A method and system for detecting the solid-liquid separation state of a drilling mud vibrating screen.

Abstract

This invention discloses a method and system for detecting the solid-liquid separation state of a drilling mud vibrating screen, relating to the field of solids control monitoring technology. The method includes: arranging fiber optic strain sensors on the screen surface and installing accelerometers on the screen support structure for self-calibration, outputting raw data; using tension distribution data and vibration signal data, constructing a physical model of mud thickness to convert local strain values ​​into mud thickness, generating a mud thickness distribution map; based on the mud thickness distribution map, extracting the natural frequency, vibration amplitude, and modal shape information of the screen surface through Fourier transform and modal recognition algorithms to form a set of elastic modal parameters; and based on the solid-liquid separation state judgment data of the screen surface, generating detection information according to decision logic and adjusting the drilling mud vibrating screen. This invention improves solid-liquid separation efficiency and the intelligent level of equipment response by constructing a multi-parameter matching model through spatial alignment with the thickness distribution map and modal-driven similarity analysis.

Description

Technical Field

[0001] This invention relates to the field of solids control monitoring technology, and in particular to a method and system for detecting the solid-liquid separation state of a drilling mud vibrating screen. Background Technology

[0002] With the increasing depth and complexity of oil drilling, drilling fluid circulation systems place higher demands on the performance of solids control equipment. Vibrating screens, in particular, as primary solid-liquid separation devices, directly affect drilling fluid recovery efficiency and drill bit operational stability. Existing monitoring technologies mainly employ accelerometer arrays to monitor vibration parameters or mud distribution detection methods based on pressure sensors. In recent years, some studies have attempted to introduce sensors such as strain and acceleration to indirectly assess the screen surface condition, while others have utilized image recognition and other methods for post-processing analysis of separation effects. In dynamic coupling analysis, fiber optic sensing technology is being applied to screen strain monitoring through vibration mode inversion of screen surface load distribution.

[0003] Furthermore, there are two main shortcomings in solid-liquid separation state monitoring. First, relying solely on single-type sensor data cannot reflect the coupling relationship between the multimodal physical states of the screen surface, and lacks the ability to comprehensively evaluate factors such as vibration behavior and mud thickness. Second, the algorithms mostly remain at the level of data recording and experience-based judgment, lacking intelligent analysis mechanisms for dynamic modal recognition and multi-parameter fusion, making it difficult to achieve rapid response and decision control under complex drilling conditions, thus restricting the ability to comprehensively grasp and optimize the operating status of the vibrating screen. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for detecting the solid-liquid separation state of a drilling mud vibrating screen to solve the problem of insufficient dynamic detection accuracy of the solid-liquid separation state of the screen surface in the working condition detection of drilling mud vibrating screens.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for detecting the solid-liquid separation state of a drilling mud vibrating screen surface, which includes arranging an optical fiber strain sensor on the screen surface, installing an acceleration sensor on the screen support structure and performing self-checking calibration, and outputting raw data.

[0008] The edge computing unit receives raw data and performs noise filtering, baseline calibration, and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data.

[0009] By using tension distribution data and vibration signal data, a physical model of mud thickness is constructed to convert local strain values ​​into mud thickness and generate a mud thickness distribution map.

[0010] Based on the mud thickness distribution map, the natural frequency, vibration amplitude and modal shape information of the screen surface are extracted by Fourier transform and modal recognition algorithm to form a set of elastic modal parameters;

[0011] By combining the mud thickness distribution map with the set of elastic modal parameters and establishing a multi-parameter correlation model to compare the matching state, the solid-liquid separation state judgment data of the screen surface is obtained.

[0012] Based on the solid-liquid separation status judgment data of the screen surface, detection information is generated according to the decision logic, and the drilling mud vibrating screen is adjusted.

[0013] As a preferred embodiment of the method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface according to the present invention, the step of constructing a physical model of mud thickness to convert local strain values ​​into mud thickness is as follows.

[0014] The tension distribution data and vibration signal data are combined into a three-dimensional tensor data pair;

[0015] Using the Bayesian inverse problem framework, a physical model of mud thickness is constructed based on three-dimensional tensor data through fractional differential constraints and pseudospectral Fourier inversion techniques.

[0016] By inverting and solving the mud thickness function, the mud thickness distribution map of the screen surface at each time moment can be reconstructed.

[0017] As a preferred embodiment of the method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface according to the present invention, the method involves jointly analyzing the mud thickness distribution map and the set of elastic modal parameters, and establishing a multi-parameter correlation model to compare the matching state. The specific steps are as follows.

[0018] Align the mud thickness distribution map with the elastic modal parameter set in a spatial coordinate system to obtain the modal-guided thickness response map;

[0019] By combining the actual modal response power spectrum with the third-order difference integral term and the nonlinear modulation function, a modal driving state similarity index is defined.

[0020] Based on the modal driving state similarity index, solid-liquid separation state determination data for the sieve surface is generated.

[0021] As a preferred embodiment of the method for detecting the solid-liquid separation state of a drilling mud vibrating screen according to the present invention, the natural frequency, vibration amplitude, and modal shape information of the screen surface are extracted by Fourier transform and modal recognition algorithms. The specific steps are as follows.

[0022] The mud thickness distribution map was subjected to a spatiotemporal Fourier transform to obtain the inherent spectral distribution on the screen surface at different spatial and temporal frequencies.

[0023] The inherent frequency points are identified using a modal recognition algorithm, and the frequency domain data is processed by spatial inverse transformation to recover the modal shape information of the sieve surface;

[0024] The response intensity at different locations near the natural frequency of the acceleration signal is extracted to form a vibration amplitude map;

[0025] As a preferred embodiment of the drilling mud vibrating screen surface solid-liquid separation state detection method of the present invention, the noise filtering, baseline calibration and time domain and frequency domain synchronization processing are obtained by dynamically selecting the effective frequency band by wavelet packet energy entropy and combining it with LSTM network to optimize noise reduction, using dynamic time warping to align data delay, and using cross power spectrum to compensate for frequency response phase difference, thereby obtaining tension distribution data and vibration signal data.

[0026] As a preferred embodiment of the method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface described in this invention, the elastic modal parameter set refers to the unified organization of natural frequency, modal shape information and vibration amplitude according to frequency index.

[0027] As a preferred embodiment of the drilling mud vibrating screen surface solid-liquid separation state detection method of the present invention, the decision logic is based on the key indicator evaluation results of the solid-liquid separation state judgment data of the screen surface, and automatically guides the adjustment of the operating parameters of the drilling mud vibrating screen through a preset optimization adjustment strategy.

[0028] Secondly, the present invention provides a system for detecting the solid-liquid separation state of a drilling mud vibrating screen, comprising a sensor acquisition module, an edge computing module, a thickness modeling module, a modal recognition module, a data fusion module, and a control and adjustment module.

[0029] The sensing and acquisition module is used to arrange the fiber optic strain sensor on the screen surface, install the acceleration sensor on the screen support structure and perform self-test calibration, and output raw data.

[0030] The edge computing module is used to receive raw data in the edge computing unit and perform noise filtering, baseline calibration and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data.

[0031] The thickness modeling module is used to establish a physical model of mud thickness using tension distribution data and vibration signal data, converting local strain values ​​into mud thickness and generating a mud thickness distribution map.

[0032] The modal recognition module is used to extract the natural frequency, vibration amplitude and modal shape information of the screen surface through Fourier transform and modal recognition algorithm based on the mud thickness distribution map, and form a set of elastic modal parameters;

[0033] The data fusion module is used to jointly analyze the mud thickness distribution map and the set of elastic modal parameters, and to establish a multi-parameter correlation model to compare the matching state and obtain the solid-liquid separation state judgment data of the screen surface.

[0034] The control and adjustment module is used to adjust the drilling mud vibrating screen based on the detection information generated by the solid-liquid separation state judgment data and decision logic of the screen surface.

[0035] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the drilling mud vibrating screen surface solid-liquid separation state detection method as described in the first aspect of the present invention.

[0036] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for detecting the solid-liquid separation state of a drilling mud vibrating screen as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: By arranging fiber optic strain sensors on the screen surface and installing accelerometers on the support structure, combined with an edge computing platform to perform noise filtering, frequency domain synchronization, and time delay alignment on tension and vibration signals, the invention effectively improves the spatiotemporal consistency and analysis accuracy of the original data. Based on the Bayesian inverse problem framework, a nonlinear mud thickness inversion model is constructed, and the adaptability of the model to complex dynamic boundary conditions is improved through fractional derivative and pseudospectral Fourier methods, achieving a high-precision mud thickness distribution map. The elastic response characteristics of the screen surface are extracted through modal recognition and Fourier transform, forming a complete set of natural frequencies, modal shapes, and vibration amplitude parameters. By performing spatial alignment and modal-driven similarity analysis with the thickness distribution map, a multi-parameter matching model is constructed, improving the solid-liquid separation efficiency and the intelligent level of equipment response. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a method for detecting the solid-liquid separation state of a drilling mud vibrating screen.

[0040] Figure 2 A schematic diagram for modeling mud thickness.

[0041] Figure 3 This is a flowchart for vibration modal analysis.

[0042] Figure 4 Adjust the drilling mud vibrating screen flow chart for intelligent decision control. Detailed Implementation

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for detecting the solid-liquid separation state of a drilling mud vibrating screen, comprising the following steps:

[0047] S1. Arrange the fiber optic strain sensor on the screen surface, install the accelerometer on the screen support structure and perform self-test calibration, and output the raw data.

[0048] Furthermore, by using finite element simulation to determine the key strain regions on the screen surface, we can achieve the minimum deployment of fiber optic strain sensors while maximizing sensitivity coverage.

[0049] Ideally, maximum principal strain distribution analysis is used to identify stress concentration areas on the screen surface under dynamic loads (such as screen edges and near support points), avoiding the blindness of experience-based deployment. The K-means clustering algorithm is used to divide the high-strain zone into N clusters, with the cluster center point serving as the sensor deployment benchmark, ensuring that the maximum sensitive area is covered with the fewest possible sensors. Through simulation and clustering algorithms, the number of sensors is reduced while maintaining high sensitivity. The load function is based on the spatiotemporal variation characteristics of the mud thickness, making the simulation results closer to actual working conditions and avoiding deployment errors caused by static models.

[0050] Specifically, input the screen surface material properties (elastic modulus, Poisson's ratio), length, width, slurry thickness, and boundary conditions (support structure). Apply dynamic loads to simulate the strain distribution when the slurry distribution is uneven. The load function is:

[0051] F(x,y,t)=ρgh(x,y,t)

[0052] Where ρ is the mud density, h is the local mud thickness, g is the gravitational acceleration, F(x,y,t) is the load function, x is the length, y is the width, and t is the time.

[0053] The maximum principal strain distribution on the sieve surface was analyzed to identify high-strain sensitive areas. The K-means clustering algorithm was used to divide the high-strain areas into N clusters. A fiber optic sensor was deployed at the center of each cluster to ensure coverage of the high-strain area. Dual fiber optic sensors (axial and transverse) were installed orthogonally at the center of each cluster to form a redundant network, eliminating directional errors. The orthogonal redundancy design improved the directional robustness of strain measurement and avoided data loss due to single sensor failure.

[0054] In this design, dual fiber optic sensors are installed orthogonally at the center of each cluster, forming a redundant network. This orthogonal deployment allows for simultaneous measurement of axial and transverse strain components, correcting for unidirectional measurement errors (such as transverse strain interference caused by screen vibration) through vector synthesis. The redundancy design ensures that strain values ​​can be reconstructed from data in the other direction even if a single sensor fails, preventing local data loss due to single-sensor failure. A single sensor can only measure strain in one direction and is easily affected by changes in the direction of screen vibration; the orthogonal redundancy design reduces overall error, and through the redundant network, maintains the integrity of most data even if a few sensors fail.

[0055] Accelerometers are mounted on the screen support structure using modal analysis, enabling dynamic self-testing.

[0056] Furthermore, modal analysis is used to determine the natural frequencies and mode shapes of the screen support structure. Accelerometers are installed at the modal nodes of the support structure, and an LMS adaptive filter is used to eliminate transmission path noise. The LMS filter dynamically adjusts its weights based on the reference signal (support structure vibration) to eliminate transmission path noise (such as motor vibration interference) while retaining the direct vibration signal from the screen surface. By monitoring the frequency domain characteristics of the acceleration signal (such as natural frequency shift), it is possible to determine in real time whether the accelerometer is abnormally loose or damaged.

[0057] Sensor data is calibrated in real time using a digital twin model, and the raw data is output to the edge computing unit.

[0058] Specifically, an LMS adaptive filter is used to eliminate noise transmitted through the support structure while retaining the direct vibration signal from the screen surface. A digital twin model is constructed based on finite element simulation results and real-time sensor data, and the sensor data is dynamically calibrated using a Kalman filter algorithm (e.g., strain drift compensation). The digital twin model interacts with the physical screen surface in real time to calibrate sensor baseline offsets caused by temperature drift and material creep. The load parameters of the twin model are adjusted according to the mud thickness distribution to ensure consistency between simulation and measured data. The monthly drift error of sensors due to environmental factors (temperature, humidity) is relatively large, but digital twin calibration can control the error to within 0.5%. Through real-time data feedback, the model can dynamically correct boundary conditions (e.g., changes in support structure stiffness), improving its generalization ability under different working conditions.

[0059] S2. The edge computing unit receives raw data and performs noise filtering, baseline calibration, and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data.

[0060] Furthermore, the effective frequency band is dynamically selected by wavelet packet energy entropy, and the wavelet coefficient threshold is optimized by combining LSTM network to reconstruct the denoised signal.

[0061] Specifically, by quantifying the value of frequency band information through energy entropy, the fixed threshold method addresses the frequency band confusion problem in scenarios where mud impact (high frequency) and mechanical vibration (low frequency) coexist. For example, it preserves mud load characteristics in frequency bands above 500Hz while suppressing mechanical noise. Utilizing an LSTM network to learn wavelet coefficient correction rules from historical data, it can dynamically adapt to changes in noise statistical characteristics caused by sudden changes in drilling conditions. This directly improves the quality of the raw data, providing high-fidelity input for subsequent tension inversion and vibration demodulation, and laying a data foundation for determining the solid-liquid separation state.

[0062] It synchronously processes time and frequency domain data to generate tension distribution data and vibration signal data.

[0063] Specifically, tension and vibration sensors deployed at multiple locations on the vibrating screen surface acquire multi-point sensor data. To address the time delay caused by differences in vibration propagation paths at different measuring points on the screen surface, a Dynamic Time Warping (DTW) algorithm is employed to align the time axes of the original data from each channel, thus resolving the issue of inconsistent timing for the same event across multiple channels. To address the frequency response differences between different sensors due to hardware characteristics, phase differences are extracted using cross-power spectral analysis and frequency domain compensation is performed to ensure consistent frequency response and prevent phase differences from affecting the accurate extraction of vibration modal parameters. The synchronized tension and vibration data are fused in both the temporal and spatial dimensions to form a joint tensor data structure, preserving the spatiotemporal correlation characteristics of tension-vibration information. Feature extraction and pattern recognition are then performed on the fused tensor using a Temporal-Convolutional Tensor Network (TCNN). The spatiotemporal tensor convolutional network structure introduces an attention mechanism to explicitly highlight the correlation between tension and vibration features, focusing on regions with clear physical meaning (such as regions with significant changes in vibration modes corresponding to high tension changes), automatically adjusting the parameters of the vibrating screen (such as frequency, angle, flow rate, etc.) to achieve closed-loop dynamic control and improve slurry separation efficiency.

[0064] A superior approach aligns the original data delay using Dynamic Time Warping (DTW) and compensates for frequency response phase differences using cross-power spectrum. By aligning the original data time axis through DTW, the data misalignment problem caused by vibration transmission delays at different locations on the screen surface is resolved. Cross-power spectrum phase compensation eliminates the influence of sensor frequency response differences on vibration modal parameter extraction. The attention mechanism focuses on the physical correlation region between tension and vibration, particularly the matching relationship between tension abrupt changes and modal characteristic changes, making the model output more physically interpretable, avoiding "black box" judgments, and improving engineering practicality. This method not only enables high-frequency real-time evaluation of solid-liquid separation efficiency but also achieves automatic control response, constructing a dynamic feedback path between separation state and control parameters, adapting to different geological, mud, and equipment conditions, and improving drilling site mud treatment efficiency. The real-time output separation efficiency index can be used to intelligently adjust the vibrating screen operating parameters, effectively avoiding over-operation or unnecessary energy consumption, and achieving energy-saving optimization of the mud treatment process.

[0065] S3. Using tension distribution data and vibration signal data, construct a physical model of mud thickness to convert local strain values ​​into mud thickness and generate a mud thickness distribution map;

[0066] Furthermore, the tension distribution data and vibration signal data are unified into a three-dimensional tensor data pair.

[0067] Specifically, strain tensor data output from fiber optic tension sensors distributed across various locations on the screen surface, along with vibration data collected by triaxial accelerometers at corresponding tension measurement points, are collected to form multi-source spatiotemporal physical observation data. Based on time synchronization and spatial coordinate consistency, strain and acceleration data at each time point and sensor location are fused to construct a three-dimensional tensor structure. The three dimensions of this three-dimensional tensor correspond to the time series, spatial sensor placement, and physical quantity channels (i.e., tension and vibration characteristics), respectively.

[0068] Ideally, a three-dimensional tensor structure preserves the spatiotemporal correlation between strain and vibration (e.g., the synchronicity between strain abrupt changes and vibration amplitude variations), avoiding information fragmentation caused by time-division or channel-division processing. Tensor data contains non-uniform rheological properties and sensor layout perturbations, providing complete input features for nonlinear inversion modeling. Linear calibration only utilizes single-point strain-thickness relationships, while three-dimensional tensors integrate spatial gradients, vibration spectra, and time series, expanding the information dimensions and improving the ability to capture local perturbations. Through multi-dimensional correlation of tensor data, anomalous noise can be eliminated based on spatial consistency and strain trends of adjacent sensors.

[0069] Using the Bayesian inverse problem framework, a physical model of mud thickness is constructed based on three-dimensional tensor data through fractional differential constraints and pseudospectral Fourier inversion techniques.

[0070] Furthermore, a nonlinear mapping model between strain-vibration and mud thickness is constructed based on the Bayesian inverse problem modeling method. Fractional-order differential constraints are employed to improve robustness to non-uniform rheological properties and sensor interference. A pseudospectral Fourier inversion technique is combined to achieve frequency domain modeling and solution, avoiding the computational complexity caused by high-dimensional nonlinear problems. Simultaneously, a dual constraint mechanism is used to limit the upper and lower boundaries of the inversion variables, improving the physical feasibility and convergence of the inversion solution. To enhance the model's adaptability, an automatic hyperparameter adjustment mechanism is adopted to dynamically adjust prior weights and observation error tolerances under different operating conditions and signal-to-noise ratio scenarios, thereby improving the overall stability of the inversion process.

[0071] Specifically, based on the physical constraints of mechanics and fluid fields, a nonlinear relationship is defined between the local mud thickness function H(x,y,t) and the strain tensor. Considering the local stiffness variation of the screen surface, vibration coupling terms, and dynamic load disturbances, a constrained optimization objective function is formed, expressed as follows:

[0072]

[0073] f(ε,A)=α·tanh(β1ε)+γ·ln(1+β2A);

[0074] Where H(x,y,t) is the mud thickness at position (x,y) on the screen surface at time t, ε is the strain tensor data collected by the fiber optic sensor, A is the acceleration vibration characteristics, including instantaneous amplitude, dominant frequency, etc., and f(ε,A) is the nonlinear mapping function. λ is the gradient of mud thickness variation, used for smoothing; α is the regularization coefficient controlling the smoothness of the thickness map; α is the fusion coefficient, representing the weight of the tension strain component in mud thickness estimation; β1 is the strain nonlinearity adjustment coefficient, controlling the response range and compressibility of tanh(β1ε) to the strain tensor data ε; γ is the fusion coefficient, representing the weight of the vibration signal component in mud thickness estimation; and β2 is the adjustment coefficient of the vibration logarithmic function, used to scale the influence range of the acceleration vibration characteristic quantity A in the logarithmic function ln(1+β2A) to prevent excessive input from causing numerical instability.

[0075] It should be noted that the Bayesian framework outputs the confidence interval for thickness estimation through prior distributions (e.g., mud thickness follows a Gamma distribution) and likelihood functions (sensor error models). Fractional derivatives can describe the long-range correlation of mud thickness gradients (e.g., thickness changes in the central region caused by accumulation at the screen edges), overcoming the limitation of integer derivatives which can only model local variations. Linear assumptions can lead to thickness estimation bias under vibration coupling, while fractional constraints reduce the modeling error of nonlinear rheological properties. The pseudospectral Fourier inversion technique transforms the physical equations into spectral coefficient constraints, avoiding convergence oscillations caused by the penalty function method and reducing the number of iterations.

[0076] By inverting and solving the mud thickness function, the mud thickness distribution map of the screen surface at each time moment can be reconstructed.

[0077] Furthermore, the objective function is projected onto the spectral coefficient space using Chebyshev polynomial expansion. Pseudo-spectral Fourier inversion technology is then used to transform the physical equations (such as non-negativity of mud thickness and maximum thickness constraints) into a feasible region of spectral coefficients. Combined with the L-BFGS algorithm, the global optimum is efficiently searched within the coefficient space. A hyperparameter adaptive mechanism (such as dynamically adjusting the fusion weights α and γ of strain and vibration based on the local tension change rate) optimizes the inversion accuracy in high-gradient regions. Finally, a mud thickness distribution map with physical consistency (value range constraint [0, Hmax]) is output, achieving robust thickness reconstruction under low signal-to-noise ratio and strong disturbance conditions. This solves the error accumulation problem caused by neglecting vibration coupling and non-uniform rheology in linear calibration methods. Specifically, the obtained mud thickness function H(x,y,t) is used to construct a thickness distribution map sequence in the screen surface space for use in downstream modal analysis and decision-making logic. The output mud thickness function H(x,y,t) has a range of [0,Hmax], where Hmax is the maximum mud thickness that the screen can bear (which can be calibrated on site, for example, 30-60mm). When H(x,y,t) < 5mm, it indicates that the screen surface has basically completed solid-liquid separation and the mud concentration is low. When H(x,y,t) > 30mm, it indicates that the solid-liquid separation efficiency is low and the mud is accumulating, and the vibration intensity and screen tilt angle need to be adjusted.

[0078] S4. Based on the mud thickness distribution map, extract the natural frequency, vibration amplitude and modal shape information of the screen surface through Fourier transform and modal recognition algorithm to form a set of elastic modal parameters;

[0079] Furthermore, a spatiotemporal Fourier transform was performed on the mud thickness distribution map to obtain the inherent spectral distribution on the screen surface at different spatial and temporal frequencies.

[0080] Specifically, using the mud thickness distribution map as key input data, a three-dimensional spatiotemporal data structure is constructed along the time axis and spatial coordinate axis to analyze the thickness changes of the screen surface at various time points. Then, a spatiotemporal joint Fourier transform is applied to the three-dimensional spatiotemporal data structure to convert the original time series and spatial distribution information into spectral expressions in the frequency and wavenumber domains, thereby obtaining the inherent spectral distribution of the screen surface at different time frequencies and spatial wavenumber dimensions. This method can simultaneously capture the dynamic response characteristics (such as vibration frequency and frequency drift) and spatial modal morphology (such as wave patterns and node distribution) of the screen surface structure, achieving high-precision extraction of vibration modal parameters.

[0081] In the spatiotemporal spectrum diagram, for the sudden changes in thickness or fluidity of the vibrating screen surface caused by mud accumulation, local blockage, etc., the spectrum separation method is used to distinguish steady-state and non-steady-state components. For example, by identifying the modulation characteristics generated by the slow flow of mud in the low-frequency region, the resonance mode of the screen surface is identified in the high-frequency region, effectively eliminating the frequency interference components caused by mud disturbance and improving the identification accuracy of the true modal frequency of the structure.

[0082] Furthermore, to prevent pseudo-modalities caused by localized accumulation or uneven distribution of mud on the screen surface, a spatial wavenumber filtering strategy is employed to identify and suppress abnormally high wavenumber components, thereby improving the stability and accuracy of modal identification results. Finally, the extracted modal frequencies, vibration amplitudes, modal shapes, and other parameters are summarized into a set of elastic modal parameters of the screen surface, serving as important inputs for subsequent state fusion analysis and judgment models.

[0083] The inherent frequency points are identified using a modal recognition algorithm, and the frequency domain data is processed by spatial inverse transformation to recover the modal shape information of the sieve surface;

[0084] Specifically, an improved Stochastic Subspace Identification (SSI) algorithm is employed, which can extract structural modal parameters based solely on the response without relying on excitation input. Simultaneously, Independent Component Analysis (ICA) is combined to perform source signal separation processing on the acquired frequency domain data, effectively separating resonant noise signals generated by external interference sources such as the support structure and drive motor from the modal response of the screen surface itself. The screen surface modal shape is then reconstructed through inverse spatial Fourier transform. ICA separates support structure transmission path noise (such as motor vibration interference) from the true screen surface modes, solving the modal confusion problem at low signal-to-noise ratios (SNR < 10 dB). Singular Value Decomposition (SVD) is used to prioritize and filter the initially extracted modal set, automatically filtering out secondary modes with low energy or indistinct features, such as some higher-order harmonics or boundary effect interference components, to ensure that the finally recovered screen surface modal shape information is representative and has engineering decision-making value.

[0085] Furthermore, the response intensity at different locations near the natural frequency of the acceleration signal is extracted to form a vibration amplitude map;

[0086] Specifically, wavelet packet transform is performed on the acceleration signal near the natural frequency to calculate the vibration energy at each location, and continuous vibration amplitude maps are generated through Kriging interpolation. The wavelet packets integrate energy within a narrow band, suppressing interference from adjacent frequency bands (such as when the motor's fundamental frequency is close to the screen's natural frequency), thus reducing energy estimation errors. Kriging interpolation, combined with sensor layout density, fills in vibration data in areas where sensors are not installed. The vibration amplitude map can quantify local stiffness loss on the screen surface, locate abnormal areas, and guide precise maintenance. By comparing continuous vibration amplitude maps corresponding to different natural frequencies, the differences between uneven mud distribution (low-frequency amplitude variations) and mechanical faults (high-frequency amplitude abrupt changes) can be distinguished.

[0087] The natural frequencies, modal shape information, and vibration amplitude diagrams are organized into an elastic modal parameter set according to a frequency index. This set is then structured into a hierarchical database with added physical metadata (such as the damping ratio of the screen material). By linking these three data points through the frequency index, a complete modal fingerprint of "frequency-mode shape-energy" is established, resolving the issues of missed modal shape anomalies due to relying solely on frequency and misjudgments caused by isolated parameters. The elastic modal parameter set can be directly input into fault diagnosis models (such as pattern matching based on Siamese neural networks), achieving a seamless connection from "data" to "decision-making." The parameter set supports modal similarity calculation with historical health states, enabling early fault warning. Through normalization, the natural frequencies are converted into ratios relative to the fundamental frequency, making the parameter set applicable to operating conditions with different vibration intensities.

[0088] S5. Combine the mud thickness distribution map with the elastic modal parameter set for analysis, and establish a multi-parameter correlation model to compare the matching state and obtain the solid-liquid separation state judgment data of the screen surface.

[0089] Furthermore, by aligning the mud thickness distribution map with the set of elastic modal parameters in a spatial coordinate system, a modally guided thickness response map is obtained.

[0090] Specifically, the mud thickness distribution map is spatially registered with the set of elastic modal parameters (frequency index, modal shape information, vibration amplitude map). Affine transformations are used to align the coordinate system through translation, rotation, and scaling to generate a modally guided thickness response map. Analyzing mud thickness and vibration modal data independently cannot correlate spatial coupling effects, such as vibration suppression caused by the overlap of mud accumulation areas with modal nodes. Spatial alignment superimposes mud thickness and mode shapes in the same coordinate system, quantifying the excitation weight of thickness distribution on specific modes. The impact of mud thickness variations in different regions of the screen surface on modal response varies significantly; for example, mud accumulation at the modal midpoint significantly alters vibration energy, and such spatial correlations cannot be quantified. The modally guided thickness response map reveals the spatial correlation between mud thickness and mode shapes, providing spatial targets for dynamic control. Affine transformations compensate for screen surface deformation and alignment errors, ensuring spatial consistency.

[0091] Based on the actual modal response power spectrum, a modal driving state similarity index is defined by using the third-order difference integral term and the nonlinear modulation function;

[0092] Furthermore, by using a third-order difference integral term and a nonlinear modulation function, a modal driving state similarity index is defined to achieve high-dimensional nonlinear coupling between the thickness distribution map and the modal domain power distribution, quantifying the degree of thickness-vibration matching under different modes. The expression is as follows:

[0093]

[0094] Among them, S i,k It is the reference power spectral density of the i-th mode under historical health state k. It is the measured average power spectral density of the i-th mode at the current time t, v i These are modal weighting coefficients, determined by the modal participation factor (e.g., v2 indicates that the second-order mode has a greater impact on the solid-liquid separation state), u i is the mud thickness-modal response coupling coefficient, W(t) is the modal driving state similarity index, and n is the total number of modes.

[0095] It should be noted that similarity indices (such as Euclidean distance and correlation coefficient) are insufficiently sensitive to nonlinear differences and do not consider the coupling effect between modal weights and thickness. This is addressed through third-order difference terms. It amplifies significant deviations from a healthy state while suppressing minor fluctuations. The nonlinear modulation function tanh(v) i ·u i The oversaturation effect in the high coupling coefficient region is compressed. This solves the problem of insufficient sensitivity to small deviations, and the second-order difference term... The change is insensitive, while the third-order term improves the detection sensitivity of minor anomalies (such as power spectrum shifts caused by early screen cracks) by 3 times. This enables early fault warnings, triggering an alert when W(t) > 0.15, via v i The contribution of key modes is dynamically adjusted to make the state determination more in line with the actual working conditions.

[0096] Based on the modal driving state similarity index, solid-liquid separation state determination data for the sieve surface is generated.

[0097] Furthermore, state levels are classified based on the threshold value of the modal-driven state similarity index.

[0098] Specifically, W(t) < 0.1 indicates normal operating conditions, with a solid-liquid separation efficiency ≥ 90%; 0.1 ≤ W(t) < 0.3 indicates a slight anomaly, with localized mud accumulation (thickness > 20 mm), suggesting an increase in amplitude; W(t) ≥ 0.3 indicates a severe anomaly, with screen damage or mud blockage, requiring shutdown for maintenance. The modal-driven state similarity index integrates multimodal parameters and thickness distribution to achieve quantitative state classification. Uniform mud thickening may lead to thickness exceeding limits despite normal vibration response, resulting in a false alarm. However, the modal-driven state similarity index remains < 0.1 because the power spectrum does not shift, avoiding false alarms. Differential commands are triggered based on the modal-driven state similarity index level (e.g., adjusting parameters only for slight anomalies, immediate maintenance for severe anomalies), improving maintenance efficiency.

[0099] The superior method quantifies the physical correlation between mud thickness and modal response through spatial alignment and coupling coefficients, overcoming the limitations of traditional single-parameter analysis. Third-order difference integrals and nonlinear modulation functions constitute a dual mechanism of sensitivity enhancement and noise suppression, improving the detection rate of small anomalies while avoiding overfitting. Under normal operating conditions, it automatically enters energy-saving mode, reducing the power consumption of the vibration motor. Early warning extends the screen replacement cycle, reducing unplanned downtime losses. It achieves high-precision, adaptive determination of the solid-liquid separation state of the screen surface, providing core algorithmic support for intelligent operation and maintenance of drilling mud treatment.

[0100] S6. Based on the solid-liquid separation state judgment data of the screen surface, generate detection information according to the decision logic and adjust the drilling mud vibrating screen.

[0101] Furthermore, based on the natural frequency, vibration amplitude, and modal shape in the solid-liquid separation state determination data, machine learning algorithms are used to train historical data, and the optimized model is used to predict the most suitable operating parameters (such as vibration frequency and amplitude) under the current state.

[0102] Specifically, by integrating data on the solid-liquid separation status of the screen surface, including natural frequency offset, vibration amplitude distribution, and modal shape distortion, with machine learning algorithms, a dynamic decision-making logic is constructed to optimize the control of the drilling mud vibrating screen. Screen stiffness degradation is identified based on the natural frequency offset, and mud blockage or screen damage areas are located by combining the spatial distribution of vibration amplitude, overcoming the limitations of traditional single thickness threshold criteria and reducing the misjudgment rate. Gradient information of modal shape is used to capture micro-cracks in the screen, guiding precise maintenance and improving efficiency compared to manual inspection. Furthermore, an LSTM neural network is used, with historical modal parameters (frequency, amplitude, mode shape) as input, to predict mud separation efficiency and dynamically optimize control parameters (such as vibration frequency adjustment ±3Hz, tilt angle adjustment ±2°), maintaining control accuracy even during sudden changes in mud density, and narrowing the screen life prediction error. Through multi-modal data fusion and adaptive optimization using machine learning, problems such as high misjudgment rate and control lag in solid-liquid separation status under complex operating conditions are solved.

[0103] This embodiment also provides a drilling mud vibrating screen surface solid-liquid separation state detection system, including: a sensor acquisition module, an edge computing module, a thickness modeling module, a modal recognition module, a data fusion module, and a control and adjustment module. The sensor acquisition module is used to arrange fiber optic strain sensors on the screen surface, install accelerometers on the screen support structure and perform self-calibration, and output raw data. The edge computing module is used to receive raw data in the edge computing unit and perform noise filtering, baseline calibration, and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data. The thickness modeling module is used to utilize the tension distribution data and vibration signals... The system comprises four modules: a physical model of mud thickness to convert local strain values ​​into mud thickness and generate a mud thickness distribution map; a modal recognition module to extract the natural frequency, vibration amplitude, and modal shape information of the screen surface based on the mud thickness distribution map using Fourier transform and modal recognition algorithms, forming a set of elastic modal parameters; a data fusion module to jointly analyze the mud thickness distribution map and the set of elastic modal parameters, and to establish a multi-parameter correlation model to compare and match the states, obtaining solid-liquid separation state determination data for the screen surface; and a control and adjustment module to adjust the drilling mud vibrating screen based on the solid-liquid separation state determination data and the detection information generated by the decision logic.

[0104] This embodiment also provides a computer device applicable to the method for detecting the solid-liquid separation state of a drilling mud vibrating screen, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the method for detecting the solid-liquid separation state of a drilling mud vibrating screen as proposed in the above embodiment.

[0105] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0106] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the method for detecting the solid-liquid separation state of a drilling mud vibrating screen as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0107] In summary, this invention effectively improves the spatiotemporal consistency and analytical accuracy of the original data by arranging fiber optic strain sensors on the screen surface and installing accelerometers on the support structure, combined with an edge computing platform for noise filtering, frequency domain synchronization, and time delay alignment of tension and vibration signals. Based on the Bayesian inverse problem framework, a nonlinear mud thickness inversion model is constructed, and the adaptability of the model to complex dynamic boundary conditions is improved through fractional derivative and pseudospectral Fourier methods, achieving a high-precision mud thickness distribution map. The elastic response characteristics of the screen surface are extracted through modal recognition and Fourier transform, forming a complete set of natural frequencies, modal shapes, and vibration amplitude parameters. By spatially aligning with the thickness distribution map and performing modal-driven similarity analysis, a multi-parameter matching model is constructed, improving solid-liquid separation efficiency and the intelligence level of equipment response.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting the solid-liquid separation state of a drilling mud vibrating screen, characterized in that: include, Fiber optic strain sensors are arranged on the screen surface, and accelerometers are installed on the screen support structure and self-calibrated to output raw data. The edge computing unit receives raw data and performs noise filtering, baseline calibration, and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data. By using tension distribution data and vibration signal data, a physical model of mud thickness is constructed to convert local strain values ​​into mud thickness and generate a mud thickness distribution map. Based on the mud thickness distribution map, the natural frequency, vibration amplitude and modal shape information of the screen surface are extracted by Fourier transform and modal recognition algorithm to form a set of elastic modal parameters; The mud thickness distribution map and the elastic modal parameter set were analyzed together, and a multi-parameter correlation model was established to compare the matching state, so as to obtain the solid-liquid separation state judgment data of the screen surface. The specific steps are as follows. Align the mud thickness distribution map with the elastic modal parameter set in a spatial coordinate system to obtain the modal-guided thickness response map; Based on the actual modal response power spectrum, a modal driving state similarity index is defined by using the third-order difference integral term and the nonlinear modulation function; ； in, It is the first Each modality in historical health state The reference power spectral density is below. It is the current moment. Next Measured average power spectral density of each mode These are modal weighting coefficients, determined by the modal participation factor. It is the mud thickness-modal response coupling coefficient. As a modal-driven state similarity index, It is the total number of modes; Based on the modal driving state similarity index, generate solid-liquid separation state determination data for the sieve surface; Based on the solid-liquid separation status judgment data of the screen surface, detection information is generated according to the decision logic, and the drilling mud vibrating screen is adjusted.

2. The method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface as described in claim 1, characterized in that: The construction of the physical model for mud thickness converts local strain values ​​into mud thickness. The specific steps are as follows. The tension distribution data and vibration signal data are combined into a three-dimensional tensor data pair; Using the Bayesian inverse problem framework, a physical model of mud thickness is constructed based on three-dimensional tensor data through fractional differential constraints and pseudospectral Fourier inversion techniques. By inverting and solving the mud thickness function, the mud thickness distribution map of the screen surface at each time moment can be reconstructed.

3. The method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface as described in claim 1, characterized in that: The natural frequency, vibration amplitude, and modal shape information of the screen surface are extracted using Fourier transform and modal recognition algorithms. The specific steps are as follows: The mud thickness distribution map was subjected to a spatiotemporal Fourier transform to obtain the inherent spectral distribution on the screen surface at different spatial and temporal frequencies. The inherent frequency points are identified using a modal recognition algorithm, and the frequency domain data is processed by spatial inverse transformation to recover the modal shape information of the sieve surface; The response intensity at different locations near the natural frequency of the acceleration signal is extracted to form a vibration amplitude map.

4. The method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface as described in claim 1, characterized in that: The noise filtering and baseline calibration are achieved by dynamically selecting effective frequency bands through wavelet packet energy entropy and combining them with LSTM network calibration to optimize noise reduction. Synchronous processing of the time and frequency domains refers to using dynamic time warping to align data delays in the time domain and using cross power spectrum compensation to compensate for frequency response phase differences in the frequency domain.

5. The method for detecting the solid-liquid separation state of a drilling mud vibrating screen as described in claim 1, characterized in that: The set of elastic modal parameters refers to the unified organization of natural frequencies, modal shape information, and vibration amplitude according to frequency index.

6. The method for detecting the solid-liquid separation state of the drilling mud vibrating screen surface as described in claim 1, characterized in that: The decision-making logic is based on the evaluation results of key indicators of the solid-liquid separation state determination data of the screen surface, and automatically guides the adjustment of the operating parameters of the drilling mud vibrating screen through a preset optimization and adjustment strategy.

7. A system for detecting the solid-liquid separation state of a drilling mud vibrating screen, based on the method for detecting the solid-liquid separation state of a drilling mud vibrating screen according to any one of claims 1 to 6, characterized in that: It includes a sensor acquisition module, an edge computing module, a thickness modeling module, a modal recognition module, a data fusion module, and a control and adjustment module. The sensing and acquisition module is used to arrange the fiber optic strain sensor on the screen surface, install the acceleration sensor on the screen support structure and perform self-test calibration, and output raw data. The edge computing module is used to receive raw data in the edge computing unit and perform noise filtering, baseline calibration and synchronous processing in the time and frequency domains to generate tension distribution data and vibration signal data. The thickness modeling module is used to establish a physical model of mud thickness using tension distribution data and vibration signal data, converting local strain values ​​into mud thickness and generating a mud thickness distribution map. The modal recognition module is used to extract the natural frequency, vibration amplitude and modal shape information of the screen surface through Fourier transform and modal recognition algorithm based on the mud thickness distribution map, and form a set of elastic modal parameters; The data fusion module is used to jointly analyze the mud thickness distribution map and the set of elastic modal parameters, and to establish a multi-parameter correlation model to compare the matching state and obtain the solid-liquid separation state judgment data of the screen surface. The control and adjustment module is used to generate detection information based on the solid-liquid separation state judgment data of the screen surface and according to the decision logic, and adjust the drilling mud vibrating screen.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the drilling mud vibrating screen surface solid-liquid separation state detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the drilling mud vibrating screen surface solid-liquid separation state detection method according to any one of claims 1 to 6.

Images