Lithology Identification Method Based on Hybrid Mode Decomposition of Drilling Signals and Fuzzy Entropy-PSO Optimization

Abstract

This invention discloses a lithology identification method based on hybrid mode decomposition and fuzzy entropy-PSO optimization of drilling signals, belonging to the field of integrated deep earth space exploration and drilling geophysical exploration technology. The method first acquires real-time seismic monitoring signals during drilling; then, it uses an empirical and variational hybrid mode decomposition (EVMD) method to decompose the original signal, removing strong noise disturbances; finally, it employs a particle swarm optimization (PSO) algorithm combined with adaptive optimization using fuzzy entropy as the objective function to obtain the optimal parameter combination [K, α] of EVMD, and decomposes it to obtain the optimal modal data and construct a multi-dimensional feature vector, completing the real-time identification and judgment of strata lithology and rock mass structural characteristics. This invention effectively solves the technical problems of traditional methods, such as difficulty in separating strong noise, fixed parameters that cannot be adaptively adapted, mode aliasing, and poor real-time performance. It achieves extremely high lithology identification accuracy and is suitable for deep earth space development, deep resource exploration, and intelligent construction of underground engineering projects.

Description

Technical Field

[0001] This invention relates to the field of integrated deep-earth space exploration and drilling geophysical exploration technology, specifically to a drilling-while-measuring intelligent lithology analysis method based on empirical and variational hybrid mode decomposition (EVMD) and particle swarm optimization (PSO). Background Technology

[0002] With the increasing demand for technological innovation in deep earth or underground space development and utilization projects such as deep earth scientific research and development, underground energy storage facilities, deep mineral resource exploration, tunnels and underground engineering, how to accurately detect and perceive the geological conditions of deep earth space in real time has become a key focus of the industry.

[0003] Integrated drilling and geophysical exploration technology, as a key means of fine exploration in deep earth space, acquires real-time information on formation physical parameters, lithological characteristics, and geological structure during drilling. It is a core technological path to achieve transparent and intelligent development of underground space. This technology can effectively overcome the limitations of traditional geophysical methods, such as low resolution, limited exploration depth, and susceptibility to surface conditions, while also compensating for the shortcomings of single drilling methods that only provide a single borehole view.

[0004] However, the development of drilling lithology identification technology is still in its early stages, and its implementation in engineering projects faces many technical bottlenecks: First, during the drilling process, the drilling rig's own vibration, the interaction between the drill bit and the rock, and the mud circulation will generate strong noise interference, which seriously contaminates the effective signal of the seismic monitoring while drilling. Traditional fixed parameter filtering methods are difficult to achieve accurate separation of the effective signal and noise components, which can easily lead to the loss of lithological characteristic information.

[0005] Second, different lithological strata have significantly different dynamic response characteristics, and as the drilling depth increases, the signal exhibits strong non-stationarity and nonlinearity. Signal processing methods with fixed parameters cannot adapt to complex and ever-changing geological conditions and have extremely poor generalization ability.

[0006] Third, existing mode decomposition methods have inherent defects. Empirical mode decomposition (EMD) is prone to mode aliasing, that is, features of multiple time scales appear in the same intrinsic mode function (IMF), which seriously interferes with the extraction of lithological feature signals. Although variational mode decomposition (VMD) can suppress mode aliasing, its core parameters, the number of modes K and the penalty parameter α, need to be preset manually and cannot be dynamically and adaptively adjusted according to the drilling signal, making it difficult to meet the real-time identification needs of unknown lithology.

[0007] Therefore, how to develop a drilling lithology analysis method that can adaptively remove strong noise interference, achieve dynamic optimization of decomposition parameters, and meet the requirements of real-time performance and high precision has become an urgent technical problem to be solved in this field. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a lithology identification method based on hybrid mode decomposition and fuzzy entropy-PSO optimization of drilling signals. This method can effectively remove strong noise interference in drilling seismic signals, solve the problems of mode aliasing and fixed parameters that prevent adaptive adaptation in traditional mode decomposition methods, and achieve real-time and accurate identification of formation lithology and rock mass structural characteristics during drilling.

[0009] To achieve the above objectives, the technical solution adopted by this invention is: a lithology identification method based on hybrid mode decomposition of drilling signals and fuzzy entropy-PSO optimization, comprising the following steps: S1. Real-time seismic monitoring signal acquisition during drilling: Acquiring real-time seismic monitoring signals during the drilling process. .

[0010] S2. Hybrid Mode Decomposition Preprocessing: The Empirical and Variational Hybrid Mode Decomposition (EVMD) method is used to decompose the seismic monitoring signal while drilling. This method combines the adaptive decomposition capability of Empirical Mode Decomposition (EMD) with the narrowband component separation capability of Variational Mode Decomposition (VMD). The core parameters of the decomposition process are the number of modes K and the penalty parameter α. Strong noise disturbances in the signal are removed to obtain several intrinsic mode functions (IMF) components.

[0011] S3. Characterize the differences of each IMF component based on fuzzy entropy: Use fuzzy entropy to measure the complexity of the IMF components after decomposition in step S2, and characterize the feature differences of the modal components corresponding to different lithologies through fuzzy entropy values.

[0012] S4. Adaptive optimization of EVMD parameters based on particle swarm optimization (PSO): The PSO algorithm is used with the fuzzy entropy in step S3 as the optimization objective function to adaptively optimize and obtain the optimal parameter combination [K,α] of EVMD.

[0013] S5. Optimal Modal Component Acquisition: Using the optimal parameter combination [K,α] obtained in step S4, the seismic monitoring signal from step S1 is decomposed using EVMD to output the optimized optimal modal data.

[0014] S6. Lithology Identification and Judgment: Based on optimal modal data, construct a multi-dimensional feature vector that reflects the lithology type and rock mass structure characteristics to complete the real-time identification and judgment of stratigraphic lithology.

[0015] Furthermore, in step S1, the seismic monitoring signal while drilling is acquired using a sensing and acquisition system mounted on an intelligent drilling rig. The intelligent drilling rig is a tracked tunnel intelligent drilling rig, and the sensing and acquisition system is an acceleration sensor deployed behind the drill bit. The seismic monitoring signal while drilling is the vibration time history signal of the interaction between the drill bit and the rock mass during drilling, and engineering parameters including the output torque, output speed, and push-pull force of the drilling rig are acquired simultaneously.

[0016] Furthermore, in step S2, the specific implementation process of the EVMD method is as follows: S21. The empirical and variational hybrid mode decomposition method is used to preprocess and decompose the seismic monitoring signal y(t) during drilling, decomposing the signal into several IMF components from high frequency to low frequency and the residual R. n (t), the decomposition expression is: In the formula, The original signal is time. The function; For the first The eigenmode function is the th eigenfunction obtained from the decomposition. There are 10 components, arranged from highest to lowest frequency; k is the total number of IMFs. For time variables, The residual is a monotonic function or constant that represents the overall trend of the signal and cannot be further decomposed into IMF components. EMD decomposition is used to initially separate the multi-scale characteristic components in the signal and suppress the pre-interference of mode mixing.

[0017] S22. For the IMF components after EMD preprocessing, construct a constrained variational model of VMD, setting the number of modes to be decomposed K and the penalty parameter α. The objective of the constrained variational model is to minimize the sum of the estimated bandwidths of each mode, and the constraint condition is that the sum of all mode components is equal to the original input signal.

[0018] S23. Introduce the Lagrange operator λ to transform the constrained variational problem into an unconstrained variational problem, construct the augmented Lagrange function, and iteratively update the modal components and center frequencies of each order through Fourier isochronous transformation until the preset allowable error ε is met, then stop the decomposition and output K IMF components.

[0019] Further, in step S22, the constraint variational expression of the constraint variational model is: In the formula, The original time-domain signal is represented by K, which represents the total number of modes obtained from the decomposition. The center frequency of the k-th mode; The k modal components obtained from the decomposition; It is the Dirac function; This is a partial derivative operation; The imaginary unit is t; t is the time variable.

[0020] Furthermore, in step S3, the calculation of fuzzy entropy specifically involves the following steps: S31. Phase Space Reconstruction of Time Series: For a time series u of IMF components of length N to be processed, set the embedding dimension m, perform phase space reconstruction on the time series to obtain N-m+1 reconstruction vectors, and perform local mean removal processing on each reconstruction vector.

[0021] S32. Vector Similarity Measurement: A fuzzy membership function is introduced to calculate the similarity between reconstructed vectors. The expression is: In the formula, The gradient parameter controls the rate at which similarity decays with distance; This represents the similarity tolerance, which is the threshold for the similarity between vectors. The similarity weight between points; The weights are Euclidean distances.

[0022] S33. Fuzzy Entropy Calculation: Calculate the average similarity of all reconstructed vector pairs, and calculate the average similarity when the embedding dimension is m and m+1 respectively. and The fuzzy entropy value is obtained as follows: In the formula, the fuzzy entropy value is positively correlated with the complexity of the time series. The smaller the entropy value, the stronger the regularity of the series and the lower the complexity; the larger the entropy value, the stronger the randomness of the series and the higher the complexity.

[0023] Furthermore, in step S4, fuzzy entropy is used as the optimization objective function. Specifically, the optimization objective is to minimize the weighted sum of the fuzzy entropy values ​​of each IMF component after decomposition in step S2, thereby maximizing the retention of effective vibration information related to lithology while removing noise. Among them, the IMF components corresponding to different lithologies have differentiated fuzzy entropy characteristics: the low-frequency dominant components of conglomerate and sandstone correspond to smaller fuzzy entropy values, the mid-frequency dominant components of mudstone correspond to medium fuzzy entropy values, and the high-frequency dominant components of granite and limestone correspond to larger fuzzy entropy values.

[0024] Furthermore, in step S4, the optimization process of the particle swarm optimization algorithm is specifically as follows: S41. Initialize particle swarm parameters: Set the particle swarm size, inertia weight, individual learning factor, social learning factor, and maximum number of iterations. The position vector of each particle corresponds to a set of core parameters [K, α], and the velocity vector corresponds to the update step size of the core parameters.

[0025] S42. Iteratively update particle state: In each iteration, the particle's velocity and position are updated based on the individual's historical best solution and the swarm's global best solution. The update formula is: In the formula, Represents particles In the Dimensional speed, Represents particles In the Dimensional position, Indicates inertia weight, Represents particles In the The optimal solution for a given individual. Indicates the entire group at the 1st The global optimal solution. and These represent individual learning factors and social learning factors, respectively. and It is a random number in the range (0, 1).

[0026] S43. Fitness Calculation and Optimal Solution Update: Calculate the fitness value of each particle using fuzzy entropy as the optimization objective function, update the individual historical optimal solution and the global optimal solution of the group until the maximum number of iterations is reached or the convergence condition is met, and output the optimal parameter combination [K, α] corresponding to the global optimal solution.

[0027] Further, in step S6, the construction of the multidimensional feature vector specifically involves: extracting at least one feature parameter from the intrinsic frequency, damping ratio, and corresponding fuzzy entropy value of the optimal modal data to construct a multidimensional feature vector; the lithology identification and judgment includes lithology type identification and rock mass structure feature judgment, wherein the lithology types include granite, limestone, mudstone, conglomerate, and sandstone, and the rock mass structure features include dense rock mass, fractured rock mass, broken rock mass, and weak interlayers.

[0028] Compared with the prior art, the present invention has the following advantages: 1. This invention innovatively proposes an empirical and variational hybrid mode decomposition method (EVMD), which integrates the advantages of empirical mode decomposition (EMD) in adaptively decomposing multi-scale signals with the characteristics of variational mode decomposition (VMD) in effectively suppressing mode aliasing and accurately separating narrowband components. It can effectively remove strong noise interference such as drilling rig vibration and mud circulation in drilling seismic monitoring signals, while completely preserving effective vibration characteristic information related to lithology, thus solving the inherent defects of traditional filtering methods and single mode decomposition methods.

[0029] 2. This invention introduces fuzzy entropy, which utilizes its strong robustness to noise and accurate measurement of signal complexity to achieve differentiated amplification of different lithological features. This provides a reliable quantitative basis for the subsequent optimization of core parameters of EVMD, while avoiding the loss of lithological features caused by over-filtering, and greatly improving the accuracy of lithological feature extraction.

[0030] 3. This invention uses the Particle Swarm Optimization (PSO) algorithm combined with fuzzy entropy as the optimization objective function to achieve adaptive global optimization of the EVMD core parameters [K, α]. It eliminates the need for manual preset of fixed parameters and can dynamically adjust the optimal parameter combination based on real-time signals during drilling. This solves the problems of traditional fixed parameter methods being unable to adapt to complex and variable geological conditions and having poor generalization ability. At the same time, the PSO algorithm has a fast convergence speed and high computational efficiency, and the processing time for a single signal can be controlled within 200ms, which fully meets the engineering requirements for real-time processing during drilling.

[0031] 4. This invention enables simultaneous and accurate identification of lithological types and rock mass structural characteristics during drilling. It can not only distinguish between different lithologies such as granite, limestone, mudstone, conglomerate, and sandstone, but also identify rock mass structural characteristics such as dense rock mass, fractured rock mass, broken rock mass, and weak interlayers. The identification accuracy rate can reach over 96%, providing key technical support for transparent exploration of deep space, intelligent construction of underground engineering, and precise development of deep resources. Attached Figure Description

[0032] Figure 1 This is the overall flowchart of the present invention.

[0033] Figure 2 This is a schematic diagram of the EVMD decomposition process in this invention.

[0034] Figure 3 This is a schematic diagram of the PSO algorithm of the present invention for adaptively optimizing EVMD parameters. Detailed Implementation

[0035] The present invention will be further described below.

[0036] Example 1:

[0037] This embodiment provides an intelligent lithology analysis method for drilling signals based on hybrid mode decomposition and PSO optimization. Specifically, it is applied to deep coal resource exploration drilling projects using tracked intelligent tunnel drilling rigs, with drilling depths ranging from 0 to 1500 meters. Figure 1 The specific implementation steps are as follows: S1. Vibration signal acquisition while drilling: A self-developed tracked intelligent tunnel drilling rig is used. A triaxial ICP accelerometer is installed on the drill rod 0.5m behind the drill bit. The sensor range is ±500g and the sampling frequency is set to 10kHz to collect the vibration acceleration signal of the interaction between the drill bit and the rock mass in real time during drilling. At the same time, nine engineering parameters are collected synchronously, including output torque, output speed, push and pull force, drilling pressure, mud flow rate, drilling pressure, torque, speed and feed rate. All collected data are stored in real time in the onboard industrial control system of the drilling rig to provide a data source for subsequent signal processing.

[0038] S2. Hybrid Mode Decomposition Preprocessing: The acquired raw acceleration vibration signal is decomposed using the EVMD method to remove strong noise disturbances, such as... Figure 2 The specific process shown is as follows: S21. EMD Preprocessing Decomposition: The original seismic monitoring signal while drilling is preprocessed using the EMD method. The signal is decomposed into eight IMF components ranging from high to low frequency and one residual component. The high-frequency noise component and the low-frequency lithology-related effective component in the initial separation of the signal are used to suppress the pre-interference of mode mixing. The decomposition expression is as follows: In the formula, The original signal is time. The function; For the first The eigenmode function is the th eigenfunction obtained from the decomposition. There are 10 components, arranged from highest to lowest frequency; k is the total number of IMFs. For time variables, The residual is a monotonic function or constant that represents the overall trend of the signal and cannot be further decomposed into IMF components. EMD decomposition is used to initially separate the multi-scale characteristic components in the signal and suppress the pre-interference of mode mixing.

[0039] S22. VMD Constrained Variational Model Construction: For the IMF components after EMD preprocessing, a VMD constrained variational model is constructed. The optimization range of the mode number K is set to 312, and the optimization range of the penalty parameter α is 100~3000. The optimization objective of the constrained variational model is to minimize the sum of the estimated bandwidths of each mode. The constraint condition is that the sum of all modal components is equal to the original input signal. The VMD constrained variational formula is: In the formula, The original time-domain signal is represented by K, which represents the total number of modes obtained from the decomposition. The center frequency of the k-th mode; The k modal components obtained from the decomposition; It is the Dirac function; This is a partial derivative operation; The imaginary unit is t; t is the time variable.

[0040] S23. Solving the variational problem: By introducing the Lagrange operator λ, the constrained variational problem is transformed into an unconstrained variational problem. An augmented Lagrange function is constructed, with the specific formula as follows: In the formula, To augment the Lagrange function; Let be the set of all modes; λ be the Lagrange multiplier; α be the penalty parameter, controlling the weights of the bandwidth and fidelity terms; and K be the number of mode decompositions. For the first One mode (time domain); It is the Dirac function; The impulse response is the Hilbert transform. To construct the filter required for analytic signals; For the partial derivative with respect to time; The square of the L2 norm; The original input signal; It is represented as an inner product; the modal components and center frequencies of each order are iteratively updated through Fourier isometric transform, with a preset tolerance error ε=1e-7. When the iteration error is less than ε, the decomposition stops and the IMF components under the corresponding parameters are output.

[0041] S3. Characterizing the differences of each IMF component based on fuzzy entropy: Fuzzy entropy is used to measure the complexity of the IMF components after decomposition in step S2. The fuzzy entropy value is used to characterize the feature differences of the modal components corresponding to different lithologies. The specific steps are as follows: S31. Fuzzy entropy calculation parameter settings: For each IMF component time series, set the embedding dimension m=2, the gradient parameter n=2, and the similarity tolerance r=0.15×SD, where SD is the standard deviation of the corresponding time series.

[0042] S32. Phase Space Reconstruction and Mean Removal: The time series is reconstructed in phase space to obtain N-m+1 reconstructed vectors. Each reconstructed vector is then subjected to local mean removal to enhance robustness against low-frequency drift and background noise. The specific formula is as follows: In the formula: u(i) is the original seismic signal sequence while drilling, i=1,2…N; N is the length of the time series, representing the total number of data points contained in the time series u; m is the dimension of the phase space reconstruction, which determines the number of data points contained in each reconstruction vector; u0(i) is the local mean of the i-th reconstruction vector, which is used for subsequent mean removal processing; i is the starting position index of the reconstruction vector; p is the internal offset of the summation.

[0043] S33. Similarity Measurement and Entropy Calculation: The similarity between reconstructed vectors is calculated using the fuzzy membership function. The average similarity of all reconstructed vector pairs is calculated, and the average similarity is calculated for embedding dimensions of 2 and 3, respectively, using the following formula: Finally, the fuzzy entropy value of the corresponding IMF component is obtained.

[0044] S4. PSO-based adaptive optimization of EVMD parameters: Using the PSO method with the fuzzy entropy from step S3 as the optimization objective function, the optimal parameter combination [K, α] for EVMD is adaptively optimized, as follows: Figure 3 The specific process shown is as follows: S41. Initialization of Particle Swarm Parameters and Construction of Optimization Objective Function: Set the particle swarm size to 20, the maximum number of iterations to 50, the inertia weight ω=0.7, the individual learning factor c1=1.5, and the social learning factor c2=1.5; the position vector of each particle corresponds to a set of EVMD parameters [K,α], and the velocity vector corresponds to the parameter update step size. The parameter range is consistent with step S22; the optimization objective is to minimize the weighted sum of the fuzzy entropy values ​​of each IMF component, where the weight of the low-frequency component related to lithology is set to 0.7, and the weight of the high-frequency component related to noise is set to 0.3. Construct the objective function for EVMD parameter optimization to ensure that effective lithology-related information is retained to the maximum extent while removing noise.

[0045] S42. Particle State Iterative Update: In each iteration, the particle's velocity and position are updated according to the velocity and position update formulas to ensure that the particle converges towards its individual historical optimal solution and the swarm's global optimal solution. The update formula is: In the formula: Represents particles In the Dimensional speed, Represents particles In the Dimensional position, Indicates inertia weight, Represents particles In the The optimal solution for a given individual. Indicates the entire group at the 1st The global optimal solution. and These represent individual learning factors and social learning factors, respectively. and It is a random number in the range (0, 1).

[0046] S43. Fitness Calculation and Optimal Solution Update: Calculate the fitness value of each particle using fuzzy entropy as the optimization objective function, and update the individual historical optimal solution and the global optimal solution of the group simultaneously. When the number of iterations reaches 50, stop the iteration and output the optimal EVMD parameter combination corresponding to the global optimal solution, which is K=6 and α=1200.

[0047] S5. Optimal Modal Component Acquisition: Using the optimal parameter combination [6,1200] obtained above, the drilling seismic signal from step S1 is finally decomposed using the EVMD method, and 6 optimized IMF optimal modal components are output, which effectively removes high-frequency noise such as drilling rig vibration and mud circulation, and fully preserves the low-frequency effective vibration components related to lithology.

[0048] S6. Lithology Identification and Judgment: Extract the center frequency, intrinsic frequency, and corresponding fuzzy entropy value of the six optimal IMF components to construct a 6-dimensional feature vector. Lithology identification and judgment are then performed using a pre-trained random forest classification model. The judgment rules are as follows: When the feature vector is dominated by low-frequency components and the fuzzy entropy value is between 0.2 and 0.5, it is identified as sandstone / conglomerate, and the rock mass structure is a dense rock mass.

[0049] When the feature vector is dominated by the mid-frequency component and the fuzzy entropy value is between 0.5 and 1.2, it is identified as mudstone, and the rock mass structure is a weak interlayer.

[0050] When the feature vector is dominated by high-frequency components and the fuzzy entropy value is between 1.2 and 1.8, it is identified as limestone / granite, and the rock mass structure is a fractured rock mass.

[0051] When the feature vector is dominated by high-frequency components and the fuzzy entropy value is between 1.8 and 2.5, it is identified as limestone / granite, and the rock mass structure is a fractured rock mass.

[0052] This embodiment was verified by core sampling at the drilling site. The accuracy rate of lithology identification reached 96.2%, the accuracy rate of rock mass structure feature identification reached 93.5%, and the processing time of a single signal was 168ms, which fully meets the engineering requirements of real-time identification while drilling.

[0053] Example 2:

[0054] The only difference between this embodiment and Embodiment 1 is that this embodiment is applied to deep exploration drilling projects in metal mines, with a drilling depth range of 0~3000m. In step S4, the parameters of the PSO algorithm are adjusted as follows: particle swarm size is 30, maximum number of iterations is 60, inertia weight ω adopts a linear decreasing strategy, with an initial value of 0.9 and a final value of 0.4, individual learning factor c1=2.0, and social learning factor c2=2.0. In step 3, the embedding dimension of fuzzy entropy is m=3, and the similarity tolerance is r=0.2×SD.

[0055] In this embodiment, the accuracy rate for identifying granite, limestone, and altered rocks in hard rock formations reached 95.8%, and the accuracy rate for identifying fracture zones and weak interlayers reached 92.7%, verifying the excellent adaptability of this method in ultra-deep drilling and hard rock formation scenarios.

Claims

1. A lithology identification method of drilling-while-signal mixed modal decomposition and fuzzy entropy-PSO optimization, characterized in that, Includes the following steps: S1. Acquire real-time seismic monitoring signals during drilling; S2. The empirical and variational hybrid mode decomposition method is used to decompose the seismic monitoring signal while drilling. The core parameters of the decomposition process are the mode number K and the penalty parameter α, and several intrinsic mode function (IMF) components are obtained. S3. Use fuzzy entropy to measure the complexity of the IMF components after decomposition in step S2. S4. Using the particle swarm optimization algorithm, with the fuzzy entropy in step S3 as the optimization objective function, the optimal parameter combination [K,α] of empirical and variational mixed mode decomposition is obtained through adaptive optimization. S5. Using the optimal parameter combination [K,α] obtained in step S4, the seismic monitoring signal from step S1 is decomposed using a hybrid empirical and variational mode decomposition method, and the optimized optimal mode data is output. S6. Based on the optimal modal data, construct a multi-dimensional feature vector that reflects the lithological type and rock mass structure characteristics to complete the real-time identification and judgment of stratigraphic lithology.

2. The method of claim 1, wherein, In step S1, the seismic monitoring signal while drilling is acquired by the sensing and acquisition system on the intelligent drilling rig. The seismic monitoring signal while drilling is the vibration time history signal of the interaction between the drill bit and the rock mass during the drilling process, and the engineering parameters including the output torque, output speed, and push-pull force of the drilling rig are acquired simultaneously.

3. The method of claim 1, wherein, In step S2, the specific implementation process of the empirical and variational mixed mode decomposition method is as follows: S21, the empirical and variational mixed mode decomposition method is used for pretreatment decomposition of the while-drilling seismic signal y(t), and the signal is decomposed into several IMF components and residual from high frequency to low frequency The decomposition expression is: wherein, is the original signal, is the first is the kth intrinsic mode function, arranged in descending order of frequency; k is the total number of IMFs; is the time variable, is the residual; S22. For the IMF components after empirical mode decomposition preprocessing, a constrained variational model for variational mode decomposition is constructed. The number of modes to be decomposed, K, and the penalty parameter α are set. The objective of the constrained variational model is to minimize the sum of the estimated bandwidths of each mode. The constraint condition is that the sum of all modal components is equal to the original input signal. S23. Introduce the Lagrange operator λ to transform the constrained variational problem into an unconstrained variational problem, construct the augmented Lagrange function, and iteratively update the modal components and center frequencies of each order through Fourier isochronous transformation until the preset allowable error ε is met, then stop the decomposition and output K IMF components.

4. The method according to claim 3, characterized in that, In step S22, the constraint variational expression of the constraint variational model is: In the formula, The original time-domain signal is represented by K, which represents the total number of modes obtained from the decomposition. The center frequency of the k-th mode; The k modal components obtained from the decomposition; It is the Dirac function; This is a partial derivative operation; The imaginary unit is t; t is the time variable.

5. The method according to claim 1, characterized in that, In step S3, the calculation steps for fuzzy entropy are as follows: S31. Phase Space Reconstruction of Time Series: For a time series u of IMF components of length N to be processed, set the embedding dimension m, perform phase space reconstruction on the time series to obtain N-m+1 reconstruction vectors, and perform local mean removal processing on each reconstruction vector. S32. Vector Similarity Measurement: Introducing a fuzzy membership function to calculate the similarity between reconstructed vectors, the expression is: In the formula, These are the gradient parameters; For similarity tolerance; The similarity weight between points; Euclidean distance weights; S33. Fuzzy Entropy Calculation: Calculate the average similarity of all reconstructed vector pairs, and calculate the average similarity when the embedding dimension is m and m+1 respectively. and The fuzzy entropy value is obtained as follows: Among them, the fuzzy entropy value is positively correlated with the complexity of the time series.

6. The method according to claim 1, characterized in that, In step S4, the fuzzy entropy is used as the optimization objective function. Specifically, the optimization objective is to minimize the weighted sum of the fuzzy entropy values ​​of each IMF component after decomposition in step S2, so as to maximize the retention of effective vibration information related to lithology while removing noise.

7. The method according to claim 1, characterized in that, In step S4, the optimization process of the particle swarm optimization algorithm is as follows: S41. Initialize particle swarm parameters: Set the particle swarm size, inertia weight, individual learning factor, social learning factor, and maximum number of iterations. The position vector of each particle corresponds to a set of core parameters [K, α], and the velocity vector corresponds to the update step size of the core parameters. S42. Iteratively update particle state: In each iteration, the particle's velocity and position are updated based on the individual's historical best solution and the swarm's global best solution. The update formula is: In the formula, Represents particles In the Dimensional speed, Represents particles In the Dimensional position, Indicates inertia weight, Represents particles In the The optimal solution for a given individual. Indicates the entire group at the 1st The global optimal solution. and These represent individual learning factors and social learning factors, respectively. and A random number in the range (0, 1); S43. Fitness Calculation and Optimal Solution Update: Calculate the fitness value of each particle using fuzzy entropy as the optimization objective function, update the individual historical optimal solution and the global optimal solution of the group until the maximum number of iterations is reached or the convergence condition is met, and output the optimal parameter combination [K, α] corresponding to the global optimal solution.

8. The method according to claim 1, characterized in that, In step S6, the construction of the multidimensional feature vector specifically involves: extracting at least one feature parameter from the intrinsic frequency, damping ratio, and corresponding fuzzy entropy value of the optimal modal data to construct a multidimensional feature vector; the lithology identification and judgment includes lithology type identification and rock mass structure feature judgment.

Images