A noise reduction algorithm for loaded coal and rock radiation signals based on ICEEMDAN-K-Means

Abstract

The present invention discloses a noise reduction algorithm for loaded coal rock radiation signals based on ICEEMDAN-K-Means, which belongs to the technical field of signal noise reduction. The method first discretizes the original radiation signal of the loaded coal rock to construct a discrete function, and then constructs E(·) and M(·) operators respectively to achieve the purpose of obtaining the EMD modal function and local mean of the input data at a specified number of times. Based on this, the discrete function of the radiation signal is decomposed by ICEEMDAN to obtain the modal function matrix and decomposition residue; the modal function matrix is ​​cut by row to form multiple vectors, and the K-Means algorithm is used to achieve clustering and obtain the merged vector DI i ; Sort each DI i Pearson correlation coefficient between the vector and the original signal, reducing irrelevant components; superimposing the filtered DI i The vector and decomposition residuals are combined to reconstruct the coal rock deformation-related signals, ultimately achieving the goal of denoising the original input radiation signal. The proposed denoising algorithm for loaded coal rock radiation signals can effectively suppress the interference of noise signals on characteristic information and improve the signal-to-noise ratio of the reconstructed signal, facilitating further research on the coupling relationship between radiation changes and coal rock deformation.

Description

Technical Field

[0001] The present invention relates to the technical field of signal noise reduction, and in particular to an ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal rock radiation signals. Background Art

[0002] Coal mining continues to advance below 1,000 meters below sea level, making coal-rock dynamic disasters more likely to occur. This has led to a growing interest in uncovering the mechanisms of deep coal-rock dynamic disasters and understanding their prevention and control methods. Changes in the state of loaded coal and rock can cause variations in electromagnetic and infrared radiation signals. Real-time monitoring and analysis of these radiation signals will help further elucidate the nature of coal-rock dynamic disasters and facilitate in-depth research on multi-information-based coal-rock dynamic disaster prevention and control strategies, reducing mining risks and protecting public life and property.

[0003] Mining conditions in deep coal and rock formations are more complex than those in shallower areas, and the underground environment impacts the stability of long-term equipment. These factors make directly collected radiation signals from loaded coal and rock more susceptible to signal interference, resulting in the inclusion of a large amount of uncorrelated noise in the environmentally-derived data. This presents challenges in accurately understanding the mechanisms of coal and rock dynamic hazards. Currently, several methods exist for denoising radiation signals, including data denoising based on wavelet or Fourier decomposition, decomposition and reconstruction methods based on traditional EMD and VMD algorithms, and data processing methods based on artificial neural networks and other intelligent algorithms. While these methods can achieve data denoising to some extent, they each suffer from various challenges, including poor nonlinear denoising effectiveness, slow decomposition speed, high algorithmic complexity, endpoint effects, and spectral aliasing. Furthermore, these methods are not designed specifically for the radiation signals of loaded coal and rock, making them inadequately adapted to practical needs. Therefore, it is crucial to develop an algorithm specifically tailored to the denoising of loaded coal and rock radiation signals that can perform this type of signal denoising in a simple and efficient manner. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a noise reduction algorithm for loaded coal rock radiation signals based on ICEEMDAN-K-Means.

[0005] The technical solution adopted by the present invention is a noise reduction algorithm for loaded coal rock radiation signals based on ICEEMDAN-K-Means. The overall process includes the following 6 steps.

[0006] Step 1: Construct a discrete function D[t] of the original radiation signal of the loaded coal rock, which is used to represent the time series of the original electromagnetic or infrared radiation characteristics obtained by direct sampling of radiation signal acquisition devices such as electromagnetic radiation meters and infrared radiation meters during the coal rock loading process.

[0007] Furthermore, the data of the discrete function D[t] of the original radiation signal in the above step 1 is composed of the coal rock deformation related signal D1[t], the natural environment interference signal D2[t], the experimental device error signal D3[t] and other irrelevant signals D4[t]. The specific structural relationship is shown in formula (1).

[0008] D[t]=D1[t]+D2[t]+D3[t]+D4[t] (1)

[0009] Step 2: Construct the operators E(·) and M(·) to respectively solve the EMD modal function and local mean of the input data “·” for the specified number of times.

[0010] Furthermore, in step 2 above, the operator E(·) is used to solve the kth modal function IMF of the input data “·” k The specific steps are as follows.

[0011] Step 2.1: Input data “·”, take k=0, and let the residual r0=·.

[0012] Step 2.2: Find each residual r k The upper and lower envelopes e MIN and e MAX , and calculate the average value m of the upper and lower envelopes.

[0013] Step 2.3: Calculate the (k+1)th mode function IMF using the following formula: (k+1) Candidate value d (k+1) =r k –m;

[0014] Step 2.4: Determine r k Whether the envelope average value m tends to 0, if the envelope average value m tends to 0, then d (k+1) is the (k+1)th modal function IMF of the input data “·” (k+1) , at this time calculate the (k+1)th residual r (k+1) , which is related to the (k+1)th mode function IMF (k+1) The relationship between them satisfies formula (2).

[0015] IMF (k+1) =·-r (k+1) (2)

[0016] On the contrary, if the envelope mean value m does not tend to 0, then d (k+1) Not the IMF (k+1) , now let d (k+1) =r k Repeat steps 2.2 to 2.4 until a candidate value d is found that satisfies the average envelope value m close to 0. (k+1)and the (k+1)th modal function IMF of the input data “·” (k+1) .

[0017] Step 2.5: When the obtained mode function IMF reaches the operator E k (·) Expect the required number of modes and stop the loop.

[0018] Furthermore, in the above step 2, the method of using the operator M(·) to calculate the local mean of the input data “·” is to subtract the first mode function IMF1 obtained by decomposing the EMD decomposition of the input data “·”, and the relationship between M(·) and E1(·) can be expressed by formula (3).

[0019] M(·)=·-E1(·) (3)

[0020] Step 3: Perform ICEEMDAN decomposition on the original radiation signal discrete function D[t] of the loaded coal rock to obtain the modal function matrix IMF composed of multiple modal functions and the decomposition residual Res. The specific decomposition steps are as follows.

[0021] Step 3.1: Take L zero-mean unit-variance white noise signals ω (i) (i=1,2,…,L), input the data of the original radiation signal discrete function D[t], and construct L original data D containing white noise (i) , the specific construction method is shown in formula (4).

[0022] D (i) =D+β0E1(ω (i) ) (4)

[0023] Where β0 is the setting coefficient, and its value can be determined by formula (5).

[0024]

[0025] Where ε0 is an artificially given coefficient, and its value must take into account the signal-to-noise ratio (SNR) requirement; the std(·) operator is used to calculate the standard deviation of the internal value “·”. The specific calculation method is shown in (6).

[0026]

[0027] Where n is the dimension of the input data “·”, is the mean of the input data “·”.

[0028] Step 3.2: D (i) Perform EMD decomposition to obtain the first residual r1, which is calculated as shown in formula (7).

[0029]

[0030] Where the <·> operator represents taking the average value of “·”.

[0031] Step 3.3: Remove the first residual from the original radiation signal discrete function D[t] to obtain the first modal function IMF1. The calculation method is shown in formula (8).

[0032] IMF1=D-r1 (8)

[0033] Step 3.4: Reconstruct the input data D using the first residual r1 (i) ', its data structure is shown in formula (9).

[0034] D (i) '=r1+β1E2(ω (i) ) (9)

[0035] Step 3.5: D (i) 'Use the M(·) operator to perform EMD decomposition to obtain the second residual r2, and thereby obtain the second modal function IMF2 of the original radiation signal discrete function D[t]. The specific calculation method is shown in formula (10).

[0036] IMF2=r1-r2=r1- <M(r1+β1E2(ω (i) ))> (10)

[0037] Step 3.6: Use the (k-1)th residual r in turn (k-1) To construct the data of the k-th input M(·) operator, and thereby obtain the k-th residual r k (k=3,…,k), and its specific calculation method is shown in formula (11).

[0038] r k = <M(r (k-1) +β (k-1) E k (ω (i) ))> (11)

[0039] Where, β (k-1) (k=2,3,…,k,k+1) and β0 mentioned above are both setting coefficients, and their calculation method is shown in formula (12).

[0040] β (k-1) =ε0std(r (k-1) ) (12)

[0041] Step 3.7: Use the (k-1)th residual r (k-1) and the kth residual r k To obtain the kth modal function IMF of the original radiation signal discrete function D[t]k , the specific method is shown in formula (13).

[0042] IMF k =r (k-1) -r k (13)

[0043] Step 3.8: Repeat steps 3.6 and 3.7. Step 3.6 obtains the (k+1)th residual r (k+1) Hou Ruor (k+1) The loop is terminated when the monotonic condition is satisfied within the domain t of the original radiation signal discrete function D[t]. At this time, the modal function matrix IMF and the decomposition residual Res of the original radiation signal discrete function D[t] can be obtained.

[0044] Furthermore, the IMF matrix structure in the above step 3.8 is shown in formula (14).

[0045]

[0046] Furthermore, the decomposition residue Res in step 3.8 above is the residue obtained from the last decomposition, that is, Res = r (k+1) .

[0047] Step 4: Based on the modal function matrix IMF data of the original radiation signal discrete function D[t], the K-Means algorithm is used to perform data clustering operations to filter out the natural environment interference signal D2[t], the experimental device error signal D3[t] and other irrelevant signals D4[t] from the original radiation signal discrete function D[t] to the greatest extent possible. The specific steps are as follows.

[0048] Step 4.1: Divide the modal function matrix IMF of the original radiation signal discrete function D[t] into k n-dimensional vectors by row, which are used to represent the modal functions obtained by separating D[t], and are labeled IMF1, IMF2, ..., IMF k , the i-th IMF i The structure of the vector is [imf i1 ,imf i2 ,…,imf in ].

[0049] Step 4.2: Generate N signal cluster center vectors according to the required accuracy, and their labels are A1, A2, ..., A N , the dimension of each cluster center vector is the same as that of the IMF vector, which is also n-dimensional. j The vector structure is [a j1 ,a j2 ,…,a jn ].

[0050] Step 4.3: Calculate IMF separately i (i=1,2,..,k) and N signal cluster centers A j The Euclidean distance L of (j=1,2,..,N) ij , and its calculation method is shown in formula (15).

[0051]

[0052] Step 4.4: Compare L i1 ~L iN The value of the Euclidean distance specifies the distance between each IMF. i Vector nearest signal cluster center vector A j , so that all IMF vectors are divided into N categories.

[0053] Step 4.5: Superimpose the w IMF vectors within each group according to the classification to obtain the discrete function D[t] component DI of the original radiation signal decomposed according to N classifications i (i=1,2,…,N), and its form is shown in formula (16).

[0054]

[0055] Step 5: Calculate N DIs for N categories i The Pearson correlation coefficient between the vector and the original radiation signal discrete function D, and the N DI i Sort from high to low.

[0056] Furthermore, the calculation method of the Pearson correlation coefficient described in step 5 is shown in formula (17).

[0057]

[0058] Step 6: Remove the 1-2 DIs with the lowest correlation coefficients based on the Pearson coefficient ranking and the actual situation. i vector, treat it as a noise component and use other DI i The decomposition residue Res of the vector and the original radiation signal discrete function D[t] is reconstructed to obtain the coal rock deformation related signal D1, thereby achieving the purpose of denoising D[t].

[0059] Furthermore, the signal composition structure of the coal rock deformation related signal D1 is shown in formula (18).

[0060]

[0061] Where v is the number of groups remaining after removing the DI vectors with lower correlation.

[0062] The beneficial effects of the above technical solution are as follows: the present invention provides a noise reduction algorithm for loaded coal rock radiation signal based on ICEEMDAN-K-Means, constructs the E(·) and M(·) specific operation operators based on EMD decomposition, and performs ICEEMDAN decomposition on the discrete function of the original radiation signal to achieve feature separation of the original signal, obtain the modal function matrix IMF and the decomposition residue Res; vectorize the IMF matrix and use the K-Means algorithm to cluster and group it according to the required accuracy, and superimpose the calculation to obtain the DI of the discrete function of the original radiation signal. i Components; using Pearson correlation coefficient for different DI i The contribution of component sorting is used to reconstruct the coal rock deformation related signal to achieve noise reduction of the original radiation signal and improve the signal-to-noise ratio of the collected radiation signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flowchart of the overall steps of the ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal and rock radiation signals in the present invention.

[0064] Figure 2 Solve the input data modal function IMF for the operator E(·) of the present invention k Step flow chart.

[0065] Figure 3 This is a flow chart of the decomposition steps of the discrete function ICEEMDAN of the original infrared radiation signal of the loaded coal rock in the present invention.

[0066] Figure 4 This is a diagram showing the ICEEMDAN decomposition result of the original infrared radiation signal according to an embodiment of the present invention.

[0067] Figure 5 This is a flow chart of the steps of clustering the IMF vectors of the original infrared radiation signal using the K-Means algorithm in an embodiment of the present invention.

[0068] Figure 6 This is a comparison chart of the noise reduction results of the original infrared radiation signal according to an embodiment of the present invention.

[0069] Among them, (a) is the original signal; (b) is the signal after noise reduction. DETAILED DESCRIPTION

[0070] The following embodiments of the present invention are described in further detail with reference to 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.

[0071] This example takes the raw data containing noise obtained by directly mining the infrared radiation temperature signal under uniaxial loading of coal rock as an example. Figure 1The six overall steps S1 to S6 shown in the figure sequentially complete the construction of the data discrete function, signal decomposition, feature sorting, and signal reconstruction, thereby achieving noise reduction processing of the original infrared radiation temperature signal of the loaded coal rock. The specific description is as follows:

[0072] Step 1: During the uniaxial loading of coal rock, infrared radiation temperature signal data of coal rock during loading is directly collected using an infrared radiometer. Based on this data, a discrete function D[t] of the original infrared radiation signal of the loaded coal rock is constructed to represent the time series of the original infrared radiation characteristics.

[0073] Furthermore, the data of the original infrared radiation signal discrete function D[t] in the above step 1 is composed of the infrared radiation change signal D1[t] caused by coal rock deformation, the ambient temperature interference signal D2[t], the infrared radiation instrument detection error D3[t] and other irrelevant signals D4[t]. The relationship between them can be expressed by formula (1).

[0074] D[t]=D1[t]+D2[t]+D3[t]+D4[t] (1)

[0075] Step 2: Construct the operators E(·) and M(·) to respectively perform EMD decomposition of the input infrared radiation data or the processed data “·” for a specified number of times to obtain the corresponding modal function or local mean.

[0076] Furthermore, the above-mentioned operation operator E(·) is used to solve the kth mode function IMF of the input infrared radiation data or the processed data “·” k There are 5 steps in total, and the steps are as follows Figure 2 As shown, the details are as follows:

[0077] Step 2.1: Input infrared radiation data or processed data “·”, set k = 0, and make the residual r0 = ·.

[0078] Step 2.2: Find each residual r k The upper and lower envelopes e MIN and e MAX , and calculate the average value of the upper and lower envelopes m=0.5×(e MIN +e MAX ).

[0079] Step 2.3: Calculate the (k+1)th mode function IMF (k+1) Candidate value d (k+1) =r k –m;

[0080] Step 2.4: Determine r k Whether the envelope average value m tends to 0, if the envelope average value m tends to 0, then d (k+1)is the (k+1)th mode function IMF of the input infrared radiation data or the processed data “·” (k+1) , at this time calculate the (k+1)th residual r (k+1) , which is related to the (k+1)th mode function IMF (k+1) The relationship between them satisfies formula (2).

[0081] IMF (k+1) =·-r (k+1) (2)

[0082] On the contrary, if the envelope mean value m does not tend to 0, then d (k+1) Not the IMF (k+1) , now let d (k+1) =r k Repeat steps 2.2 to 2.4 until a candidate value d is found that satisfies the average envelope value m close to 0. (k+1) The (k+1)th mode function IMF of the input infrared radiation data or the processed data “·” (k+1) .

[0083] Step 2.5: When the obtained mode function IMF reaches the operator E k (·) Expect the required number of modes and stop the loop.

[0084] Furthermore, the above-mentioned method of using the operation operator M(·) to obtain the local mean of the input infrared radiation data or the processed data “·” is to subtract the first mode function IMF1 obtained by EMD decomposition from the input infrared radiation data or the processed data “·”, and the relationship between M(·) and E1(·) can be expressed by formula (3).

[0085] M(·)=·-E1(·) (3)

[0086] Step 3: Perform ICEEMDAN decomposition on the original infrared radiation signal discrete function D[t] of the loaded coal rock to obtain the modal function matrix IMF and decomposition residual Res composed of multiple modal functions after the original infrared radiation signal discrete function D[t] is decomposed. The decomposition process is as follows: Figure 3 The specific steps are as follows:

[0087] Step 3.1: Take L zero-mean unit-variance white noise signals ω (i) (i=1,2,…,L), input the data of the original infrared radiation signal discrete function D[t], and construct L original infrared radiation data D containing white noise (i) , the specific construction method is shown in formula (4).

[0088] D (i) =D+β0E1(ω (i) ) (4)

[0089] Where β0 is the setting coefficient, and its value can be determined by formula (5).

[0090]

[0091] Where ε0 is an artificially given coefficient, and its value must take into account the signal-to-noise ratio (SNR) requirement; the std(·) operator is used to calculate the standard deviation of the internal value “·”. The specific calculation method is shown in (6).

[0092]

[0093] Where n is the dimension of the input data “·”, is the mean of the input data “·”.

[0094] Step 3.2: D (i) Perform EMD decomposition to obtain the first residual r1, which is calculated as shown in formula (7).

[0095]

[0096] Where the <·> operator represents taking the average value of “·”.

[0097] Step 3.3: Remove the first residual from the original infrared radiation signal discrete function D[t] to obtain the first modal function IMF1. The calculation method is shown in formula (8).

[0098] IMF1=D-r1 (8)

[0099] Step 3.4: Reconstruct the input data D using the first residual r1 (i) ', its data structure is shown in formula (9).

[0100] D (i) '=r1+β1E2(ω (i) ) (9)

[0101] Step 3.5: D (i) 'Use the M(·) operator to perform EMD decomposition to obtain the second residual r2, and thereby obtain the second modal function IMF2 of the original infrared radiation signal discrete function D[t]. The specific calculation method is shown in formula (10).

[0102] IMF2=r1-r2=r1- <M(r1+β1E2(ω (i) ))> (10)

[0103] Step 3.6: Use the (k-1)th residual r in turn (k-1) To construct the data of the k-th input M(·) operator, and thereby obtain the k-th residual r k(k=3,…,k), and its specific calculation method is shown in formula (11).

[0104] r k = <M(r (k-1) +β (k-1) E k (ω (i) ))> (11)

[0105] Where, β (k-1) (k=2,3,…,k,k+1) and β0 mentioned above are both setting coefficients, and their calculation method is shown in formula (12).

[0106] β (k-1) =ε0std(r (k-1) ) (12)

[0107] Step 3.7: Use the (k-1)th residual r (k-1) and the kth residual r k To obtain the kth modal function IMF of the original infrared radiation signal discrete function D[t] k , the specific method is shown in formula (13).

[0108] IMF k =r (k-1) -r k (13)

[0109] Step 3.8: Repeat steps 3.6 and 3.7. Step 3.6 obtains the (k+1)th residual r (k+1) Hou Ruor (k+1) The cycle is terminated when the monotonic condition is satisfied within the domain t of the original radiation signal discrete function D[t]. The decomposition result of the original infrared radiation signal is as follows: Figure 4 As shown in the figure, it can be seen that after the original infrared radiation signal is decomposed, a total of 5 modal functions IMF1, IMF2, ..., IMF5 and 1 decomposition residual are generated. At this time, the modal function matrix IMF and decomposition residual Res of the original radiation signal discrete function D[t] can be obtained.

[0110] Furthermore, the IMF matrix structure in the above step 3.8 is shown in formula (14).

[0111]

[0112] Furthermore, the decomposition residue Res in the above step 3.8 is the residue obtained by the last decomposition, that is, Res=r6.

[0113] Step 4: Based on the modal function matrix IMF data of the original infrared radiation signal discrete function D[t], the K-Means algorithm is used to perform data clustering operations to filter out the natural environment interference signal D2[t], the experimental device error signal D3[t] and other irrelevant signals D4[t] from the original radiation signal discrete function D[t] to the greatest extent. The execution flow chart is as follows: Figure 5 As shown, the specific steps are as follows:

[0114] Step 4.1: Divide the modal function matrix IMF of the original radiation signal discrete function D[t] into five 214-dimensional vectors by row, which are used to represent the modal functions obtained by separating D[t]. They are labeled IMF1, IMF2, ..., IMF5. The i-th IMF i The structure of the vector is [imf i1 ,imf i2 ,…,imf i214 ].

[0115] Step 4.2: Generate three signal cluster center vectors, which are labeled A1, A2, and A3. The dimension of each cluster center vector is the same as the dimension of the IMF vector, which is 214. The jth A j The vector structure is [a j1 ,a j2 ,…,a j214 ](j=1,2,3).

[0116] Step 4.3: Calculate IMF separately i (i=1,2,..,5) and 3 signal cluster centers A j The Euclidean distance L of (j=1,2,3) ij , and its calculation method is shown in formula (15).

[0117]

[0118] Step 4.4: Compare L i1 ~L i3 The value of the Euclidean distance specifies the distance between each IMF. i Vector nearest signal cluster center vector A j , all IMF vectors are divided into 3 categories, and the clustering results are as follows:

[0119] category number <![CDATA[IMF i ]]> <![CDATA[Class A1]]> 2 IMF1, IMF2 <![CDATA[Class A2]]> 2 IMF3、IMF4 <![CDATA[Class A3]]> 1 IMF5

[0120] Step 4.5: Superimpose the w IMF vectors within each group according to the classification to obtain the discrete function D[t] component DI of the original infrared radiation signal decomposed according to the three classifications i (i=1,2,3), and its form is shown in formula (16).

[0121]

[0122] Step 5: Calculate the 3 DIs for the 3 categories i The Pearson correlation coefficients between the vector and the original infrared radiation signal discrete function D are calculated to be 0.1505, 0.2756, and 0.7485, respectively. i Sort from high to low, the order is DI3, DI2, DI1.

[0123] Furthermore, the calculation method of the Pearson correlation coefficient described in step 5 is shown in formula (17).

[0124]

[0125] Step 6: According to the Pearson coefficient sorting and the actual situation, remove the DI1 vector with the lowest correlation coefficient and regard it as a noise component, and use other DI i The decomposition residue Res of the vector and the original radiation signal discrete function D[t] is reconstructed to obtain the coal rock deformation related signal D1, which is used to reduce the noise of D[t]. Figure 6 As shown, it can be seen that the method has a good effect on noise reduction of infrared radiation signals. Similarly, electromagnetic radiation signals have the same characteristics as infrared radiation signals and are also applicable to this method, which will not be described here.

[0126] Furthermore, the signal composition structure of the coal rock deformation related signal D1 is shown in formula (18).

[0127]

[0128] Although the present invention has been disclosed above in terms of preferred embodiments, they are not intended to limit the present invention. Anyone skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined by the claims of this application.

Claims

1. A noise reduction algorithm for loaded coal rock radiation signal based on ICEEMDAN-K-Means, characterized by: The ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal and rock radiation signals includes the following six steps: Step 1: Construct a discrete function D[t] of the original radiation signal of the loaded coal rock, which is used to represent the time series of the original electromagnetic or infrared radiation characteristics directly collected by the electromagnetic radiation meter or infrared radiation meter radiation signal acquisition device during the coal rock loading process; The data of the original radiation signal discrete function D[t] in step 1 above is composed of the coal rock deformation related signal D1[t], the natural environment interference signal D2[t], the experimental device error signal D3[t] and the irrelevant signal D4[t]. Its specific structural relationship is shown in formula (1): D[t]=D1[t]+D2[t]+D3[t]+D4[t] (1) Step 2: Construct the operators E(·) and M(·) to respectively solve the EMD mode function and local mean of the input data "·" for the specified number of times; Step 3: Perform ICEEMDAN decomposition on the original radiation signal discrete function D[t] of the loaded coal rock to obtain the modal function matrix IMF composed of multiple modal functions after the decomposition of the original radiation signal discrete function D[t] and the decomposition residual Res; Step 4: Based on the modal function matrix IMF data of the original radiation signal discrete function D[t], the K-Means algorithm is used to perform data clustering operations to filter out the natural environment interference signal D2[t], the experimental device error signal D3[t] and the irrelevant signal D4[t] from the original radiation signal discrete function D[t]; Step 5: Calculate N DIs for N categories i The Pearson correlation coefficient between the vector and the original radiation signal discrete function D[t], and the N DI i Sort from high to low; The calculation method of the Pearson correlation coefficient in step 5 above is shown in formula (2): Step 6: Remove the 1-2 DIs with the lowest correlation coefficients based on the Pearson coefficient ranking and the actual situation. i vector, treat it as a noise component and use other DI i The decomposition residue Res of the vector and the original radiation signal discrete function D[t] is reconstructed to obtain the coal rock deformation related signal D1, achieving the purpose of noise reduction for D[t]; The signal composition structure of the coal rock deformation related signal D1 in step 6 is shown in formula (3): Where v is the number of groups remaining after removing the 1-2 DI vectors with the lowest correlation coefficients.

2. The ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal and rock radiation signals according to claim 1, characterized in that: The solution method for the operators E(·) and M(·) in step 2 is as follows: In step 2 above, the operator E(·) is used to solve the kth mode function IMF of the input data "·" k The steps are as follows: Step 2.1: Input data "·", take k = 0, and set the residual r0 = ·; Step 2.2: Find each residual r k The upper and lower envelopes e MIN and e MAX , and calculate the average value m of the upper and lower envelopes; Step 2.3: Calculate the (k+1)th mode function IMF (k+1) Candidate value d (k+1) , the calculation formula is d (k+1) =r k –m; Step 2.4: Determine r k Whether the envelope average value m tends to 0, if the envelope average value m tends to 0, then d (k+1) is the (k+1)th modal function IMF of the input data "·" (k+1) , then calculate the (k+1)th modal function IMF (k+1) The margin r (k+1) , which is related to the (k+1)th mode function IMF (k+1) The relationship between them satisfies formula (4): IMF (k+1) =·-r (k+1) (4) On the contrary, if the envelope mean value m does not tend to 0, then d (k+1) Not the IMF (k+1) , now let d (k+1) =r k Repeat steps 2.2 to 2.4 until a candidate value d is found that satisfies the average envelope value m close to 0. (k+1) The (k+1)th mode function IMF of the input data "·" (k+1) ; Step 2.5: When the obtained mode function IMF reaches the operator E k (·) Expect the required number of modes and stop the loop; In step 2 above, the method of using the operator M(·) to calculate the local mean of the input data "·" is to subtract the first mode function IMF1 obtained by EMD decomposition from the input data "·". The relationship between IMF1 and E1(·) can be expressed by formula (5): M(·)=·-E1(·) (5).

3. The ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal rock radiation signals according to claim 1, characterized in that: The process of step 3 is as follows: Step 3.1: Take L zero-mean unit-variance white noise signals ω (i) , where i = 1, 2, ..., L, the data of the input original radiation signal discrete function D[t] is constructed into L original data D containing white noise (i) , the specific construction method is shown in formula (6): D (i) =D+β0E1(ω (i) ) (6) Where β0 is the setting coefficient, and its value can be determined by formula (7): Where ε0 is a manually given coefficient whose value should be determined by considering the signal-to-noise ratio (SNR) requirement. The std(·) operator is used to calculate the standard deviation of the internal value "·". The specific calculation method is shown in (8): Where n is the dimension of the input data "·", is the mean of the input data "·"; Step 3.2: D (i) Perform EMD decomposition to obtain the first residual r1, which is calculated as shown in formula (9): In the formula, the <·> operator represents taking the average value of "·"; Step 3.3: Remove the first residual from the original radiation signal discrete function D[t] to obtain the first modal function IMF1. The calculation method is shown in formula (10): IMF1=D-r1 (10) Step 3.4: Reconstruct the input data D using the first residual r1 (i) ', its data structure is shown in formula (11): D (i) '=r1+β1E2(ω (i) ) (11) Step 3.5: D (i) 'Use the M(·) operator to perform EMD decomposition to obtain the second residual r2, and thereby obtain the second modal function IMF2 of the original radiation signal discrete function D[t]. The specific calculation method is shown in formula (12): IMF2=r1-r2=r1-<M(r1+β1E2(ω (i) ))> (12) Step 3.6: Use the (k-1)th residual r in turn (k-1) To construct the data of the k-th input M(·) operator, k starts from 3 and thus obtains the k-th residual r k , and its specific calculation method is shown in formula (13): r k = <M(r (k-1) +b (k-1) E k (oh (i) ))> (13) Where, β (k-1) Same as β0 in the previous text, both are setting coefficients, and their calculation method is shown in formula (14): b (k-1) =ε0std(r (k-1) ) (14) Step 3.7: Use the (k-1)th residual r (k-1) and the kth residual r k To obtain the kth modal function IMF of the original radiation signal discrete function D[t] k , the specific method is shown in formula (15): IMF k =r (k-1) -r k (15) Step 3.8: Repeat steps 3.6 and 3.

7. Step 3.6 obtains the (k+1)th residual r (k+1) Hou Ruor (k+1) If the monotonicity condition is satisfied within the domain t of the original radiation signal discrete function D[t], then the steps 3.6 and 3.7 are repeated. At this time, the modal function matrix IMF and the decomposition residual Res of the original radiation signal discrete function D[t] can be obtained. Furthermore, the IMF matrix structure in step 3.8 above is as shown in formula (16): Furthermore, the decomposition residue Res in step 3.8 above is the residue obtained from the last decomposition, that is, Res = r (k+1) .

4. The ICEEMDAN-K-Means-based noise reduction algorithm for loaded coal and rock radiation signals according to claim 1, characterized in that: The process of step 4 is as follows: Step 4.1: Divide the modal function matrix IMF of the original radiation signal discrete function D[t] into k n-dimensional vectors by row, which are used to represent the modal functions obtained by separating D[t], and are labeled IMF1, IMF2, ..., IMF k , the i-th IMF i The structure of the vector is [imf i1 ,imf i2 ,…,imf in ]; Step 4.2: Generate N signal cluster center vectors according to the required accuracy, and their labels are A1, A2, ..., A N , the dimension of each cluster center vector is the same as that of the IMF vector, which is also n-dimensional. j The vector structure is [a j1 ,a j2 ,…,a jn ]; Step 4.3: Calculate IMF separately i With N signal cluster centers A j The Euclidean distance L ij , where i = 1, 2, ..., k, j = 1, 2, ..., N, L ij The calculation method is shown in formula (17): Step 4.4: Compare L i1 ~L iN The value of the Euclidean distance specifies the distance between each IMF. i Vector nearest signal cluster center vector A j , so that all IMF vectors are divided into N categories; Step 4.5: Superimpose the w IMF vectors within each group according to the classification to obtain the discrete function D[t] component DI of the original radiation signal decomposed according to N classifications i , where i = 1, 2, ..., N, and its form is shown in formula (18):

Images