Hydraulic support supporting whole process working condition identification method based on pressure data analysis

Abstract

The present application relates to a hydraulic support supporting whole process working condition identification method based on pressure data analysis, belonging to the technical field of intelligent coal mine. It comprises: collecting the original stand column pressure time sequence in the operation process of the target hydraulic support, sequentially carrying out noise processing and abnormal value processing on the original stand column pressure time sequence, obtaining the target stand column pressure time sequence, extracting the multi-dimensional working condition characteristics of all target pressure data points in the target stand column pressure time sequence, inputting the multi-dimensional working condition characteristics of all target pressure data points into the pre-trained hydraulic support working condition identification model, and determining the current supporting working condition corresponding to the current stand column pressure time sequence according to the target pressure data points and their working condition labels output by the hydraulic support working condition identification model. The misjudgment and omission rate caused by artificial experience judgment is reduced, and the accuracy and real-time performance of the hydraulic support supporting whole process working condition identification result are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent coal mines, and particularly relates to a hydraulic support whole-process working condition recognition method based on pressure data analysis. BACKGROUND

[0002] With the acceleration of the transformation of the coal industry towards intelligence and unmanned, intelligent fully-mechanized coal mining faces gradually become the key component of the modernization of coal mines. As the core device for roof control in fully-mechanized coal mining equipment, the running state of the hydraulic support is directly related to the safety of mine production and the efficiency of equipment operation. In the actual operation process, due to the complex underground environment, variable working conditions, and fluctuation of sensor reliability, etc., abnormal working conditions such as unqualified initial support force, abnormal resistance, abnormal working resistance, and safety valve opening often occur in the hydraulic support. These abnormal working conditions will seriously affect the continuity and safety of coal mining operations.

[0003] At present, the working condition recognition of the whole process of hydraulic support is mainly based on manual experience, which has the problems of poor real-time performance, strong subjectivity, high misjudgment and omission rate, and cannot meet the dual requirements of high accuracy and real-time of support working condition monitoring in intelligent mines. Although some mines have deployed pressure sensors to realize the collection of hydraulic support pressure data, due to the lack of effective data analysis and working condition recognition mechanism, the massive data collected cannot fully play its value. SUMMARY

[0004] To solve the above technical problems, the present application provides a hydraulic support whole-process working condition recognition method based on pressure data analysis. The technical scheme of the present application is as follows:

[0005] The present application provides a hydraulic support whole-process working condition recognition method based on pressure data analysis, which comprises:

[0006] S1, collecting the original column pressure time sequence in the running process of the target hydraulic support, wherein the original column pressure time sequence comprises the current column pressure time sequence;

[0007] S2, sequentially performing noise processing and outlier processing on the original column pressure time sequence to obtain the target column pressure time sequence;

[0008] S3, extracting the multi-dimensional working condition features of all target pressure data points in the target column pressure time sequence;

[0009] S4, inputting the multi-dimensional working condition features of all target pressure data points into a pre-trained hydraulic support working condition recognition model, wherein the input of the hydraulic support working condition recognition model is the multi-dimensional working condition features of the pressure data points, and the output is the pressure data points and their working condition labels arranged in time sequence according to the column pressure time sequence;

[0010] S5, determining the current support working condition corresponding to the current column pressure time sequence according to the target pressure data points and the working condition labels output by the hydraulic support working condition identification model.

[0011] Preferably, the noise processing in S2 comprises:

[0012] S21, calculating the pressure difference value between the previous original pressure data point and the next original pressure data point in the original column pressure time sequence to obtain a pressure difference value sequence;

[0013] S22, adding 0 at the first position of the pressure difference value sequence and shifting the rest positions backward to obtain a new pressure difference value sequence, and aligning each pressure difference value in the new pressure difference value sequence with each original pressure data point in the original column pressure time sequence;

[0014] S23, dividing all the pressure difference values into continuous positive segments and negative segments according to the positive and negative nature of adjacent pressure difference values in the new pressure difference value sequence, and accumulating the pressure difference values in each positive segment and negative segment to obtain the total pressure change difference value of each positive segment and negative segment;

[0015] S24, if the absolute value of the total pressure change difference value of any positive segment or negative segment is less than or equal to a preset pressure difference value threshold, determining the original pressure data points in the original column pressure time sequence aligned with all the pressure difference values in the positive segment or negative segment as noise and deleting them.

[0016] Preferably, in S3, the multi-dimensional working condition features of each target pressure data point include sliding window features and simple features, wherein the sliding window features include mean, maximum, minimum, standard deviation and linear regression rate, and the simple features include time difference and pressure change rate.

[0017] Preferably, when extracting the sliding window features of all target pressure data points in S3, it comprises:

[0018] S31, using a sliding window with a window size of 5 to slide in the target column pressure time sequence with a sliding step of 1 until all target pressure data points in the target column pressure time sequence are slid.

[0019] S32, for the middle target pressure data point at the middle position of any sliding window, determining the mean, maximum, minimum, standard deviation and linear regression rate of the five target pressure data points in the sliding window where the middle target pressure data point is located, and taking the mean, maximum, minimum, standard deviation and linear regression rate as the sliding window features of the middle target pressure data point.

[0020] Preferably, S5 further comprises:

[0021] aggregating the plurality of continuous target pressure data points with the same working condition label into a structured working condition segment according to the time sequence of the original stand pressure time sequence, and recording the time parameter of each working condition segment.

[0022] Preferably, the S5 further comprises:

[0023] If any working condition segment includes less than three target pressure data points, the working condition segment is determined as an isolated event and the working condition label of all target pressure data points included in the working condition segment is deleted.

[0024] Preferably, the S5 further comprises:

[0025] The pressure cumulative change amount of each working condition segment for the lifting column and the lowering column is calculated respectively for each working condition label;

[0026] For any working condition segment with the working condition label for the lifting column, if the pressure cumulative change amount of the working condition segment with the working condition label for the lifting column during the lifting column is less than the preset lifting column pressure cumulative change amount threshold, the lifting column working condition label of all target pressure data points included in the working condition segment with the working condition label for the lifting column is deleted.

[0027] For any working condition segment with the working condition label for the lowering column, if the pressure cumulative change amount of the working condition segment with the working condition label for the lowering column during the lowering column is less than the preset lowering column pressure cumulative change amount threshold, the lowering column working condition label of all target pressure data points included in the working condition segment with the working condition label for the lowering column is deleted.

[0028] Preferably, the S5 further comprises:

[0029] All working condition segments with the working condition label of abnormal resistance increase or adjacent frame moving frame resistance increase are obtained.

[0030] If the working condition label of any working condition segment is abnormal resistance increase, the position data of the coal mining machine and the hydraulic support stroke data of the target hydraulic support in a 30s time period after the end time of the working condition segment are obtained and used to determine whether the coal mining machine passes through the target hydraulic support and whether the target hydraulic support has a position movement, if the coal mining machine passes through the target hydraulic support and the target hydraulic support has a position movement, the working condition label of the working condition segment is determined as adjacent frame moving frame resistance increase, and the working condition label of all target pressure data points included in the working condition segment is modified to adjacent frame moving frame resistance increase.

[0031] If the working condition label of any working condition segment is adjacent frame moving frame increased resistance, the hydraulic support stroke data of the target hydraulic support in a 30s time period after the end time of the working condition segment is obtained, and it is determined whether the target hydraulic support has a position movement, if the target hydraulic support has no position movement, it is determined that the working condition label of the working condition segment is abnormal increased resistance, and the working condition label of all target pressure data points included in the working condition segment is modified to abnormal increased resistance.

[0032] Preferably, S5 further comprises:

[0033] All target pressure data points with the working condition label of initial support force unqualified are obtained, if the pressure of any target pressure data point with the working condition label of initial support force unqualified is greater than a preset initial support force threshold, the working condition label of the target pressure data point is deleted.

[0034] Preferably, S5 further comprises:

[0035] All target pressure data points corresponding to the end time of the working condition segment with the working condition label of column lifting are obtained, if the value of any target pressure data point corresponding to the end time of the working condition segment with the working condition label of column lifting is less than a preset initial support force threshold, the target pressure data point corresponding to the end time of the working condition segment is added with the working condition label of initial support force unqualified.

[0036] All the optional technical solutions described above can be combined arbitrarily, and the application does not provide a detailed description of the structure after combination.

[0037] Through the above scheme, the beneficial effects of the application are as follows:

[0038] By collecting the original column pressure time sequence in the running process of the target hydraulic support, the original column pressure time sequence is processed for noise and abnormal value, the target column pressure time sequence is obtained, and after extracting the multi-dimensional working condition features of all target pressure data points in the target column pressure time sequence, the multi-dimensional working condition features of all target pressure data points are input into the pre-trained hydraulic support working condition recognition model, and the current support working condition is determined according to the target pressure data points and their working condition labels output by the hydraulic support working condition recognition model, which provides a method for automatically identifying the whole process working condition of hydraulic support support based on hydraulic support column pressure data, which can reduce the strong subjectivity caused by manual experience judgment, reduce the misjudgment and omission rate caused by manual experience judgment, and improve the accuracy and real-time performance of the whole process working condition recognition result of the hydraulic support support.

[0039] The above description is only a summary of the technical solutions of the application, in order to more clearly understand the technical means of the application, and the content of the specification can be implemented, the following preferred embodiments of the application are described in detail with reference to the drawings. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of the hydraulic support support condition identification method based on pressure data analysis provided in the embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram showing the time-column pressure change during the entire hydraulic support support process.

[0042] Figure 3 This is a schematic diagram showing the time-coupling pressure change of the column during column raising, column lowering, and adjacent frame movement to increase resistance.

[0043] Figure 4 This is a diagram showing the change in column pressure during the period when the initial support force was not up to standard.

[0044] Figure 5 This is a schematic diagram showing the change in column pressure during the time of safety valve opening and abnormal resistance increase.

[0045] Figure 6 This is a schematic diagram showing the change in column pressure over time when the working resistance is abnormal.

[0046] Figure 7 This is a schematic diagram showing the distribution ratio of the average pressure value. Detailed Implementation

[0047] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0048] In the hydraulic support condition identification method based on pressure data analysis provided in this embodiment of the invention, all steps except for acquiring the original column pressure time sequence (S1) can be implemented by any electronic device with computing capabilities, such as a PC, mobile terminal, or server. The acquisition of the original column pressure time sequence can be achieved by a pressure sensor installed on the target hydraulic support column. After acquiring the original column pressure time sequence, the pressure sensor sends it to the electronic device for further analysis.

[0049] Example 1: As Figure 1 As shown, this embodiment of the invention provides a method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis, including:

[0050] S1, Collect the original column pressure time sequence during the operation of the target hydraulic support, the original column pressure time sequence including the current column pressure time sequence;

[0051] S2, noise processing and outlier processing are performed sequentially on the original column pressure time series to obtain the target column pressure time series;

[0052] S3, extract multi-dimension working condition features of all target pressure data points in the target column pressure time sequence;

[0053] S4, input the multi-dimension working condition features of all target pressure data points into a pre-trained hydraulic support working condition recognition model, wherein the input of the hydraulic support working condition recognition model is the multi-dimension working condition features of the pressure data points, and the output is the pressure data points and their working condition labels arranged in time sequence according to the column pressure time sequence;

[0054] S5, determine the current support working condition corresponding to the current column pressure time sequence according to the target pressure data points and their working condition labels output by the hydraulic support working condition recognition model.

[0055] In this embodiment, the whole support process of the hydraulic support refers to the complete operation cycle experienced by the hydraulic support in completing the roof support task during the advancing process of the coal mining machine in the fully-mechanized coal mining face, which is usually divided into column lifting (initial pressure increasing stage), normal operation (stable load bearing stage, coal cutting influence stage), adjacent support moving and resistance increasing, and column lowering (unloading and support lowering stage), as shown in Figure 2 .

[0056] In this embodiment, the target hydraulic support is any hydraulic support in the hydraulic support cluster consisting of all hydraulic supports in the intelligent fully-mechanized coal mining face that needs to be identified in the working condition of the whole support process.

[0057] In this embodiment, the original column pressure time sequence refers to a sequence arranged by the original pressure data points recorded in time sequence without processing during the operation of the target hydraulic support. The pressure sensor installed on the column of the target hydraulic support adopts a dynamic trigger sampling mechanism, that is, data collection is automatically triggered when fluctuations in the column pressure of the target hydraulic support are detected.

[0058] In this embodiment, noise processing refers to the process of filtering and cleaning irrelevant interference in the original pressure data points in the original column pressure time sequence. Abnormal value processing refers to the process of identifying and processing original pressure data points that deviate significantly from the normal range, for example, when power failure or pressure sensor failure occurs, the value of the original pressure data point is -1, at this time, linear interpolation is used for repair, that is, the average value of the two adjacent original pressure data points that are not -1 is used to fill in, to ensure the continuity of the data.

[0059] In this embodiment, the target pressure data point refers to the pressure data point in the target column pressure time sequence.

[0060] The multi-dimensional working condition features refer to a plurality of different dimensional descriptive features extracted from each target pressure data point in the target column pressure time sequence, including sliding window features and simple features.

[0061] In this embodiment, the hydraulic support working condition recognition model is a model trained by a deep learning method, which aims to recognize the working condition of the whole support process of the hydraulic support according to the pressure data points of the hydraulic support and the multi-dimensional working condition features thereof. The specific network composition of the hydraulic support working condition recognition model is not limited in the embodiment of the application. Preferably, the hydraulic support working condition recognition model selects a random forest model.

[0062] In this embodiment, the working condition label refers to the type of working condition corresponding to each target pressure data point predicted or labeled by the hydraulic support working condition recognition model. The working condition type includes normal working condition and abnormal working condition. The normal working condition includes column lifting, column lowering, and adjacent frame moving and resistance increasing. The column pressure change diagram of the normal working condition is as shown in Figure 3 The column lifting is the process of extending the column of the hydraulic support, making the top beam of the hydraulic support rise and closely adhere to the roof, thereby supporting the roof and maintaining the stability of the working face, during which the column pressure of the hydraulic support rapidly rises. The column lowering is the process of retracting the column of the hydraulic support, making the top beam of the hydraulic support descend, so as to move the frame or adjust the support position, during which the column pressure of the hydraulic support rapidly decreases. The adjacent frame moving and resistance increasing is the process that when the coal cutter cuts coal close to and passes through the current hydraulic support, the newly exposed roof after cutting coal increases the roof control area of the hydraulic support, and at the same time, is affected by the conduction of mining pressure and the unloading of adjacent frame, so that the working resistance of the current hydraulic support rapidly increases. The abnormal working condition includes unqualified initial support force, abnormal resistance increasing, abnormal working resistance, safety valve opening, etc. When the initial support force is unqualified, the hydraulic support cannot timely support the roof, and the roof strata will delaminate and sink after losing support. The column pressure change when the initial support force is unqualified is as shown in Figure 4 The abnormal resistance increasing is the phenomenon that the column pressure suddenly rises during the stable loading process of the hydraulic support, and sometimes accompanied by abnormal opening of the safety valve. The safety valve opening is that when the working resistance of the column of the hydraulic support exceeds the rated working resistance and reaches the set value of the safety valve opening, the safety valve control mechanism is triggered, and the hydraulic system protection function under the overpressure working condition is realized through active load relief. The column pressure changes of the safety valve opening and the abnormal resistance increasing are as shown in Figure 5Abnormal working resistance includes overload operation and underload operation. Underload operation is that the column pressure is low when the hydraulic support works, and long-term low working resistance operation state can cause local pressure increase to form a pressure concentration area, causing excessive roof subsidence and causing major safety accidents. Overload operation is that the column pressure is higher than the normal working pressure when the hydraulic support works, and long-term high working resistance operation state can cause the safety valve to open frequently, damage, reduce the service life of the hydraulic support, and cannot effectively control the roof subsidence. The column pressure change of the abnormal working resistance is as shown in FIG. 2. Figure 6

[0063] In this embodiment, in the training stage, the input of the hydraulic support working condition recognition model is the multi-dimensional working condition features of all training target pressure data points, and the output is all training target pressure data points and their working condition labels. In the application stage, the input of the hydraulic support working condition recognition model is the multi-dimensional working condition features of all target pressure data points, and the output is all target pressure data points and their working condition labels.

[0064] The working principle and beneficial effects of the above technical solution are: by collecting the original column pressure time sequence in the running process of the target hydraulic support, performing noise processing and outlier processing on the original column pressure time sequence, obtaining the target column pressure time sequence, and extracting the multi-dimensional working condition features of all target pressure data points in the target column pressure time sequence, the multi-dimensional working condition features of all target pressure data points are input into the pre-trained hydraulic support working condition recognition model, and the working condition of the current support is determined according to the target pressure data points and their working condition labels output by the hydraulic support working condition recognition model. A method for automatically identifying the working condition of the whole process of hydraulic support based on hydraulic support column pressure data is provided. Through automatic identification, the problem of strong subjectivity caused by manual experience judgment can be reduced, the misjudgment and omission rate caused by manual experience judgment can be reduced, and the accuracy and real-time performance of the working condition identification result of the whole process of hydraulic support are improved.

[0065] In a further example of the present application, the noise processing in S2 includes:

[0066] S21, calculating the pressure difference value between the previous original pressure data point and the previous original pressure data point in the original column pressure time sequence, obtaining a pressure difference value sequence;

[0067] S22, adding 0 at the beginning of the pressure difference value sequence and shifting the rest of the bits backward, obtaining a new pressure difference value sequence, and aligning each pressure difference value in the new pressure difference value sequence with each original pressure data point in the original column pressure time sequence;

[0068] ​S23, based on the sign of adjacent pressure differences in the new pressure difference sequence, divide all pressure differences into continuous positive and negative segments, and sum up the pressure differences of each positive and negative segment to obtain the total pressure change difference of each positive and negative segment;

[0069] S24. If the absolute value of the total pressure change difference in any positive or negative segment is less than or equal to a preset pressure difference threshold, then the original pressure data points in the original column pressure time sequence that are aligned with all pressure differences in the positive or negative segment are identified as noise and deleted.

[0070] In this embodiment, each raw pressure data point has a time node attribute. The pressure difference refers to the pressure difference between two consecutive raw pressure data points in the raw column pressure time series. The pressure difference sequence is a sequence composed of various pressure differences. The total pressure change difference refers to the cumulative value of all pressure differences within a certain positive or negative segment. The preset pressure difference threshold is a pre-set value used to determine whether the total pressure change difference of any positive or negative segment is within the normal range, and the value is generally [0.05, 0.1].

[0071] In this embodiment, a positive segment refers to a continuous segment of pressure differences in the new pressure difference sequence where all pressure differences are positive. A negative segment refers to a continuous segment of pressure differences in the new pressure difference sequence where all pressure differences are negative. The first zero value is assigned to the first positive or negative segment after the new pressure difference sequence is divided. For example, if the preset pressure difference threshold is set to 0.1, and the original column pressure time sequence is 35.1, 35.2, 35.5, 35.4, 35.6, 35.7, then the pressure difference sequence is 0.1, 0.3, -0.1, 0.2, 0.1, and the new pressure difference sequence is 0, 0.1, 0.3, -0.1, 0.2, 0.1. The segments are divided into two: 0, 0.1, and 0.3 are the first positive segment; -0.1 is the negative segment; and 0.2 and 0.1 are the second positive segment. Negative segments with a total pressure change difference less than or equal to 0.1 are deleted.

[0072] The working principle and beneficial effects of the above technical solution are as follows: After calculating the pressure difference between adjacent original pressure data points, dividing the positive and negative segments of the pressure difference and accumulating them to obtain the total pressure change difference, the noise in the original column pressure time series is processed according to the total pressure change difference, thereby reducing the impact of noise data on the extraction of multi-dimensional working condition features and ensuring that the accuracy and effectiveness of subsequent working condition identification can be improved.

[0073] Example 3: In a further example of the present invention, in S3, the multi-dimensional operating condition features of each target pressure data point include sliding window features and simple features. The sliding window features include mean, maximum value, minimum value, standard deviation and linear regression rate, and the simple features include time difference and pressure change rate.

[0074] In this embodiment, the time difference represents the time interval between a certain target pressure data point and its previous target pressure data point. The pressure change rate represents the ratio of the pressure difference between a certain target pressure data point and its previous data point to the time difference. For any target pressure data point, its time difference is calculated by formula (1), and its pressure change rate is calculated by formula (2)

[0075] (1);

[0076] (2);

[0077] In formula (1), represents the collection time of the i-th target pressure data point in the target stand pressure time sequence, represents the collection time of the i-1-th target pressure data point in the target stand pressure time sequence, represents the stand pressure of the i-th target pressure data point in the target stand pressure time sequence, represents the stand pressure of the i-1-th target pressure data point in the target stand pressure time sequence.

[0078] The working principle and beneficial effects of the above technical solution are: by taking the sliding window feature and the simple feature as the multi-dimensional working condition feature of the target pressure data point, the working condition characteristics of each dimension of each target pressure data point in the target stand pressure time sequence can be comprehensively evaluated, and the accuracy of subsequent working condition recognition based on the multi-dimensional working condition feature of each target pressure data point is improved.

[0079] In a further example of the present application, the S3 includes the following when extracting the sliding window feature of all target pressure data points:

[0080] S31, a sliding window with a window size of 5 is adopted, and the sliding window is slid in the target stand pressure time sequence with a sliding step of 1 until all target pressure data points of the target stand pressure time sequence are slid.

[0081] S32, for the intermediate target pressure data point at the intermediate position of any sliding window, the mean value, the maximum value, the minimum value, the standard deviation and the linear regression rate of the five target pressure data points in the sliding window where the intermediate target pressure data point is located are determined, and the mean value, the maximum value, the minimum value, the standard deviation and the linear regression rate are taken as the sliding window feature of the intermediate target pressure data point.

[0082] ​In this embodiment, the size of the sliding window is 5, and each target pressure data point and its left two target pressure data points and right two target pressure data points are taken as the five target pressure data points of the sliding window. The sliding step refers to the distance that the window moves each time when the sliding window is applied, and the sliding step is 1, which means that the sliding window moves 1 target pressure data point to the right each time.

[0083] Specifically, in any sliding window, if the five target pressure data points in the sliding window are denoted as x1, x2,..., x5, the column pressures corresponding to the five target pressure data points are denoted as y1, y2,..., y5, and the linear regression slope k of the five target pressure data points in the sliding window is calculated, formula (3) is used:

[0084] (3);

[0085] In formula (3), is the collection time of the i-th target pressure data point in the sliding window, is the column pressure of the i-th target pressure data point in the sliding window.

[0086] The mean value of the five target pressure data points in the sliding window is calculated as Formula (4) is used:

[0087] (4);

[0088] In formula (4), is the column pressure of the i-2-th target pressure data point in the sliding window, is the column pressure of the i-1-th target pressure data point in the sliding window, is the column pressure of the i+1-th target pressure data point in the sliding window, is the column pressure of the i+2-th target pressure data point in the sliding window.

[0089] The maximum value and the minimum value of the five target pressure data points in the sliding window are calculated as and Formula (5) and formula (6) are used:

[0090] (5);

[0091] (6);

[0092] The standard deviation of the five target pressure data points in the sliding window is calculated as Formula (7) is used:

[0093] (7);

[0094] It should be noted that if a certain intermediate target pressure data point is located at the edge of the first or second position of the starting position or the ending position of the target column pressure time sequence, since the sliding window cannot cover five target pressure data points, only the average value, extreme value (maximum value, minimum value), standard deviation and linear regression slope of three or four target pressure data points in the sliding window are calculated, and the average value, extreme value, standard deviation and linear regression slope are taken as the sliding window features of the intermediate target pressure data point.

[0095] The working principle and beneficial effects of the above technical solution are: by extracting the sliding window features of all target pressure data points in the target column pressure time sequence through the sliding window, the local change trend of the target column pressure time sequence can be obtained, so as to enhance the analysis capability of the target column pressure time sequence and improve the accuracy of the subsequent hydraulic support working condition recognition model in recognizing the working condition of the current column pressure time sequence.

[0096] In a further example of the present application, S5 is followed by:

[0097] A plurality of continuous target pressure data points with the same working condition label are aggregated into a structured working condition segment according to the time sequence of the original column pressure time sequence, and the time parameters of each working condition segment are recorded.

[0098] In this embodiment, aggregation means that a plurality of continuous target pressure data points with the same working condition label are combined into a working condition segment. The working condition segment is a data segment composed of a plurality of target pressure data points with the same working condition label in time sequence. The time parameters include at least the starting time, the ending time and the duration.

[0099] The working principle and beneficial effects of the above technical solution are: by aggregating the pressure data points with the same working condition label into working condition segments in time sequence and recording the time parameters, the pressure changes of the target pressure data points in different working conditions in the target column pressure time sequence can be structuredly represented, the working condition segments with actual production significance are generated, and an accurate data basis is provided for subsequent analysis of the working condition segments and further calculation of the working condition segments.

[0100] In a further example of the present application, S5 is followed by:

[0101] If any working condition segment includes less than three target pressure data points, it is determined that the working condition segment is an isolated event and the working condition labels of all target pressure data points included in the working condition segment are deleted.

[0102] In this embodiment, the isolated event refers to a phenomenon that is isolated, sparse or abnormal in the data. If a certain working condition segment contains less than three target pressure data points, it cannot accurately reflect the actual situation of the working condition segment, and is therefore determined as an isolated event that does not have effective analysis value.

[0103] The working principle and beneficial effects of the above technical solution are: by deleting the working condition labels of all target pressure data points of the working condition segment with less than three target pressure data points, the discrete points generated by random noise or short-term misjudgment of the hydraulic support working condition recognition model can be effectively filtered, and complete target pressure data points with actual production significance are retained, thereby ensuring the quality and reliability of the working condition segment.

[0104] In a further example of the present application, S5 further comprises:

[0105] The pressure cumulative change amount of each working condition segment with the working condition label as the lifting column and the lowering column is calculated respectively;

[0106] For any working condition segment with the working condition label as the lifting column, if the pressure cumulative change amount of the working condition segment with the working condition label as the lifting column during the lifting column period is less than the preset lifting column pressure cumulative change threshold, the lifting column working condition label of all target pressure data points included in the working condition segment with the working condition label as the lifting column is deleted.

[0107] For any working condition segment with the working condition label as the lowering column, if the pressure cumulative change amount of the working condition segment with the working condition label as the lowering column during the lowering column period is less than the preset lowering column pressure cumulative change threshold, the lowering column working condition label of all target pressure data points included in the working condition segment with the working condition label as the lowering column is deleted.

[0108] In this embodiment, the pressure cumulative change amount refers to the absolute value of the column pressure difference between the target pressure data point corresponding to the starting time and the target pressure data point corresponding to the ending time in each working condition segment.

[0109] In this embodiment, the preset lifting column pressure cumulative change threshold refers to that in any working condition segment with the working condition label as the lifting column, only when the pressure cumulative change amount reaches or exceeds the threshold, the lifting column process is considered to be effective and representative. This threshold sets a minimum pressure change amount standard to distinguish effective lifting column processes from those working condition segments with insufficient or abnormal pressure changes. The lowering column pressure cumulative change threshold refers to that in any working condition segment with the working condition label as the lowering column, only when the pressure cumulative change amount reaches or exceeds the threshold, the lowering column process is considered to be effective and representative.

[0110] The working principle and beneficial effects of the technical solution are as follows: by determining the size relationship between the pressure cumulative change amount of the working condition segment of the two working condition labels of the lifting column and the lowering column and the preset pressure cumulative change amount threshold, the working condition label of the target pressure data point of the invalid lifting and lowering column is eliminated, and accurate working condition segments for subsequent analysis of the two working conditions of lifting and lowering are provided.

[0111] In a further example of the present application, S5 further comprises:

[0112] Obtaining the working condition segments with all working condition labels being abnormal resistance increase or adjacent support moving support resistance increase;

[0113] If the working condition label of any working condition segment is abnormal resistance increase, the position data of the coal mining machine and the hydraulic support stroke data of the target hydraulic support in a 30s time period after the end time of the working condition segment are obtained and used to determine whether the coal mining machine passes the target hydraulic support and whether the target hydraulic support has a position movement, if the coal mining machine passes the target hydraulic support and the target hydraulic support has a position movement, the working condition label of the working condition segment is determined to be adjacent support moving support resistance increase, and the working condition labels of all target pressure data points included in the working condition segment are modified to adjacent support moving support resistance increase.

[0114] If the working condition label of any working condition segment is adjacent support moving support resistance increase, the hydraulic support stroke data of the target hydraulic support in a 30s time period after the end time of the working condition segment are obtained and used to determine whether the target hydraulic support has a position movement, if the target hydraulic support does not have a position movement, the working condition label of the working condition segment is determined to be abnormal resistance increase, and the working condition labels of all target pressure data points included in the working condition segment are modified to abnormal resistance increase.

[0115] In this embodiment, the position data of the coal mining machine represents the coordinate position of the coal mining machine; the passing range of the target hydraulic support is set based on the coordinate position of the target hydraulic support, if the coordinate position of the coal mining machine enters the passing range of the target hydraulic support, it is determined that the coal mining machine passes the target hydraulic support.

[0116] In this embodiment, the hydraulic support stroke data of the target hydraulic support in 30s after the end time of any working condition segment is obtained, the change of the stroke value of the target hydraulic support in the 30s is viewed, if the stroke value of the target hydraulic support changes obviously during this period, it is indicated that the target hydraulic support has a position movement; if the stroke value of the target hydraulic support does not change obviously, it is indicated that the target hydraulic support does not have a position movement.

[0117] The working principle and beneficial effects of the above technical solution are: abnormal resistance increase mostly occurs in the stable bearing and coal cutting influence stage, and adjacent frame moving frame resistance increase mostly occurs before the unloading and frame lowering stage. Because the multi-dimensional working condition characteristics of the two working conditions of abnormal resistance increase and adjacent frame moving frame resistance increase are similar, the pressure of adjacent frame moving frame resistance increase is generally higher than that of abnormal resistance increase in the normal operation of the target hydraulic support, and it is easier to distinguish; when the target hydraulic support underload operation condition occurs, the multi-dimensional working condition characteristics of adjacent frame moving frame resistance increase condition and abnormal resistance increase condition may be similar, therefore, the two conditions are distinguished in the embodiment of the application. The working condition segments of the two working condition labels of abnormal resistance increase and adjacent frame moving frame resistance increase are distinguished by using the position data of the coal mining machine and the hydraulic support stroke data of the target hydraulic support in the last 30s period based on the end time of each working condition segment, the misclassification problem of the working condition label of the target pressure data point output by the hydraulic support working condition identification model is solved, and an accurate data basis is provided for subsequent analysis and use of the working condition segments of the two working condition labels of abnormal resistance increase and adjacent frame moving frame resistance increase.

[0118] In a further example of the application, S5 is followed by:

[0119] All target pressure data points with a working condition label of initial support force unqualified are obtained, and if the pressure of any target pressure data point with a working condition label of initial support force unqualified is greater than a preset initial support force threshold, the working condition label of the target pressure data point with a working condition label of initial support force unqualified is deleted.

[0120] In this embodiment, the preset initial support force threshold is a specific pressure value set according to design standards and working conditions, which is generally 60%-85% of the rated working resistance of the hydraulic support. If the pressure of any target pressure data point is greater than this threshold, it means that the pressure has reached the normal working range and does not belong to the initial support force unqualified condition.

[0121] The working principle and beneficial effects of the above technical solution are: by comparing all target pressure data points with a working condition label of initial support force unqualified with a preset initial support force threshold, if the pressure of any target pressure data point exceeds the preset initial support force threshold, the initial support force unqualified working condition label of the target pressure data point is deleted, the misjudgment of the hydraulic support working condition identification model for the initial support force unqualified working condition label is eliminated, the engineering rationality of the determination of the target pressure data point with a working condition label of initial support force unqualified is ensured, and an accurate data basis is provided for subsequent analysis of the target pressure data point with a working condition label of initial support force unqualified.

[0122] In a further example of the application, S5 is followed by:

[0123] Obtaining a target pressure data point corresponding to the end time of the working condition segment with the working condition label of the lifting column, and if the value of the target pressure data point corresponding to the end time of the working condition segment with the working condition label of the lifting column is less than the preset initial support force threshold, adding the working condition label of the initial support force unqualified to the target pressure data point corresponding to the end time of the working condition segment.

[0124] The working principle and beneficial effects of the technical solution are: the embodiment eliminates the misjudgment of the hydraulic support working condition recognition model on the target pressure data point corresponding to the end time node of the working condition segment with the working condition label of the lifting column, ensures the engineering rationality of the working condition label of the working condition segment with the working condition label of the lifting column, and provides an accurate data basis for subsequent application of the working condition segment with the working condition label of the lifting column.

[0125] To verify the effect of the method provided by the embodiment of the application, the hydraulic support supporting whole process working condition recognition method based on pressure data analysis is verified, for example: after completing the structured reconstruction of all working condition segments, in order to ensure that the verification result is more reliable, the average pressure value of the column of all working condition segments is uniformly processed by Min-Max normalization. The verification result is shown in Figure 7 , and the distribution law shown in Figure 7 verifies the mapping relationship between the pressure characteristics and the working condition mechanism: the average pressure values of the two high-pressure working conditions of the support working resistance high and the safety valve opening are almost all concentrated in the high interval of 0.9-1.0. The working conditions of the lifting column, the lowering column, the abnormal resistance increase, and the adjacent frame moving frame resistance increase are mostly in the dynamic change of the pressure region, but their average pressure values show a certain rule: the average pressure values of the lifting column and the lowering column are mainly distributed in the medium interval of 0.2-0.5, and the distribution is relatively discrete; the average pressure values of the abnormal resistance increase and the adjacent frame moving frame resistance increase are distributed in the multiple intervals of 0.5-1.0, and the distribution is also relatively discrete. The initial support force unqualified and the support working resistance low are associated working conditions, and the initial support force unqualified working condition often appears with the support working resistance low working condition. The average pressure value of the initial support force unqualified is mainly distributed in the interval of 0.3-0.6, and partially overlaps with the interval of 0.5-0.7 of the support working resistance low.

[0126] When the hydraulic support support state recognition model adopts the random forest model, the accuracy of the random forest model is verified by the embodiment of the present application. Specifically, the target training pressure data time sequence with the working condition label and the multi-dimensional working condition characteristics is input into the random forest model for learning according to 80% of the training set, and the remaining 20% of the target training pressure data points are used as a verification set to verify the training result, and the accuracy of the verification set can reach 96%; the Hamming Loss reflects the overall error rate, the Micro F1 reflects the global classification performance, and the Macro F1 measures whether the random forest model performs balanced on each category, the Hamming Loss is 0.0129, indicating that the overall error rate of the recognition result is relatively low; the Micro F1 is 0.8574, proving that the random forest model has excellent global classification performance; and the Macro F1 is 0.6453, proving that the random forest model performs balanced when identifying each working condition.

[0127] Based on the output of the hydraulic support support state recognition model, the performance of the hydraulic support support state recognition model proposed in the above embodiment is evaluated, for example: the recognition effect of the hydraulic support working condition recognition model is verified by using the mine production data; the hydraulic support working condition recognition model identifies 11351 working conditions, of which 2721 are normal working conditions and 8630 are abnormal working conditions, and the specific identification results are shown in Table 1:

[0128]

[0129] In order to evaluate the overall performance of the hydraulic support working condition recognition model, the hypothesis test method is used to verify the recognition accuracy, the original hypothesis is that the overall accuracy of the hydraulic support working condition recognition model is less than or equal to 90%, and the alternative hypothesis is that the overall accuracy of the hydraulic support working condition recognition model is greater than 90%. It is assumed that the data of 60 hydraulic supports running for one week is split according to 1 hour as a window to obtain 10080 samples; 1000 samples are extracted from the samples by using the random sampling method, the working condition occurrence frequency identified by artificial recognition is 367 times, and the working condition recognition model identifies 354 times. The single-proportion z test is used to verify whether the accuracy of the hydraulic support working condition recognition model is greater than 90%, and the z test value is calculated by formula (8):

[0130] (8);

[0131] In formula (8), is the sample accuracy, , is the hypothesis accuracy, , is the standard error, Since z = 4.14 > 1.645 (critical value: one-sided alpha = 0.05), the null hypothesis is rejected, proving that the hydraulic support working condition identification model has good overall performance, and the identification accuracy is greater than 90%.

[0132] The pressure analysis of the hydraulic support supporting process starts from the perspective of overall trend, and can identify the macro characteristics of the hydraulic support in the stages of column lifting, column lowering, adjacent support resistance increasing and the like based on the curve of Figures 2-6 the hydraulic support working condition identification model, and then determine the seven types of typical working conditions to be identified, thereby providing a basis for the construction of the classification system. The analysis result provides clear classification standards and theoretical basis for subsequent support working condition identification based on pressure time series data.

[0133] The above only describes the preferred embodiments of the present application and is not used to limit the present application. It should be noted that, for ordinary skilled persons in the art, several improvements and modifications can be made without departing from the technical principles of the present application, and these improvements and modifications should be considered as the protection scope of the present application.

Claims

1. A method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis, characterized in that, include: S1, Collect the original column pressure time sequence during the operation of the target hydraulic support, the original column pressure time sequence including the current column pressure time sequence; S2, noise processing and outlier processing are performed sequentially on the original column pressure time series to obtain the target column pressure time series; S3, extract the multi-dimensional working condition features of all target pressure data points in the target column pressure time series; S4. Input the multi-dimensional working condition features of all target pressure data points into the pre-trained hydraulic support working condition recognition model. The input of the hydraulic support working condition recognition model is the multi-dimensional working condition features of the pressure data points, and the output is the pressure data points and their working condition labels arranged in the time sequence of the column pressure time series. S5. Determine the current support condition corresponding to the current column pressure time sequence based on the target pressure data points and their condition labels output by the hydraulic support condition identification model; aggregate multiple consecutive target pressure data points with the same condition label into structured condition segments according to the time order of the original column pressure time sequence, and record the time parameters of each condition segment.

2. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1, characterized in that, The noise processing in S2 includes: S21, calculate the pressure difference between the next original pressure data point and the previous original pressure data point in the original column pressure time series to obtain the pressure difference sequence; S22, add 0 to the first position of the pressure difference sequence and shift the remaining positions to the right to obtain a new pressure difference sequence, and align each pressure difference in the new pressure difference sequence with each original pressure data point in the original column pressure time sequence. S23, based on the sign of adjacent pressure differences in the new pressure difference sequence, divide all pressure differences into continuous positive and negative segments, and sum up the pressure differences of each positive and negative segment to obtain the total pressure change difference of each positive and negative segment; S24. If the absolute value of the total pressure change difference in any positive or negative segment is less than or equal to a preset pressure difference threshold, then the original pressure data points in the original column pressure time sequence that are aligned with all pressure differences in the positive or negative segment are identified as noise and deleted.

3. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1, characterized in that, In S3, the multi-dimensional operating condition features of each target pressure data point include sliding window features and simple features. The sliding window features include mean, maximum value, minimum value, standard deviation and linear regression rate, and the simple features include time difference and pressure change rate.

4. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 3, characterized in that, The step S3, when extracting the sliding window features of all target pressure data points, includes: S31, a sliding window with a window size of 5 is used to slide in the target column pressure time series with a sliding step size of 1 until all target pressure data points in the target column pressure time series are slid through; S32, for an intermediate target pressure data point located in the middle of any sliding window, determine the mean, maximum, minimum, standard deviation, and linear regression rate of the five target pressure data points within the sliding window where the intermediate target pressure data point is located, and use the mean, maximum, minimum, standard deviation, and linear regression rate as the sliding window features of the intermediate target pressure data point.

5. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1, characterized in that, Following S5, the following is also included: If any operating condition segment includes fewer than three target pressure data points, the operating condition segment is determined to be an isolated event, and the operating condition labels of all target pressure data points included in the operating condition segment are deleted.

6. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1 or 5, characterized in that, Following S5, the following is also included: Calculate the cumulative pressure change for each operating condition segment labeled as rising column and falling column, respectively; For any operating condition segment labeled as "lifting column", if the cumulative pressure change during the lifting column period is less than the preset cumulative pressure change threshold, then the lifting column operating condition labels of all target pressure data points included in the operating condition segment labeled as "lifting column" will be deleted. For any operating condition segment labeled "Plumb bob," if the cumulative pressure change during the plumb bob period is less than a preset threshold for cumulative pressure change during plumb bob, then the plumb bob operating condition labels for all target pressure data points included in the operating condition segment labeled "Plumb bob" will be deleted.

7. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1 or 5, characterized in that, Following S5, the following is also included: Obtain all operating condition segments labeled as abnormal resistance increase or adjacent frame movement resistance increase; If the working condition label of any working condition segment is abnormal resistance increase, obtain and determine whether the coal mining machine has passed the target hydraulic support and whether the target hydraulic support has moved in position based on the coal mining machine position data and the hydraulic support stroke data of the target hydraulic support within a 30-second time period after the end time of the working condition segment. If the coal mining machine has passed the target hydraulic support and the target hydraulic support has moved in position, then determine that the working condition label of the working condition segment is adjacent support movement resistance increase, and modify the working condition label of all target pressure data points included in the working condition segment to adjacent support movement resistance increase. If the working condition label of any working condition segment is "adjacent support movement increases resistance", the hydraulic support stroke data of the target hydraulic support within a 30-second time period after the end time of the working condition segment is obtained and it is determined whether the target hydraulic support has moved. If the target hydraulic support has not moved, the working condition label of the working condition segment is determined to be "abnormal resistance increase", and the working condition labels of all target pressure data points included in the working condition segment are modified to "abnormal resistance increase".

8. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1 or 5, characterized in that, Following S5, the following is also included: Obtain all target pressure data points with the working condition label indicating that the initial support force is unqualified. If the pressure of any target pressure data point with the working condition label indicating that the initial support force is unqualified is greater than the preset initial support force threshold, then delete the working condition label of the target pressure data point with the working condition label indicating that the initial support force is unqualified.

9. The method for identifying the working conditions of a hydraulic support throughout the entire support process based on pressure data analysis according to claim 1 or 5, characterized in that, Following S5, the following is also included: Obtain the target pressure data points corresponding to the end time of all working condition segments labeled "lifting column". If the value of the target pressure data point corresponding to the end time of any working condition segment labeled "lifting column" is less than the preset initial support force threshold, then add a working condition label indicating unqualified initial support force to the target pressure data point corresponding to the end time of the working condition segment.

Images