Autonomous learning and intelligent decision-making method for hydraulic support cluster control behaviors
By constructing the hydraulic support frame shifting step matrix and intelligent decision-making model, we independently learn to extract and adjust the actions from massive data, and solve the problem of linear arrangement of the hydraulic support automatic follower control system under uncertain working conditions, achieving efficient and intelligent adjustment , reduces labor costs and improves overall efficiency.
Patent Information
- Application Number
- CN202411236452.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-12-27
- Estimated Expiration
- 2044-09-04
AI Technical Summary
The existing hydraulic support automatic follower control system is difficult to realize the linear arrangement of hydraulic support clusters when facing uncertain working conditions, resulting in the need of manual adjustment, which increases labor costs and reduces the degree of intelligence.
By constructing a hydraulic support frame shifting step matrix and pre-trained intelligent decision-making model, we can independently learn to extract and adjust actions from massive data to realize intelligent decision-making and dynamic adjustment of hydraulic support clusters.
Save labor costs, improve the efficiency of hydraulic support moving frame adjustment, improve the degree of intelligence, and ensure that the linear arrangement of hydraulic support clusters meets production requirements.
Smart Images

Figure CN119195827A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent coal mine technology, and in particular to an autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior. Background Art
[0002] The hydraulic support follows the coal mining machine frame movement action is a key control function in coal mining production. According to the coal mining process requirements, each hydraulic support in the hydraulic support cluster should be moved in sequence after the coal mining machine cuts the coal, so as to support the surrounding rock in time and ensure the mining space. At present, the hydraulic support control system can realize the automatic follow-up operation of the hydraulic support through automatic program programming. However, in the actual application process on site, due to the influence of uncertain working conditions such as on-site geological conditions and equipment status, the frame movement control is carried out according to the stroke target value determined by the current hydraulic support automatic follow-up control system. The execution effect cannot 100% meet the requirements of the linear arrangement of the hydraulic support cluster in normal production, thereby affecting the requirements of the straightness of the overall working surface. This is also the main reason why the current hydraulic support automatic follow-up control system is difficult to operate normally. In fact, in actual production, after the hydraulic support performs the first follow-up action (i.e., the first frame movement), it is necessary to control the frame movement again to adjust the linear arrangement of the hydraulic support cluster until the production requirements are met.
[0003] At present, the initial shifting of the hydraulic support at the coal mining face is automatically controlled by the hydraulic support automatic following control system, and the secondary adjustment is performed by manual on-site observation of the linear arrangement of the hydraulic support cluster, and then the operation is adjusted. That is, the adjustment distance of the secondary adjustment is manually determined according to the on-site conditions of the coal mining face, which not only requires relatively high labor costs, but also leads to low adjustment efficiency and intelligence. Summary of the invention
[0004] In order to solve the above technical problems, the present invention provides an autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior. The technical solution of the present invention is as follows:
[0005] An autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior, comprising:
[0006] S1, mathematically quantify the hydraulic support frame moving action in the hydraulic support cluster to construct an initial hydraulic support frame moving step matrix, wherein the frame moving action performed by the hydraulic support in each coal cutting process of the coal mining machine includes basic actions and adjustment actions, and the initial hydraulic support frame moving step matrix includes the frame moving distances of each hydraulic support performing basic actions and adjustment actions in each coal cutting process of the coal mining machine;
[0007] S2, obtaining the position data of the coal mining machine, the hydraulic support frame movement control data and the hydraulic support frame travel data, and determining the value of each element in the initial hydraulic support frame movement step matrix according to the obtained coal mining machine position data, the hydraulic support frame movement control data and the hydraulic support frame travel data, and obtaining the reference hydraulic support frame movement step matrix;
[0008] S3, obtaining a frame shift reference matrix of a target hydraulic support from a frame shift step matrix of a reference hydraulic support, wherein the target hydraulic support is a hydraulic support in a hydraulic support cluster for which a frame shift distance of an adjustment action currently needs to be determined;
[0009] S4, inputting the frame shifting reference matrix of the target hydraulic support into a pre-trained frame shifting adjustment intelligent decision model, and the frame shifting adjustment intelligent decision model outputs the frame shifting distance for the target hydraulic support to perform the adjustment action;
[0010] S5, when the target hydraulic support performs the frame moving action, the target hydraulic support is controlled to perform the adjustment action according to the frame moving distance of the target hydraulic support for performing the adjustment action.
[0011] Optionally, the S1 includes:
[0012] S11, determining that the hydraulic support cluster of the coal mining face includes n hydraulic supports in total, and the coal mining machine advances m knives during coal mining;
[0013] S12, according to the frame shifting action performed by each hydraulic support in the hydraulic support cluster, including the basic action a and the adjustment action b, an initial hydraulic support frame shifting step matrix is constructed, which is expressed as:
[0014]
[0015] in, It refers to the moving distance of the i-th hydraulic support formed by the basic action a when the shearer cuts the j-th coal (1≤i≤n); It refers to the frame moving distance (1≤j≤m) formed by the adjustment action b of the No. i hydraulic support when the coal mining machine cuts the jth piece of coal.
[0016] Optionally, S2 determines the initial hydraulic support frame moving step matrix according to the acquired coal mining machine position data, hydraulic support frame moving action data and hydraulic support frame travel data. and , including:
[0017] S21, determine the time sequence set T of the coal mining machine passing through the hydraulic support according to the coal mining machine position data i , expressed as:
[0018]
[0019] Among them, Ti Each element in represents the various moments when the shearer passes through the i-th hydraulic support m times during the process of the shearer cutting m pieces of coal in the coal mining face;
[0020] S22, determining the hydraulic support frame shifting action event occurrence time sequence set t according to the hydraulic support frame shifting action control data i , expressed as;
[0021]
[0022] Wherein, t represents the time when the hydraulic support moving action event occurs; t i Each element in represents the various moments when the i-th hydraulic support moves o times during the process of cutting m pieces of coal at the coal mining face;
[0023] S23, when to A total of c+1 moments are in and Between, that is, satisfy When , it is determined that the kth to k+cth frame moving actions of the i-th hydraulic support occur at the j-th cutter;
[0024] S24, define Δl[t,t+d] as the change value of the hydraulic support stroke data within d time after time t, then and The specific values are as follows:
[0025] If c = 0, then
[0026] If c = 1, then
[0027] If c ≥ 2, then
[0028] Optionally, when the target hydraulic support is the i-th hydraulic support, when S3 acquires the frame shifting reference matrix of the i-th hydraulic support from the reference hydraulic support frame shifting step matrix, it includes:
[0029] From the reference hydraulic support frame shifting step matrix, take the frame shifting distance value within the range of [ix,i+x] and the advancement range of [jy,j] to form a vector data set of (2x+1)×(2y+1). The vector data set is the frame shifting reference matrix of the i-th hydraulic support, which is expressed as:
[0030]
[0031] Optionally, before S4, the method further includes: training a rack shift adjustment intelligent decision model, and the step of training the rack shift adjustment intelligent decision model includes:
[0032] S41, making training samples: obtaining the historical position data of the coal mining machine, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data during the coal mining process of the coal mining machine, and constructing the training hydraulic support frame moving step matrix according to the historical coal mining machine position data, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data; for hydraulic support i, in the process of the coal mining machine cutting m coal, it adjusts the frame moving distance of action b For target estimation, the distance values of the frame shifting within the frame number range [ix,i+x] and the advancement range [jy,j] are taken from the training hydraulic support frame shifting step matrix to form a vector data set of (2x+1)×(2y+1), and the vector data set of (2x+1)×(2y+1) is used as the input of the frame shifting adjustment intelligent decision model;
[0033] S42, making labels: determining the minimum accuracy of the hydraulic support frame shifting distance as δ, dividing the output of the frame shifting adjustment intelligent decision model into B output categories, and then the label of the frame shifting adjustment intelligent decision model satisfies:
[0034]
[0035] In the formula, represents the floor function; the shift distances corresponding to the B output categories are {[0,2δ),[2δ,4δ),…,[2(B-1)δ,l max ]},l max It refers to the maximum stroke value of each hydraulic support moving action;
[0036] Define the training hydraulic support step matrix If it falls into any output category interval, the label is the output category;
[0037] S43, using training samples and labels to train the intelligent decision model for shelf adjustment through machine learning.
[0038] Optionally, before determining the value of each element in the initial hydraulic support frame moving step matrix according to the acquired coal mining machine position data, the hydraulic support frame moving motion control data and the hydraulic support frame travel data, S2 further includes:
[0039] The hydraulic support stroke data is processed for fluctuation values and missing values.
[0040] All the above optional technical solutions can be combined arbitrarily, and the present invention does not provide detailed descriptions of the structures after the combinations.
[0041] By means of the above scheme, the beneficial effects of the present invention are as follows:
[0042] By acquiring the coal mining machine position data, the hydraulic support frame shifting action control data and the hydraulic support stroke data, and determining the values of each element in the initial hydraulic support frame shifting step matrix based on these data, after obtaining the reference hydraulic support frame shifting step matrix, for any target hydraulic support in the hydraulic support cluster that currently needs to determine the frame shifting distance of the adjustment action, the frame shifting reference matrix determined according to the reference hydraulic support frame shifting step matrix is input into a pre-trained frame shifting adjustment intelligent decision model, and the frame shifting adjustment intelligent decision model outputs the frame shifting distance for the target hydraulic support to perform the adjustment action as the frame shifting distance for the adjustment action during the frame shifting action, thereby providing a frame shifting adjustment intelligent decision model based on massive data of the coal mining working face for intelligent learning, and a method of intelligently deciding the frame shifting distance for the hydraulic support to perform the adjustment action through the frame shifting adjustment intelligent decision model, which can not only save labor costs, but also improve the frame shifting adjustment efficiency of the hydraulic support, and has a high degree of intelligence.
[0043] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a flow chart of an autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior provided by an embodiment of the present invention.
[0045] Figure 2 It is a schematic diagram of fluctuation of travel data of a certain hydraulic support obtained in an embodiment of the present invention.
[0046] Figure 3 It is a schematic diagram of the composition structure of the intelligent decision-making model for rack adjustment in an embodiment of the present invention.
[0047] Figure 4 It is an overall flow chart of the autonomous learning and intelligent decision-making method of the hydraulic support cluster control behavior provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0048] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.
[0049] The autonomous learning and intelligent decision-making method for the control behavior of the hydraulic support cluster provided in the embodiment of the present invention can be implemented by any electronic device with computing functions, such as a PC, mobile terminal or server. Massive on-site data of the coal mining face, such as hydraulic support stroke data, hydraulic support frame shifting action control data, and coal mining machine position data, contain information such as the hydraulic support cluster arrangement conditions and manual operation trajectories. The autonomous learning and intelligent decision-making method for the control behavior of the hydraulic support cluster provided in the embodiment of the present invention utilizes a reasonable machine learning method to autonomously learn on-site manual operation methods from massive data, construct an intelligent decision-making model for frame shifting and adjustment, and realize a full-process adaptive control mode of the hydraulic support following the machine's first basic action + re-adjustment action. Figure 1 As shown, the autonomous learning and intelligent decision-making method for the control behavior of the hydraulic support cluster provided by the embodiment of the present invention includes the following steps S1 to S6:
[0050] S1, mathematize the hydraulic support frame moving action in the hydraulic support cluster to construct an initial hydraulic support frame moving step matrix, wherein the frame moving action performed by the hydraulic support in each coal cutting process of the coal mining machine includes basic actions and adjustment actions, and the initial hydraulic support frame moving step matrix includes the frame moving distance of each hydraulic support performing basic actions and adjustment actions in each coal cutting process of the coal mining machine.
[0051] Specifically, assuming that there are a total of n hydraulic supports on the coal mining face, and after the coal mining machine advances m cuts, during each coal cutting process, the hydraulic support completes two types of frame moving actions, namely basic actions (denoted as a) and adjustment actions (denoted as b). Basic action a refers to the normal frame moving control completed by the hydraulic support according to the machine following process requirements. Generally, it is the first frame moving control with a full stroke. Most of the hydraulic support automatic machine following control systems on site implement this through a preset automation control strategy; adjustment action b refers to the hydraulic support adjusting the frame moving control strategy again according to on-site working conditions such as straightness after completing the basic action. Generally, it is the frame moving control with a small amount of stroke. Currently, most of the on-site operations are implemented after manual judgment. Let l be the frame moving distance caused by each frame moving action of the hydraulic support along the advancing direction of the working face, and the value range is [0≤l≤l max ],l max Refers to the maximum travel value of each hydraulic support frame moving action. According to the composition of the hydraulic support frame moving action in the hydraulic support cluster of the coal mining working face, the embodiment of the present invention quantifies the hydraulic support frame moving action in the hydraulic support cluster and constructs an initial hydraulic support frame moving step matrix.
[0052] Based on the above content, S1 includes: S11, determining that the coal mining face hydraulic support cluster includes n hydraulic supports in total, and the coal mining machine advances m cutters during coal mining; the m cutters here refer to the coal mining machine currently cutting the mth cutter of coal during the coal cutting process;
[0053] S12, according to the frame shifting action performed by each hydraulic support in the hydraulic support cluster, including the basic action a and the adjustment action b, an initial hydraulic support frame shifting step matrix is constructed, which is expressed as:
[0054]
[0055] in, It refers to the moving distance of the i-th hydraulic support formed by the basic action a when the shearer cuts the j-th coal (1≤i≤n); It refers to the moving distance of the No. i hydraulic support formed by adjusting action b when the shearer cuts the j-th coal (1≤j≤m); it is worth noting that or When the value is 0, it indicates that the action has not occurred. Under normal circumstances, the basic action Generally greater than 0 and close to l max Value, adjust action The value is usually 0 (no adjustment action is performed) or a smaller value (adjustment action is performed).
[0056] S2, obtain the coal mining machine position data, hydraulic support frame movement control data and hydraulic support stroke data, and determine the value of each element in the initial hydraulic support frame movement step matrix based on the obtained coal mining machine position data, hydraulic support frame movement control data and hydraulic support stroke data, and obtain the reference hydraulic support frame movement step matrix.
[0057] Specifically, various types of sensors are installed on the coal mining working surface. These types of sensors send data to the electronic device after collecting data. The electronic device can obtain data from the data collected by the corresponding sensor based on the required data type. For example, the position data of the coal mining machine can be obtained from the data collected by the positioning sensor installed on the coal mining machine; the hydraulic support stroke data can be obtained from the data collected by the hydraulic support stroke sensor installed on the hydraulic support. The hydraulic support frame shifting action control data is data that can reflect whether the hydraulic support frame shifting action occurs. In the embodiment of the present invention, the data is taken as the time when the hydraulic support frame shifting action event occurs.
[0058] In an optional specific embodiment, the hydraulic support stroke data may generate data fluctuations, data loss, and other problems during the collection, transmission, and storage process. These problems will affect the accuracy of the decision-making results. Therefore, after acquiring the hydraulic support stroke data, the embodiment of the present invention can also perform fluctuation value processing and missing value processing on the hydraulic support stroke data. The same is true for the coal mining machine position data and the hydraulic support frame movement control data.
[0059] Specifically, for the problem of data fluctuation, the embodiment of the present invention uses filtering technology to smooth the data, reduce the impact of short-term fluctuations, and make the data smoother. For the problem of data loss, the embodiment of the present invention uses previous values for filling to make the data more complete.
[0060] like Figure 2 As shown in the figure, it shows the fluctuation of the hydraulic support stroke data of hydraulic support No. 66. If the hydraulic support does not move, but the hydraulic support stroke data changes repeatedly in a small range, it is determined that the hydraulic support stroke data fluctuates, and the hydraulic support stroke data values during this period are [v1, v2, ..., v z ], v1 represents the first point when the fluctuation starts, z represents a total of z points, and the average value of the hydraulic support stroke data during the fluctuation period is v mean , will [v mean , v mean , ..., v mean ] data to replace the original fluctuation data. The mathematical formula can be expressed as:
[0061]
[0062] If the data is lost, for example, the hydraulic support stroke data value of the i-th hydraulic support in a certain period of time is empty, then the value before the hydraulic support stroke data is lost is used for filling. The value of the hydraulic support stroke data before the loss is v p , the value during the loss period is recorded as None. The mathematical relationship can be expressed as:
[0063] v p ,None,None,None,None→v p , v p , v p , v p , v p .
[0064] Through the above-mentioned fluctuation value processing and missing value processing methods, not only the integrity and accuracy of the hydraulic support stroke data are guaranteed, but also a high-quality data foundation is provided for subsequent model updates.
[0065] In a specific embodiment, S2 determines the initial hydraulic support frame moving step matrix according to the acquired coal mining machine position data, hydraulic support frame moving action data and hydraulic support stroke data. and , including:
[0066] S21, determine the time sequence set T of the coal mining machine passing through the hydraulic support according to the coal mining machine position data i , expressed as:
[0067]
[0068] Among them, T i The elements in represent the moments when the shearer passes through the i-th hydraulic support m times during the process of the shearer cutting m pieces of coal in the coal mining face.
[0069] S22, determining the hydraulic support frame shifting action event occurrence time sequence set t according to the hydraulic support frame shifting action control data i , expressed as;
[0070]
[0071] Wherein, t represents the time when the hydraulic support moving action event occurs; t i The elements in represent the moments when the i-th hydraulic support moves o times during the process of cutting m pieces of coal at the coal mining face.
[0072] S23, when to A total of c+1 moments are in and Between, that is, satisfy When , it is determined that the kth to k+cth frame moving action of the i-th hydraulic support occurs at the j-th cutter.
[0073] S24, define Δl[t,t+d] as the change value (reduction amount) of the hydraulic support stroke data within d time after time t, then and The specific values are as follows:
[0074] If c = 0, then
[0075] If c = 1, then
[0076] If c ≥ 2, then
[0077] Through the above-mentioned mathematical process, massive data such as hydraulic support travel data, hydraulic support movement control data, coal mining machine position data, etc. collected at the working face site can be converted into a mathematical expression of the spatiotemporal evolution of the hydraulic support movement process of the entire coal mining working face, providing a data basis for the next step of autonomous learning and intelligent decision-making of the hydraulic support cluster control behavior.
[0078] S3, obtaining a frame shift reference matrix of a target hydraulic support from a frame shift step matrix of a reference hydraulic support, wherein the target hydraulic support is a hydraulic support in a hydraulic support cluster for which a frame shift distance of an adjustment action currently needs to be determined.
[0079] Among them, the frame shifting reference matrix is determined by the frame shifting action between the first preset value hydraulic supports on the left and right sides of the target hydraulic support and the second preset value cutter before the coal mining machine cuts the current cutter in the reference hydraulic support frame shifting step matrix; the specific values of the first preset value and the second preset value can be determined as needed, and the embodiment of the present invention is not limited to this.
[0080] In a specific embodiment, when the target hydraulic support is the i-th hydraulic support, when S3 obtains the frame shift reference matrix of the i-th hydraulic support from the reference hydraulic support frame shift step matrix, the frame shift distance value within the range of [ix,i+x] and the advancement range of [jy,j] is taken from the reference hydraulic support frame shift step matrix to form a vector data set of (2x+1)×(2y+1). The vector data set is the frame shift reference matrix of the i-th hydraulic support, which is expressed as:
[0081]
[0082] It should be noted that if the target hydraulic support is located at the end position so that there is no hydraulic support on one side, for example, hydraulic support No. 1 has no hydraulic support on its left side, at this time, the reference matrix of the frame transfer is supplemented with the standard value on the left side according to the number of missing hydraulic supports, that is, the position of the basic action a is supplemented with l max The value of adjustment action b is added with 0, which means that no adjustment action occurs.
[0083] S4, inputting the frame shifting reference matrix of the target hydraulic support into a pre-trained frame shifting adjustment intelligent decision model, and the frame shifting adjustment intelligent decision model outputs the frame shifting distance for the target hydraulic support to perform the adjustment action.
[0084] The input of the intelligent decision-making model for frame shifting adjustment is the frame shifting reference matrix of the target hydraulic support, and the output is 0 or a positive value. Output 0 means that no adjustment action is required, and the positive output value is one of multiple classification values, each of which corresponds to a frame shifting distance.
[0085] It should be noted that, before S4, the step further includes: training the intelligent decision model for rack shifting and adjustment. Specifically, the step of training the intelligent decision model for rack shifting and adjustment includes:
[0086] S41, making training samples: obtaining the historical position data of the coal mining machine, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data during the coal mining process of the coal mining machine, and constructing the training hydraulic support frame moving step matrix according to the historical coal mining machine position data, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data; for hydraulic support i, in the process of the coal mining machine cutting m coal, it adjusts the frame moving distance of action b For target estimation, the distance values of the frame shifting within the frame number range [ix,i+x] and the advancement range [jy,j] are taken from the training hydraulic support frame shifting step matrix to form a (2x+1)×(2y+1) vector data set, which is used as the input of the frame shifting adjustment intelligent decision model.
[0087] Through the above-mentioned mathematical process, the historical hydraulic support travel data, historical coal mining machine position data, and historical hydraulic support movement control data collected at the working face site can be converted into a mathematical expression of the spatiotemporal evolution of the hydraulic support movement process of the entire coal mining working face, providing a complete available sample space for the intelligent decision-making model for the movement adjustment.
[0088] The method of obtaining historical coal mining machine position data, historical hydraulic support frame movement motion control data and historical hydraulic support frame travel data, and constructing a training hydraulic support frame movement step matrix based on the historical coal mining machine position data, historical hydraulic support frame movement motion control data and historical hydraulic support frame travel data is similar to the principle in the above-mentioned embodiment S2 and will not be repeated here.
[0089] Through the above S41, a plurality of (2x+1)×(2y+1) vector data sets can be obtained, and these vector data sets are used as sample data sets, namely, the input of the intelligent decision-making model for rack adjustment.
[0090] It should be noted that if during the training sample preparation process, the hydraulic support is located at the end position so that the number of hydraulic supports on one side does not meet the sampling conditions, for example, there is no hydraulic support on the left side of hydraulic support No. 1, then the standard value is supplemented on the left side according to the missing number of hydraulic supports, that is, the position of basic action a is supplemented with l max The value is set, and the position of the adjustment action b is supplemented with a value of 0, that is, it is considered that no adjustment action has occurred, and the vector data when i=1 is obtained as follows. If there is no bracket on the right side, the vector data is supplemented in the same way.
[0091]
[0092] S42, making labels: determining the minimum accuracy of the hydraulic support frame shifting distance as δ, dividing the output of the frame shifting adjustment intelligent decision model into B output categories, and then the label of the frame shifting adjustment intelligent decision model satisfies:
[0093]
[0094] In the formula, represents the floor function; the shift distances corresponding to the B output categories are {[0,2δ),[2δ,4δ),…,[2(B-1)δ,l max ]},l maxIt refers to the maximum stroke value of each hydraulic support frame moving action; δ is determined according to the minimum value of the hydraulic support frame moving action;
[0095] Define the training hydraulic support step matrix If it falls into any output category interval, the label is the output category.
[0096] Based on the above training hydraulic support frame shifting step matrix, multiple labels of the frame shifting adjustment intelligent decision model can be obtained to form a label database.
[0097] S43, using training samples and labels to train the intelligent decision model for shelf adjustment through machine learning.
[0098] Specifically, when training the rack adjustment intelligent decision model through training samples and labels, the training samples and their corresponding labels are first divided into a training set and a test set, and then after the rack adjustment intelligent decision model is assigned initial values, the rack adjustment intelligent decision model is autonomously learned and continuously iterated through the training set, and the rack adjustment intelligent decision model is verified through the test set. When the test is qualified, a trained rack adjustment intelligent decision model is obtained.
[0099] Since the training hydraulic support step matrix contains the spatiotemporal evolution trajectory of the basic actions and adjustment action control strategies, especially the adjustment actions are controlled and executed after manual on-site observation, which contains a lot of manual operation experience. Therefore, the embodiment of the present invention uses machine learning to use the frame moving distance of a hydraulic support adjustment action as the output of the frame moving adjustment intelligent decision model, and uses the frame moving distances of a certain number of neighbors on the left and right of the hydraulic support (spatial arrangement features) and the frame moving distances of the previous few cuts (time accumulation features) as the input of the frame moving adjustment intelligent decision model, and trains to obtain the frame moving adjustment intelligent decision model.
[0100] The embodiment of the present invention does not specifically limit the specific type of the intelligent decision-making model for rack moving and adjustment. For example, the intelligent decision-making model for rack moving and adjustment is KNN, SVM, decision tree or random forest, etc., or it can be a combination of several models.
[0101] As the on-site coal mining work continues, the amount of data on the working face will increase. That is, the training samples and labels produced by training the hydraulic support step matrix will increase, which means that more and more manual operation experience can be learned. By analyzing the characteristics of the training samples, the training samples have both time series characteristics and spatial characteristics. Therefore, the embodiment of the present invention combines convolutional networks, long short-term memory neural networks, and attention mechanisms to construct an intelligent decision-making model for frame adjustment. Figure 3As shown, it is a schematic diagram of the composition structure of the intelligent decision-making model for frame adjustment in an embodiment of the present invention. By continuously iterating and updating the model, the intelligent decision-making model for frame adjustment can autonomously learn more on-site manual operation methods. At present, the intelligent decision-making model for frame adjustment can achieve an accuracy rate of 83.84% through autonomous learning of massive data on the working surface, meeting the standards for industrial on-site applications.
[0102] Specifically, Figure 3 As shown, taking the input determined by actual needs as a (7*13) vector data set and the output as three-classification rack adjustment intelligent decision model parameters as an example, the specific rack adjustment intelligent decision model architecture diagram is as follows:
[0103] The network parameters of each layer are: Input: (7*13) vector data set, before passing through the first convolution layer, the data will be converted to 1D. The parameters of the first convolution layer are: type "Conv2d", input channel is 1, output channel is 32, convolution kernel size is (2,3), padding: 1. The parameters of the first batch normalization layer: type "BatchNorm2d", number of features is 32. The first activation function is: relu. The first random dropout layer: probability is 0.5. The parameters of the second convolution layer are: type "Conv2d", input channel is 32, output channel is 64, convolution kernel size is (2,3), padding: 1. The parameters of the second batch normalization layer: type "BatchNorm2d", number of features is 64. The second activation function is: relu. The second random dropout layer: probability is 0.3. Flattening layer: converts multi-dimensional data into one dimension. Parameters for the bidirectional LSTM layer: type "LSTM", input size 64, hidden size 100, number of layers 2; batch-first: True, bidirectional: True. First fully connected layer: type "Linear", input size: 200, output size: 50. Third batch normalization layer: type "BatchNorm1d", number of features 50. Attention layer: composed of a linear layer converted to 25 dimensions, applying the Tanh activation function, and then converted back to 1 dimension through a linear layer to calculate the weights. The weights are normalized using Softmax and then the input is weighted summed. Second fully connected layer: type "Linear", input size: 50, output size: 3.
[0104] S5, when the target hydraulic support performs the frame moving action, the target hydraulic support is controlled to perform the adjustment action according to the frame moving distance of the target hydraulic support for performing the adjustment action.
[0105] Specifically, when controlling the target hydraulic support to perform the adjustment action according to the moving distance of the target hydraulic support, the target support can be adjusted manually according to the moving distance of the target hydraulic support, or the target hydraulic support can be controlled by an automated tracking system to perform the adjustment action.
[0106] The method provided by the embodiment of the present invention, through processing of massive data of the working face, the autonomous learning and intelligent decision-making method of the control behavior of the hydraulic support cluster can intelligently decide how the hydraulic support cluster should be moved and regulated again after the first control of the hydraulic support cluster, and help adjust the linear arrangement of the hydraulic support cluster until the production requirements are met.
[0107] In summary, the autonomous learning and intelligent decision-making method for the control behavior of the hydraulic support cluster provided in the embodiment of the present invention pre-processes the massive data on the working face such as the historical coal mining machine position data, the historical hydraulic support frame movement control data and the historical hydraulic support stroke data (fluctuation value processing and missing value processing), mathematizes the hydraulic support frame movement action, and produces a training hydraulic support frame movement step matrix. Then, based on the training hydraulic support frame movement step matrix, a vector data set of size (2x+1)×(2y+1) is extracted as the sample input of the frame movement adjustment intelligent decision-making model, and the frame movement distance of the adjustment action b is used as the output of the frame movement adjustment intelligent decision-making model. By analyzing the characteristics of the data set, the data set has both time series characteristics and spatial characteristics, so the convolutional network, the long short-term memory neural network, and the attention mechanism are combined to construct the frame movement adjustment intelligent decision-making model. Based on the training hydraulic support frame movement step matrix, the frame movement adjustment intelligent decision-making model is autonomously learned and continuously iterated. In the subsequent mining process of the coal mining face, after obtaining the reference hydraulic support frame step matrix, the frame adjustment intelligent decision model can output the control behavior of the hydraulic support cluster, thereby realizing intelligent decision-making, and then accurately classifying the vector data set to achieve the intelligent decision-making effect of the hydraulic support cluster control behavior. The autonomous learning and intelligent decision-making method of the hydraulic support cluster control behavior can help underground workers adjust the hydraulic support frame step in time, thereby adjusting the straightness of the hydraulic support cluster, which is beneficial to improving the efficiency of coal mining and the overall support effect. Figure 4 An overall flow chart of the autonomous learning and intelligent decision-making method for the hydraulic support cluster control behavior provided in an embodiment of the present invention.
[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be regarded as within the scope of protection of the present invention.
Claims
1. An autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior, characterized in that: include: S1, mathematically quantify the hydraulic support frame moving action in the hydraulic support cluster to construct an initial hydraulic support frame moving step matrix, wherein the frame moving action performed by the hydraulic support in each coal cutting process of the coal mining machine includes basic actions and adjustment actions, and the initial hydraulic support frame moving step matrix includes the frame moving distances of each hydraulic support performing basic actions and adjustment actions in each coal cutting process of the coal mining machine; S2, obtaining the position data of the coal mining machine, the hydraulic support frame movement control data and the hydraulic support frame travel data, and determining the value of each element in the initial hydraulic support frame movement step matrix according to the obtained coal mining machine position data, the hydraulic support frame movement control data and the hydraulic support frame travel data, and obtaining the reference hydraulic support frame movement step matrix; S3, obtaining a frame shift reference matrix of a target hydraulic support from a frame shift step matrix of a reference hydraulic support, wherein the target hydraulic support is a hydraulic support in a hydraulic support cluster for which a frame shift distance of an adjustment action currently needs to be determined; S4, inputting the frame shifting reference matrix of the target hydraulic support into a pre-trained frame shifting adjustment intelligent decision model, and the frame shifting adjustment intelligent decision model outputs the frame shifting distance for the target hydraulic support to perform the adjustment action; S5, when the target hydraulic support performs the frame moving action, the target hydraulic support is controlled to perform the adjustment action according to the frame moving distance of the target hydraulic support for performing the adjustment action.
2. The autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior according to claim 1 is characterized in that: The S1 includes: S11, determining that the hydraulic support cluster of the coal mining face includes n hydraulic supports in total, and the coal mining machine advances m knives during coal mining; S12, according to the frame shifting action performed by each hydraulic support in the hydraulic support cluster, including the basic action a and the adjustment action b, an initial hydraulic support frame shifting step matrix is constructed, which is expressed as: in, It refers to the moving distance of the i-th hydraulic support formed by the basic action a when the shearer cuts the j-th coal (1≤i≤n); It refers to the frame moving distance (1≤j≤m) formed by the adjustment action b of the No. i hydraulic support when the coal mining machine cuts the jth piece of coal.
3. The autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior according to claim 2 is characterized in that: S2 determines the initial hydraulic support frame moving step matrix according to the acquired coal mining machine position data, hydraulic support frame moving action data and hydraulic support stroke data. and , including: S21, determine the time sequence set T of the event when the coal mining machine passes through the hydraulic support according to the position data of the coal mining machine. i , expressed as: Among them, T i Each element in represents the various moments when the shearer passes through the i-th hydraulic support m times during the process of the shearer cutting m pieces of coal in the coal mining face; S22, determining the hydraulic support frame shifting action event occurrence time sequence set t according to the hydraulic support frame shifting action control data i , expressed as; Where t represents the time when the hydraulic support moving action event occurs; t i Each element in represents the various moments when the i-th hydraulic support moves o times during the process of cutting m pieces of coal at the coal mining face; S23, when to A total of c+1 moments are in and Between, that is, satisfy When , it is determined that the kth to k+cth frame moving actions of the i-th hydraulic support occur at the j-th cutter; S24, define Δl[t,t+d] as the change value of the hydraulic support stroke data within d time after time t, then and The specific values are as follows: If c = 0, then If c = 1, then If c ≥ 2, then 4. The autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior according to claim 1, 2 or 3, characterized in that: When the target hydraulic support is the i-th hydraulic support, the step S3, when acquiring the frame shifting reference matrix of the i-th hydraulic support from the reference hydraulic support frame shifting step matrix, includes: From the reference hydraulic support frame shifting step matrix, take the frame shifting distance value within the range of [ix,i+x] and the advancement range of [jy,j] to form a vector data set of (2x+1)×(2y+1). The vector data set is the frame shifting reference matrix of the i-th hydraulic support, which is expressed as:
5. The autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior according to claim 1 is characterized in that: The step before S4 also includes: training the intelligent decision-making model for rack shifting and adjustment, and the step of training the intelligent decision-making model for rack shifting and adjustment includes: S41, making training samples: obtaining the historical position data of the coal mining machine, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data during the coal mining process of the coal mining machine, and constructing the training hydraulic support frame moving step matrix according to the historical coal mining machine position data, the historical hydraulic support frame moving motion control data and the historical hydraulic support stroke data; for hydraulic support i, in the process of the coal mining machine cutting m coal, it adjusts the frame moving distance of action b For target estimation, the distance values of the frame shifting within the frame number range [ix,i+x] and the advancement range [jy,j] are taken from the training hydraulic support frame shifting step matrix to form a vector data set of (2x+1)×(2y+1), and the vector data set of (2x+1)×(2y+1) is used as the input of the frame shifting adjustment intelligent decision model; S42, making labels: determining the minimum accuracy of the hydraulic support frame shifting distance as δ, dividing the output of the frame shifting adjustment intelligent decision model into B output categories, and then the label of the frame shifting adjustment intelligent decision model satisfies: In the formula, represents the floor function; the shift distances corresponding to the B output categories are {[0,2δ),[2δ,4δ),…,[2(B-1)δ,l max ]},l max It refers to the maximum stroke value of each hydraulic support moving action; Define the training hydraulic support step matrix If it falls into any output category interval, the label is the output category; S43, using training samples and labels to train the intelligent decision model for shelf adjustment through machine learning.
6. The autonomous learning and intelligent decision-making method for hydraulic support cluster control behavior according to claim 1 is characterized in that: Before determining the value of each element in the initial hydraulic support frame shifting step matrix according to the acquired coal mining machine position data, the hydraulic support frame shifting motion control data and the hydraulic support frame stroke data, S2 also includes: The hydraulic support stroke data is processed for fluctuation values and missing values.
Citation Information
Patent Citations
Coal mining face hydraulic support straightening system and straightening method
CN103541754A
Hydraulic support machine following action time sequence planning method in coal mining process
CN110847943A
Intelligent decision-making method for machine following propulsion behavior of hydraulic support group
CN111472827A
Vision-based automatic straightening control system and method for hydraulic support of fully mechanized coal mining face
CN114922667A
Mine intelligent mining digital twin multi-agent cooperative control method
CN115437257A
Cited By
Robot cluster individual behavior control method for space-time marching task
CN120178746A
Fully mechanized coal mining face hydraulic support multi-mode support moving control method and related equipment
CN121497399A
Hydraulic support multi-mode shifting control method for fully mechanized coal mining face and related equipment
CN121497399B
