Abnormal working condition identification method in hydraulic support following and moving process

Abstract

The invention relates to an abnormal working condition identification method in a hydraulic support following and moving process, and belongs to the technical field of coal mine intellectualization. Comprising the following steps: collecting oil cylinder stroke data, and preprocessing the oil cylinder stroke data; extracting support moving characteristics, wherein the support moving characteristics comprise a support moving action starting point, a support moving action ending point, an oil cylinder stroke maximum value, an oil cylinder stroke minimum value and a difference value between the oil cylinder strokes of the target hydraulic support and the oil cylinder strokes of the hydraulic supports on the left side and the right side; according to the frame moving action starting point, the frame moving action ending point and a frame moving abnormity recognition algorithm, the stroke fluctuation type working condition is recognized; and according to the maximum value of the oil cylinder stroke, the minimum value of the oil cylinder stroke, the difference value of the oil cylinder strokes of the target hydraulic support and the hydraulic supports on the left side and the right side and a stroke abnormity recognition model, the stroke abnormity type working condition is recognized. The invention provides the method for intelligently identifying the abnormal working condition in the hydraulic support following and moving process, the labor intensity of workers can be relieved, the production efficiency is improved, and the intelligent level of a coal mine is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine intelligentization, and particularly to a method for identifying abnormal working conditions during the process of following a shearer with hydraulic supports. Background Art

[0002] As an important part of the modern transformation of the coal mining industry, intelligent fully-mechanized coal mining faces have played a key role in improving production efficiency, ensuring safety, and reducing environmental impacts in recent years. The three-machine matching (shearer, conveyor, hydraulic support) is the core component of the efficient and safe operation of an intelligent fully-mechanized coal mining face. Among them, the hydraulic support plays a crucial role in ensuring the safety and stability during the coal mining process. However, in the actual production process, affected by equipment status, geological conditions, and operating factors, the operation of the hydraulic support is not always stable, and problems such as blocked support pushing, abnormal support, and pressure fluctuations in the hydraulic system may occur. These abnormal conditions not only affect the normal forward movement of the support but also pose a threat to the stability and safe production of the intelligent fully-mechanized coal mining face.

[0003] Currently, during the process of following a shearer with hydraulic supports, for the abnormal working conditions that occur during the process of following a shearer, it mainly relies on manual identification. This method not only increases the labor intensity of workers but also easily affects production efficiency and has certain potential safety hazards. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for identifying abnormal working conditions during the process of following a shearer with hydraulic supports. The technical solution of the present invention is as follows:

[0005] A method for identifying abnormal working conditions during the process of following a shearer with hydraulic supports, which includes:

[0006] S1, collect the cylinder stroke data of the target hydraulic support during the process of following a shearer within a period of time in chronological order, and preprocess the cylinder stroke data;

[0007] S2, extract the support moving features from the preprocessed cylinder stroke data of the target hydraulic support. The support moving features include the starting point of the support moving action, the ending point of the support moving action, the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference in the cylinder strokes between the target hydraulic support and the hydraulic supports on the left and right sides;

[0008] S3, identify the working conditions of stroke fluctuation type during the support moving process of the target hydraulic support according to the starting point of the support moving action, the ending point of the support moving action, and a preset support moving abnormal identification algorithm;

[0009] S4, identify the working conditions of stroke abnormal type during the support moving process of the target hydraulic support according to the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, the difference in the cylinder strokes between the target hydraulic support and the hydraulic supports on the left and right sides, and a pre-trained stroke abnormal identification model.

[0010] Optionally, when preprocessing the cylinder stroke data, S1 includes:

[0011] S21. Perform outlier processing on the collected cylinder stroke data to obtain a cylinder stroke sequence {x1, x2,..., x n}, and calculate the difference between two adjacent cylinder stroke data in the cylinder stroke sequence to obtain a cylinder stroke difference sequence {Δx1, Δx2,..., Δx n-1};

[0012] S22. Sequentially divide the cylinder stroke difference data with the same positive or negative sign continuously in the cylinder stroke difference sequence {Δx1, Δx2,..., Δx n-1} into the same group, and divide the corresponding cylinder stroke data of each group of cylinder stroke difference data into a group, and retain the last cylinder stroke data in each group of cylinder stroke data as the representative value of this group of cylinder stroke data. Combine the representative values of each group of cylinder stroke data to obtain an updated cylinder stroke sequence {x1′, x2′,..., x m ′};

[0013] S23. Calculate the sum of each cylinder stroke difference data in each group of cylinder stroke difference data to obtain a combined cylinder stroke difference sequence {Δx1′, Δx2′,..., Δx m ′}.

[0014] Optionally, when extracting the starting point of the support moving action, the ending point of the support moving action, the maximum cylinder stroke, and the minimum cylinder stroke from the preprocessed cylinder stroke data of the target hydraulic support, S2 includes:

[0015] S21. Obtain a preset stroke change threshold X;

[0016] S22. Sequentially search for the first cylinder stroke difference data less than -X in the combined cylinder stroke difference sequence, record its index as a1 and store it in the starting point list a of the support moving action;

[0017] S23. Continue to search backward in the combined cylinder stroke difference sequence for the first cylinder stroke difference data greater than X, record its index as b1 and store it in the list b of the ending point of the support moving action and the starting point of the scraper conveyor pushing;

[0018] S24. Continue to search backward in the combined cylinder stroke difference sequence for the last cylinder stroke difference data greater than X, record its index as c1 and store it in the list c of the highest point of the scraper conveyor pushing action;

[0019] S25. Continue to search backward in the merged cylinder stroke difference sequence for the first cylinder stroke difference data less than -X, the first cylinder stroke difference data greater than X, and the last cylinder stroke difference data greater than X in sequence until the last cylinder stroke difference data in the merged cylinder stroke difference sequence, obtaining the starting point list a of the support moving action as {a1, a2, …, a q}, the ending point list b of the support moving action and the starting point list of the scraper conveyor pushing as {b1, b2, …, b q}, and the highest point list c of the scraper conveyor pushing action as {c1, c2, …, c q};

[0020] S26. Divide the indexes with the same number in the starting point list a of the support moving action, the ending point list b of the support moving action and the starting point list of the scraper conveyor pushing, and the highest point list c of the scraper conveyor pushing action into a set of support moving actions. Take the point corresponding to a1 as the starting point of the support moving action, and take the cylinder stroke data corresponding to a1 in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} as the maximum cylinder stroke x start of the target hydraulic support, and calculate the support moving action value Δx ab and the scraper conveyor pushing action value Δx ab for each set of support moving actions;

[0021] S27. Search for the support moving action value Δx ab and the scraper conveyor pushing action value Δx bc corresponding to each set of support moving actions in the order of the numbers from front to back until a set of support moving actions with the absolute value of the difference between the support moving action value Δx ab and the scraper conveyor pushing action value Δx bc less than the set threshold is found. Denote its group number as k, take the point corresponding to b k as the ending point of the support moving action, and take the cylinder stroke data corresponding to b k in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} as the minimum cylinder stroke x end of the target hydraulic support.

[0022] Optionally, the S3 includes:

[0023] S31. Judge whether b k is in the first set of support moving actions. If b k is not in the first set of support moving actions, determine that a stroke fluctuation type working condition occurs in the support moving process of the target hydraulic support.

[0024] Optionally, the stroke fluctuation type working condition includes a shutdown fluctuation type. After the S31, it further includes:

[0025] S32. Obtain xstart corresponding to x end The corresponding time is respectively used as the starting time t of the support moving operation start and the ending time t of the support moving operation end , and obtain the column pressure data of the target hydraulic support between the starting time t of the support moving operation start and the ending time t of the support moving operation end . Preprocess the obtained column pressure data to obtain the merged column pressure difference sequence {Δx1′, Δx2′, …, Δx p ′};

[0026] S33. Determine whether there is column pressure difference data greater than or equal to the pressure fluctuation threshold in the merged column pressure difference sequence {Δx1′, Δx2′, …, Δx p ′}. If it exists, determine that the stroke fluctuation type working condition that occurs during the support moving process of the target hydraulic support is the shutdown fluctuation type.

[0027] Optionally, the stroke fluctuation type working condition includes the support moving overtime type. After S31, it further includes:

[0028] S34. Obtain the time corresponding to x start and x end The corresponding time is respectively used as the starting time t of the support moving operation start and the ending time t of the support moving operation end . Determine whether the time interval between the starting time t of the support moving operation start and the ending time t of the support moving operation end is greater than the preset time threshold. If the time interval between the starting time t of the support moving operation start and the ending time t of the support moving operation end is greater than the preset time threshold, determine that the stroke fluctuation type working condition that occurs during the support moving process of the target hydraulic support is the support moving overtime type.

[0029] Optionally, the stroke abnormal type working condition includes the stroke abnormal sudden drop type and the stroke abnormal average type. Before S4, it further includes:

[0030] Train a stroke abnormal recognition model based on the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference in the cylinder stroke between the target hydraulic support and the hydraulic supports on the left and right sides during the historical support moving process of the hydraulic support. And when training the stroke abnormal recognition model, set the stroke abnormal sudden drop type as label 1, set the stroke abnormal average type as label 2, and set the label of the remaining normal support moving as 0.

[0031] Optionally, the stroke abnormal recognition model is a decision tree model.

[0032] Optionally, for any set of support moving action sets, in S26, when calculating its corresponding support moving action value Δxab and the propelling action value Δx bc When it is, it is realized respectively through the following formula (1) and formula (2):

[0033] Δx ab = x a - x b (1)

[0034] Δx bc = x c - x b (2)

[0035] In formula (1) and formula (2), x a , x b and x c are respectively the cylinder stroke data corresponding to each index in the updated cylinder stroke sequence {x1′, x2′,..., x m ′} in the set of support moving actions.

[0036] All the above optional technical solutions can be combined arbitrarily, and the present invention does not elaborate on the structures after combination one by one.

[0037] By means of the above solutions, the beneficial effects of the present invention are as follows:

[0038] By pre - constructing a support moving anomaly recognition algorithm and training a stroke anomaly recognition model, after collecting the cylinder stroke data during the follow - up support moving process of the target hydraulic support, after pre - processing the cylinder stroke data and extracting the support moving features, and then inputting the support moving features into the support moving anomaly recognition algorithm and the stroke anomaly recognition model, it is possible to identify the stroke fluctuation type working conditions and the stroke anomaly type working conditions in the support moving process of the target hydraulic support. Thus, a method for intelligent identification of abnormal working conditions in the follow - up support moving process of hydraulic supports is provided, which can reduce the labor intensity of workers, improve production efficiency, reduce coal mine safety hazards, and improve the intelligent level of coal mines at the same time.

[0039] The above description is only an overview of the technical solution of the present invention. In order to understand the technical means of the present invention more clearly and be able to implement it according to the content of the specification, the following takes the preferred embodiments of the present invention and combines with the drawings to elaborate in detail as follows. Brief Description of the Drawings

[0040] Figure 1 is a flow chart of the method for identifying abnormal working conditions in the follow - up support moving process of the hydraulic support provided by the embodiment of the present invention.

[0041] Figure 2 is a schematic diagram of the change trend of the cylinder stroke and the column pressure during the normal follow - up support moving process of the hydraulic support in the intelligent fully - mechanized coal mining face.

[0042] Figure 3It is a schematic diagram of the change trends of the cylinder stroke and the column pressure during the abnormal following-shift process of the hydraulic support in the intelligent fully-mechanized mining face.

[0043] Figure 4 It is a schematic diagram of the change trend of the cylinder strokes of adjacent hydraulic supports under the abnormal stroke type working condition in the embodiment of the present invention.

[0044] Figure 5 It is the overall flow chart of the abnormal working condition identification method for the hydraulic support following-shift process provided in this embodiment. Detailed implementation manners

[0045] The following combines the accompanying drawings and embodiments to further describe in detail the specific implementation manners of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0046] In the abnormal working condition identification method for the hydraulic support following-shift process provided in the embodiment of the present invention, the steps other than data acquisition can be implemented by any electronic device with computing functions such as a PC, a mobile terminal or a server, etc. Data acquisition can be implemented by sensors. After the sensors collect data, they send the data to the electronic device, and the electronic device analyzes the data collected by the sensors.

[0047] During the normal following-shift process of the hydraulic support in the intelligent fully-mechanized mining face, it will go through four main processes: lowering the columns, shifting the support, raising the columns and pushing the scraper conveyor. As Figure 2 shown, the hydraulic support first performs the column lowering work. After the column pressure reaches a relatively low level, the support shifting process of the hydraulic support is carried out. During the support shifting process, the cylinder stroke of the hydraulic support will decrease sharply. After the cylinder stroke drops to a relatively low level, the support shifting process ends, and the columns of the hydraulic support start to rise, and the corresponding column pressure starts to increase. At the same time, the hydraulic support starts to perform the scraper conveyor pushing process, and the cylinder stroke gradually increases during the scraper conveyor pushing process. Under normal working conditions, the column lowering and the support shifting process of the hydraulic support are often extremely rapid in time, and the process duration is usually within several seconds, and the two processes usually have "non-coincidence" in time, that is, the support shifting of the hydraulic support will start after the column lowering is about to end or has already ended.

[0048] Based on the long-term observation of the cylinder stroke data and the column pressure data on the site of the intelligent fully-mechanized mining face in the embodiment of the present invention, two types of abnormal working conditions in the hydraulic support following-shift process are summarized, including the stroke fluctuation type working condition and the stroke abnormal type working condition. The stroke fluctuation type working condition includes the shutdown fluctuation type and the support shifting overtime type. The data performances of the shutdown fluctuation type, the support shifting overtime type and the stroke abnormal type working condition are as Figure 3 shown. Figure 3In Figure (a), during the process of following the shearer to move the support, when the oil cylinder moved about 200 mm, the upward fluctuation of the stroke occurred. By observing the data of the support column pressure, it was found that the phenomenon of raising and lowering the column occurred at this fluctuation position, reflecting that secondary intervention operations were carried out when abnormal working conditions occurred; Figure 3 In Figure (b), during the normal support moving process, when the stroke of the oil cylinder changed to 400 mm, a long-term shutdown phenomenon occurred, and then the normal following-the-shearer support moving work was restored; Figure 3 In Figure (c), during the support moving work, the highest point of the oil cylinder stroke was near 900 mm, but the lowest stroke was only 400 mm, and the total stroke difference was 500 mm. The stroke of the hydraulic support oil cylinder was much smaller than that in the normal working condition.

[0049] By analyzing the working conditions of shutdown and fluctuation type, support moving overtime type, and abnormal stroke type, their data performances and reasons are shown in Table 1.

[0050] Table 1

[0051]

[0052] Among them, for the abnormal stroke type of working conditions, combined with the action conditions of adjacent hydraulic supports, including two types: abnormal stroke sudden drop type and abnormal stroke uniform type, the oil cylinder stroke data and support column pressure data performances of these two types of abnormal stroke type of working conditions are as Figure 4 shown. Figure 4 In the abnormal stroke sudden drop type shown in Figure (a), the oil cylinder stroke of the middle hydraulic support (No. 50 hydraulic support) is about 700 mm, and there is an obvious gap with the oil cylinder strokes of the two hydraulic supports on its left and right; Figure 4 In the abnormal stroke uniform type shown in Figure (b), the oil cylinder strokes of the middle hydraulic support (No. 65 hydraulic support) and the hydraulic supports on its left and right sides are both about 200 mm, and there is no large gap between them, but the difference from the oil cylinder stroke in the normal working condition is very large.

[0053] The data performances and reasons of these two types of abnormal stroke type of working conditions, namely abnormal stroke sudden drop type and abnormal stroke uniform type, are shown in Table 2.

[0054] Table 2

[0055]

[0056] Based on the above analysis, the embodiments of the present invention respectively propose a support moving abnormal recognition algorithm based on stroke-pressure-time analysis, and a stroke abnormal recognition model based on the oil cylinder strokes of multiple hydraulic supports, and respectively identify the above-mentioned stroke fluctuation type of working conditions and abnormal stroke type of working conditions. As Figure 1 shown, the abnormal working condition recognition method for the hydraulic support following-the-shearer moving process provided by the present invention includes the following steps S1 to S4:

[0057] S1. Collect the cylinder stroke data of the target hydraulic support during the follow - up support movement process within a period of time in chronological order, and pre - process the cylinder stroke data.

[0058] Among them, the target hydraulic support is any non - first hydraulic support in the hydraulic support cluster in the intelligent fully - mechanized mining face. The cylinder stroke data can be collected by the stroke sensors installed on the target hydraulic support. The stroke sensors are communicatively connected to the electronic device. After the stroke sensors collect the cylinder stroke data, they send it to the electronic device. When collecting the cylinder stroke data within a period of time, the stroke sensors collect it according to a preset period.

[0059] In a specific embodiment, when performing pre - processing on the cylinder stroke data in S1, it includes the following steps S21 to S23:

[0060] S21. Perform outlier processing on the collected cylinder stroke data to obtain the cylinder stroke sequence {x1, x2,..., x n}, and calculate the difference between two adjacent cylinder stroke data in the cylinder stroke sequence to obtain the cylinder stroke difference sequence {Δx1, Δx2,..., Δx n-1}.

[0061] Specifically, during the operation of the target hydraulic support, there may be a failure of the stroke sensor, resulting in a large mutation within a short time in the original cylinder stroke data, manifested as a certain cylinder stroke data being 0, while the adjacent cylinder stroke data before and after it are both non - zero. For such outliers, the embodiments of the present invention first perform outlier processing on the collected cylinder stroke data. Specifically, when performing outlier processing, the linear interpolation method is used for repair, that is, the average value of the two adjacent non - zero points is used to fill the position where the 0 point is located to ensure the continuity of the cylinder stroke data.

[0062] When calculating the difference between two adjacent cylinder stroke data in the cylinder stroke sequence, subtract the previous cylinder stroke data from the latter cylinder stroke data, that is, Δx i＝ x i+1 -x i .

[0063] S22. From front to back, divide the cylinder stroke difference data with the same positive or negative nature in the cylinder stroke difference sequence {Δx1, Δx2,..., Δx n-1} into the same group, and divide the cylinder stroke data corresponding to each group of cylinder stroke difference data into a group, and retain the last cylinder stroke data in each group of cylinder stroke data as the representative value of this group of cylinder stroke data. Combine the representative values of each group of cylinder stroke data to obtain the updated cylinder stroke sequence {x1′, x2′,..., x m ′}.

[0064] It should be noted here that in the embodiments of the present invention, the positive hydraulic cylinder stroke difference data is classified into one category, and the data of 0 and negative hydraulic cylinder stroke differences are classified into one category. If there is only one hydraulic cylinder stroke data in a group of hydraulic cylinder stroke data, then the only hydraulic cylinder stroke data is used as the representative value of the group of hydraulic cylinder stroke data.

[0065] S23. Calculate the sum of each hydraulic cylinder stroke difference data in each group of hydraulic cylinder stroke difference data to obtain the combined hydraulic cylinder stroke difference sequence {Δx1′, Δx2′, …, Δx m ′}.

[0066] For example, if the hydraulic cylinder stroke sequence is {10, 11, 20, 20, 15, 13, 16}, then the hydraulic cylinder stroke difference sequence is {1, 9, 0, -5, -2, 3}. The grouped hydraulic cylinder stroke difference data is {1, 9}, {0, -5, -2} and {3}, a total of three groups. The grouped hydraulic cylinder stroke data is {10, 11, 20}, {20, 15, 13} and {16}. The updated hydraulic cylinder stroke sequence is {20, 13, 16}, and the combined hydraulic cylinder stroke difference sequence is {10, -7, 3}.

[0067] Through the local accumulation method based on the positive and negative of adjacent data differences described in S21 to S23, the embodiments of the present invention can accurately identify the subtle fluctuations of the hydraulic support during the support moving process, which is convenient for accurately extracting the starting point and ending point of the support moving action subsequently.

[0068] S2. Extract the support moving features from the preprocessed hydraulic cylinder stroke data of the target hydraulic support. The support moving features include the starting point of the support moving action, the ending point of the support moving action, the maximum hydraulic cylinder stroke, the minimum hydraulic cylinder stroke, and the differences between the target hydraulic support and the hydraulic cylinder strokes of the left and right adjacent hydraulic supports.

[0069] In a specific embodiment, when S2 extracts the starting point of the support moving action, the ending point of the support moving action, the maximum hydraulic cylinder stroke, and the minimum hydraulic cylinder stroke from the preprocessed hydraulic cylinder stroke data of the target hydraulic support, it includes the following steps S21 to S27:

[0070] S21. Obtain the preset stroke change amount threshold X.

[0071] By observing a large amount of hydraulic cylinder stroke data of the intelligent fully mechanized coal mining face, the embodiments of the present invention can preset the value of the stroke change amount threshold X to be 20 mm.

[0072] S22. Search for the first cylinder stroke difference data less than -X in the merged cylinder stroke difference sequence from front to back, record its index as a1 and store it in the list a of the starting points of the support moving actions. S23. Continue to search backward in the merged cylinder stroke difference sequence for the first cylinder stroke difference data greater than X, record its index as b1 and store it in the list b of the end points of the support moving actions and the starting points of the coal pushing actions. S24. Continue to search backward in the merged cylinder stroke difference sequence for the last cylinder stroke difference data greater than X, record its index as c1 and store it in the list c of the highest points of the coal pushing actions. S25. Continue to search backward in the merged cylinder stroke difference sequence in turn for the first cylinder stroke difference data less than -X, the first cylinder stroke difference data greater than X, and the last cylinder stroke difference data greater than X until the last cylinder stroke difference data in the merged cylinder stroke difference sequence, and obtain the list a of the starting points of the support moving actions as {a1, a2, …, a q}, the list b of the end points of the support moving actions and the starting points of the coal pushing actions as {b1, b2, …, b q}, and the list c of the highest points of the coal pushing actions as {c1, c2, …, c q}.

[0073] S26. Divide the indexes with the same number in the list a of the starting points of the support moving actions, the list b of the end points of the support moving actions and the starting points of the coal pushing actions, and the list c of the highest points of the coal pushing actions into a set of support moving actions. Take the point corresponding to a1 as the starting point of the support moving action, take the cylinder stroke data corresponding to a1 in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} as the maximum cylinder stroke x start of the target hydraulic support, and calculate the support moving action value Δx ab and the coal pushing action value Δx bc corresponding to each set of support moving actions.

[0074] Specifically, divide a1, b1, and c1 into the support moving action set 1, divide a2, b2, and c2 into the support moving action set 2, and so on to obtain the support moving action set q.

[0075] Furthermore, for any set of support moving actions, when calculating the corresponding support moving action value Δx ab and the coal pushing action value Δx bc , they are respectively implemented through the following formulas (1) and (2):

[0076] Δx ab = x a - x b (1)

[0077] Δx bc = x c - x b(2)

[0078] In Formulas (1) and (2), x a , x b and x c are respectively the cylinder stroke data corresponding to each index in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} of this set of support moving actions.

[0079] S27. Search for the support moving action value Δx ab and the coal pushing action value Δx bc corresponding to each set of support moving actions in the order of the numbers from front to back until a set of support moving actions is found where the absolute value of the difference between the support moving action value Δx ab and the coal pushing action value Δx bc is less than the set threshold m. Denote its group number as k, take the point corresponding to b k as the end point of the support moving action, and take the cylinder stroke data corresponding to b k in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} as the minimum cylinder stroke x end of the target hydraulic support.

[0080] Specifically, for the support moving action value Δx ab and the coal pushing action value Δx bc of the support moving action set 1, if |Δx ab -Δx bc |>m, then continue to judge whether the absolute value of the difference between the support moving action value Δx ab and the coal pushing action value Δx bc of the support moving action set 2 is greater than m, and so on, until a support moving action set with |Δx ab -Δx bc |<m is found, and then denote its group number as k. The cylinder stroke of the target hydraulic support is the difference between the maximum cylinder stroke x start and the minimum cylinder stroke x end .

[0081] It should be noted that when extracting the difference between the cylinder strokes of the target hydraulic support and the left hydraulic support, first extract the maximum cylinder stroke and the minimum cylinder stroke of the left hydraulic support of the target hydraulic support in the manner described in the above steps S1 and S2, then subtract the minimum cylinder stroke from the maximum cylinder stroke of the left hydraulic support to obtain the cylinder stroke of the left hydraulic support, and finally calculate the difference between the cylinder stroke of the target hydraulic support and the cylinder stroke of the left hydraulic support. When extracting the difference between the cylinder strokes of the target hydraulic support and the right hydraulic support, the method is the same, that is, calculate the difference between the cylinder stroke of the target hydraulic support and the cylinder stroke of the right hydraulic support through the above method.

[0082] S3. Identify the working conditions of the stroke fluctuation type during the support moving process of the target hydraulic support according to the starting point of the support moving action, the ending point of the support moving action, and the preset algorithm for identifying abnormal support moving.

[0083] In a specific embodiment, the said S3 includes: S31. Judge whether b k is in the first set of support moving actions. If b k is not in the first set of support moving actions, it is determined that the working conditions of the stroke fluctuation type occur during the support moving process of the target hydraulic support.

[0084] Specifically, if b k is in the first set of support moving actions, that is, b k is in the support moving action set 1, it is determined that the working conditions of the stroke fluctuation type do not occur during the support moving process of the target hydraulic support, and this support moving process is a normal working condition. If b k is in the first set of support moving actions, it is determined that the working conditions of the stroke fluctuation type occur during the support moving process of the target hydraulic support.

[0085] Since the working conditions of the stroke fluctuation type include two types: the shutdown fluctuation type and the support moving overtime type, therefore, when it is determined through step S31 that the working conditions of the stroke fluctuation type occur during the support moving process of the target hydraulic support, the embodiments of the present invention also determine which one of the shutdown fluctuation type and the support moving overtime type the working conditions of the stroke fluctuation type are through the following steps S32 to S34. The specific judgment method is as follows:

[0086] S32. Obtain the time corresponding to x start and x end respectively as the starting time t start of the support moving action and the ending time t end of the support moving action. Obtain the column pressure data of the target hydraulic support between the starting time t start and the ending time t end of the support moving action, and preprocess the obtained column pressure data to obtain the merged column pressure difference sequence {Δx1′, Δx2′,..., Δx p ′}; S33. Judge whether there is column pressure difference data greater than or equal to the pressure fluctuation threshold in the merged column pressure difference sequence {Δx1′, Δx2′,..., Δx p ′}. If it exists, it is determined that the working conditions of the stroke fluctuation type occurring during the support moving process of the target hydraulic support are the shutdown fluctuation type.

[0087] Specifically, since the lifting and lowering of the hydraulic support columns appear before and after a complete support moving action respectively, therefore, if the column lifting and lowering occur in the middle of the support moving action, it can be considered that a secondary adjustment (manual intervention) has been carried out. Therefore, the embodiments of the present invention introduce the column pressure data on the basis of the cylinder stroke criterion.

[0088] Among them, when preprocessing the obtained column pressure data, the principle is the same as that of preprocessing the oil cylinder stroke data in the above embodiment. Specifically: perform outlier processing on the obtained column pressure data to obtain a column pressure sequence {x1, x2, …, x o}, and calculate the difference between two adjacent column pressure data in the column pressure sequence to obtain a column pressure difference sequence {Δx1, Δx2, …, Δx o-1}; from front to back, divide the column pressure difference data with the same positive or negative sign continuously in the column pressure difference sequence {Δx1, Δx2, …, Δx o-1} into the same group, and moreover, divide the column pressure data corresponding to each group of column pressure difference data into a group, and retain the last column pressure data in each group of column pressure data as the representative value of this group of column pressure data, and combine the representative values of each group of column pressure data to obtain an updated column pressure sequence {x1′, x2′, …, x p ′}; calculate the sum of each column pressure difference data in each group of column pressure sequence difference data to obtain a combined column pressure difference sequence {Δx1′, Δx2′, …, Δx p ′}.

[0089] Specifically, if there is a column pressure difference data in the combined column pressure difference sequence {Δx1′, Δx2′, …, Δx p ′} that is greater than or equal to the pressure fluctuation threshold, it is determined that the stroke fluctuation type working condition occurring during the support moving process of the target hydraulic support is a shutdown fluctuation type. According to the observation results of the column pressure data collected for a long time in the intelligent fully mechanized coal mining face, the pressure fluctuation threshold is set to 5 Mpa in the embodiment of the present invention.

[0090] S34, obtain the time corresponding to x start and x end and use them as the starting time t start of the support moving action and the ending time t end of the support moving action respectively, and judge whether the time interval between the starting time t start of the support moving action and the ending time t end of the support moving action is greater than the preset time threshold. If the time interval between the starting time t start of the support moving action and the ending time t end of the support moving action is greater than the preset time threshold, it is determined that the stroke fluctuation type working condition occurring during the support moving process of the target hydraulic support is a support moving overtime type.

[0091] Among them, the specific value of the preset time threshold needs to be determined in combination with the specific situation of the intelligent fully-mechanized mining face. In the embodiment of the present invention, based on the analysis results of the data collected for a long time on the intelligent fully-mechanized mining face, the preset time threshold is set to 120s.

[0092] S4. Identify the working conditions of abnormal stroke during the support moving process of the target hydraulic support according to the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides, and the pre-trained abnormal stroke recognition model.

[0093] Among them, the abnormal stroke recognition model is a machine learning model, whose input is the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides, and the output is the probability of various types of abnormal stroke working conditions. In the embodiment of the present invention, the type with the highest probability in the output probability of the abnormal stroke recognition model is selected as the category of the abnormal stroke working condition. The abnormal stroke working conditions include the abnormal sudden drop type and the abnormal average type. In a specific embodiment, the abnormal stroke recognition model adopts a decision tree model.

[0094] It should be noted that before the step S4, it is necessary to train the abnormal stroke recognition model first. Specifically, during training, it is realized based on the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides during the historical support moving process of each hydraulic support on the intelligent fully-mechanized mining face. Therefore, it is necessary to collect the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides during the historical support moving process of each hydraulic support, and manually label the working condition labels (abnormal sudden drop type, abnormal average type, and normal support moving) corresponding to each group of data. On this basis, when training the abnormal stroke recognition model, the abnormal sudden drop type is set as the label 1, the abnormal average type is set as the label 2, and the labels of the remaining normal support moving are set as 0.

[0095] Specifically during training, 80% of the labeled data is used as the training set and the validation set, and the other 20% is used as the test set. When dividing the data, stratified sampling is adopted to ensure that the proportion of various types of sample in the training set and the test set is the same as that in the original data set. The test set does not participate in the training at all and is only used for the final performance evaluation to verify the generalization ability of the abnormal stroke recognition model.

[0096] The embodiment of the present invention also verifies the method proposed in the above embodiment. The verification results show that the abnormal support moving recognition algorithm based on the stroke-pressure-time analysis in the embodiment of the present invention has an accuracy rate of 1 for the recognition of two types of patterns, namely the shutdown fluctuation type and the support moving overtime type, and the recall rate is above 0.95.

[0097] Table 3

[0098] Pattern type Precision Recall F1 score Normal 0.9632 0.996 0.9794 Trip anomaly sudden drop type 1 0.9787 0.9892 Trip anomaly average small type 0.9529 0.81 0.8757

[0099] The precision and recall rates of the travel anomaly recognition model based on the travel of multiple hydraulic support cylinders for identifying two types of travel anomaly conditions are shown in Table 3. From the perspective of precision and recall rates, this travel anomaly recognition model performs relatively well in identifying abnormal conditions, especially in the recognition ability of sudden drop in travel anomalies. According to the results of this data test, the recall rate of this category reaches 0.9787, indicating that the vast majority of sudden drop in travel anomalies can be successfully detected by the model, with a very low false negative rate. At the same time, its precision rate is 1, meaning that all samples determined by the model to be sudden drop in travel anomalies are real abnormal samples, without false positives. Such performance shows that when the travel anomaly recognition model identifies sudden drop in travel anomalies, it can not only cover all actual abnormal situations to the greatest extent, but also ensure the accuracy of prediction, avoiding misidentifying normal samples as abnormal, thus ensuring high reliability. For the uniform small travel anomalies, the precision rate of the travel anomaly recognition model reaches 0.9529, indicating that most of the abnormal samples predicted by it are actual abnormalities, with a low false positive rate, while the recall rate is 0.81, indicating that although the coverage ability of the travel anomaly recognition model for this type of anomaly is slightly inferior to that of sudden drop in travel anomalies, it can still identify most of the uniform small abnormal samples. Generally speaking, the travel anomaly recognition model has strong capabilities in identifying both abnormal patterns. Among them, the recognition effect of sudden drop in travel anomalies is particularly superior, being able to take into account the dual advantages of recall rate and precision rate. Although the recognition of uniform small travel anomalies is slightly insufficient, it is still at a relatively high level. Overall, this travel anomaly recognition model has high accuracy and reliability in the abnormal pattern recognition task.

[0100] In summary, as Figure 5As shown in the figure, for the method provided by the embodiment of the present invention, when identifying the abnormal working conditions during the process of the target hydraulic support following the shearer for support moving, the stroke data of the cylinders of the target hydraulic support and the column pressure data are first collected. Subsequently, the cylinder stroke data and the column pressure data are preprocessed, including outlier processing, difference calculation between adjacent data, and data merging based on the positive or negative of the difference, which provides a basis for subsequent feature extraction while improving the data quality and reliability. Then, an algorithm for identifying abnormal support moving and a model for identifying abnormal stroke are established. The algorithm for identifying abnormal support moving based on stroke-pressure-time analysis is used to identify the stroke fluctuation working conditions of the shutdown fluctuation type and the support moving overtime type; the model for identifying abnormal stroke is used to identify the abnormal stroke working conditions. Specifically, when identifying, six features are extracted, including the starting point of the support moving action, the ending point of the support moving action, the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the differences between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides. The starting point and the ending point of the support moving action are input into the algorithm for identifying abnormal support moving to identify the stroke fluctuation working conditions, and further, the column pressure fluctuation and the support moving action time are respectively discriminated for the stroke fluctuation working conditions to identify the shutdown fluctuation type and the support moving overtime type; the subsequent four stroke-related features are input into the model for identifying abnormal stroke to identify two types of abnormal stroke working conditions, namely, the abnormal stroke sudden drop type and the abnormal stroke average type.

[0101] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for identifying abnormal working conditions during the hydraulic support follow-up and frame transfer process, characterized in that: include: S1, collecting the cylinder stroke data of the target hydraulic support during the machine moving process within a period of time in chronological order, and preprocessing the cylinder stroke data; S2, extracting the frame moving features from the pre-processed cylinder stroke data of the target hydraulic support, wherein the frame moving features include the starting point of the frame moving action, the ending point of the frame moving action, the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, and the difference between the cylinder stroke of the target hydraulic support and the hydraulic supports on the left and right sides; S3, identifying the stroke fluctuation type working condition of the frame moving process of the target hydraulic support according to the frame moving action starting point, the frame moving action ending point and the preset frame moving abnormality recognition algorithm; S4, identifying the abnormal stroke condition of the target hydraulic support during the moving process according to the maximum value of the cylinder stroke, the minimum value of the cylinder stroke, the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides, and the pre-trained abnormal stroke recognition model.

2. The abnormal working condition identification method of the hydraulic support during the machine shifting process according to claim 1 is characterized in that: When the S1 pre-processes the cylinder stroke data, it includes: S21, perform outlier processing on the collected cylinder stroke data to obtain the cylinder stroke sequence {x1, x2, ..., x n }, and calculate the difference between the stroke data of two adjacent cylinders in the cylinder stroke sequence to obtain the cylinder stroke difference sequence {Δx1, Δx2, …, Δx n-1 }; S22, from front to back, the cylinder stroke difference sequence {Δx1, Δx2, ..., Δx n-1 } are divided into the same group, and the cylinder stroke data corresponding to each group of cylinder stroke difference data are divided into one group, and the last cylinder stroke data in each group of cylinder stroke data is retained as the representative value of the group of cylinder stroke data, and the representative values ​​of each group of cylinder stroke data are combined to obtain the updated cylinder stroke sequence {x1′, x2′, …, x m '}; S23, calculating the sum of the cylinder stroke difference data in each group of cylinder stroke difference data, and obtaining a combined cylinder stroke difference sequence {Δx1′, Δx2′, …, Δx m ′}.

3. The abnormal working condition identification method of the hydraulic support during the machine moving process according to claim 2 is characterized in that: When extracting the starting point of the frame moving action, the ending point of the frame moving action, the maximum value of the cylinder stroke and the minimum value of the cylinder stroke from the pre-processed cylinder stroke data of the target hydraulic support, S2 includes: S21, obtaining a preset stroke change threshold value X; S22, searching the combined cylinder stroke difference sequence from front to back for the first cylinder stroke difference data that is less than -X, recording its index as a1 and storing it in the rack moving action starting point list a; S23, continue to search backwards in the merged cylinder stroke difference sequence for the first cylinder stroke difference data that is greater than X, record its index as b1 and store it in the list b of the end point of the rack moving action and the start point of the push-slide action; S24, continue to search backwards in the merged cylinder stroke difference sequence for the last cylinder stroke difference data that is greater than X, record its index as c1 and store it in the push-slide action highest point list c; S25, continue to search backward in the merged cylinder stroke difference sequence in turn for the first cylinder stroke difference data less than -X, the first cylinder stroke difference data greater than X, and the last cylinder stroke difference data greater than X, until the last cylinder stroke difference data of the merged cylinder stroke difference sequence, and obtain the starting point list a of the rack moving action as {a1, a2, ..., a q }, the list b of the end of the rack moving action and the start point of the push slide is {b1, b2, ..., b q }、The list of highest points of the push-and-slide action c is {c1, c2, ..., c q }; S26, divide the indexes with the same number in the list a of starting points of the rack moving action, the list b of the end points of the rack moving action and the start points of the push-slide action, and the list c of the highest points of the push-slide action into a set of rack moving action sets, take the point corresponding to a1 as the starting point of the rack moving action, and set a1 in the updated cylinder stroke sequence {x1′, x2′, …, x m The corresponding cylinder stroke data in ′} is used as the maximum cylinder stroke x of the target hydraulic support start , and calculate the rack moving action value Δx corresponding to each set of rack moving action sets ab and push-slide action value Δx bc ; S27, according to the order of numbers from front to back, the rack moving action values ​​Δx corresponding to each rack moving action set are calculated. ab and push-slide action value Δx bc Search until the rack moving action value Δx is found ab and the push-and-slide action value Δx bc The absolute value of the difference between the two is less than the set threshold, and its group is recorded as k, and b k The corresponding point is used as the end point of the rack moving action, and b k In the updated cylinder stroke sequence {x1′, x2′, …, x m The corresponding cylinder stroke data in ′} is used as the minimum cylinder stroke value x of the target hydraulic support end .

4. The abnormal working condition identification method of the hydraulic support during the machine shifting process according to claim 3 is characterized in that: The S3 includes: S31, judge b k Is it in the first set of rack moving actions? If b k If the frame moving process of the target hydraulic support is not in the first frame moving action set, it is determined that a stroke fluctuation type working condition occurs in the frame moving process of the target hydraulic support.

5. The abnormal working condition identification method of the hydraulic support during the machine moving process according to claim 4 is characterized in that: The stroke fluctuation type operating condition includes a shutdown fluctuation type, and after S31, the following further includes: S32, get x start With x end The corresponding time is used as the starting time t of the rack moving action. start and the end time of the rack moving action t end , obtain the target hydraulic support at the starting time t of the frame moving action start and the end time of the rack moving action t end The column pressure data between the two columns are preprocessed to obtain the combined column pressure difference sequence {Δx1′, Δx2′, …, Δx p '}; S33, determining the combined column pressure difference sequence {Δx1′, Δx2′, …, Δx p ′} whether there is column pressure difference data greater than or equal to the pressure fluctuation threshold; if so, it is determined that the stroke fluctuation type condition of the target hydraulic support during the frame moving process is a shutdown fluctuation type.

6. The abnormal working condition identification method of the hydraulic support during the machine shifting process according to claim 4 or 5 is characterized in that: The stroke fluctuation type working condition includes a rack shifting timeout type, and after S31, the following further includes: S34, get x start With x end The corresponding time is used as the starting time t of the rack moving action. start and the end time of the rack moving action t end , determine the starting time t of the rack moving action start and the end time of the rack moving action t end Is the time interval between the two greater than the preset time threshold? If the starting time of the rack moving action is t start and the end time of the rack moving action t end If the time interval between them is greater than a preset time threshold, it is determined that the stroke fluctuation condition of the target hydraulic support during the frame moving process is a frame moving overtime type.

7. The abnormal working condition identification method of the hydraulic support during the machine moving process according to claim 1 is characterized in that: The stroke abnormality type working condition includes the stroke abnormality sudden drop type and the stroke abnormality is small, and before S4, it also includes: The stroke anomaly recognition model is trained based on the maximum and minimum cylinder strokes during the historical shifting process of the hydraulic support, and the difference between the cylinder strokes of the target hydraulic support and the hydraulic supports on the left and right sides. When training the stroke anomaly recognition model, the sudden drop type of stroke anomaly is set as label 1, the small stroke anomaly is set as label 2, and the labels of the remaining normal shifting are set to 0.

8. The abnormal working condition identification method of the hydraulic support during the machine shifting process according to claim 1 or 7 is characterized in that: The trip abnormality recognition model is a decision tree model.

9. The abnormal working condition identification method of the hydraulic support during the machine moving process according to claim 3 is characterized in that: For any set of rack moving action sets, S26 calculates the corresponding rack moving action value Δx ab and push-slide action value Δx bc When , it is realized by the following formula (1) and formula (2) respectively: Δx ab =x a -x b (1) Δx bc =x c -x b (2) In formula (1) and formula (2), x a 、x b and x c are the indexes in the updated cylinder stroke sequence {x1′, x2′, …, x m ′} corresponds to the cylinder stroke data.

Images