Modular x-ray analysis apparatus and method for oil and gas geology

Abstract

The present application relates to the technical field of X-ray analysis instrument software, and particularly relates to a modular X-ray analysis device and method for oil and gas geology, wherein a multiple-target X-ray source is driven by a time-sharing control unit to alternately excite fluorescence and diffraction signals, and XRF and XRD are collected in time-sharing or synchronization by a time sequence controller; an environment adaptation module identifies sample types based on a classification model and switches between vacuum and atmospheric environment; a shock resistance unit suppresses external vibration and temperature drift interference; a dynamic strategy module triggers an energy filtering mode according to a mineral phase identification result; a monochromator improves diffraction resolution; a gradient descent algorithm iteratively updates model parameters; and target parameters and time sequence weights are synchronously adjusted. The present application solves the problems of low detection efficiency, data fragmentation and insufficient isotope identification capability caused by independent operation of XRF and XRD.

Description

Technical Field

[0001] This invention relates to the field of X-ray analysis instrument software technology, and in particular to a modular X-ray analysis device and method for oil and gas geology. Background Technology

[0002] X-ray fluorescence spectroscopy (XRF) is a non-destructive analytical technique that analyzes the elemental composition and content of a sample by measuring the characteristic fluorescence spectra emitted after a substance is excited by high-energy X-rays. Its principle is based on the characteristic peaks resulting from the energy differences in electron transitions after the atoms of different elements are excited. X-ray diffraction (XRD), on the other hand, utilizes the Bragg diffraction phenomenon generated by the interaction of X-rays with crystalline materials. By analyzing the diffraction angle and intensity distribution, it reveals the microstructural information of the material, such as its crystal structure, phase composition, and lattice parameters. These two techniques provide material characterization data from the perspectives of elemental composition and crystal structure, respectively. XRF can quickly locate elemental composition, while XRD can accurately identify compound morphologies. The two techniques complement each other in material analysis, and their combined application can effectively address complex analytical needs where individual techniques cannot distinguish allotropes or substances with the same elements but different phase structures.

[0003] In existing technologies, XRF and XRD equipment operate independently, requiring separate steps to detect elemental composition and phase structure. This results in long detection cycles, poor data correlation, and an inability to effectively distinguish samples with the same elements but different phases (such as hematite and magnetite). For example, in geological cuttings analysis, current methods require separate XRF determination of elemental content and XRD identification of mineral composition. This cumbersome process makes it difficult to simultaneously verify the correlation between elemental distribution and phase formation, affecting the efficiency and accuracy of comprehensive judgment. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a modular X-ray analysis device and method for oil and gas geology. This invention solves the problems of low detection efficiency, fragmented data on elemental composition and phase structure caused by the independent operation of existing XRF and XRD technologies, and insufficient ability to distinguish samples with the same elements or different phase structures, such as allotropes or samples with the same elements but different phase structures.

[0005] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a modular X-ray analysis device for oil and gas geology, comprising:

[0007] The acquisition module is used to trigger the sample to generate X-ray fluorescence signal and X-ray diffraction signal through a preset multi-target X-ray source, simultaneously acquire the energy and intensity distribution curve of the fluorescence signal and the diffraction angle and intensity spectrum of the diffraction signal, and transmit the acquired signals to the data processing module.

[0008] An environment adaptation module, connected to the acquisition module, is used to identify the sample type based on the energy distribution curve of the fluorescence signal and the diffraction angle pattern of the diffraction signal through a pre-trained classification model, obtain the sample type, and switch between vacuum or atmospheric environment according to the sample type.

[0009] The data processing module receives the fluorescence signal and diffraction signal transmitted by the acquisition module, performs energy channel calibration on the fluorescence signal, and performs angle and intensity normalization processing on the diffraction signal to generate a time-series correlated standardized dataset.

[0010] The intelligent analysis module calls the element-phase mapping rules in the calibration database, inputs the standardized dataset into the pre-trained random forest model, outputs the correlation between element content and mineral phase composition, and generates reservoir permeability and porosity prediction results by combining core scan data.

[0011] The dynamic strategy module generates a three-dimensional visualization report based on the correlation and reservoir parameter prediction results output by the intelligent analysis module, and adjusts the target selection or signal acquisition mode of the X-ray source in the acquisition module through mobile terminal interaction commands.

[0012] Furthermore, in the modular X-ray analysis device for oil and gas geology described in this invention, the acquisition module includes:

[0013] The time-division control unit is connected to the data processing module via a timing controller and is configured to alternately activate fluorescence signal acquisition and diffraction signal acquisition according to a preset timing instruction.

[0014] In the fluorescence signal acquisition stage, the time-division control unit drives the first target to excite the sample to generate an energy distribution curve, and transmits the energy distribution curve to the data processing module through the timing controller.

[0015] During the diffraction signal acquisition stage, the time-division control unit switches to the second target to generate a diffraction pattern, synchronously triggers the sample stage to rotate to eliminate crystal orientation deviation, and transmits the diffraction pattern to the data processing module through the timing controller.

[0016] Furthermore, in the modular X-ray analysis device for oil and gas geology described in this invention, the environment adaptation module includes:

[0017] The vacuum switching unit is connected to the time-division control unit in the acquisition module. It is used to start the vacuum environment according to the acquisition stage command of the time-division control unit when detecting light elements, and switch to atmospheric mode when detecting block samples. At the same time, it sends an environmental status signal to the data processing module to trigger timing calibration.

[0018] The anti-vibration unit is connected to the X-ray source and detector in the acquisition module through a vibration damping base. It monitors the external vibration intensity in real time. When the vibration intensity exceeds a preset threshold, dynamic damping is activated, and the temperature of the X-ray source and detector is controlled within the preset threshold through a heat dissipation system. The temperature control signal of the anti-vibration unit is synchronously transmitted to the data processing module to adjust the noise filtering parameters.

[0019] Furthermore, in the modular X-ray analysis device for oil and gas geology described in this invention, the intelligent analysis module includes:

[0020] The data fusion unit receives the standardized dataset generated by the data processing module and aligns the standardized dataset with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. The three-dimensional model serves as the input data for the pre-trained random forest model.

[0021] The verification unit, connected to the data fusion unit, receives elemental quantitative data from a third-party detection device. Based on the elemental distribution data and mineral phase ratio data in the three-dimensional model, it corrects the mineral phase ratio error output by the random forest model through an error comparison algorithm, and feeds back the corrected error value to the data fusion unit to update the spatial coordinate registration parameters.

[0022] Furthermore, in the modular X-ray analysis device for oil and gas geology described in this invention, the dynamic strategy module includes:

[0023] The mode switching unit is connected to the time-division control unit in the acquisition module and the intelligent analysis module. In response to the mobile terminal command, it adjusts the signal acquisition mode to time-division acquisition or energy filtering synchronous acquisition, and sends a mode switching command to the intelligent analysis module to update the input parameters of the pre-trained random forest model. The input parameters include the time-series correlation weights between the fluorescence signal energy distribution curve and the diffraction pattern acquired in the time-division mode.

[0024] The algorithm iteration unit, connected to the verification unit in the intelligent analysis module, updates the training parameters of the pre-trained random forest model using the gradient descent algorithm based on the mineral phase data in the regional geological database and the corrected error value output by the verification unit. This optimizes the nonlinear mapping rule between mineral phase ratio and element content, and synchronizes the updated model parameters to the intelligent analysis module.

[0025] Secondly, the modular X-ray analysis method for oil and gas geology provided by the present invention is applied to the modular X-ray analysis apparatus for oil and gas geology as described above, comprising:

[0026] Step 1: Acquire pre-processed sample data. Excite the sample with a preset multi-target X-ray source to generate X-ray fluorescence signal and X-ray diffraction signal in a time-division or synchronous manner. Based on the time-division control command, alternately acquire the energy spectrum distribution and diffraction pattern of the fluorescence signal.

[0027] Step 2: Based on the energy distribution curve of the fluorescence signal and the diffraction angle spectrum of the diffraction signal, identify the sample type through a pre-trained classification model, obtain the sample type, and switch between vacuum or atmospheric environment according to the sample type;

[0028] Step 3: Perform energy channel calibration on the collected fluorescence signal and normalize the angle and intensity of the diffraction signal to generate a time-series correlated standardized dataset.

[0029] Step 4: Align the standardized dataset with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. Input the three-dimensional model into a pre-trained random forest model to output the correlation between elemental content and mineral phase composition, as well as the prediction results of reservoir permeability and porosity.

[0030] Step 5: Generate a 3D visualization report based on the correlation, and adjust the target selection or signal acquisition mode of the X-ray source according to real-time feedback, and update the input parameters and training parameters of the random forest model simultaneously.

[0031] Furthermore, in the modular X-ray analysis method for oil and gas geology described in this invention, step 1 includes:

[0032] In time-sharing mode, the first target material is activated first to generate fluorescence signal and record full element energy spectrum according to preset timing instructions. After the acquisition is completed, the second target material is switched to generate diffraction signal through the timing controller. The angle of the goniometer is adjusted to the preset diffraction angle range and the sample stage is triggered to rotate to eliminate crystal orientation deviation.

[0033] In energy filtering mode, a monochromator is activated based on mobile terminal interaction commands to separate fluorescence and diffraction signals, and the intensity distribution of both is recorded simultaneously. The time sequence information of the fluorescence and diffraction signals is associated with timestamps.

[0034] Furthermore, in the modular X-ray analysis method for oil and gas geology described in this invention, step 2 includes:

[0035] Based on the fluorescence signal energy distribution curve acquired in step 1, the peak position shift caused by detector gain drift is eliminated by the energy channel calibration algorithm, and the calibrated energy spectrum is matched with the elemental characteristic peaks in the calibration database to determine the content range of the target element in the sample.

[0036] Based on the diffraction angle spectrum of the diffraction signal acquired in step 1, the diffraction signal is processed by the angle and intensity normalization algorithm, the lattice spacing is calculated by combining the preset Bragg diffraction equation, and the diffraction peaks are compared with those in the mineral phase standard database to identify the type and proportion of mineral phases in the sample.

[0037] Furthermore, in the modular X-ray analysis method for oil and gas geology described in this invention, step 4 includes:

[0038] The standardized dataset generated in step 3 is aligned with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. The three-dimensional model is then input into a pre-trained random forest model, and a correlation between elemental content thresholds and mineral phase ratios is established through nonlinear mapping rules.

[0039] The system receives elemental quantitative data from a third-party testing device, and based on the element-phase mapping rules in the calibration database, corrects the mineral phase ratio error output by the random forest model using an error comparison algorithm. The corrected error value is then fed back to the spatial coordinate registration algorithm to update the coordinate registration parameters of the three-dimensional model.

[0040] Furthermore, in the modular X-ray analysis method for oil and gas geology described in this invention, step 5 includes:

[0041] When allotropes or samples with the same elements but different phase structures are detected based on the mineral phase composition correlation output in step 4, the energy filtering mode is switched according to the activation command of the energy filtering mode. The resolution of the diffraction signal is improved by the monochromator, and the training parameters of the random forest model are updated by the gradient descent algorithm based on the corrected error value.

[0042] Based on the reservoir permeability and porosity prediction results generated in step 4, the target tube voltage and beam intensity in the time-division control command are dynamically adjusted to adapt to the detection requirements of different geological lithologies, and the time-series correlation weights are simultaneously optimized to match the real-time detection mode.

[0043] Beneficial effects of this invention;

[0044] The beneficial effects of this invention are reflected in the efficient acquisition and data fusion of XRF and XRD signals through the coordinated optimization of the time-division control unit and the dynamic strategy module. The time-division control unit alternately drives multiple targets to excite fluorescence and diffraction signals based on timing commands, and combines this with sample stage rotation to eliminate crystal orientation deviations, significantly shortening the detection cycle and improving data synchronization. The intelligent analysis module uses a spatial coordinate registration algorithm to align the standardized dataset with core scanning imaging data, constructing a three-dimensional model that integrates elemental distribution and mineral phases. It establishes a correlation rule between elemental content thresholds and mineral phase ratios using a random forest model, solving the problem of data separation between elemental composition and phase structure. The dynamic strategy module triggers an energy filtering mode based on the mineral phase identification results, using a monochromator to filter out non-target energy ranges to improve diffraction resolution. It iteratively optimizes model parameters using a gradient descent algorithm, enhancing the ability to identify allotropes and heterogeneous samples. Each module dynamically adjusts target parameters and acquisition strategies through a closed-loop feedback mechanism, achieving integrated analysis of oil and gas reservoir characteristics while maintaining detection efficiency. Attached Figure Description

[0045] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0046] Figure 1 A flowchart of a modular X-ray analysis device and method for oil and gas geology provided in an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The technical solutions provided by various embodiments of this invention will be described in detail below with reference to the accompanying drawings. To better understand the objectives of this invention, it will be described in further detail below.

[0048] In a first aspect, the present invention provides a modular X-ray analysis device for oil and gas geology, comprising:

[0049] The acquisition module is used to trigger the sample to generate X-ray fluorescence signal and X-ray diffraction signal through a preset multi-target X-ray source, simultaneously acquire the energy and intensity distribution curve of the fluorescence signal and the diffraction angle and intensity spectrum of the diffraction signal, and transmit the acquired signals to the data processing module.

[0050] An environment adaptation module, connected to the acquisition module, is used to identify the sample type based on the energy distribution curve of the fluorescence signal and the diffraction angle pattern of the diffraction signal through a pre-trained classification model, obtain the sample type, and switch between vacuum or atmospheric environment according to the sample type.

[0051] The data processing module receives the fluorescence signal and diffraction signal transmitted by the acquisition module, performs energy channel calibration on the fluorescence signal, and performs angle and intensity normalization processing on the diffraction signal to generate a time-series correlated standardized dataset.

[0052] The intelligent analysis module calls the element-phase mapping rules in the calibration database, inputs the standardized dataset into the pre-trained random forest model, outputs the correlation between element content and mineral phase composition, and generates reservoir permeability and porosity prediction results by combining core scan data.

[0053] The dynamic strategy module generates a three-dimensional visualization report based on the correlation and reservoir parameter prediction results output by the intelligent analysis module, and adjusts the target selection or signal acquisition mode of the X-ray source in the acquisition module through mobile terminal interaction commands.

[0054] This invention provides a modular X-ray analysis device for oil and gas geology, which achieves the functions and logical connections of each module through the following technical solutions:

[0055] The acquisition module triggers the generation of X-ray fluorescence and X-ray diffraction signals from the sample using a multi-target X-ray source. Specifically, it selects a Cu / Mo dual-target as the excitation source and dynamically adjusts the tube voltage and beam current intensity according to the sample characteristics. The fluorescence signal captures the full elemental energy spectrum (1-50 keV) using a silicon drift detector, while the diffraction signal records the diffraction angle (5°-80°) and intensity distribution using a two-dimensional array detector. The acquisition module has a built-in time-division control unit that alternately activates the fluorescence and diffraction signal acquisition process through a timing controller. During the fluorescence acquisition stage, the Mo target is prioritized to generate high-energy X-rays. After signal recording is completed, the module switches to the Cu target to excite the diffraction signal, simultaneously triggering the sample stage rotation to eliminate crystal orientation deviations. The acquired raw signals undergo preliminary noise reduction by an independent processing unit before being transmitted to the data processing module.

[0056] The environmental adaptation module identifies sample types using a pre-trained classification model trained on a joint XRF / XRD dataset of NIST standard materials and mineral standards. It extracts fluorescence peak position features and diffraction pattern angular distribution features via a convolutional neural network, outputting sample type classification results (e.g., powder, bulk, or film). Based on the classification results, the vacuum switching unit activates a vacuum environment to reduce air scattering interference during light element detection and switches to atmospheric mode for bulk sample detection to meet rapid loading requirements. The environmental adaptation module monitors external vibration intensity in real time. When vibration exceeds a preset threshold, it activates a dynamic damping system and uses a heat dissipation unit to control the temperature of the X-ray source and detector below 40°C. The temperature control signal is synchronously transmitted to the data processing module to adjust noise filtering parameters.

[0057] The data processing module calibrates the energy channels of the received fluorescence signals, correcting energy deviations caused by detector gain drift by adjusting the characteristic peak positions in the calibration database to ensure precise matching between the fluorescence energy spectrum and elemental composition. The diffraction signals are normalized for angle and intensity, and the lattice spacing is calculated based on the Bragg equation to eliminate the influence of sample surface roughness on diffraction intensity. The calibrated fluorescence data and normalized diffraction data are then correlated with timestamps to generate a standardized dataset, providing time-consistent input for subsequent analysis.

[0058] The intelligent analysis module invokes element-phase mapping rules from the calibration database to align the standardized dataset with core scan imaging data using a spatial coordinate registration algorithm, generating a 3D model that includes elemental distribution and mineral phases. This model is then fed into a pre-trained random forest model, which establishes a correlation between elemental content (e.g., Fe=15%) and mineral phase ratios (e.g., hematite:pyrite=3:1) through nonlinear mapping rules. The validation unit incorporates third-party detection data (e.g., SEM-EDS elemental quantification results), corrects the mineral phase ratio errors output by the model using an error comparison algorithm, and feeds the correction results back to the coordinate registration algorithm to optimize the accuracy of the 3D model.

[0059] The dynamic strategy module generates a visual report of superimposed elemental heatmaps and three-dimensional mineral phase distribution based on the reservoir permeability and porosity predictions output by the intelligent analysis module. The acquisition mode is adjusted in real-time via mobile interactive commands: in time-sharing mode, the temporal correlation weights are optimized to balance XRF / XRD detection efficiency; in energy filtering mode, the monochromator is activated to improve diffraction signal resolution. The algorithm iteration unit updates the training parameters of the random forest model based on the regional geological database and optimizes the nonlinear mapping rules between mineral phases and elemental content using the gradient descent algorithm, forming a closed-loop optimized reservoir characteristic analysis system.

[0060] Data is transmitted between modules via a high-speed serial bus (LVDS) and gigabit Ethernet. The timing controller ensures the synchronization of time-division acquisition and signal processing. Sample identification, environmental adaptation, data processing, and intelligent analysis form a series link. The dynamic strategy module optimizes the acquisition parameters and analysis model through a feedback mechanism, enabling efficient analysis of oil and gas geological samples and dynamic prediction of reservoir characteristics.

[0061] Specifically, the modular X-ray analysis device for oil and gas geology described in this invention includes an acquisition module comprising:

[0062] The time-division control unit is connected to the data processing module via a timing controller and is configured to alternately activate fluorescence signal acquisition and diffraction signal acquisition according to a preset timing instruction.

[0063] In the fluorescence signal acquisition stage, the time-division control unit drives the first target to excite the sample to generate an energy distribution curve, and transmits the energy distribution curve to the data processing module through the timing controller.

[0064] During the diffraction signal acquisition stage, the time-division control unit switches to the second target to generate a diffraction pattern, synchronously triggers the sample stage to rotate to eliminate crystal orientation deviation, and transmits the diffraction pattern to the data processing module through the timing controller.

[0065] The time-division control unit in the acquisition module of this invention is connected to the data processing module via a timing controller to achieve alternating acquisition of fluorescence and diffraction signals. Based on preset timing commands, the time-division control unit drives the first target material (such as a Mo target) to excite the sample during the fluorescence signal acquisition stage, generating an energy distribution curve. The Mo target operates in high-energy X-ray mode, exciting the characteristic fluorescence spectrum in the sample, and the silicon drift detector captures the energy distribution signal with a microsecond-level response time. The energy distribution curve is timestamped by the timing controller and synchronized with subsequent data processing stages, then transmitted to the data processing module for energy channel calibration.

[0066] During the diffraction signal acquisition stage, the time-division control unit switches to a second target (e.g., a Cu target) to excite the sample's diffraction signal using low-energy X-rays. The switching process is accomplished by an electromagnetic drive mechanism, with the target switching time controlled in milliseconds to avoid signal interference. A synchronously triggered sample stage rotation device, driven by a stepper motor, rotates the sample stage continuously in preset angular steps (e.g., 0.1°) to eliminate the influence of crystal orientation deviations on the diffraction pattern. A two-dimensional array detector records the diffraction intensity distribution at different rotation angles, generating spectral data containing the relationship between diffraction angle and intensity. The diffraction pattern is correlated with the timestamp of the fluorescence signal via a timing controller to ensure timing consistency before being transmitted to the data processing module.

[0067] The timing instructions for the time-division control unit are generated by the FPGA controller. Preset instructions include the fluorescence signal acquisition duration (e.g., 100 ms), diffraction signal acquisition interval (e.g., 200 ms), and the sample stage rotation angle sequence. The FPGA controller receives signal quality assessment results from the data processing module via Gigabit Ethernet and dynamically adjusts timing parameters to optimize acquisition efficiency. For example, when the fluorescence signal intensity is below a threshold, the Mo target excitation time is extended or the tube current is increased to ensure the signal-to-noise ratio of the energy distribution curve.

[0068] The timing controller acts as a data relay node, initially encapsulating the fluorescence signal energy distribution curve and diffraction pattern, adding timestamps and environmental status tags (such as vacuum level and temperature). The encapsulated data is transmitted to the data processing module via a high-speed serial bus, using the LVDS standard to reduce transmission latency. The synchronization between the sample stage rotation angle and diffraction signal acquisition is calibrated in real-time by a photoelectric encoder. The encoder signal is fed back to the time-division control unit to dynamically correct rotation step size deviations, ensuring the accuracy of diffraction angle measurements.

[0069] The interaction logic between the time-division control unit and the data processing module forms a closed-loop control link. After completing a single time-division acquisition, the data processing module sends a new round of timing commands to the time-division control unit based on the signal integrity analysis results, realizing adaptive acquisition mode switching. For example, when a high content of light elements is detected, the fluorescence signal acquisition cycle is shortened and the vacuum environment is activated first, improving the identification accuracy of light element characteristic peaks. This time-division control mechanism maximizes hardware resource utilization while ensuring the synergy between XRF and XRD data, meeting the high-throughput detection requirements of oil and gas geological samples.

[0070] Specifically, the modular X-ray analysis device for oil and gas geology described in this invention includes an environment adaptation module comprising:

[0071] The vacuum switching unit is connected to the time-division control unit in the acquisition module. It is used to start the vacuum environment according to the acquisition stage command of the time-division control unit when detecting light elements, and switch to atmospheric mode when detecting block samples. At the same time, it sends an environmental status signal to the data processing module to trigger timing calibration.

[0072] The anti-vibration unit is connected to the X-ray source and detector in the acquisition module through a vibration damping base. It monitors the external vibration intensity in real time. When the vibration intensity exceeds a preset threshold, dynamic damping is activated, and the temperature of the X-ray source and detector is controlled within the preset threshold through a heat dissipation system. The temperature control signal of the anti-vibration unit is synchronously transmitted to the data processing module to adjust the noise filtering parameters.

[0073] The vacuum switching unit in the environment adaptation module of this invention is connected to the time-division control unit and dynamically adjusts the detection environment according to the acquisition phase instructions. When the time-division control unit activates the fluorescence signal acquisition mode, the vacuum switching unit determines whether the sample type identification requires light element detection: if it is identified as a light element (such as C, O), the vacuum pump is activated to reduce the gas pressure in the detection chamber to a minimum. Below Pa, the scattering interference of air molecules on low-energy X-rays is eliminated; if the sample is identified as a blocky rock core, the system switches to atmospheric mode and activates the rapid loading mechanism of the sample chamber door, monitoring the chamber sealing status in real time via a pressure sensor. Environmental status signals (including vacuum level, pressure value, and chamber door status) are transmitted to the data processing module via the CAN bus, triggering a timing calibration program to adjust the signal acquisition interval and match the detector response time under different environments.

[0074] The seismic-resistant unit is rigidly connected to the X-ray source and detector via a vibration-damping base. The base incorporates a piezoelectric vibration sensor to collect triaxial vibration data in real time. When the vibration intensity exceeds a preset threshold (e.g., 0.5 g), the dynamic damping system adjusts the damping coefficient using magnetorheological fluid, attenuating the vibration amplitude to a safe range within 10 ms. The temperature of the X-ray source and detector is monitored in real time by thermocouples. The cooling system employs a combination of semiconductor cooling and air cooling to maintain the temperature below 40°C. The temperature control signal is converted into a digital value by an ADC module and transmitted to the data processing module. A noise filtering algorithm dynamically adjusts the filter cutoff frequency based on the temperature data to suppress the impact of thermal noise on the fluorescence spectrum baseline.

[0075] Data from the vacuum switching unit and the vibration-resistant unit are synchronously exchanged with the time-sharing control unit via a high-speed serial interface. For example, during the vacuum environment startup phase, the door closing operation is automatically delayed when vibration data exceeds a threshold to prevent mechanical impact from causing seal failure. Under high-temperature conditions, the cooling system prioritizes the activation of semiconductor cooling mode, while the time-sharing control unit extends the signal acquisition interval to reduce the load on the X-ray source. Environmental status signals and vibration temperature data are fused in the data processing module to generate an environmental quality assessment index. This index inversely controls the timing command cycle of the time-sharing control unit, forming a balance mechanism between the stability of the detection environment and data acquisition efficiency.

[0076] The collaborative workflow of each unit is as follows: After the time-sharing control unit sends the acquisition phase command, the vacuum switching unit switches the environment according to the sample type, and the vibration suppression unit simultaneously starts vibration suppression and temperature control; the environmental status signal triggers the data processing module to load the corresponding calibration parameter library (such as the energy spectrum gain correction coefficient in vacuum mode), and the vibration temperature data optimizes the noise filtering algorithm in real time; after a single detection is completed, the environmental quality assessment index updates the timing strategy of the time-sharing control unit, for example, shortening the single acquisition time and increasing the heat dissipation cycle in high temperature and high vibration scenarios. This invention maintains the stability of the X-ray source and detector under complex working conditions, ensuring the accuracy of elemental and phase analysis data.

[0077] Specifically, the modular X-ray analysis device for oil and gas geology described in this invention includes an intelligent analysis module comprising:

[0078] The data fusion unit receives the standardized dataset generated by the data processing module and aligns the standardized dataset with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. The three-dimensional model serves as the input data for the pre-trained random forest model.

[0079] The verification unit, connected to the data fusion unit, receives elemental quantitative data from a third-party detection device. Based on the elemental distribution data and mineral phase ratio data in the three-dimensional model, it corrects the mineral phase ratio error output by the random forest model through an error comparison algorithm, and feeds back the corrected error value to the data fusion unit to update the spatial coordinate registration parameters.

[0080] The data fusion unit in the intelligent analysis module of this invention receives a standardized dataset generated by the data processing module and aligns it with core scan imaging data using a spatial coordinate registration algorithm. The standardized dataset includes energy channel calibration results for fluorescence signals and angle and intensity normalized data for diffraction signals. The core scan imaging data is obtained as a high-resolution three-dimensional voxel model using a CT scanner. The spatial coordinate registration algorithm employs the Iterative Closest Point (ICP) method, using the distribution of characteristic elements (such as Fe and Si) in the fluorescence signal as reference points to map the three-dimensional coordinates of the core scan data to the spatial coordinate system of the standardized dataset, generating a three-dimensional model that integrates elemental distribution and mineral phase information. This model labels the elemental content percentage and corresponding mineral phase category (such as hematite and quartz) in voxel units, serving as the input data structure for a pre-trained random forest model.

[0081] The verification unit receives elemental quantitative data transmitted from third-party testing equipment (such as SEM-EDS) via an RS-485 interface and extracts the measured values ​​of elemental content in the target area. Based on the elemental distribution data and mineral phase proportion data of the corresponding area in the 3D model, the error comparison algorithm calculates the mean square error between the model's predicted values ​​and the measured values, identifying areas where the predicted mineral phase proportion deviation exceeds a threshold (e.g., ±5%). The correction process uses the least squares method to optimize the weight parameters of the random forest model and adjust the mineral phase classification decision boundary. The corrected error value is fed back to the data fusion unit through a backpropagation mechanism to update the registration weight coefficients in the spatial coordinate registration algorithm, for example, increasing the coordinate matching accuracy weight of high-error areas to optimize the data alignment effect in the next round of 3D model generation.

[0082] The data fusion unit and the validation unit form a closed-loop optimization chain. In a single analysis process, after the data fusion unit generates an initial 3D model, the random forest model outputs the predicted mineral phase composition. The validation unit introduces external measured data to perform error correction, and the correction results improve the accuracy of subsequent data fusion by updating the registration parameters. For example, when there is a significant difference between the predicted proportion of quartz phase in a certain area and the measured value by EDS, adjusting the registration weight coefficient will strengthen the spatial matching intensity between the core scan data and the peak position of the fluorescence spectrum in that area, correcting the coordinate offset caused by the unevenness of the sample surface.

[0083] Data interaction between units is achieved via a high-speed data bus, and the voxel data and error correction parameters of the 3D model adopt a hierarchical transmission protocol. The 3D model output by the data fusion unit is stored in a tensor structure, including the elemental content matrix, mineral phase classification matrix, and coordinate transformation parameters, and undergoes feature normalization processing before being input into the random forest model. The correction parameters of the validation unit are fed back to the data fusion unit through an independent channel to avoid conflicts with the main data stream. This invention maintains high-throughput data processing efficiency while improving the accuracy of mineral phase analysis through a dynamic parameter optimization mechanism, meeting the analytical needs of complex mineral compositions in oil and gas reservoirs.

[0084] Specifically, the modular X-ray analysis device for oil and gas geology described in this invention includes a dynamic strategy module comprising:

[0085] The mode switching unit is connected to the time-division control unit in the acquisition module and the intelligent analysis module. In response to the mobile terminal command, it adjusts the signal acquisition mode to time-division acquisition or energy filtering synchronous acquisition, and sends a mode switching command to the intelligent analysis module to update the input parameters of the pre-trained random forest model. The input parameters include the time-series correlation weights between the fluorescence signal energy distribution curve and the diffraction pattern acquired in the time-division mode.

[0086] The algorithm iteration unit, connected to the verification unit in the intelligent analysis module, updates the training parameters of the pre-trained random forest model using the gradient descent algorithm based on the mineral phase data in the regional geological database and the corrected error value output by the verification unit. This optimizes the nonlinear mapping rule between mineral phase ratio and element content, and synchronizes the updated model parameters to the intelligent analysis module.

[0087] The mode switching unit in the dynamic strategy module of this invention is connected to the time-sharing control unit and the intelligent analysis module to achieve dynamic adaptation between the acquisition mode and model parameters. The mode switching unit receives mobile terminal interaction commands via the MQTT protocol and parses the mode switching requests (such as time-sharing acquisition or energy filtering synchronous acquisition) in the commands. In time-sharing acquisition mode, the mode switching command triggers the time-sharing control unit to adjust the temporal correlation weights, for example, increasing the proportion of fluorescence signal acquisition cycles from 60% to 80%, prioritizing the acquisition of high signal-to-noise ratio element distribution data; in energy filtering synchronous acquisition mode, the monochromator is activated to filter out non-target energy range X-rays, and the intensity distribution of fluorescence and diffraction signals is recorded synchronously, with the temporal correlation weights adjusted to 1:1 to balance data integrity. The updated input parameters (including temporal weights and energy filtering thresholds) are transmitted to the intelligent analysis module via a high-speed data bus. After loading the new parameters, the pre-trained random forest model adjusts the weights of the input layer nodes to optimize the real-time response speed of mineral phase classification.

[0088] The algorithm iteration unit accesses the regional geological database via an ODBC interface to extract typical mineral facies distribution data for the target area (e.g., the quartz / clay facies ratio range in shale). Combined with the corrected error values ​​output by the validation unit (e.g., hematite prediction error ±3%), the gradient descent algorithm iteratively updates the connection weights of the hidden layer neurons in the random forest model, aiming to minimize the loss function. For example, when the clay facies prediction error in a certain area remains consistently high, the algorithm increases the feature weights of corresponding elements (e.g., Al, Si), weakens the influence coefficient of interfering elements (e.g., Fe), and optimizes the nonlinear mapping rule between mineral facies ratio and element content. The updated model parameters are encrypted and verified before being synchronized to the intelligent analysis module, replacing the original parameter file. The synchronization process uses a version control mechanism to prevent data conflicts and maintain the continuity of model iteration.

[0089] The mode switching unit and the algorithm iteration unit work collaboratively via a data bus. After a mode switch, the algorithm iteration unit dynamically adjusts the learning rate and batch size of the gradient descent algorithm based on the new input parameter structure (such as the temporal correlation weights in the energy filtering mode) to adapt to the differences in data characteristics under different acquisition modes. For example, high-frequency signals in the time-division acquisition mode require a smaller learning rate to avoid overfitting, while the full-spectrum data in the energy filtering mode uses a larger batch to improve convergence efficiency. The mineral facies data from the regional geological database and the error data from the validation unit are weighted and fused during the iteration process to form a regionally adaptive training set, enhancing the model's generalization ability to specific oil and gas reservoirs.

[0090] The interaction between the dynamic strategy module and the intelligent analysis module forms a bidirectional optimization link. The reservoir parameter prediction results output by the intelligent analysis module are fed back into the dynamic strategy module, driving the mode switching unit to optimize the acquisition strategy. For example, when the prediction results indicate high-porosity reservoir characteristics, the mode switching unit automatically increases the acquisition frequency of the energy filtering mode, improving the identification accuracy of clay mineral phases; the algorithm iteration unit simultaneously calls the mineral phase data of the corresponding region, strengthening the correlation rules between porosity and mineral phase ratios. This closed-loop optimization mechanism maintains hardware acquisition efficiency while improving the analytical accuracy of oil and gas geological samples through dynamic iteration of model parameters, meeting the real-time analysis needs of complex reservoir characteristics.

[0091] Secondly, please refer to Figure 1 The modular X-ray analysis method for oil and gas geology provided by this invention is applied to the modular X-ray analysis apparatus for oil and gas geology as described above, comprising:

[0092] Step 1: Acquire pre-processed sample data. Excite the sample with a preset multi-target X-ray source to generate X-ray fluorescence signal and X-ray diffraction signal in a time-division or synchronous manner. Based on the time-division control command, alternately acquire the energy spectrum distribution and diffraction pattern of the fluorescence signal.

[0093] Step 2: Based on the energy distribution curve of the fluorescence signal and the diffraction angle spectrum of the diffraction signal, identify the sample type through a pre-trained classification model, obtain the sample type, and switch between vacuum or atmospheric environment according to the sample type;

[0094] Step 3: Perform energy channel calibration on the collected fluorescence signal and normalize the angle and intensity of the diffraction signal to generate a time-series correlated standardized dataset.

[0095] Step 4: Align the standardized dataset with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. Input the three-dimensional model into a pre-trained random forest model to output the correlation between elemental content and mineral phase composition, as well as the prediction results of reservoir permeability and porosity.

[0096] Step 5: Generate a 3D visualization report based on the correlation, and adjust the target selection or signal acquisition mode of the X-ray source according to real-time feedback, and update the input parameters and training parameters of the random forest model simultaneously.

[0097] The modular X-ray analysis method for oil and gas geology provided by this invention achieves logical connection and functional synergy among its various steps through the following technical solutions:

[0098] In step 1, the multi-target X-ray source is driven by a time-division control unit. In time-division mode, the Mo target is preferentially activated to generate high-energy X-rays that excite the sample fluorescence signal. The silicon drift detector acquires the full elemental energy spectrum with a microsecond-level response speed. After the fluorescence signal acquisition is completed, the time-division control unit switches to the Cu target to excite low-energy X-rays, synchronously triggering the sample stage stepper motor to rotate at a preset angle step size to eliminate crystal orientation deviation. The two-dimensional array detector records the diffraction intensity distribution at different rotation angles. In energy filtering synchronization mode, the monochromator filters out X-rays in non-target energy ranges, synchronously acquires the intensity distribution of fluorescence and diffraction signals, and associates the time sequence information through timestamps.

[0099] In step 2, the pre-trained classification model is constructed based on the joint XRF / XRD dataset of NIST standard materials and mineral standards. It extracts the characteristic peak positions of fluorescence spectra and the angular distribution features of diffraction patterns using a convolutional neural network, outputting the sample type classification result (e.g., powder, block, or film). The classification result triggers the vacuum switching unit: during light element detection, the vacuum pump is activated to reduce the chamber pressure to [a certain value]. Below Pa, air scattering interference is reduced; when testing block samples, switch to atmospheric mode and activate the rapid loading mechanism. Environmental status signals are transmitted to the data processing module via the CAN bus, triggering a timing calibration program to adjust the signal acquisition interval.

[0100] In step 3, the energy channel calibration of the fluorescence signal corrects detector gain drift by adjusting the characteristic peak positions in the calibration database. For example, the Fe-Kα peak position is corrected to 6.40 keV to match the element's characteristic energy range. The angle and intensity normalization of the diffraction signal are calculated based on the Bragg equation to eliminate the influence of sample surface roughness on the intensity distribution, generating a standardized dataset containing timestamps. Timing correlation is achieved using a unified clock source, controlling the time deviation between fluorescence and diffraction data to within milliseconds.

[0101] In step 4, the spatial coordinate registration algorithm uses the distribution of characteristic elements (such as Fe and Si) in the fluorescence signal as the reference point to map the three-dimensional coordinates of the core scan CT data to the standardized dataset coordinate system. The registered three-dimensional model is labeled with the percentage of element content and mineral phase category (e.g., hematite accounts for 30%) in voxel units and input into the pre-trained random forest model. The model establishes the correlation between element content thresholds (e.g., Ca>5%) and mineral phase proportions (calcite phase ≥15%) through nonlinear decision tree rules, and outputs the predicted values ​​of reservoir permeability and porosity by combining the core pore structure data.

[0102] In step 5, the 3D visualization report overlays an elemental distribution heatmap and a 3D model of mineral phases, marking areas of permeability anomalies and pore development zones. A real-time feedback mechanism dynamically adjusts target selection and acquisition modes based on user interaction commands or model prediction results: optimizing the Mo / Cu target excitation duration ratio in time-sharing mode; adjusting the monochromator filtering range in energy filtering mode. The input parameters (such as temporal correlation weights) and training parameters (such as the number of hidden layer nodes) of the random forest model are iteratively updated using a gradient descent algorithm. The updated parameters are synchronized to the intelligent analysis module via a version control mechanism, maintaining the timeliness and regional adaptability of the model predictions.

[0103] Signals and parameters are transmitted between each step via a high-speed data bus. The interaction between the time-division control unit and the data processing module ensures synchronization between acquisition timing and signal processing. Sample type identification, environmental adaptation, data calibration, and model analysis form a series link. The dynamic strategy module optimizes the acquisition parameters and model structure based on feedback data, achieving integrated and efficient analysis of oil and gas geological sample elements, phases, and reservoir parameters.

[0104] Specifically, the modular X-ray analysis method for oil and gas geology described in this invention includes step 1 as follows:

[0105] In time-sharing mode, the first target material is activated first to generate fluorescence signal and record full element energy spectrum according to preset timing instructions. After the acquisition is completed, the second target material is switched to generate diffraction signal through the timing controller. The angle of the goniometer is adjusted to the preset diffraction angle range and the sample stage is triggered to rotate to eliminate crystal orientation deviation.

[0106] In energy filtering mode, a monochromator is activated based on mobile terminal interaction commands to separate fluorescence and diffraction signals, and the intensity distribution of both is recorded simultaneously. The time sequence information of the fluorescence and diffraction signals is associated with timestamps.

[0107] In the time-division mode of step 1 of this invention, fluorescence and diffraction signals are acquired in a time-division manner through a timing controller. The timing controller generates preset timing instructions based on an FPGA. During the fluorescence signal acquisition phase, it drives the first target (e.g., a Mo target) to excite the sample with high-energy X-rays, and the silicon drift detector captures the full-element energy spectrum with a microsecond-level response speed. After preliminary noise reduction, the energy spectrum data is transmitted to the data processing module via a high-speed serial bus. After fluorescence acquisition is completed, the timing controller sends a switching instruction to the electromagnetic drive mechanism, switching to the second target (e.g., a Cu target) to excite low-energy X-rays. Simultaneously, the goniometer stepper motor is triggered to adjust the angle to a preset diffraction angle range (e.g., 20°-80°). The sample stage is driven by a servo motor to rotate in 0.1° steps, eliminating the influence of crystal orientation deviation on the diffraction pattern.

[0108] In energy filtering mode, mobile terminal interaction commands are transmitted to the monochromator control unit via the MQTT protocol, activating the monochromator to filter out non-target energy range X-rays. The monochromator separates the fluorescence signal (5-20 keV) and the diffraction signal (Cu-Kα 8.04 keV) based on the multilayer film reflection principle. In synchronous acquisition mode, a two-dimensional array detector records the intensity distribution of both. The timing controller generates a timestamp using a GPS-synchronized clock, controlling the acquisition time deviation between the fluorescence signal energy spectrum and the diffraction angle-intensity map within milliseconds, ensuring the consistency of timing information correlation.

[0109] The switching logic between time-sharing mode and energy filtering mode is driven by the sample type identification result. When the classification model identifies the sample as a light element or complex mineral phase, it automatically switches to energy filtering mode; for homogeneous blocky samples, time-sharing mode is preferred to improve detection efficiency. The timing controller dynamically adjusts the signal acquisition period during mode switching: the fluorescence and diffraction acquisition interval is 200 ms in time-sharing mode, while the synchronous acquisition period is shortened to 50 ms in energy filtering mode. The coordination between the goniometer angle adjustment and the sample stage rotation is calibrated in real time through a photoelectric encoder. The encoder signal is converted by the ADC module and then input to the timing controller to correct the diffraction angle deviation caused by mechanical transmission errors.

[0110] Both data acquisition modes interact with the data processing module via a high-speed data bus. The time-series correlation weights (e.g., fluorescence:diffraction = 3:1) for time-division mode and the timestamp parameters for energy filtering mode are stored in non-volatile memory for subsequent standardization processing. This invention maintains hardware control precision while adapting to the analytical needs of different oil and gas geological samples through a dynamic mode switching mechanism, thereby improving the efficiency of collaborative acquisition of elemental and phase data.

[0111] Specifically, in the modular X-ray analysis method for oil and gas geology described in this invention, step 2 includes:

[0112] Based on the fluorescence signal energy distribution curve acquired in step 1, the peak position shift caused by detector gain drift is eliminated by the energy channel calibration algorithm, and the calibrated energy spectrum is matched with the elemental characteristic peaks in the calibration database to determine the content range of the target element in the sample.

[0113] Based on the diffraction angle spectrum of the diffraction signal acquired in step 1, the diffraction signal is processed by the angle and intensity normalization algorithm, the lattice spacing is calculated by combining the preset Bragg diffraction equation, and the diffraction peaks are compared with those in the mineral phase standard database to identify the type and proportion of mineral phases in the sample.

[0114] Step 2 of this invention improves the accuracy of elemental and mineral phase analysis through energy channel calibration and diffraction signal normalization. The specific implementation process is as follows:

[0115] Based on the fluorescence signal energy distribution curve acquired in step 1, the energy channel calibration algorithm eliminates peak position shifts caused by detector gain drift by using reference peak positions in the calibration database (e.g., Fe-Kα=6.40 keV, Cu-Kα=8.04 keV). The calibration database stores the nominal energy values ​​and allowable deviation ranges (±0.05 keV) of elemental characteristic peaks. The calibration algorithm uses the least squares method to fit the linear relationship between the actual peak position and the nominal value, generating a gain correction coefficient matrix. The calibrated energy spectrum is compared with the characteristic peaks in the calibration database using a peak matching algorithm. During the matching process, a signal-to-noise ratio threshold (e.g., SNR>10) is introduced to filter out noise interference peaks, determine the content range of target elements (e.g., Fe, Si, Ca) in the sample (e.g., Fe=12-18%), and calculate the confidence interval using a probability distribution model.

[0116] For the diffraction angle pattern of the diffraction signal acquired in step 1, the angle and intensity normalization algorithm first eliminates the intensity deviation caused by sample surface roughness and detector response inhomogeneity. The normalization process is based on the Bragg diffraction equation, using the Cu-Kα ray wavelength (λ=0.154 nm) as a reference, to calculate the lattice spacing (d=λ / (2sinθ)) corresponding to each diffraction angle, generating a standardized d-intensity distribution curve. The mineral phase standard sample database stores the diffraction peak positions and half-peak width parameters of typical minerals (such as quartz d=3.34 Å, calcite d=3.03 Å). The measured d-values ​​are matched with the standard peak positions in the database using a multi-peak fitting algorithm. The matching results are combined with the peak area integration method to calculate the mineral phase ratio (e.g., quartz accounts for 35%, clay minerals account for 50%).

[0117] Data from energy channel calibration and diffraction signal normalization are correlated using timestamps. The calibrated elemental content range serves as a constraint for mineral phase analysis, inputting into a multi-peak fitting algorithm. For example, when the Fe content exceeds 15%, hematite diffraction peaks are preferentially matched over magnetite. The mineral phase proportion data output by the normalization algorithm is used to inversely optimize the characteristic peak matching weights of the calibration database. For instance, the matching tolerance for high-proportion mineral phases is tightened from ±0.5% to ±0.2%, improving the detection sensitivity of low-content mineral phases.

[0118] The collaborative processing of the two parts forms a cross-validation mechanism: the elemental content range after fluorescence signal calibration constrains the mineral phase identification results of the diffraction signal, while the mineral phase proportion data inversely corrects the weighting coefficients of the elemental quantitative model. For example, when diffraction analysis identifies a high proportion of pyrite, the Fe content threshold in the fluorescence signal is automatically raised to eliminate iron oxide interference. This invention improves the analytical accuracy of complex mineral compositions through data cross-validation, meeting the technical requirements of elemental and phase coupling analysis in oil and gas reservoirs.

[0119] Specifically, in the modular X-ray analysis method for oil and gas geology described in this invention, step 4 includes:

[0120] The standardized dataset generated in step 3 is aligned with the core scanning imaging data using a spatial coordinate registration algorithm to generate a three-dimensional model containing elemental distribution and mineral phases. The three-dimensional model is then input into a pre-trained random forest model, and a correlation between elemental content thresholds and mineral phase ratios is established through nonlinear mapping rules.

[0121] The system receives elemental quantitative data from a third-party testing device, and based on the element-phase mapping rules in the calibration database, corrects the mineral phase ratio error output by the random forest model using an error comparison algorithm. The corrected error value is then fed back to the spatial coordinate registration algorithm to update the coordinate registration parameters of the three-dimensional model.

[0122] Step 4 of this invention achieves multi-dimensional correlation analysis of element-mineral phase-reservoir parameters through spatial coordinate registration and model iterative optimization. The specific implementation process is as follows:

[0123] The standardized dataset includes time-series correlated fluorescence spectroscopy calibration data and diffraction signal normalization results. The spatial coordinate registration algorithm uses the distribution of characteristic elements (such as Fe and Si) in the fluorescence signal as spatial reference points, and maps the voxel coordinates of the core scan CT imaging data to the three-dimensional coordinate system of the standardized dataset using the Iterative Closest Point (ICP) algorithm. A feature point cloud matching strategy is introduced during the registration process, using regions with gradient changes in elemental content (such as a sudden change in Fe content from 10% to 20%) as key registration points to generate a three-dimensional model that integrates elemental distribution and mineral phase information. This model labels the elemental content percentage (such as Fe = 15% ± 0.5%) and mineral phase classification labels (such as hematite and quartz) on a voxel basis. Before being input into a pre-trained random forest model, feature normalization is performed to eliminate the influence of different dimensions on classification decisions.

[0124] The random forest model establishes a correlation between element content thresholds and mineral phase proportions through nonlinear decision tree rules. For example, when the Fe content in a voxel exceeds 12% and the diffraction d-value matches 3.67 Å, ​​the decision tree classifies the mineral phase as hematite; if the Si content is simultaneously detected to be higher than 20%, a quartz phase weighting coefficient is added. Reservoir permeability and porosity predictions are based on regression models using mineral phase proportions and core pore structure data; for example, a low permeability warning is triggered when the clay mineral proportion exceeds 40%.

[0125] The verification unit receives elemental quantitative data transmitted from third-party testing equipment (such as a laser-induced breakdown spectrometer) via an RS-485 interface and extracts the measured elemental content values ​​of the target area. An error comparison algorithm calculates the mean square error between the mineral phase ratio predicted by the random forest model and the measured values, identifying voxel regions with deviations exceeding a preset threshold (e.g., ±5%). The correction process uses weighted least squares to optimize model parameters and adjust the feature weights of decision tree split nodes. The corrected error values ​​are fed back to the spatial coordinate registration algorithm via backpropagation, dynamically updating the registration weight matrix. For example, the feature point cloud density is increased in high-error areas (from 10 points per cubic millimeter to 20 points per cubic millimeter) to optimize the spatial alignment accuracy of subsequent 3D models.

[0126] Data fusion and model calibration form a closed-loop iterative chain. In a single analysis process, after the initial 3D model is generated, the random forest model outputs mineral facies prediction results; the validation unit introduces external measured data to correct model errors, and the calibration parameters are adjusted through registration weights to improve the accuracy of the next round of data fusion. For example, when the predicted proportion of quartz facies in a certain area differs significantly from the measured value, the registration algorithm strengthens the matching weight between the CT imaging porosity characteristics and fluorescence spectra of that area, correcting the coordinate offset caused by the unevenness of the core surface. The 3D model version management mechanism records the parameter changes of each iteration, supports historical data analysis and model backtesting verification, and maintains the stability of reservoir parameter predictions.

[0127] Data interaction between steps is achieved through a hierarchical transmission protocol. The voxel data of the 3D model is stored in a tensor structure, containing an element matrix, a mineral phase classification matrix, and coordinate transformation parameters. Principal component analysis is performed for dimensionality reduction before inputting it into the random forest model. Error correction parameters are fed back to the registration algorithm via an independent data channel to avoid conflicts with the main data stream. This invention maintains high-throughput data processing capabilities while improving the accuracy of mineral phase analysis in complex oil and gas reservoirs through dynamic parameter optimization.

[0128] Specifically, in the modular X-ray analysis method for oil and gas geology described in this invention, step 5 includes:

[0129] When allotropes or samples with the same elements but different phase structures are detected based on the mineral phase composition correlation output in step 4, the energy filtering mode is switched according to the activation command of the energy filtering mode. The resolution of the diffraction signal is improved by the monochromator, and the training parameters of the random forest model are updated by the gradient descent algorithm based on the corrected error value.

[0130] Based on the reservoir permeability and porosity prediction results generated in step 4, the target tube voltage and beam intensity in the time-division control command are dynamically adjusted to adapt to the detection requirements of different geological lithologies, and the time-series correlation weights are simultaneously optimized to match the real-time detection mode.

[0131] Step 5 of this invention improves the analytical capability of complex samples through dynamic mode switching and parameter optimization mechanisms. The specific implementation process is as follows:

[0132] When the mineral phase composition correlation output in step 4 detects allotropes (such as hematite and magnetite) or samples with the same elements but different phase structures, the intelligent analysis module generates an energy filtering mode activation command. This command is transmitted to the monochromator drive unit via the control bus, activating the multilayer film reflective monochromator to filter out non-target energy range X-rays, improving the resolution of target diffraction signals (such as characteristic peaks of Fe oxides). The monochromator filtering range is dynamically adjusted according to the mineral phase type identified in step 4; for example, for hematite, the 8-9 keV energy range is preferentially retained to suppress background noise. The fluorescence and diffraction signals acquired synchronously in energy filtering mode are correlated through timestamps to generate a high-precision time-series correlation dataset, which is then input into a random forest model for secondary analysis.

[0133] The training parameters of the random forest model are updated based on the corrected error value (e.g., hematite proportion error ±3%). The gradient descent algorithm aims to minimize the loss function and iteratively adjusts the feature weights of the decision tree split nodes. For example, when the prediction error of allotropic regions remains high, the algorithm increases the weight coefficient of the diffraction signal d-value, weakens the influence of Fe content in the fluorescence signal, and optimizes the mineral phase classification rules. The updated model parameters are encrypted and verified before being synchronized to the intelligent analysis module, replacing the historical parameter file. The parameter version control mechanism prevents data conflicts and maintains the continuity of model iteration.

[0134] Based on the reservoir permeability and porosity prediction results generated in step 4, the dynamic strategy module adjusts the target tube voltage and beam intensity in the time-sharing control commands using a PID control algorithm. For example, in high-porosity reservoirs, the predicted trigger voltage for Cu targets is increased from 40 kV to 50 kV to enhance low-energy X-ray penetration; in low-permeability reservoirs, the Mo target beam intensity is increased to 20 mA to improve the fluorescence signal-to-noise ratio. The time-series correlation weights are dynamically optimized according to the real-time detection mode: in energy filtering mode, the weights are set to 1:1 to balance signal integrity; in time-sharing mode, they are adjusted to 3:1 to prioritize the acquisition of elemental distribution data.

[0135] The dynamic strategy module and the time-sharing control unit interact in real time via a data bus, and parameter adjustment commands are embedded in the FPGA logic unit of the timing controller. For example, when the tube voltage adjustment causes a change in the X-ray source load, the time-sharing control unit automatically extends the signal acquisition interval (from 100 ms to 150 ms) to match the thermal balance cycle of the cooling system. The optimization results of the timing-related weights are stored in non-volatile memory for subsequent standardized dataset generation, forming a historical benchmark for parameter iteration.

[0136] The closed-loop technical logic of step 5 is reflected in: mineral phase identification results driving acquisition mode switching, and model parameter optimization improving classification accuracy; reservoir parameter prediction is fed back to the hardware control unit, dynamically adjusting target parameters and acquisition strategies. For example, high clay mineral proportion prediction triggers high-frequency activation of the energy filtering mode, simultaneously optimizing the model's sensitivity to layered silicates; low porosity results prioritize the use of a time-sharing mode to extend the diffraction acquisition cycle, improving the accuracy of lattice parameter measurement. This invention achieves a dynamic balance between oil and gas geological sample detection efficiency and analytical accuracy through software and hardware co-optimization.

[0137] In specific embodiments of the present invention, in the context of oil and gas geological sample analysis, the following technical solutions are used to achieve collaborative detection and data fusion of XRF and XRD:

[0138] The acquisition module employs a Cu / Mo dual-target X-ray source, driven by a time-division control unit to alternately activate fluorescence and diffraction signal acquisition in time-division mode. During fluorescence signal acquisition, high-energy X-rays are generated from the Mo target (tube voltage 50 kV, beam current 20 mA), and a silicon drift detector captures the full elemental energy spectrum (1-50 keV) with a period of 100 ms. After acquisition, the module switches to the Cu target (tube voltage 30 kV, beam current 15 mA) to excite diffraction signals, simultaneously triggering the sample stage stepper motor to rotate in 0.1° steps. A two-dimensional array detector records the diffraction angle (5°-80°) and intensity distribution. The time-division controller encapsulates the acquired data into time-stamped standardized signal packets via an LVDS bus and transmits them to the data processing module. The environmental adaptation module identifies sample types based on a classification model: for light elements (such as C and O), a vacuum pump is activated to reduce the chamber pressure to [a value missing]. When the pressure is below Pa, the block sample detection switches to atmospheric mode and the rapid loading mechanism is activated. At the same time, external vibrations are suppressed by the shock-absorbing base and dynamic damping (threshold 0.5 g). The heat dissipation system controls the X-ray source temperature below 40°C, and the temperature signal is fed back to the noise filtering algorithm in real time to optimize the signal-to-noise ratio.

[0139] The data processing module calibrates the fluorescence signal by energy channels, corrects detector gain drift by using characteristic peak positions in the calibration database (e.g., Fe-Kα = 6.40 keV), and generates elemental distribution curves. It performs angle-intensity normalization on the diffraction signal and calculates the lattice spacing (d = λ / (2sinθ), λ = 0.154 nm) based on the Bragg equation to eliminate the influence of surface roughness. The calibrated data is aligned with core scan CT images using a spatial coordinate registration algorithm. Using the Fe and Si elemental distributions as reference points, a three-dimensional model is generated using the Iterative Closest Point (ICP) algorithm. Voxel units are labeled with elemental content (e.g., Fe = 15% ± 0.5%) and mineral phase types (e.g., quartz, hematite). A random forest model establishes a correlation between elemental thresholds (Ca > 5%) and mineral phase proportions (calcite ≥ 15%) based on nonlinear decision tree rules. Combined with third-party detection data (e.g., SEM-EDS), an error comparison algorithm corrects prediction errors, and the error values ​​are fed back to the coordinate registration parameters to update the accuracy of the three-dimensional model.

[0140] The dynamic strategy module triggers an energy filtering mode based on mineral phase identification results. The monochromator filters out non-target energy ranges (e.g., retaining 8-9 keV to enhance hematite diffraction peaks), simultaneously acquiring fluorescence and diffraction signals and associating them with time-series information via timestamps. The gradient descent algorithm iteratively updates model parameters based on a regional geological database (e.g., the quartz / clay phase ratio in shale) and correction errors (±3%), optimizing mineral phase classification weights. Reservoir parameter prediction results inversely control time-sharing commands: for high-porosity reservoirs, the Cu target tube voltage is increased to 50 kV to enhance penetration; for low-permeability reservoirs, the Mo target beam current is increased to 25 mA to optimize the fluorescence signal-to-noise ratio. The time-series association weights are dynamically adjusted to a 3:1 time-sharing mode or a 1:1 energy filtering mode. This scheme is validated through hardware collaborative control and data closed-loop verification.

[0141] This invention addresses the problems caused by the independent operation of existing XRF and XRD technologies through the following technical solutions: First, to address the issue of low detection efficiency, the device achieves time-division or synchronous acquisition of XRF and XRD signals through the collaborative work of a time-division control unit and a dynamic strategy module. The time-division control unit generates timing instructions based on an FPGA, alternately driving multiple target X-ray sources to activate the fluorescence and diffraction signal acquisition process. The timing controller compresses the fluorescence signal acquisition cycle to the millisecond level and eliminates crystal orientation deviations by synchronously adjusting the sample stage rotation and goniometer angle. The dynamic strategy module adjusts the target tube voltage and beam current intensity based on real-time feedback, optimizing the timing correlation weights in the time-division mode, and shortening the detection cycle while ensuring data integrity.

[0142] Secondly, to address the issue of data separation between elemental composition and phase structure, the intelligent analysis module employs a spatial coordinate registration algorithm to align the standardized dataset with core scan imaging data, generating a 3D model that integrates elemental distribution and mineral phase information. The data fusion unit uses the distribution of characteristic elements (such as Fe and Si) as a reference point and achieves spatial matching of cross-modal data through an iterative nearest-point algorithm. The random forest model establishes a correlation between elemental content thresholds and mineral phase ratios based on nonlinear mapping rules. Combined with third-party detection data introduced by the validation unit to correct errors, a multi-dimensional data closed-loop validation mechanism for elements, phases, and reservoir parameters is formed.

[0143] Finally, to enhance the identification capability of allotropes or phase heterogeneous samples, the dynamic strategy module triggers an energy filtering mode based on the correlation between mineral phase composition. A monochromator filters out non-target energy range X-rays, enhancing the resolution of the target diffraction signal, and iteratively updates the training parameters of the random forest model using a gradient descent algorithm. For example, for the identification of hematite and magnetite, the algorithm increases the weight coefficient of the diffraction signal d-value, weakens the interference of Fe content in the fluorescence signal, and optimizes the mineral phase classification decision boundary. This invention achieves accurate identification of complex mineral phases and dynamic prediction of reservoir characteristics through hardware and software co-optimization.

Claims

1. A modular X-ray analysis device for oil and gas geology, characterized in that, include: The acquisition module is used to trigger the sample to generate X-ray fluorescence signal and X-ray diffraction signal through a preset multi-target X-ray source, simultaneously acquire the energy and intensity distribution curve of the fluorescence signal and the diffraction angle and intensity spectrum of the diffraction signal, and transmit the acquired signals to the data processing module. An environment adaptation module, connected to the acquisition module, is used to identify the sample type based on the energy distribution curve of the fluorescence signal and the diffraction angle pattern of the diffraction signal through a pre-trained classification model, obtain the sample type, and switch between vacuum or atmospheric environment according to the sample type. The data processing module receives the fluorescence signal and diffraction signal transmitted by the acquisition module, performs energy channel calibration on the fluorescence signal, and performs angle and intensity normalization processing on the diffraction signal to generate a time-series correlated standardized dataset. The intelligent analysis module calls the element-phase mapping rules in the calibration database, inputs the standardized dataset into the pre-trained random forest model, outputs the correlation between element content and mineral phase composition, and generates reservoir permeability and porosity prediction results by combining core scan data. The dynamic strategy module generates a three-dimensional visualization report based on the correlation and reservoir parameter prediction results output by the intelligent analysis module, and adjusts the target selection or signal acquisition mode of the X-ray source in the acquisition module through mobile terminal interaction commands. The acquisition module includes: The time-division control unit is connected to the data processing module via a timing controller and is configured to alternately activate fluorescence signal acquisition and diffraction signal acquisition according to a preset timing command. In the fluorescence signal acquisition stage, the time-division control unit drives the first target to excite the sample to generate an energy distribution curve, and transmits the energy distribution curve to the data processing module through the timing controller. During the diffraction signal acquisition stage, the time-division control unit switches to the second target to generate a diffraction pattern, synchronously triggers the sample stage to rotate to eliminate crystal orientation deviation, and transmits the diffraction pattern to the data processing module through the timing controller; The environment adaptation module includes: The vacuum switching unit is connected to the time-division control unit in the acquisition module. It is used to start the vacuum environment according to the acquisition stage command of the time-division control unit when detecting light elements, and switch to atmospheric mode when detecting block samples. At the same time, it sends an environmental status signal to the data processing module to trigger timing calibration. The anti-vibration unit is connected to the X-ray source and detector in the acquisition module through a vibration damping base. It monitors the external vibration intensity in real time. When the vibration intensity exceeds a preset threshold, dynamic damping is activated, and the temperature of the X-ray source and detector is controlled within the preset threshold through a heat dissipation system. The temperature control signal of the anti-vibration unit is synchronously transmitted to the data processing module to adjust the noise filtering parameters.

2. The modular X-ray analysis device for oil and gas geology of claim 1, wherein, The intelligent analysis module includes: A data fusion unit receives the standardized data set generated by the data processing module, aligns the standardized data set with the core scanning imaging data through a spatial coordinate registration algorithm, and generates a three-dimensional model containing element distribution and mineral phase, which is used as input data of the pre-trained random forest model; A verification unit, connected to the data fusion unit, receives element quantitative data from a third-party detection device, corrects the mineral phase proportion error output by the random forest model based on the element distribution data and mineral phase proportion data in the three-dimensional model through an error comparison algorithm, and feeds back the corrected error value to the data fusion unit to update the spatial coordinate registration parameters.

3. The modular X-ray analysis device for oil and gas geology of claim 1, wherein, The dynamic strategy module includes: A mode switching unit, connected to the time-sharing control unit in the acquisition module and the intelligent analysis module, adjusts the signal acquisition mode to time-sharing acquisition or energy filtering synchronous acquisition in response to a mobile terminal instruction, and sends a mode switching instruction to the intelligent analysis module to update the input parameters of the pre-trained random forest model, including the time sequence correlation weight of the energy distribution curve of the fluorescence signal and the diffraction pattern acquired in the time-sharing mode; An algorithm iteration unit, connected to the verification unit in the intelligent analysis module, updates the training parameters of the pre-trained random forest model based on the mineral phase data in the regional geological database and the corrected error value output by the verification unit, optimizes the nonlinear mapping rule of mineral phase proportion and element content, and synchronizes the updated model parameters to the intelligent analysis module.

4. A modular X-ray analysis method for petroleum geology, applied to the modular X-ray analysis device for petroleum geology according to any one of claims 1 to 3, characterized in that, It includes: Step 1: Obtain the preprocessed sample data, generate X-ray fluorescence signal and X-ray diffraction signal by time-sharing or synchronous excitation of the sample through a pre-set multi-target X-ray source, and alternately acquire the energy spectrum distribution of the fluorescence signal and the diffraction pattern based on time-sharing control instructions; Step 2: Based on the energy distribution curve of the fluorescence signal and the diffraction angle pattern of the diffraction signal, identify the sample type through a pre-trained classification model to obtain the sample type, and switch the vacuum or atmospheric environment according to the sample type; Step 3: Calibrate the energy channel of the acquired fluorescence signal, and normalize the angle and intensity of the diffraction signal to generate a time-correlated standardized data set; Step 4: Align the standardized data set with the core scanning imaging data through a spatial coordinate registration algorithm to generate a three-dimensional model containing element distribution and mineral phase, and input the three-dimensional model into a pre-trained random forest model to output the correlation between element content and mineral phase composition and the prediction results of reservoir permeability and porosity; Step 5: Generate a three-dimensional visualization report based on the correlation, and adjust the target material selection of the X-ray source or the signal acquisition mode according to real-time feedback to update the input parameters and training parameters of the random forest model.

5. The method for modular X-ray analysis of oil and gas geology according to claim 4, characterized in that, The step 1 includes: In the time-sharing mode, the first target material is activated to generate a fluorescent signal according to a preset timing instruction, and a full-element energy spectrum is recorded. After the collection is completed, the timing controller is switched to the second target material to generate a diffraction signal, the goniometer angle is adjusted to a preset diffraction angle range, and the sample table is rotated to eliminate the crystal orientation deviation; In the energy filtering mode, the monochromator is activated to separate the fluorescent signal and the diffraction signal based on a mobile terminal interaction instruction, the intensity distribution of the two signals is recorded synchronously, and the timing information of the fluorescent signal and the diffraction signal is associated through a time stamp.

6. The method for modular X-ray analysis of oil and gas geology according to claim 5, characterized in that, The step 2 comprises: Based on the fluorescent signal energy distribution curve collected in step 1, the peak shift caused by the detector gain drift is eliminated through an energy channel calibration algorithm, and the calibrated energy spectrum is matched with the element characteristic peaks in the calibration database to determine the content range of the target element in the sample; Based on the diffraction angle spectrum of the diffraction signal collected in step 1, the diffraction signal is processed through an angle and intensity normalization algorithm, the lattice spacing is calculated based on a preset Bragg diffraction equation, and the diffraction peaks in the mineral phase standard sample database are compared to identify the type and proportion of the mineral phase in the sample.

7. The method for modular X-ray analysis of oil and gas geology according to claim 4, characterized in that, The step 4 comprises: The standardized data set generated in step 3 is aligned with the core scanning imaging data through a spatial coordinate registration algorithm to generate a three-dimensional model containing element distribution and mineral phase, and the three-dimensional model is input into a pre-trained random forest model to establish an association between the element content threshold and the mineral phase proportion through a nonlinear mapping rule; The element quantitative data of the third-party detection equipment is received, the mineral phase proportion error output by the random forest model is corrected through an error comparison algorithm based on the element and phase mapping rule in the calibration database, and the corrected error value is fed back to the spatial coordinate registration algorithm to update the coordinate registration parameters of the three-dimensional model.

8. The method for modular X-ray analysis of oil and gas geology of claim 4, wherein, The step 5 comprises: When detecting the same element but different mineral phase structure samples based on the mineral phase composition association relationship output in step 4, the energy filtering mode is switched to the energy filtering mode according to the activation instruction of the energy filtering mode, the diffraction signal resolution is improved through the monochromator, and the training parameters of the random forest model are updated through a gradient descent algorithm based on the corrected error value; According to the reservoir permeability and porosity prediction results generated in step 4, the target tube voltage and beam intensity in the time-sharing control instruction are dynamically adjusted to adapt to the detection needs of different geological lithology, and the timing correlation weight is optimized synchronously to match the real-time detection mode.

Images