KR20220132818A - AI Gas Leak Detection System with Self-Diagnosis Function and operating Method thereof - Google Patents

Abstract

The present invention relates to an AI gas leak detection system having a self-diagnosis function. By periodically diagnosing whether the gas leak detection system is operating normally and continuously storing the diagnostic data in the database, the accumulated information obtained is used as machine learning data to improve the reliability of the gas leak detection system, which is greatly increasing in complexity, and even system administrators who lack expertise can easily recognize the state of gas leakage and reduce adverse functions such as fatigue due to excessive alarms (warning). According to the present invention, there are effects of efficiently inspecting all sensors in a system in which numerous sensors are installed and operated at all times and quickly identifying problems such as an alarm system not operating normally even though there is a gas leak or an error alarm generated due to a malfunction even though there is no gas leak. In addition, since replacement cycles are predicted and guided for numerous sensors in advance, there is an effect that the improvement of management efficiency of the gas leak detection system can be expected, and an effect of improving user convenience while resolving inconvenience caused by excessive alarms such as alarm (alarm) errors.

Description

AI Gas Leak Detection System with Self-Diagnosis Function and operating Method thereof

The present invention relates to an AI (artificial intelligence) gas leak detection system having a self-diagnosis function, and more particularly, to periodically diagnose whether the gas leak detection system operates normally and to continuously store and secure the diagnostic data in a database By utilizing the accumulated information as machine learning data, the reliability of the gas leak detection system, which is greatly increasing in complexity, is improved, and even system administrators with insufficient expertise can easily recognize the gas leak state, but feel fatigue due to excessive alarms (alarms). It relates to an AI gas leak detection system with a self-diagnosis function that can reduce adverse functions such as

In the gas leak detection system used in industrial sites or at home, various sensor signals that detect gas leaks are transmitted to the control server through wired and wireless communication means, and through the analysis and control process of the control server, the abnormality of the gas facilities is checked by the central control center or manager. It is common to design the terminal so that it can be notified visually (display), auditory (speaker), tactile sense (vibration motor), etc. Recently, as the importance of gas safety management has been greatly emphasized, the types of gas sensors have been diversified and the functions of the control server have become more advanced. . In particular, gas facility managers have no choice but to react sensitively to alarm (alarm) errors that may occur due to abnormal operation of the gas leak detection system. There is an urgent need for an alarm (alarm) in a state that guarantees the normal operation of the system.

Reflecting such realistic needs, technology development with a self-diagnosis function to check whether the gas leak detection system is operating normally is continuously being made. Patent No. 10-1084482 (Prior Patent 1) discloses a gas leak blocking control device capable of self-diagnosis and a diagnostic method thereof. A function that can indirectly check whether the gas circuit breaker operates normally through the detection unit or the display unit according to the operation signal is exemplified. However, the reliability level of the alarm (alarm) was not up to the expectations as the function to diagnose whether the gas leak alarm was operating normally in the absence of an operation signal from the manager was lacking.

In addition, in Utility Model Registration No. 20-278916 (Prior Patent 2), the time when the gas leak alarm occurred is stored in the memory in units of year, month, day, and hour, and when the user presses the OK key, the stored information is collectively stored. It is configured so that it can be checked and output on the screen, so that the gas leakage cycle can be predicted and the corresponding gas facility can be maintained in advance according to the expected cycle. However, in Patent 2 above, it is simply that the operator predicts the gas leak cycle by checking the alarm history of the gas leak alarm, and clearly distinguishes whether the alarm is caused by the normal operation or malfunction of the gas alarm. The problem was that it couldn't be done.

After all, in the above prior patents 1 and 2, in that the diagnosis is made artificially through intentional manipulation of the administrator, rather than systematically automatically diagnosing whether the gas leak detection system is operating normally, the The possibility of an alarm (alarm) error cannot be completely excluded, and there is still a limit to fundamentally solving adverse functions such as fatigue and discomfort caused by excessive alarm (alarm).

Patent No. 10-1084482 Utility model registration No. 20-278916

The present invention has been made in view of the problems with the above background art,

The purpose of this is to provide an AI gas leak detection system with improved user convenience and an operation method thereof while resolving the inconvenience caused by excessive alarms such as alarm errors.

In addition, it has a function of self-diagnosing whether there is an abnormality in the system, and by using the accumulated information obtained through the diagnosis function as learning data, an optimal system operating environment is set to increase the reliability of the alarm (alarm). The purpose is to provide a detection system and its operating method.

In the AI gas leak detection system according to a preferred embodiment of the present invention, a plurality of gas leak detection sensors, gas leak alarms and alarm means installed in a gas use area and a gas supply area are connected to a control server through a communication network, and the control server is further provided with a self-diagnosis unit, the self-diagnosis unit includes a database for storing self-diagnosis information, a machine learning algorithm, and a display, and the self-diagnosis unit includes a transmission capable of transmitting a diagnostic signal for a simulation test to each detection sensor It is characterized in that it is provided with means and a receiving means capable of receiving a feedback signal corresponding to the diagnostic signal for the simulation test from each detection sensor.

In addition, the AI gas leak detection system of the present invention sets the transmission cycle of the simulation test diagnostic signal variably according to the usage time of the gas facility, and the machine learning algorithm uses the accumulated information of self-diagnosis stored in the database to generate the AI prediction model. It is characterized in that it is configured to create.

In addition, the AI gas leak detection system of the present invention has a function to predict whether each sensor needs zero adjustment or a failure in the AI prediction model, a function to determine whether an alarm signal is generated when a gas leak is determined, and a replacement time for the gas sensor Among the functions for predicting, the function to suggest the correction time and desired correction value of the initially set reference value in response to the aging curve of the gas sensor, and the function to determine the priority of the alarm (alarm) according to the severity of the gas leak It is characterized in that it is configured to include at least one or more functions.

On the other hand, the method of operating the AI gas leak detection system of the present invention recognizes whether a preset self-diagnosis period has been reached, transmits a diagnostic signal for a simulation test to each sensor, and generates in response to a diagnostic signal for a simulation test from each sensor receiving a feedback signal, comparing the feedback signal with a reference value stored in advance, determining whether the feedback signal is out of the range of the reference value, and if it is within the range of the reference value, it is determined as normal and the diagnosis result is stored in the database (DB) ), when the feedback signal is outside the reference value range, it is determined as abnormal, and the corresponding feedback signal is input to the AI prediction model. Step, generating an alarm when the abnormal determination by the AI prediction model is outside the permissible limit of the sensor function, and if the abnormal determination by the AI prediction model is not outside the permissible limit of the sensor function, without generating an alarm It characterized in that it consists of the step of displaying the character guide through the display.

In addition, the operating method of the AI gas leak detection system of the present invention sets the transmission cycle of the simulation test diagnostic signal variably according to the use time of the gas facility, and the machine learning algorithm uses AI through accumulated information of self-diagnosis stored in the database. It is characterized in that the step of generating a predictive model is further provided.

According to the present invention, all sensors can be efficiently checked in a system in which numerous sensors are installed and operated all the time, and the alarm system does not operate normally even when there is actually a gas leak, or an error alarm due to a malfunction is provided even when there is no gas leak. It has the effect of solving problems such as those that occur.

In addition, since the replacement cycle of numerous sensors is predicted and guided in advance, the management efficiency of the gas leak detection system can be improved, and the user convenience is improved while eliminating the inconvenience caused by excessive alarms such as alarm (alarm) errors. Of course, it is possible to reduce the occurrence of unnecessary alarms by dualizing the alarm method for self-diagnosis results, so it is effective to prevent the manager from relaxing the alarm due to frequent alarms (alarms).

1 is a configuration diagram for explaining a conventional gas leak detection system.
2 is a block diagram for explaining the AI gas leak detection system of the present invention.
3 is an exemplary diagram for explaining the configuration and function of a self-diagnosis unit based on machine learning premised in the present invention.
4 is an exemplary diagram for explaining a prediction model targeted in the present invention.
5 is an exemplary view for explaining the operation method of the AI gas leak detection system of the present invention.

Hereinafter, an AI gas leak detection system and an operating method thereof according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

For reference, in describing the present invention, when it is determined that a detailed description of a known configuration or function may unnecessarily obscure the gist of the present invention, a detailed description thereof will be omitted. And it should be noted that the technical terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. In addition, the technical terms used in the present invention should be interpreted as meanings generally understood by those of ordinary skill in the art to which the present invention belongs, unless otherwise defined in the present invention, and excessively comprehensive It should not be construed as a human meaning or an excessively reduced meaning.

1 shows the configuration of a conventional gas leak detection system. In general, since gas facilities can cause enormous damage even with the slightest negligence, it is accepted as a natural duty to build safety facilities for gas facilities and to operate the safety facilities at all times. The gas leak detection system of FIG. 1 is provided with a plurality of sensors and a leak alarm, and the leak alarm compares the detection signal received from each sensor with a preset reference value to determine whether there is a gas leak, and then an alarm according to the determination result is configured to occur. The system illustrated in FIG. 1 is divided into a 'use area alarm' that alerts whether there is a leak in an indoor gas facility and a 'supply area alarm' that alerts whether a leak occurs in a gas supply facility located outdoors. Signal transmission is made in one-way communication between sensor → leak alarm → alarm means (or control center). It is operated in a way that supports the rapid recognition of leaks.

However, in the conventional system of FIG. 1, the manager takes safety measures only when an alarm (alarm) occurs while using the setting conditions in the initial construction stage, or manually through the user's operation to check whether the system is operating normally. is done For this reason, it was not easy to efficiently check and manage all sensors in a system in which numerous sensors were installed and operated all the time. This meant that there was a limit to the manager's ability to quickly respond to problems such as the alarm system not operating normally even though there was actually a gas leak, or an alarm caused by a malfunction even though there was no gas leak. Furthermore, until now, sensors have been separated from the equipment to check for defects, and the verification process has a strong tendency to depend only on the manager's subjective judgment or experience, making it difficult to quickly and accurately check various types of sensors. In addition, many sensors were not able to meet the expected level in terms of efficient system management because they were replaced after a measurement error was discovered due to performance degradation, rather than being replaced before an error occurred by judging whether the performance deteriorated in advance. .

The present invention is to overcome the above problems of the prior art, Figure 2 shows a preferred embodiment of the present invention. The gas leak detection system of FIG. 2 is characterized in that the communication network is configured to enable bidirectional communication in terms of hardware, and a self-diagnosis unit 100 and a database (DB) are added to the control server. By continuously accumulating the data of the self-diagnosis results together with the diagnosis function in the DB, and using the accumulated data as machine learning data, optimal alarm generation (AI alarm generation) is possible and excessive alarm generation Or, there is a technical feature to prevent adverse functions such as fatigue due to the occurrence of a malfunction alarm.

In general, when an alarm error occurs due to a sensor malfunction or a connection fault in the system wiring, or when an alarm does not work even when an alarm needs to be generated according to an emergency situation, the reliability of the detection system is greatly affected.

Therefore, the present invention configures a communication network to enable two-way communication, sends a diagnostic simulation signal to each sensor and leak alarm, etc. through the self-diagnosis unit 100 provided in the management manual, and provides a feedback signal in response to the simulated diagnostic signal. It is configured so that it can be received and determined whether the system is operating normally. At this time, the normal operation is determined by comparing the feedback signal with a preset reference value, and the diagnosis result is continuously stored in the DB of the control server to be utilized as learning data for the machine learning algorithm 110 .

Here, the self-diagnosis unit 100 may be manufactured as an independent product by including the machine learning algorithm 110 and the display 120 . This will have the advantage that it can be marketed as a single unit because it can be simply attached to the leak alarm or control server of the conventional system as shown in FIG. 1 .

3 is an exemplary diagram for explaining the configuration and function of the self-diagnosis unit 100 based on machine learning premised in the present invention. The self-diagnosis unit 100 of FIG. 3 finds an AI prediction model through supervised learning in a state where the correct answer of alarm occurrence is given, and analyzes the characteristics of the new diagnosis data through the AI prediction model to determine the safety measures. designed to guide needs.

Machine learning consists of a machine learning algorithm 110, in which the system learns certain signals and patterns from data without human help based on mathematical optimization and statistical analysis techniques, predicts the next situation based on it, and makes appropriate decisions. However, since various open sources that are already widely used are available, there will be no special technical difficulties in implementing such an algorithm 110 . In particular, as is well known, the core of the AI prediction model generated by the machine learning algorithm 110 is the amount of data. It would be desirable to design the system of the present invention in such a way that the quality of the predictive model is improved by maximally utilizing the accumulated data.

4 illustrates a prediction model targeted in the present invention. The prediction model of FIG. 4 recognizes the output of various sensors and predicts whether the sensor needs zero adjustment or failure, a function to determine whether an alarm signal is generated when judging a gas leak, and a function to predict the replacement time of the gas sensor At least among the function, the function to present the desired correction value and the correction time of the initially set reference value in response to the aging curve of the gas sensor, and the function to determine the priority of the alarm (alarm) according to the severity of the gas leak situation It would be desirable to design to include more than one function. At this time, the machine learning algorithm 110 for deriving the AI prediction model may be installed in any of the self-diagnosis unit 100 of the leak alarm or the control server in consideration of the size of the gas facility or the installation environment of the system.

Since most gas facilities are exposed to harsh environments such as dust, moisture, corrosion, and high pressure, the reliability of the sensor signal for such environmental factors must be secured. Because signal distortion may occur, self-diagnosis of each sensor should be performed periodically. At this time, it would be desirable to set the self-diagnosis cycle to a long cycle such as monthly intervals in the initial stage of construction of gas facilities, and to design so that the cycle is gradually shortened as time passes by weekly or daily units.

If the result data of such periodic self-diagnosis is stored to be continuously accumulated in the database (DB) of the control server and the accumulated information is used as the learning data of the machine learning algorithm 110, as the amount of accumulated information increases, the machine learning algorithm ( 110) will have the effect of further improving the quality of the AI prediction model.

On the other hand, when there is a need to guide the safety measures of gas facilities, an auditory means such as a speaker is used for an alarm (alarm) with a high priority, and an alarm (alarm) with a low priority is sent to the self-diagnosis unit 100 . It would be desirable to design to guide the text message through the provided display 120 . In particular, high-priority alarms set a narrow reference value range so that a quick alarm sound is generated through an auditory means such as a speaker, and low-priority alarms set a wide reference value range, such as the display 120. If it is designed to be guided through visual means, it will be helpful in relieving the manager's fatigue and adverse functions such as relaxation of alertness due to excessive alarms or alarm (alarm) errors.

In addition, the system is designed to inform the administrator by periodically scanning the data stored in the DB to determine whether there are any unconfirmed alarm occurrences or whether there is a history of omitting alarm occurrences due to system errors even when an alarm (alarm) is required. This will contribute to enhancing reliability.

In addition, when judging whether the detection signal from the sensor deviates from the reference value (threshold) through the self-diagnosis function, the reference value can be automatically corrected in consideration of environmental factors such as daytime, nighttime, or season at the time of self-diagnosis. If designed to do so, it can be expected that the reliability of the gas leak detection system will be improved. For example, it would be a good example to measure the concentration of gas around a gas facility to calculate an average value, and to design a reference value to be automatically adjusted within a certain range of the average value.

5 is a flowchart for explaining the operation method of the gas leak detection system of the present invention step by step.

The present invention is premised on periodic self-diagnosis of the AI gas leak detection system. At this time, the self-diagnosis cycle is set to be long at the initial stage of construction in consideration of the aging curve of the gas facility, and to become shorter after a certain period of time has elapsed. For example, it is designed to set the diagnosis cycle on a monthly basis for about 1 year in the initial construction period, which is usually an AS period, and to shorten the cycle gradually, such as weekly after 1 year and daily after 2 years.

In order to perform the self-diagnosis of the gas leak system of the present invention, first, the self-diagnosis unit 100 recognizes whether a preset self-diagnosis period has been reached, and transmits a diagnostic signal for a simulation test to each sensor. In this case, the diagnostic signal for the simulation test uses a signal designed through a simulation process in advance.

Next, the self-diagnosis unit 100 receives a feedback signal generated in response to a diagnostic signal for a simulation test from each sensor, and compares the feedback signal with a pre-stored reference value. In this case, it is preferable to design the feedback signal of each sensor to be compared with the initial reference value at the beginning of construction, and to be compared with the corrected reference value after a certain period of time has elapsed.

In the next step, it is determined whether the feedback signal is out of the range of the reference value, and if it is within the range of the reference value, it is determined as normal, and the diagnosis result is stored in the database (DB). The diagnosis results stored in the DB are sequentially accumulated each time a diagnosis is performed, providing a basis that can be utilized as machine learning data. And if the feedback signal is out of the reference value range, it is determined to be abnormal and the corresponding feedback signal is input to the AI prediction model. After that, if the analysis result by the AI prediction model is predicted as a priority alarm target, an alarm is generated to inform that the operation of the gas leak detection system is abnormal. In this case, it would be desirable to use an auditory means such as a speaker or siren to inform the severity of the abnormality. In addition, if it is predicted that the analysis result is not a priority alarm target, the next step is to determine whether the sensor function is abnormal, and if there is an abnormality to the extent that the sensor function is out of the allowable limit, an alarm is generated.

On the other hand, if the abnormality of the sensor function is not beyond the permissible limit, if it is predicted as a simple alarm error, or the replacement time of the sensor or the time of correction of the reference value is predicted to be imminent, the prediction result is not generated by the manager A character guide is displayed on the display 120 so that the user can recognize it. In this way, if the alarm (alarm) method for the self-diagnosis result is operated in two ways. Unnecessary alarm occurrence can be reduced, and the inconvenience caused by an alarm (alarm) error or frequent alarm (alarm) can be eliminated, and it will have the advantage of preventing the relaxation of the manager's alertness.

According to the present invention as described above, it is possible to efficiently check all sensors in a system in which numerous sensors are installed and operated at all times, and the alarm system does not operate normally even when there is actually a gas leak, or an error caused by a malfunction even when there is no gas leak It can be expected that the administrator can quickly identify problems such as alarms.

In addition, it is possible to improve the efficiency of the management of the gas leak detection system because it predicts and guides the replacement cycle for a number of sensors in advance.

Although the preferred embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and any person skilled in the art may make various modifications within the scope without departing from the gist of the present invention. It will be possible. However, it should be noted that whether it departs from the gist of the present invention is determined solely by the description of the claims.

100: self-diagnosis unit
110: machine learning algorithm
120: display

Claims (10)

In the AI gas leak detection system comprising a plurality of gas leak detection sensors, gas leak alarms and alarm means installed in a gas use area and a gas supply area are connected to a control server through a communication network,
The control server further includes a self-diagnosis unit 100,
The self-diagnosis unit 100 is configured to include a database for storing self-diagnosis information, a machine learning algorithm 110 and a display 120,
The self-diagnosis unit 100 is configured to further include a transmission means for transmitting a diagnostic signal for a simulation test to each detection sensor, and a reception means for receiving a feedback signal corresponding to the diagnostic signal for a simulation test from each detection sensor. AI gas leak detection system featuring According to claim 1,
AI gas leak detection system, characterized in that the transmission period of the diagnostic signal for the simulation test is set variably according to the period of use of the gas facility According to claim 1,
AI gas leak detection system, characterized in that the machine learning algorithm 110 is configured to be installed in the gas leak alarm or control server 3. The method of claim 1 or 2,
AI gas leak detection system, characterized in that the machine learning algorithm 110 is configured to generate an AI prediction model by inputting accumulated information of self-diagnosis stored in the database 5. The method of claim 4,
The AI prediction model has a function of predicting whether each sensor needs zero adjustment or a failure, a function of determining whether an alarm signal is generated when a gas leak is determined, a function of predicting the replacement time of the gas sensor, and the aging curve of the gas sensor It is characterized in that it includes at least one function among a function of presenting a correction time and a desired correction value of the initially set reference value in response to the aging curve, and a function of determining the priority of an alarm (alarm) according to the severity of the gas leak situation. AI gas leak detection system A method of operating an AI gas leak detection system in which a plurality of gas leak detection sensors, gas leak alarms and alarm means installed in a gas use area and a gas supply area are connected to a control server through a communication network, the method comprising:
Recognizing whether a preset diagnostic cycle has been reached, and transmitting a diagnostic signal for a simulation test to each sensor;
Receiving a feedback signal generated in response to a diagnostic signal for a simulation test from each sensor;
comparing the feedback signal with a pre-stored reference value;
determining whether the feedback signal is out of the range of the reference value, determining whether the feedback signal is within the range of the reference value, determining that it is normal, and storing the diagnosis result in a database (DB);
When the feedback signal is out of the reference value range, determining that it is abnormal and inputting the corresponding feedback signal to the AI prediction model;
generating an alarm when abnormal judgment by the AI prediction model is predicted as a priority alarm target;
generating an alarm when the abnormal judgment by the AI prediction model is outside the allowable limit of the sensor function;
Operation of an AI gas leak detection system, characterized in that the step of displaying text guidance through the display 120 without generating an alarm if the abnormal determination by the AI prediction model is not beyond the allowable limit of the sensor function Way 7. The method of claim 6,
The operation method of the AI gas leak detection system, characterized in that the transmission of the diagnostic signal for the simulation test is set variably according to the period of use of the gas facility. 8. The method of claim 7,
The operating method of the AI gas leak detection system, characterized in that the diagnostic signal for the simulation test uses a signal of a predetermined form through a simulation process 7. The method of claim 6,
The AI prediction model has a function of predicting whether each sensor needs zero adjustment or a failure, a function of determining whether an alarm signal is generated when a gas leak is determined, a function of predicting the replacement time of the gas sensor, and the aging curve of the gas sensor Prediction result generated from at least one function among the function of presenting the correction time and the desired correction value of the initially set reference value in response to the aging curve, and the function of determining the priority of the alarm (alarm) according to the severity of the gas leak situation Operation method of AI gas leak detection system, characterized in that using 7. The method of claim 6,
The step of comparing the feedback signal with the pre-stored reference value comprises comparing the feedback signal of each sensor with an initial reference value at the initial stage of construction of the gas facility, and comparing it with a corrected reference value after a certain period of time has elapsed. operation method

Images