CN-122020356-A - Coal mine safety monitoring alarm studying and judging method and system based on data fusion

CN122020356ACN 122020356 ACN122020356 ACN 122020356ACN-122020356-A

Abstract

The application provides a coal mine safety monitoring alarm studying and judging method and system based on data fusion, wherein the method comprises the steps of establishing an identification framework for representing an alarm cause; the method comprises the steps of determining first supporting probabilities of various alarm factors in an identification frame according to data of an alarm sensor, determining second supporting probabilities of various alarm factors in the identification frame according to data of the same type of sensors located at the upstream and downstream of the alarm sensor, determining third supporting probabilities of various alarm factors in the identification frame according to data of different types of sensors located in the same area as the alarm sensor, fusing the first supporting probabilities, the second supporting probabilities and the third supporting probabilities of the various alarm factors based on an evidence theory, obtaining occurrence probabilities representing the various alarm factors, and judging the causes of the alarm event according to the occurrence probabilities. According to the technical scheme provided by the application, the accuracy and reliability of alarm reason identification are obviously improved while the research and judgment efficiency is improved.

Inventors

FENG YING
HUA DONG
MA JIAN
ZHANG DESHENG
SHAO TIANTIAN
HUANG SIQI
HUANG ZENGBO
JIA XIAODI
WANG XIN
DAI WANBO

Assignees

煤炭科学技术研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (10)

1. The coal mine safety monitoring alarm studying and judging method based on data fusion is characterized by comprising the following steps of: establishing an identification framework for representing the alarming cause; acquiring self data of an alarm sensor corresponding to an alarm event, sensor data of the same type positioned at the upstream and downstream of the alarm sensor and sensor data of different types positioned in the same area as the alarm sensor; Determining a first support probability of various alarm causes in the identification framework according to the data of the alarm sensor; Determining a second supporting probability of various alarm causes in the identification frame according to the sensor data of the same type positioned at the upstream and downstream of the alarm sensor; Determining a third supporting probability of various alarm causes in the identification frame according to different types of sensor data in the same area as the alarm sensor; And fusing the first support probability, the second support probability and the third support probability of the various alarm causes based on the evidence theory to obtain the occurrence probability for representing the various alarm causes, and studying and judging the causes of the alarm event according to the occurrence probability.
2. The method of claim 1, wherein identifying each type of alarm cause in the framework comprises: True alarm, false alarm caused by calibration process, false alarm caused by sensor interference or fault.
3. The method of claim 2, wherein said determining a first supporting probability of each type of alarm cause in said identification framework based on said alarm sensor's own data comprises: Judging whether the data change process of the alarm sensor is normal or not based on the self data of the alarm sensor at each moment in the history period; when the data change process of the alarm sensor is abnormal, judging that the alarm cause of the alarm sensor is false alarm caused by the sensor being interfered or failed, wherein the first support probability corresponding to the true alarm is 0, the first support probability corresponding to the false alarm caused by the calibration process is 0, and the first support probability corresponding to the false alarm caused by the sensor being interfered or failed is 1; When the data change process of the alarm sensor is normal, judging a preliminary alarm cause of the alarm sensor according to the preset standard gas concentration for sensor calibration, and determining first support probabilities of various alarm causes in the identification frame according to the preliminary alarm cause of the alarm sensor.
4. A method according to claim 3, wherein said determining whether the data change process of the alarm sensor is normal based on the own data comprises: Determining the maximum value and the minimum value of the self data of the alarm sensor in the historical working period based on the self data of the alarm sensor at each moment in the historical period; Searching self data rising time periods and self data falling time periods based on self data at each moment in the history time period, searching first self data corresponding to a difference minimum value of self data maximum values in the history working time period in the self data rising time periods, determining time corresponding to the first self data, searching second self data corresponding to a difference minimum value of self data maximum values in the history working time period in the self data falling time periods, and determining time corresponding to the second self data; Taking a time period formed by the time corresponding to the first self data and the time corresponding to the second self data as a time period in which abnormal data exist; Judging whether the absolute value of the difference value of the self data at any two adjacent moments is greater than or equal to 90% (X max -x max ) in the time period when the abnormal data exist, if yes, judging that the data change process of the alarm sensor is abnormal, wherein the alarm cause of the alarm sensor is false alarm caused by interference or faults of the sensor, the first support probability corresponding to real alarm is 0, the first support probability corresponding to false alarm caused by calibration process is 0, the first support probability corresponding to false alarm caused by interference or faults of the sensor is 1, otherwise, judging that the data change process of the alarm sensor is normal, X max is the maximum value of the alarm sensor in the time period when the abnormal data exist, and X max is the maximum value of the self data of the alarm sensor in the historical working time period.
5. The method of claim 4, wherein said determining the preliminary alarm cause of the alarm sensor based on the preset standard gas concentration for sensor calibration, and determining the first support probability of each type of alarm cause in the identification framework based on the preliminary alarm cause of the alarm sensor, comprises: Judging whether rho (1-10%) < X max ++rho (1+10%) and t q -t p +.60deg.S are satisfied, if satisfied, judging that the preliminary alarm of the alarm sensor is a false alarm caused by the calibration process and/or a real alarm, wherein the first support probability corresponding to the real alarm is 0.2, the first support probability corresponding to the false alarm caused by the calibration process is 0.8, the first support probability corresponding to the false alarm caused by the sensor being interfered or failed is 0, if not satisfied, judging that the preliminary alarm cause of the alarm sensor is a real alarm, the first support probability corresponding to the real alarm is 1, the first support probability corresponding to the false alarm caused by the calibration process is 0, and the first support probability corresponding to the false alarm caused by the sensor being interfered or failed is 0; Wherein ρ is a preset standard gas concentration for calibrating the sensor, t q is a time corresponding to less than X max X95% for the first time in the data falling process in the self data in the history period, and t p is a time corresponding to more than or equal to X max X95% for the first time in the data rising process in the self data in the history period.
6. The method of claim 5, wherein said determining a second supporting probability of each type of alarm cause in the identification framework based on the same type of sensor data located upstream and downstream of the alarm sensor comprises: analyzing the same type sensor data at each moment in a first preset time period in the same type sensor data at the upstream and downstream of the alarm sensor, judging whether calibration data exist in the same type sensor data at each moment in the first preset time period, if so, the second support probability corresponding to real alarm is 0, the second support probability corresponding to false alarm caused by the calibration process is 1, the second support probability corresponding to false alarm caused by the sensor interfered or fault is 0, if not, the second support probability corresponding to real alarm is equal to the first support probability corresponding to real alarm, the second support probability corresponding to false alarm caused by the calibration process is equal to the first support probability corresponding to false alarm caused by the calibration process, and the second support probability corresponding to false alarm caused by the sensor interfered or fault is equal to the first support probability corresponding to false alarm caused by the sensor interfered or fault.
7. The method of claim 6, wherein the determining a third support probability for each type of alarm cause in the identification framework based on different types of sensor data in the same area as the alarm sensor comprises: When the alarm sensor is a methane sensor, judging whether the air volume sensor data of the air duct is abnormal or not based on different types of sensor data in the same area as the alarm sensor, if so, judging that the third support probability corresponding to the real alarm is 1, the third support probability corresponding to the false alarm caused by the calibration process is 0, the third support probability corresponding to the false alarm caused by the sensor interfered or fault is 0, otherwise, the third support probability corresponding to the real alarm is equal to the first support probability corresponding to the real alarm, the third support probability corresponding to the false alarm caused by the calibration process is equal to the first support probability corresponding to the false alarm caused by the calibration process, and the third support probability corresponding to the false alarm caused by the sensor interfered or fault is equal to the first support probability corresponding to the false alarm caused by the sensor interfered or fault; When the alarm sensor is a carbon monoxide sensor, judging whether the temperature sensor data and/or the smoke sensor data are abnormal based on different types of sensor data in the same area as the alarm sensor, if so, determining that the third support probability corresponding to the real alarm is 1, the third support probability corresponding to the false alarm caused by the calibration process is 0, the third support probability corresponding to the false alarm caused by the sensor interfered or failed is 0, otherwise, determining that the third support probability corresponding to the real alarm is equal to the first support probability corresponding to the real alarm, determining that the third support probability corresponding to the false alarm caused by the calibration process is equal to the first support probability corresponding to the false alarm caused by the calibration process, and determining that the third support probability corresponding to the false alarm caused by the sensor interfered or failed is equal to the first support probability corresponding to the false alarm caused by the sensor interfered or failed.
8. The method of claim 7, wherein the fusing the first support probability, the second support probability, and the third support probability for each type of alarm cause based on evidence theory to obtain the occurrence probability characterizing each type of alarm cause comprises: using the formula Determining the occurrence probability of real alarm; using the formula Determining the occurrence probability of false alarm caused by the calibration process; using the formula Determining the occurrence probability of false alarm caused by interference or faults of the sensor; Wherein, the For the probability of occurrence of a true alarm, For a first support probability corresponding to a true alarm, A second support probability corresponding to a true alarm, A third support probability for a true alarm correspondence, To calibrate the probability of false alarm occurrence caused by the process, For a first support probability corresponding to a false alarm caused by the calibration process, For a second support probability corresponding to a false alarm caused by the calibration process, For a third supported probability of false alarm correspondence caused by the calibration process, For the probability of false alarm occurrence caused by a disturbance or failure of the sensor, For a first support probability corresponding to false alarms caused by disturbances or faults to the sensor, For a second support probability corresponding to false alarms caused by disturbances or faults to the sensor, For a third support probability corresponding to false alarms caused by disturbances or faults in the sensor, As a first parameter of the first set of parameters, As a second parameter, the first parameter is, , 。
9. The method of claim 8, wherein the determining the cause of the alarm event based on the probability of occurrence comprises: sequencing the real alarm occurrence probability, the false alarm occurrence probability caused by the calibration process and the false alarm occurrence probability caused by the sensor interference or fault from left to right from large to small to form a first sequence, and taking a cause corresponding to a first value at the left side in the first sequence as a cause of an alarm event; When the probability of occurrence of the real alarm, the probability of occurrence of the false alarm caused by the calibration process and the probability of occurrence of the false alarm caused by the sensor interference or the fault are equal, updating the probability of occurrence of the real alarm to be equal to the first support probability corresponding to the real alarm, the probability of occurrence of the false alarm caused by the calibration process to be equal to the first support probability corresponding to the false alarm caused by the calibration process, and the probability of occurrence of the false alarm caused by the sensor interference or the fault to be equal to the first support probability corresponding to the false alarm caused by the sensor interference or the fault.
10. The utility model provides a colliery safety monitoring warning research judgement system based on data fusion which characterized in that, the system includes: the building module is used for building an identification framework for representing the alarming cause; the acquisition module is used for acquiring the data of the alarm sensor corresponding to the alarm event, the data of the same type of sensor positioned at the upstream and downstream of the alarm sensor and the data of different types of sensor positioned in the same area with the alarm sensor; the first determining module is used for determining first supporting probabilities of various alarm causes in the identification frame according to the data of the alarm sensor; The second determining module is used for determining second supporting probabilities of various alarm causes in the identification frame according to the sensor data of the same type positioned at the upstream and downstream of the alarm sensor; the third determining module is used for determining third supporting probabilities of various alarm causes in the identification frame according to different types of sensor data in the same area as the alarm sensor; and the judging module is used for fusing the first supporting probability, the second supporting probability and the third supporting probability of various alarm causes based on the evidence theory to obtain the occurrence probability for representing various alarm causes, and judging the causes of the alarm event according to the occurrence probability.

Description

Coal mine safety monitoring alarm studying and judging method and system based on data fusion Technical Field The application relates to the technical field of coal mine safety, in particular to a coal mine safety monitoring alarm studying and judging method and system based on data fusion. Background Coal mine safety is a major concern in mine production. Complex safety monitoring systems are commonly deployed in modern coal mines, and all-weather real-time monitoring is carried out on dangerous gas concentration and environmental parameters through various sensors (such as methane, carbon monoxide, wind speed, temperature sensors and the like) distributed underground to form massive monitoring data. These systems have become a critical technical barrier to the prevention of major accidents such as gas explosion, fire, asphyxia, etc. In actual operation, the existing coal mine safety monitoring system has the following outstanding technical problems, namely, the early warning efficiency and the automation level are seriously restricted, 1. False alarm frequently occurs, the underground production environment is extremely complex, and various adverse factors such as high humidity, high dust, electromagnetic interference, equipment vibration and the like exist. The sensor itself may generate an abnormal signal due to aging, malfunction or instantaneous strong interference. In addition, the periodic calibration (calibration) of the sensor according to the safety regulations also generates data meeting alarm thresholds in the system, forming a large number of unrealistic "process alarms". These "false alarms" triggered by environmental disturbances, equipment failures and calibration operations are mixed with "true alarms" generated by true dangerous symptoms, forming a massive, true-false, and indistinguishable alarm information stream. 2. The research and judgment is highly dependent on manual work, and the efficiency and the instantaneity are low, so that the screening, the cause analysis and the verification of alarm information are mainly carried out by the manual experience of the on-duty personnel of the monitoring center at present. The manual screening workload is huge and the research and judgment efficiency is extremely low in the face of hundreds of thousands to millions of alarm data possibly generated every day. The mode is difficult to meet the strict requirement of coal mine safety monitoring on real-time response, response delay to real dangerous situations is extremely easy to cause, or important police situations are ignored by operators on duty due to alarm fatigue, so that huge potential safety hazards exist. 3. The intelligent multisource information collaborative analysis capability is lacking, namely the existing system collects multiple sensor data, but alarm judgment is usually based on threshold value out-of-limit of a single sensor and belongs to single-point judgment. When the sensor data is abnormal, the system lacks an effective mechanism to automatically call and comprehensively analyze the data of the sensors of the same type at the upstream and downstream of the sensor to verify trend, and fails to cooperatively verify the data of the sensors of different types (such as a methane sensor and an air quantity sensor, a carbon monoxide sensor and a temperature and smoke sensor) with physical causal logic in the same area. The data island phenomenon ensures that the system cannot carry out cross verification and comprehensive research and judgment on alarm events from multiple angles and multiple layers, so that the alarm reason identification accuracy is low, and reliable automatic support cannot be provided for emergency decision. Therefore, a scheme capable of automatically, rapidly and accurately performing intelligent research and judgment on the alarm information of the coal mine safety monitoring system is needed, so that the technical defects of dependence on manpower, low efficiency, high false alarm rate and lack of multi-source collaborative analysis are overcome, and the intelligent level and the early warning reliability of coal mine safety monitoring are truly improved. Disclosure of Invention The application provides a coal mine safety monitoring alarm studying and judging method and system based on data fusion, which at least solve the technical problems of dependence on manpower, low efficiency, high false alarm rate and lack of multi-source collaborative analysis in the prior art. An embodiment of a first aspect of the present application provides a coal mine safety monitoring alarm studying and judging method based on data fusion, the method comprising: establishing an identification framework for representing the alarming cause; acquiring self data of an alarm sensor corresponding to an alarm event, sensor data of the same type positioned at the upstream and downstream of the alarm sensor and sensor data of different types positioned in the same area as the alar