verification and adjustment of the reliability of metadata can be done by entering the incident data and tracking the reliability of the received data in relation to the entered incident data. In this case, if the received data differs from reliable, then the degree of influence of the data on each other and the formed relationships between these data are adjusted. If there is no data, then the degrees of influence of the data on each other and new relationships between the data are added. For example, in the case of entering the description of the incident, the cause and the action related to eliminating the cause of the incident can be determined and, if they do not correspond to reliable data, the data can be adjusted and / or the necessary data added.

Entering incident data involves entering data about the incident and provides the ability to correlate incident data with metadata. After that, the degree of influence of the incident data on the metadata can be determined and the corresponding relationships from the metadata determined. If, after data entry, no corresponding relationships were found, then the nearest metadata can be selected for the incident data. Thus, for example, when entering an incident description, one or more causes of the incident and / or one or more measures to eliminate the causes of the incident can be detected.

Obtaining the results of an incident investigation implies choosing the most appropriate metadata links that provide the opportunity to identify the cause of the incident, and measures to eliminate the cause of the incident.

Prevention of a recurrence of incidents implies the identification of any actions aimed at preventing circumstances that may cause similar incidents, as well as monitoring the effectiveness of the actions taken. Prevention of recurrence of incidents can be carried out by means of an incident prevention module, which can be represented by the system server.

Prevention of recurrence of incidents may include the development and implementation of an action plan aimed at reducing the likelihood of a recurrence of an incident. The development of the plan can be carried out individually for each module of the system or comprehensively for all modules, while the sequence of actions inside each module can be determined, as well as the sequence of actions of the modules relative to each other. Implementation of the plan means the automatic transfer of the developed plan to all modules of the system, while the implementation of the plan can be carried out simultaneously in all modules of the system, or in a specific sequence to each module individually. After the implementation of the plan, individual and / or comprehensive verification of the effectiveness of incident prevention can be carried out. An individual check implies an assessment of the performance of each individual module, and a comprehensive check implies an assessment of the effectiveness of several modules. In this case, adjustments can be made to the actions of each individual module after each individual action or to several modules after several actions.

Prevention of the recurrence of incidents can be made by adjusting the actions when forming user tolerances. For example, the reasons for the occurrence of an incident can be taken into account and users have limited access to the source of the incident, or users who were involved in the occurrence of an incident, for example, due to illness, may lose access to objects during the illness, or users who were involved in events to eliminate the incident may be directed to the facility where, according to the results of the investigation, an incident may occur.

Prevention of recurrence of incidents can be made by adjusting actions when observing the occurrence of incidents. For example, the data on users who were participants in the incidents can be taken into account, and the observation can be concentrated on them, or the data on the objects at which the greatest number of incidents took place can be taken into account, and the monitoring of these objects can be strengthened.

Additionally, in order to reduce the risk of incidents after preventing the recurrence of incidents, an inspection of the actions taken can be carried out.

The group of inventions is characterized by a new, previously unknown essential feature, which consists in the use of machine learning to investigate incidents and prevent the recurrence of incidents while ensuring work safety. This eliminates the participation of a person in the process of investigating incidents and eliminates the possible negative impact of the human factor on the process of investigating incidents, thereby increasing the accuracy, speed and objectivity of the analysis of incident data, thereby increasing the accuracy and speed of obtaining the results of investigation of incidents, as well as by quickly to obtain accurate results of incident investigations, the speed and accuracy of the development and implementation of a plan to prevent recurrence and tsidentov, which ensures the achievement of the technical result is to increase the effectiveness of the investigation of incidents and to prevent recurrence of incidents, thereby improving the quality of security operations.

The presence of a new distinctive essential feature indicates the compliance of the group of inventions with the patentability criterion of "novelty."

The use of machine learning when performing operations involving the processing of large volumes of input data is known from the prior art and is widely used in various industries. However, the use of machine learning to investigate incidents and prevent the recurrence of incidents while ensuring work safety is not known in the art. This is confirmed by the fact that it is not known from the prior art to process a large amount of any type of input data about an incident and to obtain non-standard and non-obvious results of an incident investigation, on the basis of which a step-by-step or comprehensive action plan is automatically developed to prevent circumstances that may cause a re-occurrence incident. Also, it is not known from the prior art to implement the developed action plan, which is carried out taking into account the effectiveness at each individual stage and automatically adjusts the developed action plan in real time. Based on this, we can conclude that the group of inventions meets the patentability criterion of "inventive step".

The group of inventions can be implemented using known means, materials and technologies, which indicates its compliance with the patentability criterion of "industrial applicability".

The group of inventions is united by a single inventive concept, consisting in the use of machine learning to investigate incidents and prevent the recurrence of incidents while ensuring work safety, which allows us to conclude that it meets the criterion of “unity of invention”.

Below are examples of the implementation of the group of inventions that serve as confirmation of the possibility of achieving a technical result, can be changed and supplemented in any way, and cannot serve as a limitation of the scope of legal protection determined by the independent claims.

The group of inventions is illustrated by the following schemes.

Figure 1 - General diagram of a work safety system.

The work safety system consists of a base of 1 facilities, a base of 2 workers, a module 3 of workers' access to facilities, a module 4 for monitoring incidents at facilities, a module 5 for investigating incidents that have occurred at the facilities, and a module 6 for preventing the recurrence of incidents.

The way to ensure the safety of work is implemented as follows.

Admission module 3 distributes workers from base 2 to facilities from base 1. Workers begin to perform work duties, and incident occurrence monitoring module 4 monitors and records the data of incidents. An incident database for the past 15 years has been uploaded to Incident Investigation Module 5, based on which metadata has been generated.

After an incident occurs, incident investigation module 5 from modules 3 and 4 receives the description of the incident being investigated, after which they are analyzed and compared with metadata. After that, the appropriate causes of the incident and measures to eliminate the cause of the incident are determined and selected.

Incident investigation module 5 transmits information about the cause of the incident, as well as the event to eliminate the cause of the incident, to incident prevention module 6, which develops a plan for preventing the recurrence of incidents and passes it to modules 3 and 4.

Example 1

The responsible person entered the description of the incident under investigation via a tablet computer: “There was a fire in workshop No. 1”. After that, module 5 determines the cause: "Lathe wiring failure" and the event: "Replace the lathe wiring." Incident repetition prevention module 6 develops a plan that includes actions to replace the wiring of all machines of this type and check the wiring of all machines in the enterprise and transfer it to all modules of the system. The admission module 3 determines the repair team from base 2 to all workshops of the enterprise from the base 1 and sends it to conduct work to replace the wiring, and the observation module 4 monitors the execution of work through time sheets. Upon completion of work, modules 3 and 4 inform module 6 of the completion of the plan and module 6 monitors data related to the recurrence of the incident within 3 months. However, such incidents no longer occur.

Example 2

Surveillance cameras recorded the incident, which was transmitted to the incident investigation module 5 in the form of images from the scene of the incident, from which it follows that “The worker was injured while in workshop No. 1”. Module 5 defines the reason: “The employee was not at his workplace” and the events: “Conduct an explanatory conversation with the employee”, “Observe the employee for 2 months.” Incident repetition prevention module 6 develops a plan that includes the steps to conduct a test of professional suitability of all employees whose work experience is less than 6 months, and to check the checkpoints at the enterprise, and passes it to all system modules. The admission module 3 restricts the access of employees from base 2 who have not passed the test to the workshop of the enterprise from base 1. The observation module 4 monitors the passage of tests by employees of the enterprise and reports the test results to admission module 3, which allows access to employees who successfully pass the test. Monitoring module 4 prepares a report on violations at checkpoints and passes it to module 3 of the admission, which introduces a system of fines for violations at checkpoints. Modules 3 and 4 inform module 6 of the completion of the plan and module 6 checks the incident data to identify such incidents. However, no such incidents were detected and module 6 established a repeat of this plan every 12 months.

Example 3

The responsible person recorded an incident using a mobile phone, which was transferred to the incident investigation module 5 in the form of a video file, from which it follows that “the drive belt of the milling machine is torn in workshop No. 3”. Module 5 determines the reason: “In the workshop No. 3 there was no scheduled maintenance of the milling machines for more than 2 years” and the event “Perform scheduled maintenance of the milling machines in the workshop No. 3”. Module 6 for preventing the recurrence of incidents develops a plan that includes actions to conduct an unscheduled inspection of milling machines of the entire enterprise and distributing the workload between the milling machines of workshop No. 3 and 4 during maintenance of the milling machines in workshop No. 3, and transfers it to all system modules. Admission module 3 defines a team of workers from base 2 for work in the enterprise’s workshops from base 1 and distributes the milling machine operators from base 2 to workshops Nos. 3 and 4. Monitoring module 4 collects data on the implementation of the plan and, after the plan is completed, transfers it to module 6 that checks incident data to identify similar incidents. However, no such incidents were detected and module 6 recorded a successful result of the developed plan. After 12 months, the facilities of the enterprise were inspected, as a result of which no violations in the operation of the equipment were detected.

Example 4

At the end of the month, it was revealed that “In workshops Nos. 1,2,3 and 4 there were a large number of incidents involving injuries by employees due to loss of consciousness”, which were transferred to module 5 of the incident investigation in the form of text documents, and also video and audio files. Module 5 defines the reason: “The temperature and humidity in the shops No. 1,2,3 and 4 exceeds the norm by 30%” and the event “Install in the workshops No. 1,2,3 and 4 air-conditioning systems with increased productivity”. Module 6 for preventing the recurrence of incidents develops a plan that includes actions to limit the admission of workers to workshops whose air temperature exceeds 38 ° C and reduce the working hours of employees whose age exceeds 50 years on days when the air temperature exceeds 25 ° C. The admission module 3 defines a third-party team of workers to install the air conditioning system, enters them into base 2 and defines them for installation of the air conditioning system in workshops Nos. 1,2,3 and 4, and also restricts the access of employees in accordance with the plan developed module 6. Monitoring module 4 monitors the situation in workshops Nos. 1,2,3 and 4, at all other facilities of the enterprise and transmits data related to the reduction of the occurrence of such incidents by 80% to module 6. Module 6 recorded a successful result, and module 4ablyudeniya after 1 month transmitter module 6 the results of medical examinations, which showed a significant improvement in the health of employees, and after a further 1 month was not recorded such incidents.

Thus, a technical result is achieved consisting in increasing the efficiency of incident investigation and preventing the recurrence of incidents, thereby improving the quality of work safety.

Claims (11)

1. The use of machine learning to investigate incidents and prevent the recurrence of incidents while ensuring work safety. 2. A method for preventing the recurrence of incidents while ensuring the safety of work, implemented by the system according to claim 11, including generating user permissions, monitoring incidents, investigating incidents and preventing the recurrence of incidents, characterized in that the investigation of incidents is carried out using machine learning, and Prevention of recurrence of incidents is carried out taking into account the data obtained during the investigation of incidents. 3. The method according to claim 2, characterized in that the investigation of incidents using machine learning includes compiling an incident database, analyzing an incident database, creating metadata, verifying and adjusting the accuracy of metadata, entering incident data and receiving the results of an incident investigation. 4. The method according to claim 3, characterized in that after compiling the database, the database preprocessing is performed. 5. The method according to claim 3, characterized in that the analysis of the database of incidents is performed by correlating the data with each other to determine the frequency of mentioning the data in the database and determine the degree of influence of the data in the database on each other. 6. The method according to claim 5, characterized in that the creation of metadata is performed by comparing data having a certain degree of influence on each other, to track the relationship between the degree of influence of one data on the change of other data. 7. The method according to claim 2, characterized in that the prevention of the recurrence of incidents is done by automatically determining actions aimed at reducing the likelihood of a recurrence of an incident. 8. The method according to claim 7, characterized in that the prevention of the recurrence of incidents is done by automatically transferring actions aimed at reducing the likelihood of a recurrence of an incident. 9. The method according to claim 2, characterized in that the prevention of the recurrence of incidents is made by adjusting the actions when forming user tolerances. 10. The method according to claim 2, characterized in that the prevention of the recurrence of incidents is made by adjusting actions when observing the occurrence of incidents. 11. A system for ensuring the recurrence of incidents while ensuring work safety, consisting of a database of objects, a database of users of objects, a module for user access to objects, a module for monitoring the occurrence of incidents, a module for investigating incidents and a module for preventing the recurrence of incidents, while the module for monitoring the occurrence of incidents incidents is connected to the module for investigating incidents with the ability to transmit incident data, the module for monitoring incidents in it is complete with the possibility of generating a report and transmitting it to the user access module to the objects, the incident recurrence warning module is configured to develop an action plan to prevent the recurrence of incidents and transfer the developed action plan to prevent the recurrence of incidents to all system modules, the tolerance module is configured to extract data about objects from the database of objects and user data from the database of user objects ntov configured to obtain data on the completion of the plan of action to prevent recurrence of incidents on users of the module tolerances objects and observation unit for an incident, wherein the incident investigation unit is equipped with a means of machine learning.

Images