A method for extracting mine pressure features from hydraulic support working resistance monitoring data
By mapping and standardizing the fields of monitoring data of hydraulic supports in coal mines, valid peaks and peak inflection points are identified and verified, solving the problems of chaotic and scattered data formats. This enables accurate extraction of representative points of high load on supports and comparison of working face mine pressure, supporting reliable assessment of roof pressure changes and support load-bearing status.
Patent Information
- Application Number
- CN202611257783.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-19
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, the monitoring data of working resistance of hydraulic supports in coal mines suffers from problems such as inconsistent field names, chaotic formats, interference from invalid records, scattered data, and difficulty in forming a scale comparison of the working face, which makes it difficult to accurately reflect the high load and pressure state of the supports.
Data is processed through field mapping, date standardization, and stent number standardization. Candidate peaks are identified and valid peaks are verified and screened. Combined with peak inflection points and high-level supplementation mechanisms, representative high-load points are extracted daily. The average daily representative working resistance of a single stent is calculated, and the average load index of each zone is calculated by group.
It can adapt to data from different coal mine systems, accurately extract representative points of high load on the supports, avoid interference from ordinary fluctuation data, and provide comparative data of mine pressure at the working face spatial scale, which facilitates the analysis of roof pressure changes and the assessment of support bearing status.
Smart Images

Figure CN123115753A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of coal mine pressure monitoring data processing technology, specifically to a method for extracting mine pressure features from hydraulic support working resistance monitoring data. Background Technology
[0002] During the advancement of a fully mechanized mining face, hydraulic supports play a crucial role in roof support and the transfer of loads from the surrounding rock. The working resistance of these supports is a vital monitoring indicator reflecting changes in roof pressure, the support's load-bearing capacity, and the characteristics of mine pressure manifestation at the working face. Currently, coal mine sites can typically export large amounts of data in Excel spreadsheets from the support electro-hydraulic control system or mine pressure monitoring system. This data includes information such as acquisition time, support number, support working resistance, and pressure on the left and right columns.
[0003] However, raw mine pressure monitoring data often presents several problems before it can be directly used for analysis. First, field names exported from different coal mines and monitoring systems are inconsistent; the time column, support number column, and support working resistance column may have multiple header formats. Second, data from different supports on the same working face may be scattered across multiple files, or data from multiple supports and sampling times may exist on the same date. Third, the original time format may be inconsistent due to inconsistent spreadsheet cell formats, dates stored as serialized values, or differences in export formats from different monitoring systems; support working resistance data may also contain null values, text values, or invalid values. Fourth, directly averaging all raw data will mix in data such as support lowering, support shifting, low-resistance platforms, and ordinary fluctuations, resulting in daily-scale representative values that cannot fully reflect the high-load pressure state of the supports. Fifth, if single-support results are not further categorized by date and spatial location, it is difficult to form an overall mine pressure comparison at the working face scale.
[0004] Traditional manual processing methods typically rely on manually selecting peak values, copying and pasting date data, and manually organizing support numbers, which is prone to omissions and inconsistencies in standards. While general data cleaning methods exist to handle format conversion and table merging, they often directly use the cleaned data for statistical averaging, failing to distinguish between effective peaks, peak inflection points, and high-level supplementary points in the hydraulic support working resistance curve, and lacking a processing chain from single-support feature extraction to working face partition summarization. Therefore, a method is needed that can extract high-load representative points from the original support working resistance data according to fixed rules, and further generate daily representative values for single supports and comparative data for working face partitions. Summary of the Invention
[0005] To address the technical problems mentioned above, this invention provides a method for extracting mine pressure features from hydraulic support working resistance monitoring data, comprising the following steps: S1. Obtain several raw data files of hydraulic support working resistance. Perform field mapping, date standardization, support number standardization, and support working resistance numerical processing on the raw data files to obtain cleaned data. Based on the cleaned data, group the data according to support number or installation location to obtain single support time series data. S2. Based on the time series data of a single support, candidate peaks are identified for the time series of each support. The candidate peaks are verified and screened to extract the effective peaks. Starting from the effective peaks, peak inflection points are identified according to preset conditions. The effective peaks, peak inflection points, and representative points obtained after high-level supplementation are merged to obtain a set of daily high-load representative points. S3. Calculate the daily average representative working resistance of a single support based on the daily high-load representative point set; calculate the group average value according to the support group based on the daily average representative working resistance of a single support, and obtain the average load index of the working face zone.
[0006] Preferably, in step S1, the field mapping step includes: Standardize the original column headers by removing leading and trailing spaces, unifying full-width and half-width characters, unifying Chinese and English brackets, ignoring spaces and common separators, and ignoring capitalization. Following the order of user-filled column header priority, built-in alias exact matching, and fuzzy keyword matching, the original column headers are mapped to internal standard fields, which include at least the data time, installation location, and bracket working resistance fields.
[0007] Preferably, in step S1, the steps of date standardization, stent number standardization, and stent working resistance numericalization include: Convert the data time to a unified date and time format and extract the natural date; convert the stent working resistance to numerical data and discard records that cannot be converted to numerical values; Extract the bracket number from the installation location field or file name, compatible with formats including hash symbols, formats without hash symbols, and formats containing a mixture of Chinese characters and numbers. The output is uniformly formatted as n# or the corresponding numerical bracket number.
[0008] Preferably, in step S2, the candidate peak identification step includes: Scan each sampling point in chronological order. When the working resistance of the support at a certain sampling point is greater than the previous different value and greater than the next different value, this sampling point is identified as a candidate peak. If a series of consecutive high-value plateaus appear, the middle position of the consecutive equal segment is selected as the candidate peak position.
[0009] Preferably, in step S2, the step of verifying and screening candidate peaks and extracting valid peaks includes: Establish time windows of preset time lengths before and after each candidate peak time. If within a time window there is at least one comparison point that is less than or equal to a preset proportion of the candidate peak value, and there is no other comparison point that is greater than or equal to a preset proportion of the candidate peak value, then the candidate peak value passes the window verification. Calculate the daily high reference value, which is the average of the top 10% of the effective support working resistance data for the day. Candidate peaks that pass the window verification and whose peak value is not lower than 80% of the daily high reference value are identified as valid peaks.
[0010] Preferably, in step S2, the step of identifying the peak inflection point according to preset conditions includes: Starting from the effective peak, scan backwards and calculate the decrease between two adjacent sampling points; When the adjacent decrease does not exceed the preset descent threshold, continue scanning until the adjacent decrease exceeds the preset descent threshold or an invalid value is encountered, at which point the scanning stops. When the adjacent descent within a peak does not exceed 8%, the actual duration of the peak is greater than 4 hours, the peak end value is not less than 80% of the effective peak start value, the descent of the next sampling point after the peak end is not less than 15% relative to the peak end, and the peak contains at least 2 sampling points, the current continuous segment is identified as a peak, the peak start is output as the effective peak value, and the end of the continuous segment of the peak is output as the peak inflection point.
[0011] Preferably, in step S2, the high-level padding step includes: If the number of representative points of valid peaks and peak inflection points identified on the same day is insufficient, then the following steps are taken in sequence: supplementing invalid candidate peaks at higher levels, verifying the daily maximum value, and using higher quantile values as a safety net. The supplementary high-level invalid candidate peaks are selected from candidate peaks that have never been identified as valid peaks. The candidate peaks with values not lower than 90% of the high reference value of the day and at least 1 hour apart from the selected representative points are selected as supplementary representative points. The daily maximum value check is performed if the maximum value among the selected representative points is lower than 95% of the maximum value of the source data for that day, then the daily maximum value is added. The high percentile value fallback uses the 95th percentile and 90th percentile values as fallback supplementary points.
[0012] Preferably, in step S3, the step of calculating the group mean according to the stent group includes: Users can divide the support into three groups: upper, middle, and lower, based on the layout of the work surface. The groups can be entered in either range or list format. For a given date and a given group, the group average is calculated based on the daily average working resistance of valid and non-zero stents within that group. Stents with missing stents, invalid stent working resistance records, and stents with a daily average working resistance of 0 are not included in the calculation. The output records the group name, the range of stents in the group, the number of stents involved, and the exclusion status.
[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention automatically adapts to multi-format data exported from different coal mines and monitoring systems by performing field mapping, date standardization, support number standardization, and support working resistance numerical processing on the original Excel data. This solves the problems of inconsistent field names, format confusion, and interference from invalid records in the original mine pressure monitoring data. By identifying candidate peaks and extracting valid peaks in the single-support time series, combined with peak inflection point identification and high-level supplementation mechanisms, the extracted daily high-load representative points fully reflect the sustained high-load and unloading transition processes experienced by the supports under roof pressure. This avoids the problem of directly averaging the original full-day data and mixing in support descent, shifting, low-resistance platforms, and ordinary fluctuation data, which would otherwise prevent the daily-scale representative values from fully reflecting the high-load pressure state of the supports. Furthermore, by calculating the daily average representative working resistance of a single support and summarizing by date and upper, middle, and lower groups, the single-support feature extraction results can be expanded into mine pressure comparison data at the working face spatial scale. This facilitates comparison of the load-bearing differences of supports in different zones on the same date, providing a reliable data foundation for analyzing roof pressure changes and assessing the support load-bearing status. Attached Figure Description
[0014] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0015] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation
[0016] 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 skilled in the art without creative effort are within the scope of protection of the present invention.
[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0018] Example 1 like Figure 1 As shown, this embodiment of the invention provides a method for extracting mine pressure features from hydraulic support working resistance monitoring data, including the following steps: S1. Obtain several raw data files of hydraulic support working resistance. Perform field mapping, date standardization, support number standardization, and support working resistance numerical processing on the raw data files to obtain cleaned data. Based on the cleaned data, group the data according to support number or installation location to obtain single support time series data. This step transforms the raw Excel data exported from the coal mine site into standardized, foundational data that can be used for subsequent peak value identification. This step does not determine the mine pressure peak; it primarily addresses whether the data can be read, whether the supports can be assigned, and whether the date and support working resistance can be included in the calculations.
[0019] S1.1 Obtain one or more raw data files of hydraulic support working resistance. The files are in .xls or .xlsx format, or are table files exported from different mine pressure monitoring systems. Each raw data file contains at least three types of core information: data time, support number or installation location, and support working resistance. When the raw data is already a single support time series, it must at least contain the data time and support working resistance. The support number is obtained by parsing the file name.
[0020] To accommodate data exported from different mine pressure monitoring systems, this step standardizes the original column headers, including removing leading and trailing spaces, unifying full-width and half-width characters, unifying Chinese and English parentheses, ignoring spaces and common separators, and ignoring capitalization. Then, following the order of "user-filled column headers first, built-in alias exact matching second, and fuzzy keyword matching last," the original column headers are mapped to internal standard fields. These internal standard fields include at least data time, installation location, and support working resistance fields. The "whole support (KN)" field from the original table is used as support working resistance data in the calculation; the left column, right column, and operating status fields are optional and do not affect the core calculations based on time and support working resistance if they are not identified.
[0021] S1.2 After field mapping, the data time is uniformly converted to date and time format, and the natural date is further extracted as the basis for subsequent daily calculations. For date display anomalies caused by inconsistent spreadsheet cell formats, dates stored as serialized values, or differences in export formats from different monitoring systems, the data is converted to the true date according to a unified date parsing rule and written out in a unified date format during the output stage to avoid inconsistent display of the same date in different files or different processing stages.
[0022] The stent working resistance is converted into numerical data; records that cannot be converted into numerical data are discarded or recorded based on whether they affect the core calculation; records with empty time, unresolved time, missing stent number, or missing stent working resistance are considered invalid records and will not participate in the subsequent selection of valid peaks.
[0023] The bracket number is extracted from the installation location field or file name, and is compatible with formats such as "5", "5#", "5# bracket", "5# bracket", "3103-21-5#", and "32183-30#32183". The output is uniformly formatted as n# or the corresponding numeric bracket number, ensuring that bracket data from different sources can be arranged in the actual bracket order.
[0024] (1) Classified by number of flights The cleaned data is grouped according to the support number or installation location, with each group corresponding to the time series data of one hydraulic support. For data of the same support, it is sorted by data time; when there are the same timestamps, the sorting is maintained according to the original file order and original line number to avoid the order of records at the same time point being disordered in subsequent peak identification.
[0025] The output results categorized by flight are single-stent time series files. Each file stores the stent working resistance change process of a certain stent within the monitoring period, and this result serves as the input for step S2.
[0026] (2) Classified by time In addition to categorizing by frame count, the data from the cleaned multi-supports is also categorized by date: an Excel file is output for each N-day group, with each sheet in each file corresponding to one day. Within each sheet, data is sorted by support number and data time, retaining the original or post-cleaning monitoring data for each support on that day. The significance of categorizing by time is to provide a basic table for subsequent manual verification, comparison of mine pressure conditions on a given day at the working face, and graphical visualization, while also enabling horizontal observation of data from different supports on the same day.
[0027] (3) Functional representation The functional representation involved in this step is as follows: Let the original data table be A={a_k}, where a_k represents the k-th original record, which includes the original time field, the original support field, and the original support working resistance field.
[0028] ① Column header normalization function: N(c)=lower(remove_symbol(unify(c))) Where 'c' is the original column header, 'unify(c)' means unifying full-width / half-width characters and Chinese and English parentheses, 'remove_symbol(·)' means removing spaces, underscores, hyphens, colons and common separators, and 'lower(·)' means ignoring case differences.
[0029] ② Field mapping function: Let A_j be the set of aliases for the j-th standard field, and K_j be the set of keywords. Then the field mapping function is expressed as: F_j = first{c∈C|There exists a∈A_j such that N(c) = N(a), or there exists q∈K_j such that q⊂N(c)}, Where C is the original set of column headers, N(·) is the column header normalization function, and first means taking the first column header that meets the conditions in the order of "user-filled column header priority, built-in alias exact matching, and fuzzy keyword matching"; the mapping result includes at least three types of fields: data time, bracket number or installation location, and bracket working resistance.
[0030] ③ Time analysis function: T_k=parse_time(x_k),D_k=date(T_k) Where x_k is the original time value, T_k is the standard date and time, and D_k is the natural date; records that cannot be parsed into valid time will not be included in the subsequent peak identification.
[0031] ④ Stent number parsing function: S_k=parse_support(y_k) When y_k contains "#" or "#", the numbers before the hash symbol are extracted first; otherwise, the last valid digit string in the text is extracted and uniformly represented as n#.
[0032] ⑤ Numerical function of stent working resistance: R_k = parse_num(z_k), Where z_k is the working resistance of the support or the value of the "whole support (KN)" field in the original table; if R_k cannot be converted into a numerical value, the record will not participate in the core calculation of step S2.
[0033] ⑥ Output set according to bracket: A_i={(T_k,D_k,S_k,R_k)|S_k=i, T_k and R_k are valid}; Output collection by time: B_d={(T_k,D_k,S_k,R_k)|D_k=d,T_k and S_k are valid}.
[0034] The above A_i is used as the input for step S2, and B_d is used for the verification of basic data of multiple stents on the same day. When B_d is used as the basic verification table, invalid records of stent working resistance R_k can be retained and marked with status. When the records in B_d are used in subsequent calculations of valid peaks, representative points or averages, R_k should be a valid value.
[0035] S2. Based on the time series data of a single support, candidate peaks are identified for the time series of each support. The candidate peaks are verified and screened to extract the effective peaks. Starting from the effective peaks, peak inflection points are identified according to preset conditions. The effective peaks, peak inflection points, and representative points obtained after high-level supplementation are merged to obtain a set of daily high-load representative points. This step involves extracting the characteristics of mine pressure on a single support. Instead of directly averaging the original daily data, it extracts representative points of high load from the single support's working resistance curve. Representative characteristic points include effective peak values, peak inflection points, supplementary points for high-level invalid candidate peaks, daily maximum value verification points, and supplementary points for high-level quantile values.
[0036] S2.1 For the time series of working resistance of a single support, first scan each sampling point in chronological order; when the working resistance of the support at a certain sampling point is greater than the previous different value and greater than the next different value, the point is identified as a candidate peak.
[0037] If a series of consecutive equal high-value plateaus appear in the time series, and the entire consecutive equal segment is higher than the values before and after it, then not all points within the plateau are selected as candidate peaks. Instead, the middle position of the consecutive equal segment is selected as the candidate peak position to avoid the candidate peak being repeated due to consecutive equal value points.
[0038] S2.2. Candidate peaks only represent local high-value points and cannot be directly equated with effective peaks in the sense of mine pressure. This step establishes time windows of preset lengths before and after each candidate peak time; in this embodiment, the time windows are 3 hours before and after.
[0039] Let the candidate peak value be P_i. Within this time window, if there is at least one comparison point less than or equal to 0.94P_i, and no comparison point greater than or equal to 1.06P_i, then the candidate peak is verified through the local time window. The engineering implications are: a significant drop near the candidate peak indicates that the point has high load characteristics; simultaneously, the absence of higher load points within the window indicates that the point is not covered by stronger peak values.
[0040] For data with stable sampling intervals, a location slicing window is used to improve computational efficiency; for data with unstable sampling intervals, a true time window is preferred to ensure the accuracy of the temporal significance of the selected effective peaks. After verification through a local time window, the daily high reference value is further calculated, which is the average of the top 10% of the highest values in the daily effective support working resistance data. Candidate peaks must also meet the requirement that their peak value is not lower than 80% of the daily high reference value to be considered effective peaks, in order to filter out small local protrusions on the low-level platform and avoid misjudgment.
[0041] S2.3 In this embodiment, a peak refers to a load evolution segment in the hydraulic support's working resistance time series, starting from a pressure peak that has passed effective peak verification and meets the high load limit. After the peak, the load does not immediately and rapidly unload, but rather remains relatively high or slowly decreases within a subsequent continuous sampling period, culminating in a significant sudden drop at the end of this continuous period. A peak is not a single peak point, nor is it a simple plateau of equal values; rather, it is a mine pressure response structure composed of "effective peak start point – high-level maintenance or slow decline segment – peak end turning point – subsequent significant sudden drop segment." The mine pressure significance of the peak lies in reflecting the sustained high load and subsequent unloading turning point experienced by the support under the action of the roof. The effective peak emphasizes a single point of high load, the peak emphasizes the sustained high load process, and the peak turning point emphasizes the location where this sustained high load process ends and significant unloading is imminent.
[0042] S2.4 Peak identification must start with a valid peak. The program arranges all valid peak positions in chronological order and scans backward from each valid peak. During the scan, the descent rate between two adjacent sampling points is calculated: Adjacent descent range = (current point support working resistance - next point support working resistance) / current point support working resistance.
[0043] When the adjacent descent amplitude does not exceed the preset descent threshold, the next point is considered to still belong to the peak continuum segment; when the adjacent descent amplitude exceeds the preset descent threshold or an invalid value is encountered, the scanning of the current continuum segment is stopped. In this embodiment, the adjacent descent amplitude within the peak does not exceed 8%, the actual duration of the peak is greater than 4 hours, the peak end value is not less than 80% of the effective peak start value, the next sampling point after the peak end has a descent amplitude of not less than 15% relative to the peak end, and the peak contains at least 2 sampling points.
[0044] When the above conditions are met simultaneously, the continuous segment is identified as a peak. The starting point of the peak is output as the effective peak value, and the end of the continuous peak segment is output as the peak inflection point. The duration of the peak and the ratio of the peak end to the starting point are recorded. Although the peak inflection point is not necessarily the local highest point, it represents the end of the high-load continuous segment and the position where significant unloading is about to occur. Therefore, it is used as an independent representative point of mine pressure characteristic in the subsequent calculation of the daily representative working resistance of the support.
[0045] S2.5. For a given day, if insufficient effective peak values and peak inflection points are identified, instead of using median values, 35% values, or other low-to-mid-range statistical points for forced supplementation, representative points are supplemented around the high-load range of that day. First, the high-level reference value for the day is calculated, prioritizing the retention of identified effective peak values and peak inflection points. If the number of representative points is insufficient, the process proceeds sequentially to supplement invalid candidate peaks at high levels, verify the daily maximum value, and use high-level quantile values as a fallback. The purpose of supplementing daily high-load representative points is to ensure that the samples used in step S3 still originate from the higher load range of that day. If the number is still insufficient after all high-level supplementation, existing representative points are retained, and the date is marked "Insufficient high-load representative points for the day" for use in subsequent calculation result verification.
[0046] S2.6. For a given day for a specific stent, if a valid peak or peak inflection point has been identified, but the number of representative records is insufficient, daily high-load representative points should be added. Priority should be given to selecting points that still have high-load representative significance from candidate peaks that have never been identified as valid peaks. The supplementary points should come from a local candidate peak set, and their values should not be lower than 90% of the high reference value for the day. Furthermore, the supplementary point should be at least one hour apart from the selected representative points and should not overlap with the time and value of the selected records.
[0047] If the number of invalid candidate peaks is still insufficient after supplementation, the daily maximum value verification will continue: if the maximum value among the selected representative points for the day is lower than 95% of the maximum value of the source data for the day, additional maximum values for the day will be added as verification points. This verification does not change the determination of valid peaks, but is only used to prevent the highest load of the day from not being covered by representative points.
[0048] If, after verification using effective peaks, peak inflection points, high-level invalid candidate peaks, and daily maximum values, there are still fewer than three representative points, then the 95% and 90% high-level values are used as fallback supplementary points. The program calculates the corresponding high-level quantile value in the effective support working resistance data for the day and finds the record in the original data that is closest to the target quantile value as a supplementary point. If there are still fewer than three points after all high-load supplementation steps are completed, then mid-to-low-level statistical points are no longer introduced; instead, existing representative points are retained and marked "Insufficient high-load representative points for the day".
[0049] S2.7. Let the effective time series of a certain support i be X_i={(t_k,R_k)}, where t_k is the k-th sampling time, R_k is the working resistance of the corresponding support, and the series is arranged in chronological order. The effective peak, peak plateau, and daily high load representative point in this step are represented by the following function rules.
[0050] (1) Candidate peak function: C_i={k|R_k>R_l and R_k>R_r}, Where l is the position of the nearest valid sampling point that is not equal to R_k before k, and r is the position of the nearest valid sampling point that is not equal to R_k after k; if the continuous equal value high-level segment is [a,b], then the candidate peak position is taken as k=a+floor((b-a+1) / 2), that is, the middle position of the continuous equal value segment is taken; when the length of the continuous segment is even, the sampling point slightly behind the middle is taken.
[0051] (2) Time window function: W_k(τ)={m|m≠k,|t_m-t_k|≤τ} The preferred value is τ=3h. The condition for candidate peak k to pass the window verification is: there exists m∈W_k(τ) such that R_m≤0.94R_k, and there does not exist m∈W_k(τ) such that R_m≥1.06R_k.
[0052] (3) Daily high reference value function: When n_{i,d}^{all}=0, H_{i,d} is not calculated, and the support does not enter the effective peak and high load compensation calculation for that day; when n_{i,d}^{all}>0... H_{i,d}=mean(Top_{q}{R_k|D_k=d,R_k is valid}), Where q=max(1,ceil(0.10n_{i,d}^{all})), n_{i,d}^{all} is the number of effective working resistance records of stent i within date d; that is, the average of the top 10% of the effective stent working resistance of the stent on that day is taken, and at least one high value is taken when there are few effective records.
[0053] (4) Effective peak function: P_{i,d}={k∈C_i|D_k=d, Where k is verified through the W_k(3h) window and R_k≥0.80H_{i,d}}; local candidate peaks that do not meet this high load line are not considered valid peaks.
[0054] (5) Function for the descent amplitude inside the peak: δ_m=(R_m-R_{m+1}) / R_m Scanning backward from the effective peak k, if any adjacent sampling points within the continuous segment [k,j] satisfy δ_m≤0.08, it is considered to still be in the high load maintenance or slow decline segment.
[0055] (6) Peak persistence and end-point constraints: Δt_{k,j}=t_j-t_k, L_{k,j}=R_j / R_k; The preferred conditions are: Δt_{k,j}>4h, L_{k,j}≥0.80 And the number of sampling points is no less than 2.
[0056] (7) The function of sudden drop after the peak: η_j=(R_j-R_{j+1}) / R_j When there is a valid sampling point at j+1 and η_j≥0.15, [k,j] is identified as a peak, and the valid peak corresponding to k and the peak inflection point corresponding to j are output.
[0057] (8) High-level invalid candidate peak supplementation function: In the process of supplementing high-level invalid candidate peaks, verifying daily maximum values, and providing a safety net for high-level quantile values, E^{cur}_{i,d} represents the set of representative points selected before the start of the current step. U_{i,d}={k∈C_i|D_k=d,k∉P_{i,d},R_k≥0.90H_{i,d},[(E^{cur}_{i,d}=∅) or min_{s∈E^{cur}_{i,d}}|t_k-t_s|≥1h],and (t_k,R_k) is not repeated with the selected records in E^{cur}_{i,d}}}. This function is used to supplement high-load representative points in local candidate peaks that have never been identified as valid peaks.
[0058] (9) Daily maximum value verification function: R^{max}_{i,d}=max{R_k|D_k=d,R_k is valid}, m^*_{i,d}=argmax{R_k|D_k=d,R_k is valid}. If E^{cur}_{i,d} is empty, or max{R_s|s∈E^{cur}_{i,d}}<0.95R^{max}_{i,d}, and (t_{m^*},R_{m^*}) does not repeat the selected records in E^{cur}_{i,d}, then M_{i,d}={m^*_{i,d}}; otherwise M_{i,d}=∅. M_{i,d} represents the set of supplementary points determined by the daily maximum value verification, which does not change the determination of valid peaks.
[0059] (10) High quantile catch-all function: q^{val}_{i,d,p}=quantile_p{R_k|D_k=d,R_k is valid}, where p∈{0.95,0.90}; the position of the supplementary point is taken as k^*_{i,d,p}=argmin_{k:D_k=d,R_k is valid}|R_k-q^{val}_{i,d,p}|. For p∈{0.95,0.90}, the supplementary points k^*_{i,d,p} are determined respectively, and the points that do not overlap with the selected representative points in E^{cur}_{i,d} form the high quantile supplementary point set Q_{i,d}.
[0060] (11) Set of representative high-load points for the day: The final set of representative high-load points for the day is denoted as E_{i,d}=P_{i,d}∪T_{i,d}∪U_{i,d}∪M_{i,d}∪Q_{i,d}, where T_{i,d} is the set of peak turning points, M_{i,d} is the set of daily maximum value verification points, and Q_{i,d} is the set of high quantile value supplementary points. If |E_{i,d}|<3, then record I_{i,d}=1, indicating that there are not enough representative high-load points for the day; otherwise, I_{i,d}=0.
[0061] S3. Calculate the daily average representative working resistance of a single support based on the daily high-load representative point set; calculate the group average value according to the support group based on the daily average representative working resistance of a single support, and obtain the average load index of the working face zone.
[0062] This step further transforms the high-load representative points obtained in step S2 into daily-scale representative values and extends the single-support results to the working face spatial scale. This step includes the calculation of daily representative values for single supports and the summarization of multiple supports by date and upper, middle, and lower groups. It is a link connecting the single-support mine pressure feature extraction results with the working face mine pressure comparison analysis results.
[0063] S3.1. For all representative feature points of the same stent within the same natural date, calculate the daily average representative working resistance of the stent on that day using the arithmetic mean method. Let the i-th stent have n representative feature points within date d, and their stent working resistances be P_{i,d,1}, P_{i,d,2}, ..., P_{i,d,n}, then the daily average representative working resistance of the stent is calculated as follows: The average daily representative working resistance of the stent = the sum of the working resistance of the stent at all representative characteristic points on that day / the number of representative characteristic points on that day.
[0064] This value represents the load-bearing status of a support structure on a specific date, as characterized by a high-load representative point. It is not equivalent to the daily maximum value, nor is it equivalent to the ordinary arithmetic mean of all monitoring points throughout the day. Its calculation basis is the effective peak, peak inflection point, and high-level supplementary point. When outputting the daily average representative working resistance of the support structure, the sample size is also retained to indicate how many representative feature points were used to calculate the daily representative value, facilitating subsequent data reliability checks.
[0065] S3.2 Classification by Frame Number uses the frame number as the main line, continuously arranging the daily average representative working resistance of the same frame on different dates to form a single-frame daily-scale time series. This classification method facilitates the observation of the working resistance change trend of a particular frame as the working face advances, and also facilitates corresponding analysis with the spatial location of the frame, roof conditions, and pressure stage. When outputting by frame classification, each frame forms a separate sheet or result file, with fields including date, frame number, daily average representative working resistance of the frame, sample size, and necessary data status information.
[0066] S3.3. Classification by time uses natural dates as the main line, summarizing the daily average representative working resistance of all supports within the same day into a single table. This step is not a simple file export, but rather organizes the daily representative values of a single support into multi-support comparative data for a specific date on the working face, used to observe the load-bearing differences between different support positions on the same day. In this embodiment, an Excel file is output every N days, with each sheet in each file corresponding to one day; each sheet lists the daily average representative working resistance of all identified supports for that day. When a support has no data for a given day or its working resistance is invalid, the support is not directly deleted, but rather retained and marked as missing or invalid for subsequent verification of data integrity.
[0067] S3.4 This step expands the calculation results of a single support into comparative data on mine pressure at the working face spatial scale. Since supports at different locations on the working face may be affected differently by the roof structure, end effect, coal seam dip angle, and advance disturbance, this method allows users to classify supports into three categories: upper, middle, and lower, based on the actual working face layout.
[0068] Support groupings can be entered in range or list format, such as "5-20" or "5,10,15,20", or a mixed format. Since the number of supports and spatial partitions vary across different work surfaces, this method does not fix a specific support range; instead, it is set by the user based on the location relationships of the work surfaces on site.
[0069] For a given date and group, the group average is calculated based on the daily representative working resistance of valid and non-zero supports within that group. Supports with missing or invalid support working resistance records, and supports with a daily representative working resistance of 0, are excluded from the calculation. The denominator is the actual number of valid supports involved in the calculation. The output also records the group name, the range of supports in the group, the number of participating supports, and exclusion criteria. The upper, middle, and lower group averages are used to convert the daily representative value of each support into the average bearing capacity index for the working face zone. This result can be used to compare the bearing capacity differences of upper, middle, and lower supports on the same date, and can also be used to subsequently plot the working face zone mine pressure change curve.
[0070] S3.5 Let E_{i,d} be the set of high-load representative points obtained by the i-th support in step S2 within date d, and R_s be the support working resistance corresponding to the s-th representative point in the set. The calculation in this step is expressed as the following function.
[0071] (1) The daily representative working resistance function of a single stent: V_{i,d}=(1 / n_{i,d})Σ_{s∈E_{i,d}}R_s, where n_{i,d}=|E_{i,d}|; if n_{i,d}=0, then there is no daily representative value to be calculated for that stent.
[0072] (2) Sample number function: N_{i,d}=|E_{i,d}|, which is output synchronously with V_{i,d} and is used to explain how many high-load representative points are used to calculate the daily representative value.
[0073] (3) Multi-staff matrix by date: A_d={(i,V_{i,d},N_{i,d},status_{i,d})|i∈S}, where S is the set of identified staffs, and status_{i,d} is used to mark the status such as normal, missing, invalid or containing invalid records.
[0074] (4) Working face group set: G_g={i|i belongs to the g-th group set by the user}, g∈{upper part, middle part, lower part}; the group is determined by a continuous range or a list of discrete support numbers.
[0075] (5) Grouped mean function: V_{g,d}=Σ_{i∈G_g}V_{i,d}·I(valid(V_{i,d})∧V_{i,d}≠0) / Σ_{i∈G_g}I(valid(V_{i,d})∧V_{i,d}≠0), Where I(·) is the indicator function, which takes the value 1 if the condition is met, and 0 otherwise; when the denominator is 0, the mean value is not output for the date and the group is marked as having no valid participants.
[0076] (6) Exclude the stent number function: N^{excl}_{g,d}=|G_g|-Σ_{i∈G_g}I(valid(V_{i,d})∧V_{i,d}≠0), This value is used to record the number of stents that were not included in the mean calculation due to missing, invalid, or zero values in a specific date and group.
[0077] S3.6 The method described in this embodiment is further implemented as a graphical operation software. This software provides functions such as selecting the original data folder, filling in field mappings, setting the output path, extracting valid peaks and peak tiers, calculating the daily representative value for a single support, summarizing multiple supports by date, and calculating the average value of upper, middle, and lower groupings. The graphical operation software is only used to call the aforementioned field mapping function, peak filtering function, peak tier identification function, representative value calculation function, and grouping summary function. The software code, interface layout, and button styles are not the focus of protection in this method.
[0078] Example 2 This embodiment will demonstrate the correspondence between core calculation rules and formula functions. To facilitate understanding of the technical solution of this method, the following uses textual calculation rules and formula functions to explain each core step. The code module is only used as a carrier for implementing this method.
[0079] (1) Raw data processing: N(c)=lower(remove_symbol(unify(c))); S_k = parse_support(y_k); R_k = parse_num(z_k); A_i={(T_k,D_k,S_k,R_k)|S_k=i}.
[0080] (2) Candidate peaks and valid peaks: C_i = {k|R_k>R_l and R_k>R_r}; W_k(3h)={m|m≠k, |t_m-t_k|≤3h}; Valid peaks must meet the window validation condition and R_k≥0.80H_{i,d}.
[0081] (3) Peak identification: δ_m=(R_m-R_{m+1}) / R_m≤0.08, Δt_{k,j}>4h, R_j / R_k≥0.80, And when there is a valid sampling point at j+1 and η_j=(R_j-R_{j+1}) / R_j≥0.15, the peak inflection point j is output.
[0082] (4) Daily high load representative points: The final set of daily high load representative points is as follows E_{i,d}=P_{i,d}∪T_{i,d}∪U_{i,d}∪M_{i,d}∪Q_{i,d} Where P is the effective peak, T is the peak inflection point, U is the high-level invalid candidate peak, M is the set of daily maximum value verification points, and Q is the set of high-level quantile value supplementation points; in the process of gradually supplementing U, M, and Q, E^{cur}_{i,d} represents the set of representative points that have been selected before the start of the current step.
[0083] (5) Daily representative value of a single stent and the mean value of each group: V_{i,d}=(1 / n_{i,d})Σ_{s∈E_{i,d}}R_s; V_{g,d}=Σ_{i∈G_g}V_{i,d}I(valid(V_{i,d})∧V_{i,d}≠0) / Σ_{i∈G_g}I(valid(V_{i,d})∧V_{i,d}≠0), where a denominator of 0 indicates no valid scaffolding.
[0084] In a preferred embodiment of this example, the effective peak verification window is 3 hours before and after the candidate peak; the effective peak determination adopts a double constraint of being 6% below and 6% above the peak value; the daily high reference value is the average of the top 10% of the effective support working resistance values of the day; the effective peak must not be lower than 80% of the daily high reference value; the threshold for the adjacent descent within the peak is 8%; the actual duration of the peak is greater than 4 hours; the value at the end of the peak is not lower than 80% of the effective peak starting value; the threshold for the sudden drop after the peak ends is 15%; the high invalid candidate peak must not be lower than 90% of the daily high reference value and must be at least 1 hour apart from the selected point; the daily maximum value verification coverage ratio is 95%; the high percentile value is based on the 95% and 90% high percentile values.
[0085] The above parameters can be appropriately adjusted according to different coal mine monitoring frequencies, rated working resistance of supports, coal mining technology and roof conditions. In specific implementation, they are described as preset thresholds or preferred thresholds, and the specific values used by the current software are given in the embodiments.
[0086] This invention processes raw support resistance monitoring data into daily representative values for a single support and summarized results for different working face zones through three levels and four stages. Its main technical advantages include: First, the raw data processing layer maps the time column, support number column, and support working resistance column exported from different mine pressure monitoring systems into a unified field, and outputs basic data according to support and date.
[0087] Second, the single-support mine pressure feature extraction layer first extracts the effective peak, peak inflection point and high-level supplementary point, and then uses these points as input for Step 3, avoiding the direct averaging of all monitoring points throughout the day.
[0088] Third, the representative value calculation and working face summary hierarchy first obtain the daily representative value of a single support, and then Step 4 generates a multi-support table by date and the average values of the upper / middle / lower groups, so that the results can be used for spatial scale comparison of the working face.
[0089] Fourth, the output retains the sample count, missing status, invalid status, and necessary logs, so that the mean of a certain day, a certain stent, or a certain group can be retrieved to find the corresponding input records and representative point sets.
[0090] Fifth, the graphical software is used to perform path selection, parameter input, and result export in Steps 1 to 4 without changing the data processing method itself.
[0091] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for extracting mine pressure features from hydraulic support working resistance monitoring data, characterized in that, Includes the following steps: S1. Obtain several raw data files of hydraulic support working resistance. Perform field mapping, date standardization, support number standardization, and support working resistance numerical processing on the raw data files to obtain cleaned data. Based on the cleaned data, group the data according to support number or installation location to obtain single support time series data. S2. Based on the time series data of a single support, candidate peaks are identified for the time series of each support. The candidate peaks are verified and screened to extract the effective peaks. Starting from the effective peaks, peak inflection points are identified according to preset conditions. The effective peaks, peak inflection points, and representative points obtained after high-level supplementation are merged to obtain a set of daily high-load representative points. S3. Calculate the daily average representative working resistance of a single support based on the daily high-load representative point set; calculate the group average value according to the support group based on the daily average representative working resistance of a single support, and obtain the average load index of the working face zone.
2. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S1, the field mapping steps include: Standardize the original column headers by removing leading and trailing spaces, unifying full-width and half-width characters, unifying Chinese and English brackets, ignoring spaces and common separators, and ignoring capitalization. Following the order of user-filled column header priority, built-in alias exact matching, and fuzzy keyword matching, the original column headers are mapped to internal standard fields, which include at least the data time, installation location, and bracket working resistance fields.
3. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S1, the steps of date standardization, stent number standardization, and stent working resistance numericalization include: Convert the data time to a unified date and time format and extract the natural date; convert the stent working resistance to numerical data and discard records that cannot be converted to numerical values; Extract the bracket number from the installation location field or file name, compatible with formats including hash symbols, formats without hash symbols, and formats containing a mixture of Chinese characters and numbers. The output is uniformly formatted as n# or the corresponding numerical bracket number.
4. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S2, the candidate peak identification step includes: Scan each sampling point in chronological order. When the working resistance of the support at a certain sampling point is greater than the previous different value and greater than the next different value, this sampling point is identified as a candidate peak. If a series of consecutive high-value plateaus appear, the middle position of the consecutive equal segment is selected as the candidate peak position.
5. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S2, the steps of verifying and screening candidate peaks and extracting valid peaks include: Establish time windows of preset time lengths before and after each candidate peak time. If within a time window there is at least one comparison point that is less than or equal to a preset proportion of the candidate peak value, and there is no other comparison point that is greater than or equal to a preset proportion of the candidate peak value, then the candidate peak value passes the window verification. Calculate the daily high reference value, which is the average of the top 10% of the effective support working resistance data for the day. Candidate peaks that pass the window verification and whose peak value is not lower than 80% of the daily high reference value are identified as valid peaks.
6. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S2, the step of identifying the peak inflection point according to preset conditions includes: Starting from the effective peak, scan backwards and calculate the decrease between two adjacent sampling points; When the adjacent decrease does not exceed the preset descent threshold, continue scanning until the adjacent decrease exceeds the preset descent threshold or an invalid value is encountered, at which point the scanning stops. When the adjacent descent within a peak does not exceed 8%, the actual duration of the peak is greater than 4 hours, the peak end value is not less than 80% of the effective peak start value, the descent of the next sampling point after the peak end is not less than 15% relative to the peak end, and the peak contains at least 2 sampling points, the current continuous segment is identified as a peak, the peak start is output as the effective peak value, and the end of the continuous segment of the peak is output as the peak inflection point.
7. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S2, the high-level padding step includes: If the number of representative points of valid peaks and peak inflection points identified on the same day is insufficient, then the following steps are taken in sequence: supplementing invalid candidate peaks at higher levels, verifying the daily maximum value, and using higher quantile values as a safety net. The supplementary high-level invalid candidate peaks are selected from candidate peaks that have never been identified as valid peaks. The candidate peaks with values not lower than 90% of the high reference value of the day and at least 1 hour apart from the selected representative points are selected as supplementary representative points. The daily maximum value check is performed if the maximum value among the selected representative points is lower than 95% of the maximum value of the source data for that day, then the daily maximum value is added. The high percentile value fallback uses the 95th percentile and 90th percentile values as fallback supplementary points.
8. The method for extracting mine pressure features from hydraulic support working resistance monitoring data according to claim 1, characterized in that, In step S3, the step of calculating the group mean according to the stent group includes: Users can divide the support into three groups: upper, middle, and lower, based on the layout of the work surface. The groups can be entered in either range or list format. For a given date and a given group, the group average is calculated based on the daily average working resistance of valid and non-zero stents within that group. Stents with missing stents, invalid stent working resistance records, and stents with a daily average working resistance of 0 are not included in the calculation. The output records the group name, the range of stents in the group, the number of stents involved, and the exclusion status.
