Mine sensor array self-calibration method and system for gravel curtain layer wind field monitoring

Abstract

The application discloses a mine sensor array self-calibration method and system for gravel curtain layer wind field monitoring, and particularly relates to the technical field of sensor measurement data calibration, and is used for solving the problem that the existing self-calibration method based on a virtual reference value is invalid in a strong spatial heterogeneous flow field due to the lack of physical meaning of the reference value; by acquiring sensor array readings and generating an initial virtual reference value and node reading residual, the measurement points are divided into local consistency clusters according to the similarity of the spatial position and the reading residual, the consistency of the topological structure and the physical space layout of each cluster is analyzed, and only when each cluster constitutes an independent and connected spatial region, it is determined that the condition is met and the final virtual reference value with clear physical representation is generated based on all sensor readings, which is used as a reference to calibrate each sensor in the array.

Description

Technical Field

[0001] This invention relates to the field of sensor measurement data calibration technology, and more specifically, to a self-calibration method and system for mine sensor arrays for monitoring wind fields in gravel curtain layers. Background Technology

[0002] In the field of underground ventilation safety monitoring in mines, to accurately grasp the airflow status in key areas, it is common to deploy multiple wind speed sensors in an array to monitor the wind field in complex structural areas such as gravel curtain layers. Due to the harsh mine environment, sensors are prone to drift or performance degradation, thus requiring regular calibration. In existing technologies, a common self-calibration approach is to utilize the monitoring data from multiple sensors within the array, without the possibility of setting an external reference sensor, to generate a unified virtual reference value through a data fusion algorithm. This virtual reference value is then used as a benchmark to evaluate and correct the readings of individual sensors.

[0003] However, it faces limitations when applied to special scenarios with highly heterogeneous wind fields, such as gravel curtain layers: the complex disturbance of wind flow by the gravel curtain structure can lead to significant and uncertain spatial differences in the actual wind speed values ​​of each measuring point in the monitoring area. This makes the virtual reference values ​​generated by conventional fusion algorithms often lack clear physical meaning and are difficult to represent the real wind conditions at any actual spatial location. Using them as a calibration benchmark not only fails to effectively correct the sensor's own errors, but may also introduce new biases, leading to a decrease in the reliability of the entire array monitoring data. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a self-calibration method and system for mine sensor arrays for monitoring wind fields in gravel curtain layers to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A self-calibration method for mine sensor arrays used for monitoring wind fields in gravel curtain layers includes the following steps:

[0007] S1. Obtain the readings of all sensors in the sensor array within the set monitoring period;

[0008] S2. Generate an initial virtual reference value based on the readings of all sensors, and calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residuals at each measuring point;

[0009] S3. Based on the spatial relationship between each measuring point and the node reading residual of each measuring point, divide all measuring points in the sensor array into at least one locally consistent cluster.

[0010] S4. Analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array to obtain the topology consistency analysis results;

[0011] S5. Based on the topology consistency analysis results, determine whether the preset virtual reference value generation conditions are met;

[0012] S6. When the condition is met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated.

[0013] Furthermore, the readings of all sensors in the sensor array within a set monitoring period are obtained, including:

[0014] Acquire sensor readings for each sensor in the sensor array at multiple consecutive sampling times within a set monitoring period;

[0015] Time synchronization processing is performed on sensor readings from multiple consecutive sampling times to form a set of sensor readings at the same time.

[0016] Perform fault value verification and marking on each sensor reading in the sensor reading set.

[0017] Furthermore, an initial virtual reference value is generated based on the readings of all sensors, and the difference between the sensor reading at each measuring point and the initial virtual reference value is calculated as the nodal reading residual at each measuring point, including:

[0018] Based on the sensor readings at multiple consecutive sampling times within a set monitoring period, obtain the sensor reading sequence corresponding to each measuring point sensor;

[0019] Analyze the changing trends of each sensor reading sequence within the set monitoring period, and group the sensors at each measuring point according to the consistency of the changing trends;

[0020] An initial virtual reference value is generated by weighted averaging the sensor readings of sensors at various measuring points within the same group.

[0021] The sensor readings at each measuring point are compared with the initial virtual reference value to obtain the node reading residuals for each measuring point.

[0022] Furthermore, based on the spatial relationship between each measuring point and the nodal reading residuals of each measuring point, all measuring points in the sensor array are divided into at least one locally consistent cluster, including:

[0023] Based on the nodal reading residuals of each measuring point, the similarity of reading residuals between each measuring point is calculated;

[0024] Based on the spatial relationship between each measuring point, calculate the spatial distance between each measuring point;

[0025] Based on the similarity of reading residuals among the measuring points and the spatial distance between the measuring points, identify a set of measuring points with similar reading residuals and adjacent spatial locations.

[0026] The set of measurement points that meet the preset similarity conditions and preset spatial distance conditions will be merged or divided into at least one locally consistent cluster.

[0027] Furthermore, the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array is analyzed to obtain the topology consistency analysis results, including:

[0028] For each of at least one locally consistent cluster, check whether all the measurement points constituting the corresponding locally consistent cluster form a spatially connected region in the physical spatial layout of the sensor array.

[0029] Analyze the overall distribution pattern of at least one local consistency cluster in the physical spatial layout of the sensor array, and determine whether there is an overlap or inclusion relationship between the spatial connected regions corresponding to each local consistency cluster.

[0030] Based on the connectivity check results and distribution pattern analysis results of the spatial connectivity regions corresponding to each local consistency cluster, topology consistency analysis results are generated.

[0031] Furthermore, the check of whether all the measuring points constituting the corresponding local consistency cluster form a spatially connected region in the physical spatial layout of the sensor array is achieved in the following way: obtaining the three-dimensional coordinates of all the measuring points constituting the corresponding local consistency cluster, determining the spatial adjacency relationship between each measuring point based on the preset adjacent measuring point determination rules, and checking whether all measuring points are interconnected through spatial adjacency relationships to form a whole region.

[0032] Furthermore, the determination of whether there is an overlap or inclusion relationship between the spatially connected regions corresponding to each local consistency cluster is achieved in the following way: obtain the three-dimensional spatial range information of the spatially connected regions corresponding to each local consistency cluster, calculate the positional relationship between each spatially connected region, and determine whether any two spatially connected regions partially overlap in three-dimensional space or whether one spatially connected region is completely located inside another spatially connected region.

[0033] Furthermore, based on the topology consistency analysis results, it is determined whether the preset virtual reference value generation conditions are met, including:

[0034] When the topology consistency analysis results indicate that each local consistency cluster constitutes a spatially connected region and there is no overlap or inclusion relationship between the spatially connected regions corresponding to each local consistency cluster, it is determined that the preset virtual reference value generation conditions are met.

[0035] Otherwise, it is determined that the preset virtual reference value generation conditions are not met.

[0036] Furthermore, when the conditions are met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated, including:

[0037] When the preset virtual reference value generation conditions are met, the final virtual reference value is calculated and generated based on the verified and time-synchronized sensor readings, according to the set rules.

[0038] The readings of each sensor in the sensor array that have been calibrated and time-synchronized are compared with the final virtual reference value to obtain the reading deviation of the corresponding sensor.

[0039] Based on the reading deviation of each sensor, the calibration parameters of the corresponding sensors in the sensor array are compensated and adjusted to complete the calibration of the sensor array.

[0040] On the other hand, the present invention provides a mine sensor array self-calibration system for monitoring wind fields in gravel curtain layers, comprising the following modules:

[0041] The reading acquisition module is used to acquire the readings of all sensors in the sensor array within a set monitoring period;

[0042] The residual calculation module is used to generate an initial virtual reference value based on the readings of all sensors, and to calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residuals at each measuring point.

[0043] The measurement point division module is used to divide all measurement points in the sensor array into at least one locally consistent cluster based on the spatial positional relationship of each measurement point and the node reading residual of each measurement point.

[0044] The results analysis module is used to analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array, and obtain the topology consistency analysis results.

[0045] The condition generation module is used to determine whether the preset virtual reference value generation conditions are met based on the topology consistency analysis results.

[0046] The calibration execution module is used to generate a final virtual reference value based on the readings of all sensors and calibrate each sensor in the sensor array when the conditions are met.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. By introducing a topology verification mechanism based on spatial consistency, the physical representativeness and reliability of virtual reference values ​​generated in strongly heterogeneous flow fields are significantly improved. First, the nodal reading residuals of each measuring point are separated using the initial virtual reference values. These residuals more directly reflect the sensor's own errors and local specific disturbances. Then, combined with the spatial positional relationship of the measuring points, measuring points with similar residual characteristics and spatial proximity are clustered into locally consistent clusters, so that the similarity of data features and the proximity of physical space can be correlated. Subsequently, by analyzing whether the topological structure of these locally consistent clusters constitutes an independent and connected region in physical spatial layout, it is determined whether the current data state represents a spatially continuous and structurally clear wind field structure. Only when all clusters are spatially connected and have no overlap or inclusion relationship with each other are they determined to meet the generation conditions. This makes the finally generated virtual reference values ​​based on a clear and internally consistent physical space, solving the fundamental problem of the ambiguity of the physical meaning of virtual reference values ​​in heterogeneous flow fields in traditional methods.

[0049] 2. Improved accuracy of calibration results and overall reliability of array data: Since the generation of the final virtual reference value and the execution of calibration actions strictly depend on the positive judgment of the topology consistency analysis results, this essentially constitutes an adaptive quality control checkpoint. Only when the monitoring data of the sensor array can map a clear and reasonable spatial physical structure will the system trigger the calibration process based on the final virtual reference value generated by all sensor readings. This effectively avoids the risk of introducing secondary errors caused by blindly using virtual reference values ​​that lack representativeness for calibration in the case of chaotic flow structure or improper sensor grouping. The calibration process is constrained to be carried out in a high-confidence scenario supported by both data quality and spatial logic, thereby ensuring the accuracy and effectiveness of sensor parameter compensation and adjustment. Ultimately, this makes the output data of the entire sensor array more consistent in space and more stable and reliable in time. Attached Figure Description

[0050] Figure 1 This is a flowchart of the self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to the present invention.

[0051] Figure 2 This is a schematic diagram of the structure of the mine sensor array self-calibration system for monitoring wind fields in gravel curtain layers according to the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1: Figure 1 This invention presents a self-calibration method for mine sensor arrays used in monitoring wind fields in gravel curtain layers, comprising the following steps:

[0054] S1. Obtain the readings of all sensors in the sensor array within the set monitoring period;

[0055] S2. Generate an initial virtual reference value based on the readings of all sensors, and calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residuals at each measuring point;

[0056] S3. Based on the spatial relationship between each measuring point and the node reading residual of each measuring point, divide all measuring points in the sensor array into at least one locally consistent cluster.

[0057] S4. Analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array to obtain the topology consistency analysis results;

[0058] S5. Based on the topology consistency analysis results, determine whether the preset virtual reference value generation conditions are met;

[0059] S6. When the condition is met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated.

[0060] S1. Obtain the readings of all sensors in the sensor array within the set monitoring period. The specific implementation is as follows:

[0061] The monitoring period is determined based on the specific monitoring task requirements and the system's data processing capabilities. For example, a typical monitoring period could be set to 10 minutes. Within this monitoring period, the system drives each sensor in the array to perform measurements at a constant sampling frequency. This sampling frequency needs to be set according to the typical timescale of the wind field changes in the measured gravel curtain layer; for example, it could be set to 1 Hz, meaning one sample per second. Therefore, within a 10-minute monitoring period, each sensor will obtain raw readings from 600 consecutive sampling moments. The physical time point corresponding to each sampling moment is precisely recorded as a timestamp and stored in conjunction with the sensor readings acquired at that moment, forming an independent, chronologically ordered sequence of raw readings for each sensor. This step provides the most basic raw data input for all subsequent analyses.

[0062] To synchronize sensor readings from multiple consecutive sampling times into a set of sensor readings at the same moment, the following method is used. Due to slight differences in the hardware clocks of each sensor and varying delays in the data transmission links, the directly acquired sensor reading sequences are not strictly aligned on the time axis. To address this issue, a unified standard time axis needs to be established. This time axis uses the start point of the set monitoring period as zero and divides a series of standard alignment times according to a fixed time interval. For example, if the sampling frequency is 1 Hz, the standard alignment times can be set to 0 seconds, 1 second, 2 seconds, and so on, until the end of the period. For each standard alignment time, the processing system iterates through the raw reading sequences of all sensors. If a sensor has a raw reading with the same timestamp at a certain standard alignment time, that reading is directly adopted. If there is no perfectly matching timestamp, an interpolation algorithm is used for estimation: the two nearest actual sampling points before and after the standard alignment time in the raw reading sequence of the sensor are found, and the estimated reading for the standard alignment time is calculated through a linear relationship based on the reading values ​​and timestamps of these two actual sampling points. After performing the above operations on all sensors, each standard alignment time corresponds to a data set containing the readings or estimated readings of all sensors at that time. This data set is the set of sensor readings for the same moment. This process ensures the comparability of all sensor data across the time dimension.

[0063] Each sensor reading in the sensor reading set is checked for and marked as a bad value. This is achieved through the following method: A bad value refers to an abnormal data point that deviates significantly from physical reality due to instantaneous sensor anomalies, transient signal interference, or other reasons. The verification process first establishes a dynamic effective reading range threshold for each sensor. This threshold is obtained through statistical learning based on historical data of the sensor's recent normal operation. Specifically, it can extract all effective readings of the sensor after time synchronization processing over a past period, such as the past hour, and calculate the average and standard deviation of these readings. Based on this, the lower limit of the sensor's current effective range is set to the average value minus N times the standard deviation, and the upper limit is the average value plus N times the standard deviation, where N is a multiplier set based on experience, such as typically 3. After setting the threshold, each sensor reading set generated at the same time is traversed, and each reading in the set is checked to see if it is within the dynamic effective range threshold corresponding to its sensor. If a reading is lower than its sensor's lower effective range threshold or higher than its sensor's upper effective range threshold, the reading is determined to be a bad value, and it is explicitly marked in the data storage structure, for example, by setting its validity flag to invalid. All readings marked as bad values ​​will be automatically identified and excluded from the calculation data source by the system in subsequent steps, such as generating initial virtual reference values. This ensures the reliability of the input data quality for subsequent core data processing flows and avoids abnormal data contamination.

[0064] S2. Generate an initial virtual reference value based on the readings of all sensors, and calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residual for each measuring point. The specific implementation is as follows:

[0065] First, based on a set of sensor readings generated at multiple simultaneous moments within a defined monitoring period, a corresponding sensor reading sequence is constructed for each measuring point. Specifically, the readings of each sensor at each standard alignment moment are extracted and arranged in chronological order. For example, if there are 600 standard alignment moments within a defined monitoring period, each sensor will correspond to a sequence containing 600 readings arranged in chronological order, i.e., a sensor reading sequence. This sensor reading sequence completely records the continuous process of wind field changes at that measuring point within the current monitoring period, serving as the basis for subsequent trend analysis and grouping. This step directly uses data that has undergone time synchronization processing and out-of-value marking, ensuring the temporal accuracy and basic quality of the input data.

[0066] The analysis of the changing trends of each sensor reading sequence within a set monitoring period, and the grouping of sensors at each measuring point based on the consistency of these trends, is achieved through the following method. The changing trend refers to the overall direction and shape of the sensor readings changing over time. To quantify this trend, one approach is to perform linear fitting on the sensor reading sequence for each sensor and calculate the slope of its trend line. For example, using the time point number as the independent variable and the corresponding sensor reading as the dependent variable, a straight line is obtained through least squares fitting. The slope of this line characterizes whether the wind speed reading at that measuring point shows an upward trend, a downward trend, or remains stable within the monitoring period. After obtaining the trend line slopes for all sensors, all measuring point sensors are clustered based on the similarity of these slope values. Here, a trend consistency threshold needs to be set to determine whether the trends are consistent. For example, the trend consistency threshold can be set to the absolute value of the difference between slope values ​​being less than a certain specific value, such as 0.05 seconds per second. Sensors whose slope difference between each other's trend lines is less than this trend consistency threshold are grouped into the same group; if the slope difference between a sensor and any member of the group exceeds this threshold, it may be grouped into another group or become a separate group. This method allows sensors that may be spatially dispersed but are influenced by similar flow structures and exhibit synchronous changing trends to be grouped together.

[0067] An initial virtual reference value is generated by weighted averaging of sensor readings from all measuring points within the same group. This is achieved through the following methods: Weighted averaging means that different sensor readings are assigned different importance, i.e., weighting factors, when calculating the average. The weighting factors must be set based on the reliability of the sensor or its representativeness within the current group. One method for setting weighting factors is based on the stability of the sensor reading sequence. For example, the standard deviation of each sensor reading sequence can be calculated. The smaller the standard deviation, the smaller the fluctuation in the sensor readings within the monitoring period, and the more stable and reliable it is likely to be, thus assigning it a higher weighting factor; conversely, a higher standard deviation is assigned a lower weighting factor. Another method is based on the fit of the trend fit, i.e., calculating the fit error between the actual reading sequence of each sensor and its fitted trend line. The smaller the error, the higher the weighting factor. In practice, for each sensor, its stability and trend fit can be combined into a score, and this score can be normalized and used as its weighting factor. After determining the weighting factor for each sensor within the same group, for each standard alignment time, the readings of all sensors within that group at that time are taken, multiplied by their respective weighting factors, summed, and then divided by the sum of all weighting factors to obtain the weighted average reading of that group at that time. This calculation is repeated for each standard alignment time within a set monitoring period, resulting in a time-varying sequence of readings, which is the initial virtual reference value generated based on that group. If the sensor array is divided into multiple groups, each group will independently generate an initial virtual reference value sequence representing the common variation characteristics of that group.

[0068] The node reading residuals for each measuring point are obtained by subtracting the initial virtual reference value from the sensor readings at each measuring point. This step is performed independently at each standard alignment time. For a given measuring point, its sensor group is first determined, and the initial virtual reference value generated by that group at the current time is obtained. Then, the sensor reading at the current time is subtracted from the corresponding initial virtual reference value; the difference is the node reading residual for that measuring point at that time. The node reading residual is a signed numerical value, and its physical meaning represents the degree of deviation of the actual reading at that measuring point from the overall trend level of its group. This calculation process is applied to each measuring point in the sensor array and iterates through each standard alignment time within the set monitoring period, ultimately generating a node reading residual sequence corresponding to the time series for each measuring point. These node reading residual sequences are key data products. They strip away the common trends determined by local flow structures and more purely reflect the inherent sensor errors, local specific disturbances, and random noise that may be contained in the readings of each measuring point. This provides direct input for subsequent spatial consistency-based analysis and sensor calibration.

[0069] S3. Based on the spatial relationship of each measuring point and the nodal reading residuals of each measuring point, all measuring points in the sensor array are divided into at least one locally consistent cluster, specifically implemented as follows:

[0070] First, the similarity between measuring points in terms of reading residuals is calculated based on the node reading residuals of each measuring point. The node reading residuals are numerical sequences calculated for each measuring point in previous steps and arranged chronologically within a set monitoring period. Calculating the similarity between two measuring points in terms of reading residuals assesses the degree of consistency in the variation patterns of the two node reading residual sequences. One specific implementation method is to calculate the Pearson correlation coefficient. This calculation uses the residual values ​​of the two sequences at the same time point as input data pairs, and mathematical operations yield a numerical result ranging from -1 to +1. The closer the result is to +1, the more synchronized the variation trends of the two sequences are, the stronger the positive correlation, and the higher the similarity; the closer the result is to 0, the weaker the correlation and the lower the similarity. Performing this calculation on all possible measuring point pairs in the sensor array yields a similarity matrix, where each element represents the reading residual similarity value between the corresponding two measuring points. For subsequent judgment, a clear similarity threshold needs to be set. The similarity threshold can be set based on the numerical distribution characteristics of the entire calculated similarity matrix. For example, the median of the similarity values ​​for all measurement points can be calculated, and a similarity threshold can be set to a specific value higher than that median. A more specific method is to first calculate the upper quartiles of all similarity values, and then set the similarity threshold to be equal to or slightly higher than that upper quartile value, for example, setting the similarity threshold to 0.75. This means that in subsequent steps, when it is necessary to determine whether the reading residuals of any two measurement points are similar, their similarity values ​​will be compared with this preset similarity threshold of 0.75. If the similarity value is greater than or equal to the similarity threshold of 0.75, the two measurement points are considered similar in terms of reading residuals; if the similarity value is less than the similarity threshold of 0.75, they are considered dissimilar in terms of reading residuals.

[0071] Based on the spatial relationships of each measuring point, the spatial distance between them is calculated using the following method. Each measuring point has a unique, pre-measured coordinate in the three-dimensional space of the mine, typically including east-facing, north-facing, and altitude values. Calculating the spatial distance between two measuring points is equivalent to calculating the straight-line distance between these two three-dimensional coordinate points, also known as the Euclidean distance. The specific calculation process involves three steps: first, calculating the difference in the east-facing coordinate values, the difference in the north-facing coordinate values, and the difference in altitude values; then, squaring each of these three differences; next, summing the three squared results; and finally, taking the square root of the sum. The result is the spatial distance between the two measuring points, expressed in units of length. Performing this calculation on all possible pairs of measuring points in the sensor array yields a distance matrix, where each element represents the spatial distance between the corresponding two measuring points. For subsequent judgment, a specific spatial distance threshold needs to be set. The setting of the spatial distance threshold needs to consider the actual deployment density of the sensor array and the physical spatial scale of the monitoring area. One specific method involves first calculating the average spatial distance between all pairs of measuring points to reflect the average spacing of the sensors. Then, based on the spatial correlation scale of the wind field characteristics in the monitored area, a spatial distance threshold is set as a multiple of this average spacing. For example, if the calculated average spacing is 5 meters, empirically, the spatial distance threshold can be set to twice the average spacing, i.e., 10 meters. This means that in subsequent steps, when it is necessary to determine whether any two measuring points are spatially adjacent, their spatial distance is compared with this preset spatial distance threshold of 10 meters. If the spatial distance is less than or equal to the spatial distance threshold of 10 meters, the two measuring points are determined to be spatially adjacent; if the spatial distance is greater than the spatial distance threshold of 10 meters, they are determined to be spatially non-adjacent.

[0072] Based on the similarity of reading residuals between measuring points and the spatial distance between them, a set of measuring points with similar reading residuals and spatially adjacent locations is identified. This is achieved through the following method. This step requires the comprehensive application of the two conditions calculated and set with thresholds previously. The specific identification process is carried out iteratively or traversally. Any measuring point in the sensor array can be used as a starting point to examine the relationship between this starting point and all other measuring points in the array. For each other measuring point, two judgments need to be performed simultaneously: the first judgment is to check whether the similarity value of the reading residuals between the starting point and the other measuring point meets the preset similarity condition, i.e., whether the similarity value is greater than or equal to the previously set similarity threshold, such as 0.75; the second judgment is to check whether the spatial distance value between the starting point and the other measuring point meets the preset spatial distance condition, i.e., whether the spatial distance value is less than or equal to the previously set spatial distance threshold, such as 10 meters. Only when both judgments for a certain other measuring point are yes, is that other measuring point marked as having similar reading residuals to the starting point and being spatially adjacent. After comparing with all other measuring points, the starting point itself and all other measuring points marked as satisfying the dual conditions are grouped together to form an initial set of measuring points. Each measuring point in this initial set satisfies both the conditions of similar reading residuals and spatial adjacency with at least one other measuring point in the set; these other measuring points are usually the starting points during the initial construction. By repeating the above process, using measuring points in the array that have not yet been assigned to any initial set as new starting points, multiple such initial sets of measuring points can be identified.

[0073] The measurement point sets that meet the preset similarity and spatial distance conditions are merged or divided into at least one locally consistent cluster, achieved as follows: The multiple initial measurement point sets identified in the previous step may not be the final ideal independent clusters because they may overlap or have excessively large spatial extents within a single set. Therefore, merging and division operations are necessary. The merging operation targets initial sets that are spatially closely related or have highly overlapping measurement point members. Specifically, a merging overlap threshold can be set to guide the merging process. For example, the proportion of common measurement points between any two initial sets to the total number of measurement points in the smaller set is calculated. If this proportion exceeds the merging overlap threshold, for example, exceeding 60%, then the two initial sets are considered to actually represent the same spatially continuous region, and they should be merged into a new, larger measurement point set. The division operation targets initial sets with large internal spatial spans to ensure the spatial compactness of each final cluster. For this purpose, a maximum intra-cluster spatial span threshold needs to be set. This threshold can be set based on the overall size of the monitoring area and the desired degree of intra-cluster spatial continuity, for example, 20 meters. For each initial measurement point set, the maximum spatial distance between any two measurement points within the set is calculated. If this maximum value exceeds the preset maximum spatial span threshold of 20 meters within a cluster, it indicates that the initial set covers too wide a spatial range and may contain sub-regions that are not physically directly connected. In this case, the set needs to be divided based on the spatial proximity between measurement points. One division method is hierarchical clustering based on spatial distance: select the two spatially closest measurement points from all measurement points in the set as seed points, and gradually assign other measurement points to the corresponding sub-clusters based on their spatial distance from the already formed sub-clusters, until the maximum spatial distance within all sub-clusters does not exceed the maximum spatial span threshold of 20 meters within the cluster. After multiple rounds of merging, division, and processing, several sets of measurement points with no common measurement points and relatively compact internal spatial structures are finally formed. Each set of measurement points is defined as a locally consistent cluster. All measurement points within each locally consistent cluster not only exhibit high similarity in their node reading residual sequences but also are close to each other in three-dimensional physical space, forming a spatially coherent region. All measurement points in the entire sensor array are assigned to a local consistency cluster, thus completing the clustering of measurement points from the dual dimensions of data feature similarity and spatial proximity.

[0074] S4. Analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array to obtain the topology consistency analysis results. The specific implementation is as follows:

[0075] First, for each of the at least one locally consistent clusters obtained in the previous steps, a spatial connectivity check is performed. This check examines whether all the measuring points constituting the corresponding locally consistent cluster form a spatially connected region within the physical spatial layout of the sensor array. This check begins by obtaining the precise three-dimensional coordinates of all the measuring points constituting the locally consistent cluster, which define the specific location of each measuring point in the mine space. The core of the check is determining the spatial adjacency relationship between measuring points within the cluster, which requires a preset adjacent measuring point determination rule. This rule is based on the concept of proximity in physical space; for example, a specific adjacent determination distance threshold can be set. The setting of this adjacent determination distance threshold needs to refer to the typical deployment spacing of measuring points in the sensor array. One specific method is to calculate the minimum and average distances between all pairs of measuring points in the entire array, and set the adjacent determination distance threshold to a value between the minimum and average distances, for example, 80% of the average distance. According to this rule, for any two measuring points within the locally consistent cluster being checked, the three-dimensional spatial distance between them is calculated. If this distance is less than or equal to the preset adjacent determination distance threshold, then the two measuring points are determined to be spatially adjacent. After identifying all pairs of measurands within a cluster that satisfy the adjacency relationship, a connectivity check is performed using graph theory: all measurands are treated as nodes in the graph, and spatial adjacency relationships are considered as edges connecting the nodes, constructing the adjacency graph corresponding to the locally consistent cluster. Then, starting from any node in the graph, a depth-first or breadth-first search algorithm is used to traverse all nodes reachable by edges. If all nodes have been visited after the traversal, the graph is considered connected, meaning that all measurands constituting the locally consistent cluster are interconnected through spatial adjacency relationships, forming an indivisible whole region. In this case, the locally consistent cluster is determined to constitute a spatially connected region. Conversely, if there are unvisited nodes, the locally consistent cluster is determined not to constitute a spatially connected region, and its measurands are physically scattered. This connectivity check needs to be performed independently for each locally consistent cluster, and the check result for each cluster must be recorded.

[0076] After completing the spatial connectivity check of each locally consistent cluster, it is necessary to further analyze the overall distribution pattern of at least one locally consistent cluster in the physical spatial layout of the sensor array. The core is to determine whether there is an overlap or inclusion relationship between the spatially connected regions corresponding to each locally consistent cluster. This analysis presupposes mapping each locally consistent cluster—that is, the set of measurement points it contains regardless of whether it passes the connectivity check—to its corresponding spatial occupancy range in physical space, i.e., the spatially connected region. For clusters that have passed the connectivity check, their spatially connected region is the continuous spatial range formed by connecting all their measurement points through adjacency relationships. To quantify this range for comparison, it is necessary to obtain the three-dimensional spatial range information of each spatially connected region. One specific method is to calculate the three-dimensional axial bounding box of each region, i.e., to find the maximum and minimum values ​​of the coordinates of all measurement points within the region in the east, north, and elevation directions. These six values ​​together define a cuboid spatial range that can completely enclose all measurement points in the region. After obtaining the three-dimensional spatial range information of each region, the positional relationship between these ranges is calculated to determine whether any two spatially connected regions overlap or contain each other. To determine overlap, we can check whether the 3D bounding boxes of two regions overlap in all three coordinate axes. For example, for regions A and B, if the minimum eastward coordinate of region A is less than the maximum eastward coordinate of region B, and the maximum eastward coordinate of region A is greater than the minimum eastward coordinate of region B, and this condition also holds true in the north and elevation directions, then we can determine that the 3D bounding boxes of the two regions overlap, and further infer that their corresponding spatially connected regions are highly likely to have partial overlap in 3D space. To determine containment, we need to check whether the 3D bounding box of one region is completely enclosed by the 3D bounding box of another region. Specifically, if the minimum eastward coordinate of region A is greater than or equal to the minimum eastward coordinate of region B, and the maximum eastward coordinate of region A is less than or equal to the maximum eastward coordinate of region B, and this condition also strictly holds true in the north and elevation directions, then we can determine that the bounding box of region A is completely inside the bounding box of region B, and further infer that the spatially connected region corresponding to region A may be completely inside the spatially connected region corresponding to region B. This analysis requires performing the analysis on all pairs of spatially connected regions corresponding to locally consistent clusters, and recording which cluster pairs have overlapping relationships and which cluster pairs have inclusion relationships.

[0077] Finally, based on the connectivity check results and distribution pattern analysis results of the spatial connectivity regions corresponding to each locally consistent cluster, a topology consistency analysis result is generated. This result is not a single numerical value, but a structured set of judgment conclusions. Its generation logic is based on predefined topology consistency criteria. A typical criterion is: ideally, each locally consistent cluster should correspond to an independent, internally connected, and mutually non-interfering physical spatial region. Therefore, the topology consistency analysis result needs to explicitly answer several key questions. First, has every locally consistent cluster passed the spatial connectivity check, i.e., has it all constituted a spatially connected region? Second, are there any overlapping relationships among all locally consistent clusters that constitute spatially connected regions? Third, are there any inclusion relationships among all locally consistent clusters that constitute spatially connected regions? Based on the answers to these questions, a general topology consistency analysis result can be generated. For example, if all locally consistent clusters constitute spatially connected regions, and there is neither an overlapping nor an inclusion relationship between any two spatially connected regions, then the generated topology consistency analysis result can be described as having high topology consistency. If some locally consistent clusters do not constitute spatially connected regions, or if, although all clusters are connected, there are overlapping or containment relationships between the spatially connected regions of some cluster pairs, the generated topological consistency analysis result can be described as low topological consistency. Furthermore, the result can list in detail which clusters are not connected and what types of spatial relationship conflicts exist between cluster pairs. This topological consistency analysis result directly reflects the degree of agreement between the sensor grouping based on data features and the real physical spatial structure, providing crucial decision-making basis for the next step of determining whether the conditions for generating reliable virtual reference values ​​are met.

[0078] S5. Based on the topology consistency analysis results, determine whether the preset virtual reference value generation conditions are met. The specific implementation is as follows:

[0079] The core of this step is to execute a logical judgment process based on the topology consistency analysis results generated in the previous step to determine whether the sensor data status within the current monitoring period supports the generation of a reliable final virtual reference value. The preset virtual reference value generation conditions are explicitly defined in this method as a set of structural criteria regarding the spatial topology consistency of the sensor array data, the specific content of which is directly related to the key information contained in the topology consistency analysis results. The input to the judgment process is the topology consistency analysis results, which are a structured data object or information set that explicitly records at least the following three types of information: the first type is the spatial connectivity check conclusion for each local consistency cluster, i.e., whether each local consistency cluster constitutes a spatially connected region; the second type is the conclusion regarding whether there is an overlap relationship between the spatially connected regions corresponding to any two local consistency clusters; and the third type is the conclusion regarding whether there is an inclusion relationship between the spatially connected regions corresponding to any two local consistency clusters. These conclusions have been determined when generating the topology consistency analysis results; the task of this step is to review and comprehensively evaluate these conclusions according to the preset rules.

[0080] First, the logic checks the spatial connectivity status of all locally consistent clusters in the topology consistency analysis results. Specifically, it reads the connectivity check conclusion of each locally consistent cluster recorded in the results one by one. During this process, a clear connectivity pass criterion needs to be established for the final summary judgment. This criterion requires that, among all locally consistent clusters covered by the topology consistency analysis results, the check conclusion of each locally consistent cluster must be that it constitutes a spatially connected region. This means that from the first locally consistent cluster to the last, no individual cluster can have a conclusion that it does not constitute a spatially connected region. If at least one locally consistent cluster is found to not constitute a spatially connected region, it means that some measurement points in the sensor array, while similar in data characteristics, are discrete or fragmented in the real physical space. This mismatch between data and space suggests strong heterogeneity in the flow structure or potential irrationality in sensor grouping. Therefore, the basic spatial structure conditions required to generate reliable virtual reference values ​​are not met. In this case, the judgment branch for not meeting the conditions can be directly triggered without further spatial relationship checks.

[0081] The decision logic proceeds to the next step, checking the spatial relationships between spatially connected regions, only after all locally consistent clusters have been confirmed as constituting spatially connected regions. This stage requires sequentially reviewing the two types of spatial relationship conclusions recorded in the topology consistency analysis results. First, the conclusions regarding overlap are reviewed. The decision logic needs to traverse all records in the results regarding whether pairs of spatially connected regions overlap. The preset virtual reference value generation conditions require that, ideally, each spatially connected region should represent an independent region in physical space that does not interfere with each other. Therefore, the rule regarding overlap in the generation conditions is: there is no overlap between any two spatially connected regions among all locally consistent clusters confirmed as constituting spatially connected regions. That is, for each pair of regions in the record, the overlap relationship conclusion must be non-overlapping. If at least one pair of spatially connected regions is found to have an overlapping relationship, it indicates that the jurisdiction of different sensor clusters divided based on data characteristics overlaps in physical space. This overlap may mean that different flow structures are mixed at the boundary, or that the current clustering fails to clearly distinguish spatially adjacent but different regions. Such a situation is considered not to meet the conditions for generating a single virtual reference value with a clear physical meaning.

[0082] Secondly, after confirming that there is no overlap between all regions, the conclusions regarding containment relationships need to be further examined. The judgment logic also traverses all records in the results regarding whether there is a containment relationship between pairs of spatially connected regions. The preset generation conditions are more stringent regarding containment relationships: requiring that no two spatially connected regions have a containment relationship. That is, there is no situation where one spatially connected region is completely located inside another spatially connected region. If the topology consistency analysis results show the existence of such a containment relationship, for example, a small region consisting of only a few internal measurement points is completely surrounded by a large region consisting of a large number of peripheral measurement points, this may indicate a hierarchical data feature or a core-periphery structure. However, this nesting relationship makes defining a single virtual reference value representing the entire array complex and may mask local characteristics. Therefore, it is also considered to fail to meet the preset virtual reference value generation conditions.

[0083] The final determination is a result of logical synthesis. Only when the topology consistency analysis result simultaneously satisfies the following three sub-conditions can the system ultimately determine that the preset virtual reference value generation conditions are met: the first sub-condition is that all locally consistent clusters constitute spatially connected regions; the second sub-condition is that there is no overlap between all spatially connected regions; and the third sub-condition is that there is no inclusion relationship between all spatially connected regions. All three sub-conditions must be true; none can be missing. Once the judgment logic confirms that all three sub-conditions are true, the system outputs the final conclusion that the preset virtual reference value generation conditions are met. Conversely, if any one or more of the above three sub-conditions are false—that is, at least one locally consistent cluster is not connected, or any pair of regions overlaps, or any pair of regions contains each other—the system outputs the final conclusion that the preset virtual reference value generation conditions are not met. This final conclusion will be passed to subsequent steps as a clear instruction to decide whether to perform sensor calibration, thus completing the logical transformation and closed loop from spatial topology analysis to calibration decision.

[0084] S6. When the condition is met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated. The specific implementation is as follows:

[0085] First, the process of generating the final virtual reference value is triggered. The input data for this process consists of high-quality, calibrated, and time-synchronized sensor readings obtained after complete preprocessing in step one. These readings are organized as a set of sensor readings at the same moment, covering the entire set monitoring period. The setting rules used to generate the final virtual reference value need to be clearly defined. A basic and robust rule is the arithmetic mean rule: for each standard alignment moment, sum all valid, calibrated, and time-synchronized sensor readings in the sensor array at that moment, and then divide by the total number of valid readings to obtain the final virtual reference value for that moment. Valid readings refer to readings that have not been marked as bad values. Another, more refined setting rule is the weighted average rule. Under this rule, a weighting factor needs to be assigned to each sensor in the sensor array. The weighting factor can be set based on the sensor's long-term reliability history. For example, the frequency with which each sensor has been marked as bad over multiple monitoring periods can be counted; sensors with lower bad value frequencies are considered more reliable, and their weighting factors should be set higher. In specific configuration, a baseline reliability index can be defined. The reciprocal of the out-of-range frequency of each sensor is taken and normalized, ensuring the sum of the weighting factors for all sensors is 1, thus obtaining the specific weighting factor value for each sensor. When applying the weighted average rule, for each standard alignment time, the effective reading of each sensor at that time is multiplied by its weighting factor, and all products are summed to obtain the final virtual reference value for that time. Regardless of the configuration rule used, the above calculation is repeated for each standard alignment time within the set monitoring period, ultimately generating a final virtual reference value sequence corresponding to the time series. This sequence represents a consensus-based, highly reliable estimate of the monitored gravel layer wind field by the entire sensor array under ideal conditions that pass topological consistency verification.

[0086] Next, the sensor reading deviation is calculated. This step requires comparing the raw measurement data of each sensor with the newly generated authoritative benchmark point by point. Specifically, for each sensor in the sensor array, every standard alignment time within the set monitoring period is traversed. At each time point, the calibrated and time-synchronized sensor reading is acquired, along with the corresponding final virtual reference value. The sensor reading is subtracted from the final virtual reference value at that time; the difference is the instantaneous reading deviation of the sensor at that specific time. The instantaneous reading deviation is a positive or negative value; a positive deviation indicates that the sensor reading is higher than the final virtual reference value, and a negative deviation indicates that it is lower than the final virtual reference value. To obtain the overall deviation characteristics of each sensor within a monitoring period, statistical analysis of the instantaneous reading deviations at all time points is required. A typical approach is to calculate the arithmetic mean of all valid instantaneous reading deviations of the sensor over the entire set monitoring period; this average is defined as the sensor's reading deviation. When calculating the average, only those instantaneous deviation values ​​where the sensor reading itself is valid and the corresponding final virtual reference value also exists are used. This reading deviation is a single scalar value that broadly represents the degree of systematic deviation of the sensor from the array consensus value throughout the monitoring period. Its dimensions are the same as those of the sensor reading, such as wind speed in meters per second.

[0087] Finally, based on the reading deviation of each sensor, the calibration parameters of the corresponding sensors in the sensor array are compensated and adjusted. This step assumes that there is a known, adjustable mathematical relationship between the measurement output of each sensor and its internal calibration parameters, such as a linear correction relationship: the true value equals the sensor's original reading multiplied by a calibration coefficient plus a zero-point offset. The reading deviation directly reflects the systematic error generated by using existing calibration parameters in the current measurement environment. Therefore, the goal of calibration is to adjust these parameters so that, after applying new parameters, the expected reading value of the sensor will approach the final virtual reference value. The adjustment strategy needs to be formulated based on the magnitude and direction of the reading deviation. First, a reading deviation significance threshold can be set to determine whether a sensor's reading deviation is large enough to require calibration parameter adjustment. This threshold can be set according to the measurement accuracy requirements and the natural amplitude of wind speed variations, for example, it can be set to 0.1 m / s. If the absolute value of a sensor's reading deviation is less than this reading deviation significance threshold, its deviation is considered insignificant and may fall within the range of random fluctuations, and its calibration parameters are not adjusted for the time being. For sensors with an absolute reading deviation greater than or equal to the threshold, parameter adjustment is initiated. A direct linear compensation adjustment method is as follows: Assuming the sensor's output model is linear, its reading deviation mainly originates from zero-point offset drift. Therefore, the calculated reading deviation of the sensor can be directly used as the correction amount for the zero-point offset parameter. Specifically, the original zero-point offset parameter value of the sensor is added to the currently calculated reading deviation value. Note the sign: if the reading is a positive deviation, the offset parameter should be reduced, resulting in a new, compensated zero-point offset parameter. If the deviation is determined to mainly originate from changes in the calibration coefficients, the coefficients can be adjusted according to the deviation ratio. The adjusted new calibration parameters are then updated and stored in the configuration unit or database corresponding to the sensor, replacing the old parameter values. After updating the parameters of all sensors in the sensor array that require adjustment, the self-calibration process of the sensor array based on topological consistency verification for the current monitoring cycle is complete. The calibrated sensors will use the new parameters for measurements in subsequent monitoring cycles to obtain more accurate and spatially consistent readings, providing a reliable data foundation for the precise monitoring of wind fields in mine gravel curtain layers.

[0088] Example 2: Figure 2 A schematic diagram of the self-calibration system for mine sensor arrays used for monitoring wind fields in gravel curtain layers is provided. The self-calibration system for mine sensor arrays used for monitoring wind fields in gravel curtain layers includes the following modules:

[0089] The reading acquisition module is used to acquire the readings of all sensors in the sensor array within a set monitoring period;

[0090] The residual calculation module is used to generate an initial virtual reference value based on the readings of all sensors, and to calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residuals at each measuring point.

[0091] The measurement point division module is used to divide all measurement points in the sensor array into at least one locally consistent cluster based on the spatial positional relationship of each measurement point and the node reading residual of each measurement point.

[0092] The results analysis module is used to analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array, and obtain the topology consistency analysis results.

[0093] The condition generation module is used to determine whether the preset virtual reference value generation conditions are met based on the topology consistency analysis results.

[0094] The calibration execution module is used to generate a final virtual reference value based on the readings of all sensors and calibrate each sensor in the sensor array when the conditions are met.

[0095] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0096] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0097] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0098] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0099] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0100] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0101] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A self-calibration method for mine sensor arrays used for monitoring wind fields in gravel curtain layers, characterized in that, Includes the following steps: S1. Obtain the readings of all sensors in the sensor array within the set monitoring period; S2. Generate an initial virtual reference value based on the readings of all sensors, and calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the nodal reading residual for each measuring point, including: Based on the sensor readings at multiple consecutive sampling times within a set monitoring period, obtain the sensor reading sequence corresponding to each measuring point sensor; Analyze the changing trends of each sensor reading sequence within the set monitoring period, and group the sensors at each measuring point according to the consistency of the changing trends; An initial virtual reference value is generated by weighted averaging the sensor readings of sensors at various measuring points within the same group. The sensor readings at each measuring point are subtracted from the initial virtual reference value to obtain the node reading residuals at each measuring point. S3. Based on the spatial relationship between each measuring point and the node reading residual of each measuring point, divide all measuring points in the sensor array into at least one locally consistent cluster. S4. Analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array, and obtain the topology consistency analysis results, including: For each of at least one locally consistent cluster, check whether all the measurement points constituting the corresponding locally consistent cluster form a spatially connected region in the physical spatial layout of the sensor array. Analyze the overall distribution pattern of at least one local consistency cluster in the physical spatial layout of the sensor array, and determine whether there is an overlap or inclusion relationship between the spatial connected regions corresponding to each local consistency cluster. Based on the connectivity check results and distribution pattern analysis results of the spatial connectivity regions corresponding to each local consistency cluster, topological consistency analysis results are generated. S5. Based on the topology consistency analysis results, determine whether the preset virtual reference value generation conditions are met; S6. When the condition is met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated.

2. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, Acquire readings from all sensors in the sensor array within a set monitoring period, including: Acquire sensor readings for each sensor in the sensor array at multiple consecutive sampling times within a set monitoring period; Time synchronization processing is performed on sensor readings from multiple consecutive sampling times to form a set of sensor readings at the same time. Perform fault value verification and marking on each sensor reading in the sensor reading set.

3. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, Based on the spatial relationship of each measuring point and the nodal reading residuals of each measuring point, all measuring points in the sensor array are divided into at least one locally consistent cluster, including: Based on the nodal reading residuals of each measuring point, the similarity of reading residuals between each measuring point is calculated; Based on the spatial relationship between each measuring point, calculate the spatial distance between each measuring point; Based on the similarity of reading residuals among the measuring points and the spatial distance between the measuring points, identify a set of measuring points with similar reading residuals and adjacent spatial locations. The set of measurement points that meet the preset similarity conditions and preset spatial distance conditions will be merged or divided into at least one locally consistent cluster.

4. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, The method for checking whether all the measuring points constituting the corresponding local consistency cluster form a spatially connected region in the physical spatial layout of the sensor array is as follows: obtain the three-dimensional coordinates of all the measuring points constituting the corresponding local consistency cluster, determine the spatial adjacency relationship between each measuring point based on the preset adjacent measuring point determination rules, and check whether all measuring points are interconnected through spatial adjacency relationship to form a whole region.

5. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, The determination of whether there is an overlap or inclusion relationship between the spatially connected regions corresponding to each local consistency cluster is achieved by: obtaining the three-dimensional spatial range information of the spatially connected regions corresponding to each local consistency cluster; calculating the positional relationship between each spatially connected region; and determining whether any two spatially connected regions partially overlap in three-dimensional space or whether one spatially connected region is completely located inside another spatially connected region.

6. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, Based on the topology consistency analysis results, determine whether the preset virtual reference value generation conditions are met, including: When the topology consistency analysis results indicate that each local consistency cluster constitutes a spatially connected region and there is no overlap or inclusion relationship between the spatially connected regions corresponding to each local consistency cluster, it is determined that the preset virtual reference value generation conditions are met. Otherwise, it is determined that the preset virtual reference value generation conditions are not met.

7. The self-calibration method for mine sensor arrays for monitoring wind fields in gravel curtain layers according to claim 1, characterized in that, When the conditions are met, a final virtual reference value is generated based on the readings of all sensors, and each sensor in the sensor array is calibrated, including: When the preset virtual reference value generation conditions are met, the final virtual reference value is calculated and generated based on the verified and time-synchronized sensor readings, according to the set rules. The readings of each sensor in the sensor array that have been calibrated and time-synchronized are compared with the final virtual reference value to obtain the reading deviation of the corresponding sensor. Based on the reading deviation of each sensor, the calibration parameters of the corresponding sensors in the sensor array are compensated and adjusted to complete the calibration of the sensor array.

8. A mine sensor array self-calibration system for monitoring wind fields in gravel curtain layers, used to implement the mine sensor array self-calibration method for monitoring wind fields in gravel curtain layers as described in any one of claims 1-7, characterized in that, Includes the following modules: The reading acquisition module is used to acquire the readings of all sensors in the sensor array within a set monitoring period; The residual calculation module is used to generate an initial virtual reference value based on the readings of all sensors, and to calculate the difference between the sensor readings at each measuring point and the initial virtual reference value as the node reading residuals at each measuring point. The measurement point division module is used to divide all measurement points in the sensor array into at least one locally consistent cluster based on the spatial positional relationship of each measurement point and the node reading residual of each measurement point. The results analysis module is used to analyze the consistency between the topology of at least one locally consistent cluster and the physical spatial layout of the sensor array, and obtain the topology consistency analysis results. The condition generation module is used to determine whether the preset virtual reference value generation conditions are met based on the topology consistency analysis results. The calibration execution module is used to generate a final virtual reference value based on the readings of all sensors and calibrate each sensor in the sensor array when the conditions are met.

Images