Mining fire monitoring method and system based on sound-light alarm

Abstract

The invention relates to the technical field of alarm detection, in particular to a mining fire monitoring method and system based on sound-light alarm. The method comprises the following steps: acquiring a historical time period of each acquisition moment of a mine environment, and dividing the historical time period into a left subset and a right subset corresponding to each dimension by using a data value of each dimension corresponding to each acquisition moment in the historical time period; obtaining the importance of each dimension at the acquisition moment through the data fluctuation difference in the left and right subsets of each dimension at the acquisition moment; calculating the selectable degree of each collection moment in the historical time period; and after the historical time period is segmented based on the data value of each dimension in the collection moment corresponding to the maximum value of the selectable degree to obtain a segmentation tree, the product of the abnormal score of each dimension at each collection moment calculated in the RRCF algorithm is recorded as the abnormal score of the collection moment, so that mining fire monitoring is realized. And the accuracy of mine fire monitoring alarm is effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of alarm detection, and in particular to a method and system for monitoring a mine fire based on sound and light alarms. Background Art

[0002] Mining can provide necessary raw materials for industrial production and is an important area of ​​energy and raw material supply. In the mining process, as the mining depth increases, the air circulation in the mine will gradually decrease. If a fire occurs, it will not only cause huge economic losses, but also affect the safety of the workers.

[0003] In order to reduce economic losses and ensure the safety of workers, many mine fire alarm methods are provided in the prior art. For example, the patent application document with publication number CN119028078A discloses a mine fire identification and alarm method based on flame and smoke images. This method installs a near-infrared camera underground to collect and identify the suspected fire image features in the monitored area in real time, and identifies and alarms mine fires by monitoring the irregular morphology of early smoke and mid- and late-stage flame combustion.

[0004] The above-mentioned existing technology can realize mine fire warning by collecting images in the mine. However, in the mine environment, there may be various influencing factors such as electromagnetic interference, dust, water vapor, etc. The image recognition method may be interfered by light, obstructions, dust, etc., resulting in low accuracy of image recognition fire warning.

[0005] Based on this, how to accurately realize mine fire warning is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0006] In order to solve the technical problem of how to accurately realize early warning of mine fire, the present invention provides a mine fire monitoring method and system based on sound and light alarm.

[0007] In a first aspect, the present invention provides a method for monitoring a mine fire based on an acoustic and visual alarm, which adopts the following technical solution: A mine fire monitoring method based on sound and light alarm includes the following steps: Obtain the historical period of each collection moment of the mine environment, and use the data values ​​of each dimension corresponding to each collection moment in the historical period to divide the historical period into two left and right subsets corresponding to each dimension; obtain the importance of each dimension at the collection moment through the data fluctuation difference in the left and right subsets of each dimension at the collection moment; ; is the selectability of the i-th collection moment in the historical period, , are the first Dimension, The importance of dimensions, is the length of the neighbor period, , are respectively the first The collection time Dimension, The data value of each dimension; after the historical period is segmented to obtain the segmentation tree based on the data value of each dimension in the collection moment corresponding to the maximum optional degree, the product of the anomaly scores of each dimension at each collection moment calculated in the RRCF algorithm is recorded as the anomaly score of the collection moment to realize mine fire monitoring.

[0008] The present invention calculates the abnormal score of each dimension in the collection time through the RRCF algorithm, and can accurately realize the mine fire monitoring alarm. In this process, the present invention takes into account that the RRCF algorithm randomly selects parameter data of any dimension as a segmentation point to construct a segmentation tree, which will cause the algorithm to be difficult to distinguish between normal fluctuations and data fluctuations caused by fire; based on this, the present invention uses each dimension of each collection time as a segmentation point for subset segmentation, and evaluates the segmentation effect of each segmentation point based on the data fluctuation difference between subsets, obtains the importance of the collection time, selects the segmentation point based on the importance of the collection time to construct a segmentation tree, and can accurately identify abnormal data in the mine environment parameters. On this basis, the present invention also takes into account that normal fluctuations will affect the calculation of the importance of the collection time. Therefore, the present invention corrects the importance of the collection time by obtaining the correlation between the two dimensions, and can accurately obtain the importance of the collection time, so that the monitoring alarm of the mine fire can be accurately realized based on the abnormal data, and the accuracy of the mine fire monitoring alarm is effectively improved.

[0009] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the historical time period of each collection moment of the mine environment is obtained, including: preprocessing after obtaining the mine environment parameters of each dimension at each collection moment to obtain an environmental data set; presetting the length of the historical time period, and obtaining the historical time period of the current collection moment in the environmental data set.

[0010] The present invention takes into account that the quality of the originally collected data is poor and the dimensions of data of different dimensions are different. Therefore, the dimensions are eliminated through preprocessing to improve the overall quality of the data and prepare for subsequent data processing.

[0011] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the data values ​​of each dimension corresponding to each collection moment in the historical time period are used to divide the historical time period into two left and right subsets corresponding to each dimension, including: using the data value of one of the dimensions corresponding to the collection moment to divide the historical time period, dividing the collection moments greater than the data value into the right subset of the dimension, and dividing the collection moments not greater than the data value into the left subset of the dimension.

[0012] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the importance of each dimension at the collection time is obtained by the data fluctuation difference in the left and right subsets of each dimension at the collection time, including: recording the absolute value accumulation sum of the difference between each data value in one of the subsets of the dimension at the collection time and the mean of the subset as the first feature of the subset, and recording the extreme difference value of the collection time in the subset as the second feature; normalizing the absolute value of the difference between the ratio of the first feature and the second feature of the left subset and the right subset of the dimension to obtain the importance of the dimension.

[0013] The present invention provides an accurate method for calculating the importance of each dimension at the time of collection. By analyzing the data fluctuation difference between the left and right subsets of the dimension division, the importance of each dimension can be accurately obtained. The greater the data fluctuation difference, the better the division effect of each dimension at the time of collection, and the higher the corresponding importance.

[0014] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the method for obtaining the adjacent time period of the acquisition time includes: presetting the adjacent time period length of the acquisition time , obtained on both sides at the time of collection other collection moments to construct the neighboring time periods of the collection moment.

[0015] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the historical time period is segmented based on the data values ​​of each dimension at the collection moment corresponding to the maximum value of the optional degree to obtain a segmentation tree, including: using the data value of one of the dimensions at the collection moment corresponding to the maximum value of the optional degree of the historical time period to divide the historical time period into a left subtree and a right subtree; selecting the data value of the dimension at the collection moment corresponding to the maximum value of the optional degree in the left subtree or the right subtree, and continuing to segment the left subtree or the right subtree, and repeating this cycle until a preset stop condition is reached to obtain a segmentation tree corresponding to the dimension.

[0016] The present invention selects the segmentation point according to the selectivity of the collection moment. Each time the segmentation point is selected, a collection moment with a higher selectivity is selected. The higher the selectivity, the smaller the segmentation effect of the corresponding collection moment, thereby reducing the influence of normal fluctuations of mine environmental parameters on abnormal data identification.

[0017] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the method for obtaining the abnormal score of each dimension at each collection moment includes: taking the sum of the path lengths of all collection moments in the historical period segmentation tree of the collection moment as the complexity of the segmentation tree; inserting the data value of each dimension at the collection moment into the corresponding segmentation tree to obtain the complexity of the segmentation tree after insertion; and recording the absolute value of the difference in the complexity of the segmentation tree before and after the insertion as the abnormal score of the collection moment in the dimension.

[0018] According to the mine fire monitoring method based on sound and light alarm provided by the present invention, the implementation of mine fire monitoring includes: if the abnormality score at the time of collection is greater than a preset threshold, the fire monitoring result is abnormal; otherwise, the fire monitoring result is normal; in response to the fire monitoring result being abnormal, an sound and light alarm is issued externally.

[0019] The present invention takes into account that an abnormality in the mine environment may cause huge economic losses and affect the safety of workers. Therefore, an audible and visual alarm is emitted to the outside to prompt the workers to handle the situation in time, thereby reducing safety hazards and losses.

[0020] In a second aspect, the present invention provides a mine fire monitoring system based on sound and light alarm, which adopts the following technical solution: The mine fire monitoring system based on sound and light alarm comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the mine fire monitoring method based on sound and light alarm is implemented.

[0021] By adopting the above technical solution, the above-mentioned mine fire monitoring method based on sound and light alarm is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is made according to the memory and the processor for easy use.

[0022] The present invention has the following technical effects: Based on the above technical solution, when realizing the monitoring and alarm of mine fire, the present invention calculates the abnormal score of each dimension in the collection time by the RRCF algorithm, and can accurately realize the monitoring and alarm of mine fire. In this process, the present invention takes into account that the RRCF algorithm randomly selects parameter data of any dimension as a segmentation point to construct a segmentation tree, which will cause the algorithm to be difficult to distinguish between normal fluctuations and data fluctuations caused by fire; based on this, the present invention uses each dimension of each collection moment as a segmentation point for subset segmentation, evaluates the segmentation effect of each segmentation point based on the data fluctuation difference between subsets, obtains the importance of the collection moment, and corrects the importance of the collection moment by obtaining the correlation between the two dimensions, so as to accurately obtain the importance of the collection moment, reduce the influence of normal data fluctuations in the mine environment on the identification of abnormal data, and thus can accurately realize the monitoring and alarm of mine fire based on abnormal data, effectively improving the accuracy of mine fire monitoring and alarm. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0024] Figure 1 A schematic flow chart of a method for monitoring fire conditions in a mine based on sound and light alarms is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0026] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0027] If a fire occurs during mining, it will not only cause huge economic losses, but also affect the safety of workers. Therefore, it is necessary to monitor and alarm the mine fire. In the mine environment, there may be various influencing factors such as electromagnetic interference, dust, and water vapor. The development of a fire usually includes the invisible smoke stage, the visible smoke stage, the visible flame stage, and the intense combustion stage. Image recognition methods can often only effectively identify fires in the visible smoke or flame stage, and at this time the fire may have been large, resulting in low accuracy of image recognition fire warning.

[0028] Robust Random Cut Forest (RRCF) is an algorithm that can realize multi-dimensional parameter anomaly identification. It constructs a segmentation tree by randomly selecting random parameters of data points in the data set as segmentation points, and then evaluates the anomaly score of the real-time data point through the complexity change caused by the insertion of the real-time data point.

[0029] Based on this, an embodiment of the present invention discloses a mine fire monitoring method based on sound and light alarm. The method divides the mine environmental parameter data into a segmentation tree to calculate the abnormality score at each collection moment, and obtains the fire monitoring situation based on the abnormality score and issues an alarm.

[0030] For details, please refer to Figure 1 As shown, Figure 1 A flow chart of a method for monitoring a mine fire based on sound and light alarms is provided in an embodiment of the present invention. The method specifically includes the following steps.

[0031] S1: Get the historical period for each acquisition moment in the mine environment.

[0032] The dimensions of the mine environment parameters may be the temperature and smoke concentration of the environment; they may be specifically set according to actual needs, and the embodiments of the present invention do not impose too many limitations thereon.

[0033] It should be noted that in the early stages of a mine fire, the temperature inside the mine will gradually rise due to the heating and decomposition of the burning materials. By monitoring the temperature changes, the development stage and trend of the fire can be determined. In addition, as the fire spreads and the intensity of combustion increases, the smoke concentration will continue to rise. Abnormal changes in mine environmental parameters may indicate potential fire risks.

[0034] Based on this, the embodiment of the present invention obtains temperature data and smoke concentration data in the mine environment for analysis, identifies abnormal environmental parameters based on their development and change trends, and thus accurately obtains fire monitoring results.

[0035] Specifically, after a high-sensitivity smoke concentration sensor and a temperature sensor are installed in the mine, the temperature data and smoke concentration data of the mine are collected at the same collection frequency and collection time, and the data of the two dimensions correspond one to one.

[0036] Among them, the collection frequency can be set to once every 3 seconds; the collection frequency and collection duration can be set according to actual needs.

[0037] It should be further explained that the conventional RRCF algorithm constructs a segmentation tree by randomly selecting parameter data as segmentation points, which makes it difficult for the algorithm to distinguish between normal fluctuations and fire conditions. In addition, the randomly selected parameter segmentation points may not effectively reduce the uncertainty of the segmentation tree, that is, if the data volatility in the segmented subsets is not much different, then the segmentation point is not an important segmentation point, resulting in the segmentation tree needing to be segmented multiple times and the segmentation effect is poor.

[0038] Based on this, the embodiment of the present invention divides each dimension at each collection moment as a segmentation point to obtain two subsets corresponding to each dimension at each collection moment, analyzes the data fluctuation differences between the subsets, and obtains the importance of the dimension.

[0039] It should be further explained that the RRCF algorithm defines anomalies based on the changes in the complexity of the segmentation tree caused by the insertion of real-time data points. If the insertion of real-time data points significantly increases the complexity of the segmentation tree, the data point is considered an anomaly, where the complexity of the segmentation tree is determined based on the depth of each data point on the segmentation tree. In this process, if a large number of real-time data points are inserted into the segmentation tree, the subtree nodes in the segmentation tree may split because they can no longer accommodate more data points, forming new subtree nodes, thereby increasing the number of subtree nodes, and the complexity of the segmentation tree will also increase, reducing the accuracy of the segmentation tree in determining abnormal data.

[0040] Based on this, the embodiment of the present invention obtains the historical period within the recent period of each collection time. In the case of no fire, the data change of the historical period is usually relatively stable. Based on the historical period within the recent period of the collection time, a segmentation tree is constructed to distinguish whether the real-time input data is an abnormal point, which can effectively improve the accuracy of identifying abnormal data.

[0041] By way of example, in an embodiment of the present invention, the historical time period of the current collection moment of the mine environment is obtained, including: obtaining the mine environment parameters of each dimension at each collection moment and performing preprocessing to obtain an environmental data set; presetting the length of the historical time period, and obtaining the historical time period of the current collection moment in the environmental data set.

[0042] Among them, the preprocessing method can be maximum and minimum normalization processing of each dimensional data, data denoising, missing data interpolation, etc., which can be specifically set according to actual needs, and the embodiment of the present invention does not impose too many restrictions here.

[0043] For example, the length of the historical period can be preset to 20. The length of the historical period is the number of collection moments included in the historical period. The length can be set according to actual needs. The historical period of the obtained collection moment does not include the collection moment itself.

[0044] It is understandable that when parameter data of different dimensions are processed simultaneously, deviations in the calculation results may occur. Therefore, the embodiment of the present invention converts the mine environment data into dimensionless values ​​through maximum and minimum normalization processing to eliminate the influence of the dimension. In addition, there may be noise interference or other factors in the process of collecting mine environment data, resulting in low accuracy of the collected data and data missing. Therefore, the embodiment of the present invention processes the mine environment data through methods such as data denoising and missing data interpolation, which can effectively improve the overall quality of the data and prepare for subsequent data processing.

[0045] After obtaining the historical time periods of each collection moment based on the above steps, continue to perform the following steps.

[0046] S2: Use the data values ​​of each dimension corresponding to each collection moment in the historical period to divide the historical period into two left and right subsets corresponding to each dimension; obtain the importance of each dimension at the collection moment through the data fluctuation difference in the left and right subsets of each dimension at the collection moment.

[0047] By way of example, in an embodiment of the present invention, the historical period is divided into two left and right subsets corresponding to each dimension using the data values ​​of each dimension corresponding to each collection moment in the historical period, including: dividing the historical period using the data value of one of the dimensions corresponding to the collection moment, dividing the collection moments greater than the data value into the right subset of the dimension, and dividing the collection moments not greater than the data value into the left subset of the dimension.

[0048] It should be noted that by analyzing the fluctuation difference between the left and right subsets obtained by dividing the historical period of the collection period by the dimension of each collection moment, the importance of each dimension in the collection moment can be accurately obtained.

[0049] By way of example, in an embodiment of the present invention, the importance of each dimension at the collection moment is obtained by the data fluctuation difference in the left and right subsets of each dimension at the collection moment, including: recording the absolute value accumulation of the difference between each data value in one of the subsets of the dimension at the collection moment and the mean of the subset as the first feature of the subset, and recording the collection moment extreme difference value in the subset as the second feature; normalizing the absolute value of the difference between the ratio of the first feature and the second feature of the left subset and the right subset of the dimension to obtain the importance of the dimension.

[0050] The data values ​​in the subset obtained after each dimension is divided are the same dimension as the dimension. Each collection moment corresponds to the data values ​​of two dimensions, so the number of data values ​​in the dimension subset is the number of collection moments in the subset.

[0051] Take an example to illustrate the subset division process: select the temperature dimension in the current collection moment to divide the historical period of the current collection moment, divide the temperature values ​​in each collection moment in the historical period that are greater than the temperature value at the current collection moment into the right subset of the temperature dimension in the current collection moment, and divide the temperature values ​​in the historical period that are not greater than the temperature value at the current collection moment into the left subset of the temperature dimension in the current collection moment. The subsets obtained after the temperature dimension division all contain the temperature values ​​corresponding to each collection moment.

[0052] For ease of understanding, the embodiment of the present invention is described by taking the calculation of the importance of one dimension in the collection time as an example, but it does not mean that the embodiment of the present invention is limited to this.

[0053] For example, in the embodiment of the present invention, the first The importance of dimensions can be found in the following relationship: ; is the first The importance of dimensions, is the first The number of data values ​​in the left subset of the dimension, is the first The j-th data value in the left subset of the dimension, is the first The left subset mean of the dimension, is the first The extreme value of the collection time corresponding to the temperature value in the left subset of the dimension, is the first The number of data values ​​in the right subset of the dimension, is the first The first data values, is the first The right subset mean of the dimension, is the first The extreme value of the collection time corresponding to the temperature value in the right subset of the dimension, is the standard normalization function, is the absolute value symbol.

[0054] In the above formula, is the first characteristic of the left subset, is the second characteristic of the left subset, is the first characteristic of the right subset, is the second characteristic of the right subset.

[0055] It is the ratio of the first feature to the second feature of the left subset. The larger the value is, the greater the difference between each data value in the left subset and the mean and the smaller the range of the collection time corresponding to all data is. The more drastic the change of the data in the left subset in a shorter time span, the mine environment may have changed significantly, and the greater the data fluctuation in the left subset.

[0056] Similarly, It is the ratio of the first feature to the second feature of the right subset. The larger the value, the greater the difference between each data value and the mean in the right subset and the smaller the range of the collection time corresponding to all data. The more drastic the change of the data in the right subset in a shorter time span, the greater the corresponding data fluctuation.

[0057] is the first The data fluctuation difference between the left and right subsets of the dimension. The larger the value, the greater the data fluctuation difference between the left and right subsets, and the greater the data performance difference, that is, the data performance in one subset is relatively stable, while the data performance in the other subset is more significant. The dimension divides the historical period of the i-th collection time into two different trend expressions. The higher the importance of the dimension.

[0058] Based on the above steps, we can obtain the The importance of the dimension and the After you determine the importance of the dimension, continue with the following steps.

[0059] S3: Calculate the selectivity of each collection moment in the historical period.

[0060] It should be noted that the embodiment of the present invention can obtain the importance of each dimension at each collection moment by analyzing the fluctuation degree between the subsets divided by the data of each dimension as the segmentation point based on the above steps, and the importance of the current collection moment can be obtained by combining the importance of each dimension.

[0061] When a fire occurs, the temperature will gradually increase and the smoke density will increase synchronously. However, the temperature may also rise in mine operations, such as heating caused by long-term operation of machinery, blasting operations, etc. Obviously, such temperature changes are normal changes, not temperature changes caused by fire, so the smoke density is still in normal fluctuations. If the temperature is abnormal but the smoke density is normal at the collection time randomly selected by the RRCF algorithm, the algorithm will overestimate the abnormality of the current collection time, resulting in misjudgment.

[0062] Based on this, the embodiment of the present invention can obtain the selectivity of the current collection moment as a segmentation point by further obtaining the correlation between the changes in temperature data and smoke density data at the current collection moment.

[0063] It is understandable that only the data values ​​of each dimension corresponding to the current collection moment cannot accurately reflect the correlation of changes between dimensional data. Therefore, an embodiment of the present invention obtains the correlation of dimensional changes in the adjacent time period of the current collection moment by acquiring the data changes in the adjacent time period of the current collection moment.

[0064] For example, in an embodiment of the present invention, a method for obtaining a time period adjacent to a collection time includes: presetting a length of a time period adjacent to a collection time , obtained on both sides at the time of collection other collection moments to construct the neighboring time periods of the collection moment.

[0065] The length of the neighbor period can be set to 7, which can be set according to actual needs, and the embodiment of the present invention does not impose too many restrictions on this.

[0066] Specifically, if the number of collection moments on one side of the current collection moment is insufficient, it can be supplemented on the other side, and the final neighboring time period includes the current collection moment itself.

[0067] For example, in the embodiment of the present invention, the selectivity of each collection time in the historical period is calculated, and the specific details can be referred to the following relationship: ; is the selectability of the i-th collection moment in the historical period, is the first The importance of dimensions, is the first The importance of dimensions, is the length of the neighbor period, is the number of neighboring time periods at the i-th collection time. The collection time The data value of the dimension, is the number of neighboring time periods at the i-th collection time. The collection time The data value of the dimension.

[0068] In the above formula, Indicates the first Dimensions and The importance product of the dimension. The larger the value, the more Dimensions and The better the division effect of the dimension as a segmentation point, the more obvious the change characteristics between the obtained subsets, the higher the importance of the corresponding i-th collection moment, and the higher the degree of selectivity.

[0069] It represents the standard deviation between the parameters of the two dimensions in the adjacent time period of the i-th collection moment. The smaller the value, the stronger the correlation between the change patterns of the parameters of the two dimensions in the adjacent time period, and the consistent change trends of temperature and smoke concentration. Therefore, the i-th collection moment is more important for the abnormal fire monitoring results, and the corresponding selectivity is also higher.

[0070] It is understandable that the mine environmental parameters have two dimensions, namely temperature and smoke concentration. is the temperature, then is the smoke concentration; if is the smoke concentration, then For temperature.

[0071] After obtaining the selectivity of each acquisition time based on the above formula, continue to perform the following steps.

[0072] S4: After the historical period is segmented to obtain the segmentation tree based on the data values ​​of each dimension in the collection moment corresponding to the maximum optional degree, the product of the anomaly scores of each dimension at each collection moment calculated in the RRCF algorithm is recorded as the anomaly score of the collection moment.

[0073] It should be noted that after obtaining the selectivity corresponding to each collection moment based on the above steps, the dimensions in the collection moment can be selected based on the selectivity corresponding to each collection moment, and the segmentation trees corresponding to each dimension can be constructed respectively. The anomaly scores of each collection moment in each dimension segmentation tree can be combined to obtain the anomaly scores of each collection moment.

[0074] By way of example, in an embodiment of the present invention, a historical period is segmented based on the data values ​​of each dimension at the collection moment corresponding to the maximum value of the optional degree to obtain a segmentation tree, including: using the data value of one of the dimensions at the collection moment corresponding to the maximum value of the optional degree of the historical period to divide the historical period into a left subtree and a right subtree; selecting the data value of the dimension at the collection moment corresponding to the maximum value of the optional degree in the left subtree or the right subtree, and continuing to segment the left subtree or the right subtree, and repeating this cycle until a preset stop condition is reached to obtain a segmentation tree corresponding to the dimension.

[0075] Among them, the preset stop condition can be a preset number of segmentation times or the number of collection moments in the subtree is less than a preset number threshold; the values ​​of the preset number of segmentation times and the preset number threshold can be set according to actual needs, and the embodiment of the present invention does not impose too many restrictions on this.

[0076] Specifically, when the historical period is divided into a left subtree and a right subtree using the data value of one dimension of the collection time corresponding to the maximum optional degree value, the collection time greater than the data value can be divided into the right subtree of the dimension, and the collection time not greater than the data value can be divided into the left subtree of the dimension.

[0077] After obtaining the segmentation trees corresponding to the two dimensions based on the above method, the anomaly scores in the two segmentation trees at each acquisition time can be calculated based on the following steps.

[0078] By way of example, in an embodiment of the present invention, a method for obtaining anomaly scores of each dimension at each collection moment includes: taking the sum of the path lengths of all collection moments in a historical period segmentation tree of the collection moment as the complexity of the segmentation tree; inserting the data values ​​of each dimension at the collection moment into the corresponding segmentation tree to obtain the complexity of the segmentation tree after insertion; and recording the absolute value of the difference in complexity of the segmentation tree before and after insertion as the anomaly score of the collection moment in the dimension.

[0079] The specific steps of calculating the complexity of each segmentation tree can be obtained through the RRCF algorithm, which will not be described in detail in the embodiment of the present invention.

[0080] Take an example to illustrate the steps of obtaining the abnormal score in the temperature dimension at the current collection time: calculate the complexity of the temperature segmentation tree of the historical period at the current collection time, insert the temperature value at the current collection time into the temperature segmentation tree, obtain the complexity of the temperature segmentation tree after the insertion at the current collection time, and use the absolute value of the difference in complexity between the temperature segmentation tree before and after the insertion as the abnormal score in the temperature dimension at the current collection time.

[0081] Based on the above steps, the anomaly score of each dimension at each collection moment can be obtained. After the anomaly score of each collection moment is obtained by multiplying the anomaly score of each dimension at each collection moment, the following steps are continued.

[0082] S5: Based on the comparison result between the abnormal score at the time of collection and the preset threshold, the mine fire monitoring is realized.

[0083] By way of example, in an embodiment of the present invention, mine fire monitoring is implemented based on the comparison result between the abnormality score at the time of collection and a preset threshold, including: if the abnormality score at the time of collection is greater than the preset threshold, the fire monitoring result is abnormal; otherwise, the fire monitoring result is normal; in response to the fire monitoring result being abnormal, an audible and visual alarm is issued externally.

[0084] The preset threshold may be set to 0.7; the preset threshold may be set specifically according to actual needs, and the embodiment of the present invention does not impose too many restrictions on this.

[0085] It is understandable that if the abnormality score at the collection moment is greater than the preset threshold, it means that the temperature and smoke concentration at the current collection moment are abnormal, and there may be a fire. Therefore, an audible and visual alarm can be issued to facilitate timely processing by staff.

[0086] It can be seen that in the embodiment of the present invention, when realizing the mine fire monitoring based on sound and light alarm, the historical period of each collection moment of the mine environment can be obtained, and the historical period is divided into two left and right subsets corresponding to each dimension using the data value of each dimension corresponding to each collection moment in the historical period; the importance of each dimension at the collection moment is obtained through the data fluctuation difference in the left and right subsets of each dimension at the collection moment; ; is the selectability of the i-th collection moment in the historical period, , are the first Dimension, The importance of dimensions, is the length of the neighbor period, , are respectively the first The collection time Dimension, The data value of each dimension; after the historical period is segmented to obtain the segmentation tree based on the data value of each dimension in the collection moment corresponding to the maximum optional degree, the product of the anomaly scores of each dimension at each collection moment calculated in the RRCF algorithm is recorded as the anomaly score of the collection moment to realize mine fire monitoring.

[0087] In this way, the embodiment of the present invention calculates the abnormal score of each dimension in the collection time through the RRCF algorithm, and can accurately realize the mine fire monitoring alarm. In this process, the embodiment of the present invention takes into account that the RRCF algorithm randomly selects the parameter data of any dimension as the segmentation point to construct the segmentation tree, which will make it difficult for the algorithm to distinguish between normal fluctuations and data fluctuations caused by fire; based on this, the embodiment of the present invention uses each dimension of each collection time as a segmentation point for subset segmentation, and evaluates the segmentation effect of each segmentation point based on the data fluctuation difference between the subsets to obtain the importance of the collection time, and selects the segmentation point based on the importance of the collection time to construct the segmentation tree, which can accurately identify the abnormal data in the mine environmental parameters. On this basis, the embodiment of the present invention also takes into account that normal fluctuations will affect the calculation of the importance of the collection time. Therefore, the embodiment of the present invention corrects the importance of the collection time by obtaining the correlation between the two dimensions, and can accurately obtain the importance of the collection time, so that the monitoring alarm of the mine fire can be accurately realized based on the abnormal data, effectively improving the accuracy of the mine fire monitoring alarm.

[0088] An embodiment of the present invention also discloses a mine fire monitoring system based on sound and light alarms, including a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a mine fire monitoring method based on sound and light alarms provided by the present invention is implemented.

[0089] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0090] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM, a dynamic random access memory DRAM, a static random access memory SRAM, an enhanced dynamic random access memory EDRAM, a high bandwidth memory HBM, a hybrid memory cube HMC, etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0091] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0092] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A mine fire monitoring method based on sound and light alarm, characterized in that: include: Obtain the historical period of each collection moment of the mine environment, and use the data values ​​of each dimension corresponding to each collection moment in the historical period to divide the historical period into two left and right subsets corresponding to each dimension; The importance of each dimension at the time of collection is obtained by the data fluctuation difference between the left and right subsets of each dimension at the time of collection; ； is the selectability of the i-th collection moment in the historical period, , are the first Dimension, The importance of dimensions, is the length of the neighbor period, , are respectively the first The collection time Dimension, The data value of the dimension; After the historical period is segmented to obtain the segmentation tree based on the data values ​​of each dimension in the collection moment corresponding to the maximum optional degree, the product of the anomaly scores of each collection moment in each dimension calculated in the RRCF algorithm is recorded as the anomaly score of the collection moment to realize mine fire monitoring.

2. The method for monitoring fire conditions in a mine based on sound and light alarm according to claim 1, characterized in that: The historical period of obtaining each acquisition moment of the mine environment includes: After obtaining the mine environmental parameters of each dimension at each collection moment, preprocessing is performed to obtain an environmental data set; the length of the historical period is preset, and the historical period of the current collection moment is obtained in the environmental data set.

3. The method for monitoring fire in a mine based on sound and light alarm according to claim 1, characterized in that: The historical period is divided into two left and right subsets corresponding to each dimension using the data values ​​of each dimension corresponding to each collection moment in the historical period, including: The historical period is divided using the data value of one of the dimensions corresponding to the collection time. The collection time greater than the data value is divided into the right subset of the dimension, and the collection time not greater than the data value is divided into the left subset of the dimension.

4. The method for monitoring fire conditions in a mine based on sound and light alarm according to claim 1, characterized in that: The importance of each dimension at the collection time is obtained by using the data fluctuation difference in the left and right subsets of each dimension at the collection time, including: The absolute value cumulative sum of the difference between each data value in one subset of the dimension of the collection time and the mean of the subset is recorded as the first feature of the subset, and the collection time extreme difference value in the subset is recorded as the second feature; The absolute value of the difference between the ratio of the first feature and the second feature of the left subset and the right subset of the dimension is normalized to obtain the importance of the dimension.

5. The method for monitoring fire conditions in a mine based on sound and light alarm according to claim 1, characterized in that: The method for obtaining the adjacent time period of the collection moment includes: The length of the neighboring period of the preset collection time , obtained on both sides at the time of collection other collection moments to construct the neighboring time periods of the collection moment.

6. The method for monitoring fire conditions in a mine based on sound and light alarm according to claim 1, characterized in that: The segmentation tree is obtained by segmenting the historical period based on the data value of each dimension at the collection time corresponding to the maximum value of the optional degree, including: Use the data value of one of the dimensions at the collection time corresponding to the maximum value of the optional degree of the historical period to divide the historical period into a left subtree and a right subtree; select the data value of the dimension at the collection time corresponding to the maximum value of the optional degree in the left subtree or the right subtree, and continue to split the left subtree or the right subtree, and repeat this cycle until the preset stop condition is reached to obtain the segmentation tree corresponding to the dimension.

7. The method for monitoring fire conditions in a mine based on sound and light alarm according to claim 6, characterized in that: The method for obtaining the anomaly score in each dimension at each collection moment includes: The sum of the path lengths of all collection moments in the historical period segmentation tree of the collection moment is taken as the complexity of the segmentation tree; the data values ​​of each dimension at the collection moment are inserted into the corresponding segmentation tree to obtain the complexity of the segmentation tree after insertion; The absolute value of the difference in complexity of the partition tree before and after insertion is recorded as the anomaly score of the dimension at the time of collection.

8. The method for monitoring fire in a mine based on sound and light alarm according to claim 1, characterized in that: The implementation of mine fire monitoring includes: If the abnormality score at the time of collection is greater than a preset threshold, the fire monitoring result is abnormal; otherwise, the fire monitoring result is normal; in response to the fire monitoring result being abnormal, an audible and visual alarm is issued.

9. A mine fire monitoring system based on sound and light alarm, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a mine fire monitoring method based on sound and light alarm according to any one of claims 1 to 8 is implemented.

Images