Coal rock cutting state detection method

Abstract

The invention relates to a coal rock cutting state detection method, and belongs to the technical field of intelligent coal mines. Comprising the following steps: acquiring coal rock cutting state signals acquired by a plurality of sensors when the coal cutter currently cuts coal rock; analyzing the coal rock cutting state signal acquired by each sensor through an improved cyclic spectrum to obtain a modal component of the signal acquired by each sensor; calculating a sample entropy characteristic value of a modal component of a signal collected by each sensor; according to the sample entropy characteristic value of the modal component of each sensor acquisition signal, constructing an ICSCM matrix of the plurality of sensor acquisition signals; and inputting the ICSCM matrix into a pre-trained coal rock cutting state detection classifier, and determining the current coal rock cutting state of the coal mining machine according to an output result of the coal rock cutting state detection classifier. The method can reduce the calculation cost, and improves the accuracy and reliability of a detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent coal mines, and in particular to a method for detecting a coal-rock cutting state. Background Art

[0002] Coal rock is an important mineral resource, and its quality and reserves are of great significance to the mining and utilization of coal mines. With the continuous development of science and technology, coal rock sensor technology has been widely used, resulting in a large amount of multi-sensor data. Traditional coal rock cutting state detection methods often rely on single sensor data, such as geological exploration, geophysical exploration, etc., which have the problems of incomplete information acquisition and low accuracy. In order to more accurately detect and process coal rock information, it is necessary to use a variety of sensor data, such as seismic wave data, electromagnetic data, geochemical data, etc., to comprehensively analyze the physical, chemical, and geological characteristics of coal rock, so as to achieve comprehensive detection and processing of coal rock. Therefore, coal rock detection and processing of multi-sensor data has become a hot and difficult issue in current research. By comprehensively utilizing multi-sensor data, the accuracy and reliability of coal rock cutting state detection can be improved, providing better technical support for the mining and utilization of coal mines.

[0003] Today's coal and rock sensor data processing requires the collected multi-sensor feature groups to be segmented into a feature vector to achieve sensor data fusion. This process requires reasonable optimization of parameters, which will increase the computational cost. In addition, the current multi-sensor fusion algorithm cannot make good use of the interaction and interdependence between the data collected by each sensor when fusing the data collected by each sensor, resulting in low accuracy and reliability of the detection results. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for detecting the coal-rock cutting state. The technical solution of the present invention is as follows: A method for detecting coal-rock cutting state, comprising: S1, obtaining coal and rock cutting state signals collected by multiple sensors when the coal mining machine is currently cutting coal and rock; S2, analyzing the coal-rock cutting state signal collected by each sensor through an improved cyclic spectrum to obtain the modal component of the signal collected by each sensor; S3, calculating the sample entropy eigenvalue of the modal component of the signal collected by each sensor; S4, constructing an ICSCM matrix of the signals collected by multiple sensors according to the sample entropy eigenvalue of the modal component of each sensor collected signal; S5, inputting the ICSCM matrix into a pre-trained coal-rock cutting state detection classifier, and determining the current coal-rock cutting state of the coal mining machine according to the output result of the coal-rock cutting state detection classifier.

[0005] Optionally, any sensor is determined to be a target sensor, and when S2 analyzes the coal-rock cutting state signal acquired by the target sensor through an improved cyclic spectrum to obtain a modal component of the signal acquired by the target sensor, it includes: S21, the coal-rock cutting state signal collected by the target sensor It is divided into several components by the window function, each component includes L sampling points, and the spectrum of each component is calculated by formula (1):

[0006] In formula (1), Represents discrete frequency The i The spectrum of the components, i =1,……, Q , Q Indicates the number of time shifts of the window function; represents the window function; n represents the index of the data point in the window; L Indicates the length of the window function, that is, the number of sampling points included in each component; R Indicates the step size of the window function; discrete frequency , k =0,……, L -1, , Indicates frequency resolution; F s Indicates the sampling frequency; S22, analyzing the spectrum of each component by using the improved cyclic spectrum formula (2) to obtain the modal components of the target sensor acquisition signal;

[0007] In formulas (2) and (3), α represents the cycle frequency, f Indicates frequency; Represents the square norm of the window function; M express The peak moment, that is, the peak point of the window function, Satisfies the symmetry, that is .

[0008] Optionally, the S3 includes: S31, determine any sensor as the target sensor, and calculate the average value of the effective peak value of the modal component of the target sensor acquisition signal by formula (4) :

[0009] In formula (4), △g is Frequency resolution in direction; e represents the index, ranging from 2 to D, where D is the total number of valid peaks; f =1, 2, ⋯, O, where O is the frequency index range obtained by discretization of △g; S32, Reconstruction The phase space of the reconstructed phase space matrix Y is defined as:

[0010] In formula (5), m is the embedding dimension in the phase space, indicating the number of sampling points contained in each vector; G is the total number of modal components; K Represents the total number of valid vectors of the reconstructed phase space matrix Y, that is, the reconstructed phase space matrix Y number of rows; and They represent the first and valid vectors; S33, randomly select a vector from the reconstructed phase space matrix Y as a template vector; S34, determining vectors similar to the template vector within a preset similarity threshold, and determining the number of vectors similar to the template vector; S35, for the embedding dimension m, calculating the ratio of the number of vectors that are similar to the template vector within a preset similarity threshold; S36, repeating steps S33 to S35 to obtain the ratio of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector when the embedding dimension is m; S37, return to S32, calculate the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimension is m+1; S38, determining the sample entropy eigenvalue of the modal component of the signal collected by the target sensor according to the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1.

[0011] Optionally, the S34 includes: S341, calculating the maximum difference between any vector and the corresponding element in the template vector; S342, determining that any vector whose maximum difference with the corresponding element in the template vector is less than a preset similarity threshold is a vector similar to the template vector, and the number of the vectors is used as the number of vectors similar to the template vector.

[0012] Optionally, the S35 includes: For the embedding dimension m, the ratio of the number of vectors that are similar to the template vector within the preset similarity threshold is calculated using formula (6): :

[0013] In formula (6), δ represents the preset similarity threshold, It represents the number of vectors that are similar to the template vector within the preset similarity threshold δ.

[0014] Optionally, the S38 includes: S381, using formula (7) to calculate the average value of the ratio of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector when the embedding dimension is m ; The same is true when the embedding dimension is m+1:

[0015] S382, according to the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1, the sample entropy eigenvalue of the modal component of the target sensor acquisition signal is determined by formula (8):

[0016] In formula (8), It indicates the ratio of the number of vectors that are similar to the template vector within the preset similarity threshold when the embedding dimension is m+1.

[0017] Optionally, the S4 includes: S41, constructing a high-dimensional feature matrix of the signals collected by multiple sensors according to the sample entropy eigenvalues ​​of the modal components of the signals collected by each sensor; S42, constructing an ICSCM matrix of the signals collected by the multiple sensors according to the high-dimensional feature matrix of the signals collected by the multiple sensors.

[0018] Optionally, the S41 includes: S411, selecting J modal components with the largest sample entropy eigenvalues ​​from all modal components of the signal collected by each sensor; S412, constructing a high-dimensional feature matrix of signals collected by multiple sensors based on the modal components selected by each sensor HDF :

[0019] In formula (9), HDF represents a high-dimensional feature matrix with row vector q and column vector J. HDF i represents the i-th column of the high-dimensional feature matrix, q represents the number of sensors, SEij Represents the sample entropy eigenvalue of the jth modal component of the ith sensor.

[0020] Optionally, the S42 includes: constructing an ICSCM matrix of the signals collected by the multiple sensors according to the high-dimensional feature matrix of the signals collected by the multiple sensors by using formula (10) and formula (11):

[0021] in, u represents the average value of each column of the high-dimensional feature matrix, and ICSCM represents the ICSCM matrix.

[0022] The coal-rock cutting state detection method according to claim 1 is characterized in that the coal-rock cutting state detection classifier is an extreme learning machine.

[0023] 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.

[0024] By means of the above scheme, the beneficial effects of the present invention are as follows: By analyzing the coal-rock cutting state signal collected by each sensor through an improved cyclic spectrum, key features that are beneficial and sensitive to the detection of the coal-rock cutting state can be obtained; by calculating the sample entropy eigenvalue of the modal component of the signal collected by each sensor, and constructing the ICSCM matrix of multiple sensor collection signals according to the sample entropy eigenvalue of the modal component of the signal collected by each sensor, the coal-rock cutting state detection process not only does not require parameter optimization, greatly reducing the calculation cost, but the ICSCM matrix can fully retain the interaction between different sensors, making the detection results more accurate and reliable.

[0025] 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

[0026] Figure 1 It is a flow chart of the coal rock cutting state detection method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] 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.

[0028] The coal-rock cutting state detection method provided in the embodiment of the present invention can be implemented by any electronic device with computing function, such as a PC, a mobile terminal or a server. Figure 1As shown, the coal-rock cutting state detection method provided by the embodiment of the present invention includes the following steps S1 to S5.

[0029] S1, obtaining coal and rock cutting status signals collected by multiple sensors when the coal mining machine is currently cutting coal and rock.

[0030] Among them, the embodiment of the present invention does not specifically limit the types of multiple sensors, which can be sensors of the same type or sensors of different types, as long as the data collected by these sensors are related to the coal and rock cutting status. For example, the multiple sensors can be multiple current sensors, specifically three current sensors, which are respectively installed on the cutting motor of the coal mining machine and correspond to the three-phase current respectively. For another example, the multiple sensors include current sensors, vibration sensors, sound sensors, etc. In order to ensure that the coal and rock cutting status signals collected by multiple sensors can be comprehensively utilized to accurately detect the coal and rock cutting status, the number of sensors can be at least three.

[0031] Specifically, each sensor is electrically connected to the electronic device. When each sensor collects a coal rock cutting status signal during the current coal rock cutting process of the coal mining machine, it is sent to the electronic device in real time. The electronic device obtains the coal rock cutting status signal collected by each sensor by receiving the signal collected by each sensor.

[0032] S2, analyzing the coal-rock cutting state signal collected by each sensor through the improved cyclic spectrum to obtain the modal component of the signal collected by each sensor.

[0033] In order to realize data fusion of multiple sensors, the embodiment of the present invention uses improved cyclic spectrum analysis to decompose the coal rock cutting state signal collected by each sensor into a series of modal components. The modal components can be understood as the signal components decomposed in each window.

[0034] For ease of description, in the subsequent embodiments of the present invention, any sensor is defined as a target sensor. On this basis, when the coal-rock cutting state signal collected by the target sensor is analyzed by the improved cyclic spectrum to obtain the modal component of the target sensor collection signal, the step S2 includes steps S21 and S22: S21, the coal-rock cutting state signal collected by the target sensor The window function is used to divide the spectrum into several components, each of which includes L sampling points, and the spectrum of each component is calculated by the following formula (1):

[0035] In formula (1), Represents discrete frequency The i The spectrum of the components, i=1,……, Q , Q Indicates the number of time shifts of the window function; represents the window function; n represents the index of the data point in the window; L Indicates the length of the window function, that is, the number of sampling points included in each component; R Indicates the step size of the window function; discrete frequency , k =0,……, L -1, , Indicates frequency resolution; F s Indicates the sampling frequency;

[0036] Specifically, when the window function is divided into several components, the window overlap rate of the window function sliding on the coal-rock cutting state signal collected by the target sensor is L / 2.

[0037] S22, the spectrum of each component is analyzed by the following improved cyclic spectrum (ICS) formula (2) to obtain the modal components of the target sensor acquisition signal;

[0038] In formulas (2) and (3), α represents the cycle frequency, f Indicates frequency; Represents the square norm of the window function; M express The peak moment, that is, the peak point of the window function, Satisfies the symmetry, that is . M The value of is generally L / 2. Formula (3) is used to calculate The total energy in the time domain, that is, the sum of the squares of all sampling points, affects the distribution of the window function in the frequency domain, and thus affects the frequency resolution.

[0039] S3, calculating the sample entropy eigenvalue of the modal component of the signal collected by each sensor.

[0040] In a specific embodiment, the step S3 includes steps S31 to S38: S31, calculate the average value of the effective peak value of the modal component of the target sensor acquisition signal by the following formula (4): :

[0041] In formula (4), △g is Frequency resolution in direction; erepresents the index, ranging from 2 to D, where D is the total number of valid peaks; f =1, 2, ⋯, O, where O is the frequency index range obtained by discretization of △g; S32, Reconstruction The phase space of the reconstructed phase space matrix Y is defined as:

[0042] In formula (5), m is the embedding dimension in the phase space, indicating the number of sampling points contained in each vector (each row in the reconstructed phase space matrix Y is a vector). For example, if m=2, each vector will contain two consecutive sampling points. G is the total number of modal components; K Represents the total number of valid vectors of the reconstructed phase space matrix Y, that is, the reconstructed phase space matrix Y the number of rows, K =G−m+1 is calculated based on a sliding window to ensure the reconstructed phase space matrix Y Each row of can contain m complete modal components, and They represent the first and A valid vector.

[0043] Phase space matrix Y Each row in represents a vector (time series vector), that is, under the embedding dimension m, from the time series A segment generated by a sliding window in is used to describe the local dynamic characteristics of the time series. The phase space matrix Y Each column in represents the modal component at the same position in all vectors, which is the value of a specific relative position point across all sliding windows. Therefore, the phase space matrix Y The row direction in is the sequence of sliding windows, and the column direction is the position component in the vector.

[0044] S33, randomly select a vector from the reconstructed phase space matrix Y as a template vector.

[0045] S34, determining vectors similar to the template vector within a preset similarity threshold δ, and determining the number of vectors similar to the template vector.

[0046] Specifically, the S34 includes: S341, calculating the maximum difference between any vector and the corresponding element in the template vector.

[0047] This step is expressed by the formula: ;in, represents the template vector, represents the vector γ, Representation vector With template vector The maximum difference between corresponding elements in .

[0048] S342, determining that any vector whose maximum difference with the corresponding element in the template vector is less than a preset similarity threshold is a vector similar to the template vector, and the number of the vectors is used as the number of vectors similar to the template vector.

[0049] Specifically, if If it is less than δ, then Y(y) is determined to be the same as the template vector Similar vectors. The statistically reconstructed phase space matrix Y is All vectors similar to the template vector The number of similar vectors .

[0050] S35, for the embedding dimension m, calculating the ratio of the number of vectors that are similar to the template vector within a preset similarity threshold.

[0051] Specifically, for the embedding dimension m, S35 calculates the number ratio of vectors similar to the template vector within a preset similarity threshold by the following formula (6): :

[0052] In formula (6), δ represents the preset similarity threshold, It represents the number of vectors that are similar to the template vector within the preset similarity threshold δ.

[0053] S36, repeating steps S33 to S35 to obtain the ratio of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector when the embedding dimension is m.

[0054] Specifically, through the above steps S33 to S35, a randomly selected vector is obtained. When used as a template vector, the ratio of the number of vectors that are similar to the template vector within a preset similarity threshold. When each other vector in the reconstructed phase space matrix Y is used as a template vector, the above S33 to S35 are calculated, that is, when the embedding dimension is m, when each vector is used as a template vector, the ratio of the number of vectors that are similar to the template vector within a preset similarity threshold is obtained.

[0055] S37, return to S32, calculate the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimension is m+1.

[0056] Specifically, when the embedding dimension is m+1, the reconstruction The phase space is obtained to obtain the reconstructed phase space matrix Y with the embedding dimension of m+1, and the calculation is continued through S33 to S36 to obtain the number ratio of vectors similar to the template vector within the preset similarity threshold when the embedding dimension is m+1.

[0057] S38, determining the sample entropy eigenvalue of the modal component of the signal collected by the target sensor according to the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1.

[0058] Specifically, the S38 includes steps S381 and S382: S381, calculate the average value of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector by using the following formula (7) when the embedding dimension is m: :

[0059] Specifically, the average value of the ratio of the number of vectors similar to the template vector within the preset similarity threshold when each vector is used as a template vector is calculated by formula (12) when the embedding dimension is m+1: :

[0060] S382, according to the number ratio of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1, the sample entropy eigenvalue of the modal component of the target sensor acquisition signal is determined by the following formula (8):

[0061] In formula (8), It indicates the ratio of the number of vectors that are similar to the template vector within the preset similarity threshold when the embedding dimension is m+1.

[0062] From formula (8), we can get the value of the embedding dimension m and the preset similarity threshold It will affect the final calculated sample entropy eigenvalue. Through experiments, it is shown that when m=2, When , the detection effect is the best, and SD represents the standard deviation of the sensor acquisition signal.

[0063] For each modal component of the signal collected by each sensor, the above content is used for calculation, and the sample entropy eigenvalue of the modal component of the signal collected by each sensor can be obtained.

[0064] S4, constructing ICSCM matrices of multiple sensor acquisition signals according to the sample entropy eigenvalues ​​of the modal components of each sensor acquisition signal.

[0065] In a specific embodiment, the S4 includes steps S41 and S42: S41, constructing a high-dimensional feature matrix of the signals collected by multiple sensors according to the sample entropy eigenvalues ​​of the modal components of the signals collected by each sensor.

[0066] Specifically, in order to comprehensively present the coal-rock cutting characteristics reflected by the coal-rock cutting state signals collected by multiple sensors, the embodiment of the present invention designs high-dimensional features according to the number of sensors.

[0067] The S41 includes: S411, selecting J modal components with the largest sample entropy eigenvalues ​​from all modal components of the signal collected by each sensor. The embodiment of the present invention does not limit the value of J, for example, the value of J is 10. S412, constructing the following high-dimensional feature matrix of the signals collected by multiple sensors according to the modal components selected by each sensor HDF :

[0068] In formula (9), Represents a row vector , the column vector is The high-dimensional feature matrix, HDF i represents the i-th column of the high-dimensional feature matrix, represents the number of sensors, SE ij Represents the sample entropy eigenvalue of the jth modal component of the ith sensor.

[0069] S42, constructing an ICSCM matrix of the signals collected by the multiple sensors according to the high-dimensional feature matrix of the signals collected by the multiple sensors.

[0070] The ICSCM matrix is ​​used to fuse the coal-rock cutting state signals collected by multiple sensors. Specifically, S42 includes: constructing the ICSCM matrix of the signals collected by multiple sensors according to the high-dimensional feature matrix of the signals collected by multiple sensors through the following formulas (10) and (11):

[0071]

[0072] in, u represents the average value of each column of the high-dimensional feature matrix (i.e., the sample entropy eigenvalue of each modal component), and ICSCM represents the ICSCM matrix.

[0073] HDF i -u is the deviation of the ith modal component from its mean vector, is the transpose of the bias vector. When computing the HDF i -u and its transpose When we multiply the product of , we are actually calculating the outer product of each modal component with its own deviation, which gives a J×J matrix, in which the elements represent the covariance between different modal components. Covariance can quantify the linear relationship between different sensor features, and can identify which sensor features are interdependent when changing and which are independent of each other. This helps to understand how different sensors respond to coal-rock cutting states together. Data fusion in this way can improve the accuracy and robustness of coal-rock cutting state detection. Different sensors may capture different states of coal-rock cutting, and fusing this information can provide a more comprehensive state view. Therefore, the elements in ICSCM illustrate the interaction between different sensor features, and each element represents the covariance between different sensor features, which reflects the interaction and interdependence between features, and provides a basis for subsequent coal-rock cutting feature classification detection.

[0074] S5, inputting the ICSCM matrix into a pre-trained coal-rock cutting state detection classifier, and determining the current coal-rock cutting state of the coal mining machine according to the output result of the coal-rock cutting state detection classifier.

[0075] It should be noted that before step S5, the coal-rock cutting state detection classifier needs to be trained. For the specific method of training the coal-rock cutting state detection classifier, refer to the existing neural network model training method, which is not elaborated in detail in the embodiment of the present invention. Regarding the specific network structure of the coal-rock cutting state detection classifier, the embodiment of the present invention does not specifically limit it. Preferably, the coal-rock cutting state detection classifier is an extreme learning machine (ELM).

[0076] Specifically, the output result of the coal-rock cutting state detection classifier can characterize whether the current cutting state of the coal mining machine is coal or rock.

[0077] In summary, the coal-rock cutting state detection method provided by the embodiment of the present invention has the following characteristics: 1. The coal-rock cutting state signal collected by each sensor is analyzed by improved cyclic spectrum (ICS) to demodulate the different cutting characteristics of coal and rock. ICS has better demodulation performance than cyclic spectrum and power spectrum, and is superior in obtaining sensitive features, so it can obtain features that are beneficial to the detection of coal-rock cutting state.

[0078] 2. By constructing the ICSCM matrix of signals collected by multiple sensors, the interaction and interdependence between the coal-rock cutting state signals collected by different sensors are fully utilized to provide a relatively accurate data basis for coal-rock structural state detection. The method of coal-rock cutting state identification based on the ICSCM matrix not only does not require parameter optimization, but also fully retains the interaction between different sensors, greatly reducing the computational cost. At the same time, under the same method, the use of multi-sensor fusion is more accurate than single sensor identification. The ICSCM matrix is ​​superior to CSCM and PSCM in retaining the interaction between different sensors and reliability performance, and is more suitable for data fusion analysis of multiple sensors in coal-rock cutting state detection.

[0079] 3. Using extreme value learning machine (ELM) as the coal-rock cutting state detection classifier has the advantages of strong noise resistance, few parameters, and strong adaptability, which can improve the accuracy of the detection results.

[0080] 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. A method for detecting coal-rock cutting status, characterized in that: include: S1, obtaining coal and rock cutting state signals collected by multiple sensors when the coal mining machine is currently cutting coal and rock; S2, analyzing the coal-rock cutting state signal collected by each sensor through an improved cyclic spectrum to obtain the modal component of the signal collected by each sensor; S3, calculating the sample entropy eigenvalue of the modal component of the signal collected by each sensor; S4, constructing an ICSCM matrix of the signals collected by multiple sensors according to the sample entropy eigenvalue of the modal component of each sensor collected signal; S5, inputting the ICSCM matrix into a pre-trained coal-rock cutting state detection classifier, and determining the current coal-rock cutting state of the coal mining machine according to the output result of the coal-rock cutting state detection classifier.

2. The coal-rock cutting state detection method according to claim 1, characterized in that: Determine any sensor as a target sensor, and when S2 analyzes the coal-rock cutting state signal collected by the target sensor through an improved cyclic spectrum to obtain a modal component of the signal collected by the target sensor, it includes: S21, the coal-rock cutting state signal collected by the target sensor It is divided into several components by the window function, each component includes L sampling points, and the spectrum of each component is calculated by formula (1): In formula (1), Represents discrete frequency The i The spectrum of the components, i =1,……, Q , Q Indicates the number of time shifts of the window function; represents the window function; n represents the index of the data point in the window; L Indicates the length of the window function, that is, the number of sampling points included in each component; R Indicates the step size of the window function; discrete frequency , k =0,……, L -1, , Indicates frequency resolution; F s Indicates the sampling frequency; S22, analyzing the spectrum of each component by using the improved cyclic spectrum formula (2) to obtain the modal components of the target sensor acquisition signal; In formulas (2) and (3), α represents the cycle frequency, f Indicates frequency; Represents the square norm of the window function; M express The peak moment, that is, the peak point of the window function, Satisfies the symmetry, that is .

3. The coal-rock cutting state detection method according to claim 2, characterized in that: The S3 includes: S31, determine any sensor as the target sensor, and calculate the average value of the effective peak value of the modal component of the target sensor acquisition signal by formula (4) : In formula (4), △g is Frequency resolution in direction; e represents the index, ranging from 2 to D, where D is the total number of valid peaks; f =1, 2, ⋯, O, where O is the frequency index range obtained by discretization of △g; S32, Reconstruction The phase space of the reconstructed phase space matrix Y is defined as: In formula (5), m is the embedding dimension in the phase space, indicating the number of sampling points contained in each vector; G is the total number of modal components; K Represents the total number of valid vectors of the reconstructed phase space matrix Y, that is, the reconstructed phase space matrix Y number of rows; and They represent the first and valid vectors; S33, randomly select a vector from the reconstructed phase space matrix Y as a template vector; S34, determining vectors similar to the template vector within a preset similarity threshold, and determining the number of vectors similar to the template vector; S35, for the embedding dimension m, calculating the ratio of the number of vectors that are similar to the template vector within a preset similarity threshold; S36, repeating steps S33 to S35 to obtain the ratio of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector when the embedding dimension is m; S37, return to S32, calculate the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimension is m+1; S38, determining the sample entropy eigenvalue of the modal component of the signal collected by the target sensor according to the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1.

4. The coal-rock cutting state detection method according to claim 3, characterized in that: The S34 includes: S341, calculating the maximum difference between any vector and the corresponding element in the template vector; S342, determining that any vector whose maximum difference with the corresponding element in the template vector is less than a preset similarity threshold is a vector similar to the template vector, and the number of the vectors is used as the number of vectors similar to the template vector.

5. The coal-rock cutting state detection method according to claim 4, characterized in that: The S35 includes: For the embedding dimension m, the ratio of the number of vectors that are similar to the template vector within the preset similarity threshold is calculated using formula (6): : In formula (6), δ represents the preset similarity threshold, It represents the number of vectors that are similar to the template vector within the preset similarity threshold δ.

6. The coal-rock cutting state detection method according to claim 5, characterized in that: The S38 includes: S381, using formula (7) to calculate the average value of the ratio of the number of vectors similar to the template vector within a preset similarity threshold when each vector is used as a template vector when the embedding dimension is m ; The same is true when the embedding dimension is m+1: S382, according to the ratio of the number of vectors similar to the template vector within a preset similarity threshold when the embedding dimensions are m and m+1, the sample entropy eigenvalue of the modal component of the target sensor acquisition signal is determined by formula (8): In formula (8), It indicates the ratio of the number of vectors that are similar to the template vector within the preset similarity threshold when the embedding dimension is m+1.

7. The coal-rock cutting state detection method according to claim 1, characterized in that: The S4 includes: S41, constructing a high-dimensional feature matrix of the signals collected by multiple sensors according to the sample entropy eigenvalues ​​of the modal components of the signals collected by each sensor; S42, constructing an ICSCM matrix of the signals collected by the multiple sensors according to the high-dimensional feature matrix of the signals collected by the multiple sensors.

8. The coal-rock cutting state detection method according to claim 7, characterized in that: The S41 includes: S411, selecting J modal components with the largest sample entropy eigenvalues ​​from all modal components of the signal collected by each sensor; S412, constructing a high-dimensional feature matrix of signals collected by multiple sensors based on the modal components selected by each sensor HDF : In formula (9), HDF represents a high-dimensional feature matrix with row vector q and column vector J. HDF i represents the i-th column of the high-dimensional feature matrix, q represents the number of sensors, SE ij Represents the sample entropy eigenvalue of the jth modal component of the ith sensor.

9. The coal-rock cutting state detection method according to claim 8, characterized in that: The S42 includes: constructing an ICSCM matrix of the signals collected by the multiple sensors according to the high-dimensional feature matrix of the signals collected by the multiple sensors by using formula (10) and formula (11): in, u represents the average value of each column of the high-dimensional feature matrix, and ICSCM represents the ICSCM matrix.

10. The coal-rock cutting state detection method according to claim 1, characterized in that: The coal-rock cutting state detection classifier is an extreme learning machine.

Images