Artificial intelligence-based multi-physics monitoring data fusion interpretation method

Abstract

The application provides a kind of multi-physical field monitoring data fusion interpretation method based on artificial intelligence, comprising: carrying out indoor rock physical hydraulic fracturing experiment, and carrying out multi-physical field monitoring;Construct training data set: different rock samples are subjected to the experiment respectively, obtain data, constitute total data set;Construct data fusion interpretation module, the data fusion interpretation module includes two parts, respectively as the feature extraction part of multi-physical field data and fusion interpretation part;Training detection network using training data set;The data obtained by all actual monitoring are subjected to the pretreatment of the same step, input CT image, acoustic emission cumulative distribution diagram, the velocity diagram of ultrasound not containing piezoelectric ceramic sensor artifact interference, output fusion image;Update data set and data fusion interpretation module.The multi-physical field monitoring data fusion interpretation method based on artificial intelligence of the application mines the internal characteristics and connection of multi-physical field data, and more accurately evaluates the fracture damage area.

Description

Technical Field

[0001] This invention relates to the field of "machine learning" technology, specifically "computation based on a specific computational model." In addition to machine learning, the multiphysics monitoring of this invention relates to "obtaining an image of the interior of an object by emitting ultrasonic waves or sound waves through it." Furthermore, this invention relates to the field of "geophysics," and also to the field of "extracting oil, gas, water, soluble or fusible substances, or mineral mud from wells." Specifically, this invention relates to a method for the fusion and interpretation of multiphysics monitoring data based on artificial intelligence. Background Technology

[0002] Hydraulic fracturing technology involves injecting high-pressure fluid into shale reservoirs to create complex artificial fractures, increasing reservoir connectivity and improving single-well production. Monitoring and evaluating different stages of hydraulic fracturing reservoir stimulation are prerequisites for efficient development and safe production. However, in hydraulic fracturing field monitoring, the complex conditions at the well site, the limited range of the observation system, and the inability to obtain accurate background data of the actual work area (velocity structure, rock mechanical parameters, etc.) make it difficult to verify the accuracy of fracture inversion results during hydraulic fracturing through field monitoring.

[0003] The laboratory's rock physics hydraulic fracturing experiments can provide stable and controllable stress loading conditions and signal measurement conditions, enabling the accurate acquisition of various property parameters of rock samples. This allows for detailed analysis and research on the changes in rocks during hydraulic fracturing, further providing guidance and assistance for actual hydraulic fracturing.

[0004] Indoor rock physics hydraulic fracturing experiments can perform multi-physics monitoring (ultrasound, acoustic emission, CT). In laboratory rock physics experiments, CT, ultrasound, and acoustic emission monitoring can be performed simultaneously. Computed Tomography (CT) is an imaging technique that acquires information about the object being tested in a non-destructive manner; CT imaging uses excited X-ray beams to perform tomographic scanning of rock samples, obtaining information about the internal structure of the rock sample. Active-source ultrasound uses piezoelectric ceramic sensors (PZTs) to excite seismic waves, which propagate within the rock sample. By inverting the received seismic waves, information about the internal medium, such as velocity, can be obtained. Passive-source acoustic emission uses piezoelectric ceramic sensors to continuously and passively acquire acoustic emission signals generated by internal rock fracturing. Analysis of these acoustic emission signals can obtain source parameters related to internal rock fracture. In these monitoring methods, piezoelectric ceramic sensors are installed on the rock surface and arranged in a specific array of detectors. These PZT sensors use a rapid automatic switching system to switch between receiving and transmitting functions, allowing for both passive acquisition of acoustic emission signals and active excitation of ultrasonic signals.

[0005] Multiphysics monitoring presents challenges in data fusion and interpretation. Geophysical inversion itself is inherently non-unique; single geophysical observation data often only reflects a portion of the target's physical properties, resulting in inversion results with significant ambiguity and uncertainty. Geophysical inversion targets generally possess multiple physical characteristics. By comprehensively utilizing multiphysics observation methods based on electrical, optical, and magnetic properties, information fusion, complementary advantages, and reduced ambiguity can be achieved, leading to a more comprehensive and accurate characterization of the target. For example, joint inversion techniques combining various electromagnetic data with gravity, magnetic, and seismic data enable high-precision imaging of subsurface media; coupled joint inversion of rock properties based on rock physical parameters (such as resistivity, density, and velocity) improves the accuracy of interpretation and prediction.

[0006] However, for the same observed target, geophysical field observation data of different physical properties have different dimensions, different sensitivities to attributes, and significant differences in data structure characteristics, resulting in inconsistent observational uncertainties. How to mine and fuse information from these multi-source heterogeneous data is the core issue of multiphysics monitoring. Basically, information fusion models can be abstracted into three levels: data-level fusion, feature-level fusion, and decision-level fusion. These fusion levels increase in complexity, reduce computational load and accuracy, and increase information loss, but improve robustness and flexibility. Data-level fusion mainly refers to directly fusing data collected by sensors on the same medium; feature-level fusion refers to first extracting features from the raw data and then fusing them; decision-level fusion involves a higher level of fusion of the preliminary results to arrive at a more comprehensive and systematic decision.

[0007] In rock physics experiments, we can obtain real-time multiphysics monitoring data of the rock physics-hydraulic fracturing process. Ultrasonic data is three-dimensional, reflecting medium velocity information; acoustic emission is one-dimensional, reflecting fracture-related source information; and CT is three-dimensional, reflecting medium structure information. However, methods for fusing and interpreting this multiphysics data are currently lacking, making it impossible to fully utilize the data to effectively constrain the multi-attribute characteristics of rock samples. This patent application proposes an artificial intelligence-based multiphysics monitoring data fusion and interpretation method to comprehensively evaluate the fracture damage area and assess the fracturing effect. Summary of the Invention

[0008] The purpose of this invention is to at least partially overcome the shortcomings of the prior art and provide a method for the fusion and interpretation of multi-physics monitoring data based on artificial intelligence.

[0009] The present invention also aims to provide a method for the fusion and interpretation of multi-physics field monitoring data based on artificial intelligence, which can uncover the inherent characteristics and relationships of multi-physics field data.

[0010] The present invention also aims to provide a method for the fusion and interpretation of multi-physics field monitoring data based on artificial intelligence, so as to more accurately assess the fracture and damage area.

[0011] To achieve the above-mentioned objectives or one of them, the technical solution of the present invention is as follows:

[0012] A method for interpreting multiphysics monitoring data based on artificial intelligence, the method comprising:

[0013] Indoor rock physical hydraulic fracturing experiments were conducted, and multi-physics field monitoring was performed.

[0014] Constructing a training dataset: The experiment was performed on different rock samples to obtain data, which constituted the total dataset; the data included: CT structural images with piezoelectric ceramic sensor metal artifacts removed, acoustic emission cumulative distribution maps, and ultrasonic velocity maps;

[0015] A data fusion interpretation module is constructed, comprising two parts: a feature extraction part for multi-physics data and a fusion interpretation part. In the feature extraction part, a UNet is used independently for each type of physics data to extract its own features, using a total of 5 sets of convolutional layer units. The parameters of convolutional layer units 1-5 are (256×3×3×3), (128×3×3×3), (64×3×3×3), (256×3×3×3), and (256×3×3×3), respectively, where the first dimension is the number of convolutions and the latter three dimensions are the convolution size. Spatial cross-attention is used for feature extraction within convolutional layers of the same depth. In the fusion interpretation part, the extracted features are concatenated, and then further feature extraction is performed. The features obtained from the ultrasound image are then concatenated with the features from the CT image and the microseismic distribution map, and then further feature fusion is performed through several connection layers.

[0016] Train the detection network using the training dataset;

[0017] All data obtained from actual monitoring are preprocessed using the same steps. Input CT images, acoustic emission cumulative distribution maps, and ultrasonic velocity maps that do not contain piezoelectric ceramic sensor artifacts, and output fused images.

[0018] Update the dataset and data fusion interpretation modules.

[0019] According to a preferred embodiment of the present invention, the "conducting indoor rock physical hydraulic fracturing experiments and performing multi-physics field monitoring" includes:

[0020] Rock samples of different sizes were collected according to the research objectives. The rock samples were covered with a prefabricated rubber sleeve with multiple probe holes. No piezoelectric ceramic sensors were attached to the holes.

[0021] A single CT data was collected from a rock sample without a piezoelectric ceramic sensor attached using an indoor hydraulic fracturing experimental setup. This CT data did not contain metal artifacts from the piezoelectric ceramic sensor and served as the first-stage CT data.

[0022] The piezoelectric ceramic sensor is placed in the probe hole and then bonded to the surface of the rock sample.

[0023] The indoor hydraulic fracturing test apparatus was used to collect CT data on a rock sample with a piezoelectric ceramic sensor attached. The spatial position of the piezoelectric ceramic sensor was then determined using the CT data. The CT data included metal artifacts of the piezoelectric ceramic sensor and was used as the second-stage CT data.

[0024] Different loading strategies were employed to pressurize the rock samples. The pressurization phase included an isotropic loading phase, an axial pressure increase phase, a water injection phase to increase pore pressure, and a pressure unloading phase. During the water injection phase to increase pore pressure, several CT data points were acquired at set intervals. These CT data points, containing piezoelectric ceramic sensor metal artifacts, served as the third-stage CT data. During the pressure unloading phase, one CT data point was acquired, which also contained piezoelectric ceramic sensor metal artifacts, serving as the fourth-stage CT data. Finally, the piezoelectric ceramic sensor was removed from the rock sample, and another CT data point was acquired. This new CT data point, free of piezoelectric ceramic sensor metal artifacts, served as the fifth-stage CT data.

[0025] All acquired CT data were processed.

[0026] According to a preferred embodiment of the present invention, the rock sample has the following specifications: a diameter of 50 mm and a length of 125 mm, and is cylindrical.

[0027] Rock samples include two types: ordinary rock samples and disturbed rock samples;

[0028] The common rock sample types include sandstone and shale; they are further divided into five categories based on whether they contain bedding and the direction of bedding: bedding directions of 0±20°, 45±20°, 90±20°, 135±20°, and homogeneous media without bedding; and two categories based on axial compressive strength: 50±10MPa and 90±10MPa. Five rock samples were selected from each category, for a total of 100 rock samples.

[0029] The interfering rock samples include two types of rocks: sandstone and shale. They are homogeneous media without bedding and contain sediments. Five rock samples of each type are selected, for a total of 10 rock samples.

[0030] According to a preferred embodiment of the present invention, the dataset is generated from real rock sample experimental data.

[0031] According to a preferred embodiment of the present invention, the generation of real rock sample experimental data includes:

[0032] Indoor rock physical hydraulic fracturing experiments were conducted using ordinary rock samples. Multiple experiments were completed, and multiple experimental data were obtained.

[0033] For any k-th experiment, the start time of the water injection stage to increase pore pressure is set as the initial value T0, the total duration is Tn, and data is collected every dt seconds, obtaining a total of Secondary data; in all the multiple experimental data, different rock types led to differences in the experimental process, and thus Tn and There will be differences.

[0034] According to a preferred embodiment of the present invention, during training, the input at the left input terminal is a CT image, an acoustic emission cumulative distribution map, and an ultrasound velocity map that do not contain piezoelectric ceramic sensor artifact interference, and the output at the right output terminal is a fused image; the loss function is the structural similarity measure and mutual information of the three modes.

[0035] According to a preferred embodiment of the present invention, in the step of “training the detection network using the training dataset”, the dataset is divided into a training set and a test set in a ratio of 8:2; the data fusion interpretation module adopts the stochastic gradient descent optimization method, sets a dynamic learning rate, sets the initial value to 0.0001, reduces it by half every 50 times, sets the batch size to 40, and sets the number of iterations to 200; the training of the data fusion interpretation module is performed on a GPU image processing unit.

[0036] According to a preferred embodiment of the present invention, the training dataset is updated when the following conditions are triggered:

[0037] The first CT image after the rock fractured in this experiment was selected. The CT image The first CT image after rock fracturing in any experiment in the dataset. Compare and calculate the image similarity values ​​of the two images. When this value is greater than 10%.

[0038] According to a preferred embodiment of the present invention, the indoor hydraulic fracturing experimental apparatus includes a pressure vessel, a loading system, an acoustic emission counting and waveform acquisition system, and a CT monitoring system.

[0039] According to a preferred embodiment of the present invention, during the pressurization of rock samples using different loading strategies, active source ultrasonic data are collected at set times. During active source ultrasonic monitoring, some piezoelectric ceramic sensors are used as transmitting probes to excite ultrasonic signals, and the remaining piezoelectric ceramic sensors are used as receiving probes to receive ultrasonic signals.

[0040] During the pressurization of rock samples using different loading strategies, the piezoelectric ceramic sensor is used as both a transmitting and receiving probe during active source ultrasonic acquisition, and at other times it serves as a receiving probe to receive acoustic emission signals generated by changes in the rock sample.

[0041] The present invention provides an artificial intelligence-based multiphysics monitoring data fusion and interpretation method to mine the intrinsic characteristics and relationships of multiphysics data, improve the interpretation capability and efficiency of fractured rock samples, and thus more accurately assess the fractured damage area. Attached Figure Description

[0042] Figure 1 The arrangement of a rock sample and a piezoelectric ceramic sensor according to an embodiment of the present invention is shown. The left side shows the rock sample, and the right side shows the arrangement of PZT on the surface of the rock sample.

[0043] Figure 2 The modular architecture of a multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to an embodiment of the present invention is shown. Detailed Implementation

[0044] Exemplary embodiments of the present invention are described in detail below with reference to the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements. Furthermore, in the following detailed description, numerous specific details are set forth for ease of explanation to provide a thorough understanding of the embodiments disclosed herein. However, it will be apparent that one or more embodiments may be practiced without these specific details. In other instances, well-known structures and apparatuses are illustrated to simplify the drawings.

[0045] This application proposes a multi-physics monitoring data fusion and interpretation method based on artificial intelligence (AI), applying AI (machine learning) to fusion at different levels. Artificial intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Deep learning technology in AI brings new ideas to information fusion and inversion, improving inversion accuracy by training models to learn the complex mapping relationships between multi-physics fields, and can be applied in geophysical multi-physics monitoring.

[0046] The following describes the specific process of the multiphysics monitoring data fusion and interpretation method based on artificial intelligence, according to an embodiment of the present invention.

[0047] 1. Basic situation of the rock physics experiment involved in this invention

[0048] 1.1 Preparation of experimental samples:

[0049] Experimental samples are generally derived from rocks from specific regions. The size of the rock samples can vary depending on the actual research purpose. The standard size used for conventional indoor hydraulic fracturing experiments is a cylinder with a diameter of 50 mm and a length of 125 mm.

[0050] Two types of rock samples were set up: ordinary rock samples and disturbed rock samples.

[0051] Common rock sample types include: rock type (sandstone, shale, etc., 2 types), bedding with different orientations (bedding directions are approximately 0±20°, 45±20°, 90±20°, 135±20°, etc., and homogeneous media without obvious bedding, etc., 5 types), and axial compressive strength (50±10MPa, 90±10MPa, etc., 2 types); 5 rock samples are selected from each type, for a total of 100 rock samples.

[0052] Interfering rock sample types include: rock types (sandstone, shale, etc.), homogeneous media without obvious bedding, and rock samples containing obvious sediments, such as heavy mineral impurities and clastic particles. These sediments are internal characteristics of the rock samples and will be visible in CT images regardless of whether a PZT probe is attached to the surface, thus increasing the generalizability of the rock samples. Five rock samples from each type were selected, for a total of 10 rock samples.

[0053] The rock sample is covered with a prefabricated rubber sleeve containing 24 PZT probe holes (the number is not limited to 24; it depends on the sleeve's design). The PZT probes are placed in the holes, and the bottom of the PZT probes is glued to the rock sample surface. The positions of the PZT probe holes on the rubber sleeve are shown below. Figure 1 As shown, four columns of probes are arranged along the rock sample at 0°, 90°, 180°, and 270°, with each column having a different probe height to ensure that the four columns of probes do not appear at the same height simultaneously. PZT is a metallic material, which can produce severe metal artifacts on the rock sample surface, affecting the CT imaging quality of the rock sample; this design will reduce the metal artifacts caused by multiple PZT probes at any height (e.g., ...). Figure 1 ).

[0054] 1.2 Experimental System Composition:

[0055] The experimental system is an indoor hydraulic fracturing experimental device, mainly consisting of a pressure vessel, a loading system, and an acoustic emission counting and waveform acquisition system.

[0056] (1) The pressure vessel is a high-pressure resistant metal container, and the loaded rock sample is placed inside the container.

[0057] (2) The loading system generally adopts a triaxial compressive stress loading system. This system can provide axial pressure, injection pressure and confining pressure to the inside of the pressure vessel to simulate the in-situ formation conditions of the rock sample.

[0058] (3) The acoustic emission counting and waveform acquisition system consists of a PZT, a preamplifier and a high-speed acquisition device.

[0059] (4) The CT monitoring system is composed of CT units, with a field of view greater than 125mm, and the scanning mode is plain scan, usually 256-slice CT.

[0060] 1.3 Experimental Procedure:

[0061] The experimental system can employ different loading strategies depending on the actual research objective. The experimental procedure of this patent is as follows:

[0062] (1) Place the rock sample in a rubber sleeve and load it into a pressure vessel; place it on a CT scanner and collect CT data once. The CT data collected in this process does not contain interference from PZT metal artifacts.

[0063] (2) Remove the rock sample from the container and attach the PZT (or PZT probe) to the PZT probe hole in the outer rubber sleeve of the rock sample.

[0064] (3) Place the rock sample with PZT (or PZT probe) attached inside a pressure vessel; place it on a CT scanner and acquire CT data once. The CT data acquired during this process includes interference from PZT metal artifacts. Calibrate the spatial position Sta(x,y,z) of all probes. (Since the aperture of the PZT probe is larger than the probe diameter, there will be spatial deviations when installing the PZT probe in each experiment. Therefore, it is necessary to accurately extract the position through CT images after installing the PZT probe.)

[0065] (4) Isotropic loading stage (containment pressure).

[0066] (5) Add axial compression stage (axial compression).

[0067] (6) Water injection to increase pore pressure (injection pressure) until the rock fracturing stage and unloading stage. During this process, the CT scanner is turned on and CT data is collected according to the set time. The CT data collected during this process includes interference from PZT metal artifacts.

[0068] (7) Pressure unloading stage (injection pressure, confining pressure and axial pressure unloaded to 0), collect CT data once. The CT data collected in this process includes interference from PZT metal artifacts (after the rock sample is fractured, there are internal cracks); take the rock sample out of the container, remove the PZT probe pasted in the PZT probe hole of the outer rubber sleeve of the rock sample, and place the rock sample in the pressure container; place it on the CT machine and collect CT data once. The CT data collected in this process does not include interference from PZT metal artifacts.

[0069] In the above (4)-(6) process, active source ultrasound data are collected according to the set time. When performing active source ultrasound monitoring, some PZT probes are used as transmitting probes (excitation probes) to excite ultrasound signals, and the remaining probes are used as receiving probes to receive ultrasound signals.

[0070] During the above (4)-(6) process, due to the loading of external forces, the process will be accompanied by acoustic emission events. Except when the PZT probe is used as the transmitting probe during active source ultrasonic acquisition, it is used as the receiving probe for the rest of the time to receive the acoustic emission signals generated by the changes in the rock sample.

[0071] 1.4 Main data analysis contents:

[0072] The CT data processing workflow includes: data acquisition, CT imaging, PZT metal artifact suppression, and joint analysis with acoustic emission results.

[0073] The acoustic emission data processing flow includes: valid event picking (picking data segments with obvious phases), first arrival picking (extracting the first arrival time of the waveform from the valid event data), source location (obtaining the source location of the event based on the picked first arrival time and the spatial position of the probe), source mechanism analysis (inverting the source mechanism of the event based on the waveform of the event to obtain the rupture characteristics of the event), magnitude calculation (inverting the magnitude of the event based on the waveform of the event), and stress field analysis (based on the waveform of the event or the source mechanism of the event), etc.

[0074] The ultrasonic data processing workflow includes: ultrasonic event identification (identifying signals received by other channels based on ultrasonic excitation time), ultrasonic first arrival pickup (obtaining the first arrival time of the received signal), and velocity inversion analysis (inverting the velocity model of the rock sample based on the first arrival time, waveform, and other information of the signal).

[0075] 2. Construct the training dataset:

[0076] 2.1 Overview of the dataset.

[0077] Using the common rock samples prepared in 1.1, the experimental procedure in 1.3 can be completed to obtain a dataset for one experiment; using different common rock samples from 1.1, the experiment can be repeated. This experiment constitutes the total dataset.

[0078] In the k-th experiment, the following data is included:

[0079] (1) CT structural image after removing PZT metal artifacts 3D data This includes three-dimensional space. This experiment collected a total of Second-rate.

[0080] (2) Accumulated distribution map of acoustic emission 3D data This includes three-dimensional space. This experiment collected a total of Second-rate.

[0081] (3) Ultrasonic velocity diagram 3D data This includes three-dimensional space. This experiment collected a total of Second-rate.

[0082] 2.2 Dataset generation.

[0083] The dataset was generated from real rock sample experimental data.

[0084] Using the ordinary rock samples prepared in 1.1, and following the experimental procedure in 1.3, complete the following tasks. This experiment yielded... Experimental data.

[0085] In this case, for any k-th experiment, the initial value T0 is the starting time of process 1.3 (6), the total duration is Tn, and ultrasound is collected once every dt seconds, for a total of This data was collected every dt seconds by CT scan, for a total of [number] data points. This data. In all of them... In this experiment, different rock types led to differences in the experimental process, which in turn affected Tn and... There will be differences.

[0086] (1) CT structural image after removing PZT metal artifacts

[0087] At time T0+dt*t, the original CT image containing PZT metal artifacts is acquired. Metal artifact suppression methods (such as matched subtraction) are used to remove the metal artifacts from the CT image, thus obtaining the image at time T0+dt*t. .

[0088] (2) Acoustic emission distribution map

[0089] Acoustic emission events are monitored within the time interval T0 to T0+dt*t. Source parameters for all acoustic emission signals are calculated, and a magnitude range is selected. Level to If the level is specified, then the source parameters of any sound emission are: , where XYZ is the location, it is the excitation time, and Mg is the magnitude.

[0090] The magnitude is converted from the exponential domain to the decimal domain as follows: The magnitude within the decimal domain is normalized using the following method: Then any launch event information is Convert acoustic emission event information into data space. That is, in Within, each acoustic emission event is determined by its source location. For coordinates, magnitude The coordinate amplitude is used. Each acoustic emission event in the data space is transformed into a probability distribution map, that is, for any event in the data space... Gaussian smoothing with a standard deviation of 1 and a window size of 5×5×5 is performed to obtain the Gaussian distribution map of any event in the data space.

[0091] Using the initial value T0 as a reference, the Gaussian distribution plots of all acoustic emission events within the time interval from T0 to T0+dt*t are summed to obtain the acoustic emission distribution plot at time T0+dt*t. .

[0092] (3) Velocity diagram derived from ultrasonic inversion

[0093] Using the initial value T0 as a reference, all ultrasonic data collected between T0+dt*(t-1) and T0+dt*t are used to obtain the velocity map of the rock sample data space at time T0+dt*t using velocity inversion methods (such as linear inversion). .

[0094] 2.3 Data preprocessing.

[0095] All data undergo the same preprocessing steps.

[0096] 3. Construct a method for fusing and interpreting multiphysics monitoring data:

[0097] 3.1 Module architecture.

[0098] Multiphysics monitoring data fusion interpretation methods, such as Figure 2 As shown, it includes experimental setup, multiphysics data acquisition module, data processing module, data fusion and interpretation module, and parameter update module.

[0099] The experimental setup and multiphysics data acquisition module mainly includes the experimental section in section 1.3, which corresponds to this section.

[0100] The data processing module includes a multiphysics data computation module, including the computation of acoustic emission data, CT data, and ultrasound data, and mainly includes the methods involved in dataset generation in section 2.2.

[0101] The data fusion interpretation module contains the main structure of the multiphysics monitoring data fusion interpretation method, mainly consisting of section 3.2.

[0102] 3.2 Data Fusion and Interpretation Module

[0103] The data fusion and interpretation module mainly consists of two parts: feature extraction of multiphysics data and fusion and interpretation.

[0104] In the feature extraction section, since the input multiphysics data have the same spatial size but represent significantly different physical dimensions, a separate UNet was used to extract features for each type of physics data. A total of five convolutional layer units were used, with parameters for convolutional layer units 1-5 as follows: (256×3×3×3), (128×3×3×3), (64×3×3×3), (256×3×3×3), and (256×3×3×3), where the first dimension represents the number of convolutions and the latter three dimensions represent the convolution size. Furthermore, due to the closer relationship between CT images and microseismic distribution maps, a spatial cross-attention module was used in convolutional layers of the same depth to further extract features from these two types of data.

[0105] In the fusion interpretation section, the closer relationship between CT images and microseismic distribution maps is also fully considered. First, the features extracted from both are connected, and then further feature extraction is performed. Next, the features obtained from the ultrasound image are connected with the features from the CT image and the microseismic distribution map. Then, further feature fusion is performed through several connection layers.

[0106] During training, the input on the left side is a CT image that does not contain PZT artifact interference. (denoted as A), Acoustic emission distribution map (Referred to as B), Supersonic velocity diagram (Denoteed as C), the output on the right is the fused image y. The loss function is a measure of structural similarity and mutual information across the three modalities. Specifically: .

[0107] (1) Structural similarity measurement

[0108] The three modalities of structural similarity index measure (SSIM) model the information loss and distortion during the fusion process and reflect the structural similarity between the fused image and the source image.

[0109]

[0110] and These represent image blocks within a sliding window, representing the source image and the fused image, respectively. This represents the covariance between the source image and the fused image. and These represent the standard deviations of the source image and the fused image, respectively. and represents the mean of the source image and the fused image, respectively. , and It is a constant used to prevent division by zero. The structural similarity measure of the three modes of SSIM, with three different modal data A, B, and C as input and y as output, is:

[0111]

[0112] The larger the value, the less information loss and distortion the fusion algorithm will have.

[0113] (2) Mutual information

[0114] Mutual information (MI) is an information theory-based measure of the amount of information transferred from the source image to the fused image.

[0115]

[0116] in, , and These represent information transferred from source images A, B, and C of three different modalities to the fused image y, respectively.

[0117]

[0118] and The edge histograms of the source image X and the fused image F are shown below. This is the joint histogram of the source image and the fused image. The larger the value, the more information the fusion algorithm transfers from the source image to the fused image.

[0119] 4. Train the detection network using the training dataset.

[0120] 4.1 Divide the dataset into a training set and a test set in a ratio of 8:2.

[0121] 4.2 This data fusion and interpretation module uses the stochastic gradient descent optimization method; a dynamic learning rate is set, with an initial value of 0.0001, which is halved every 50 iterations. The batch size is set to 40; the number of iterations is 200.

[0122] 4.3 The training of the data fusion and interpretation module is carried out on the GPU image processing unit.

[0123] 5. Analysis and processing of actual monitoring data:

[0124] 5.1. Actual data preprocessing.

[0125] All data obtained from actual monitoring undergo the same preprocessing steps.

[0126] 5.2 Actual data analysis.

[0127] Following the dataset generation method in 2.2, preprocessed CT images, acoustic emission distribution maps, and ultrasound images that do not contain PZT artifact interference are obtained and input into the data fusion and interpretation module to obtain the corresponding fused interpretation images.

[0128] 5.3 Dataset Update Mechanism

[0129] The data collected in this experiment will be processed when the following conditions are triggered. Following the dataset generation method in section 2.2, the generated data is updated in the training dataset:

[0130] The first CT image after rock fracturing in 1.3(6) of this experiment was selected. The DCT image was compared with the first CT image after rock fracturing in any experiment 1.3(6) of the dataset. Compare and calculate the image similarity values ​​of the two images. When this value is greater than 10%.

[0131] Image similarity calculation:

[0132]

[0133] 6. Parameter update module and parameter update

[0134] 6.1 Module update mechanism.

[0135] The module triggers an update when the following conditions are met.

[0136] (1) After the experiment starts, the number of experiments that meet the requirements for updating is greater than 5 or the number of updated experiments exceeds 20% of the number of experiments in the existing dataset.

[0137] (2) 5.3 In any triggering condition, the similarity difference value Greater than 20%.

[0138] 6.2 Data fusion interpretation module update method.

[0139] Based on the parameters of the aforementioned data fusion and interpretation module, this experiment uses newly added data after the start of the experiment, namely the data updated in section 5.3 (both sets of data remain the same). To maintain consistent data quantity, the data fusion and interpretation module is fine-tuned to improve training speed and maintain update efficiency.

[0140] 6.3 Update the parameters of the data fusion interpretation module.

[0141] The present invention provides an artificial intelligence-based multiphysics monitoring data fusion and interpretation method to mine the intrinsic characteristics and relationships of multiphysics data, improve the interpretation capability and efficiency of fractured rock samples, and thus more accurately assess the fractured damage area.

[0142] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that variations may be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for fusing and interpreting multiphysics monitoring data based on artificial intelligence, characterized in that, The method includes: Indoor rock physical hydraulic fracturing experiments were conducted, and multi-physics field monitoring was performed. Constructing a training dataset: The experiment was performed on different rock samples to obtain data, which constituted the total dataset; the data included: CT structural images with piezoelectric ceramic sensor metal artifacts removed, acoustic emission cumulative distribution maps, and ultrasonic velocity maps; A data fusion interpretation module is constructed, comprising two parts: a feature extraction part for multi-physics data and a fusion interpretation part. In the feature extraction part, a UNet is used independently for each type of physics data to extract its own features, using a total of 5 sets of convolutional layer units. The parameters of convolutional layer units 1-5 are (256×3×3×3), (128×3×3×3), (64×3×3×3), (256×3×3×3), and (256×3×3×3), respectively, where the first dimension is the number of convolutions and the latter three dimensions are the convolution size. Spatial cross-attention is used for feature extraction within convolutional layers of the same depth. In the fusion interpretation part, the extracted features are concatenated, and then further feature extraction is performed. The features obtained from the ultrasound image are then concatenated with the features from the CT image and the microseismic distribution map, and then further feature fusion is performed through several connection layers. Train the detection network using the training dataset; All data obtained from actual monitoring are preprocessed using the same steps. Input CT images, acoustic emission cumulative distribution maps, and ultrasonic velocity maps that do not contain piezoelectric ceramic sensor artifacts, and output fused images. Update the dataset and data fusion interpretation modules.

2. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 1, characterized in that, The phrase "conducting indoor rock physical hydraulic fracturing experiments and conducting multi-physics field monitoring" includes: Rock samples of different sizes were collected according to the research objectives. The rock samples were covered with a prefabricated rubber sleeve. The rubber sleeve had multiple probe holes, but no piezoelectric ceramic sensors were attached to the holes. A single CT data was collected from a rock sample without a piezoelectric ceramic sensor attached using an indoor hydraulic fracturing experimental setup. This CT data did not contain metal artifacts from the piezoelectric ceramic sensor and served as the first-stage CT data. The piezoelectric ceramic sensor is placed in the probe hole and then bonded to the surface of the rock sample. The indoor hydraulic fracturing test apparatus was used to collect CT data on a rock sample with a piezoelectric ceramic sensor attached. The spatial position of the piezoelectric ceramic sensor was then determined using the CT data. The CT data included metal artifacts of the piezoelectric ceramic sensor and was used as the second-stage CT data. Different loading strategies were employed to pressurize the rock samples. The pressurization phase included an isotropic loading phase, an axial pressure increase phase, a water injection phase to increase pore pressure, and a pressure unloading phase. During the water injection phase to increase pore pressure, several CT data points were acquired at set intervals. These CT data points, containing piezoelectric ceramic sensor metal artifacts, served as the third-stage CT data. During the pressure unloading phase, one CT data point was acquired, which also contained piezoelectric ceramic sensor metal artifacts, serving as the fourth-stage CT data. Finally, the piezoelectric ceramic sensor was removed from the rock sample, and another CT data point was acquired. This new CT data point, free of piezoelectric ceramic sensor metal artifacts, served as the fifth-stage CT data. All acquired CT data were processed.

3. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 2, characterized in that: The rock sample has the following dimensions: 50 mm in diameter and 125 mm in length; it is cylindrical. Rock samples include two types: ordinary rock samples and disturbed rock samples; The common rock sample types include sandstone and shale; they are further divided into five categories based on whether they contain bedding and the direction of bedding: bedding directions of 0±20°, 45±20°, 90±20°, 135±20°, and homogeneous media without bedding; and two categories based on axial compressive strength: 50±10MPa and 90±10MPa. Five rock samples were selected from each category, for a total of 100 rock samples. The interfering rock samples include two types of rocks: sandstone and shale. They are homogeneous media without bedding and contain sediments. Five rock samples of each type are selected, for a total of 10 rock samples.

4. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 3, characterized in that: The dataset was generated from real rock sample experimental data.

5. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 4, characterized in that, The generation of real rock sample experimental data includes: Indoor rock physical hydraulic fracturing experiments were conducted using ordinary rock samples. Multiple experiments were completed, and multiple experimental data were obtained. For any k-th experiment, the start time of the water injection stage to increase pore pressure is set as the initial value T0, the total duration is Tn, and data is collected every dt seconds, obtaining a total of Secondary data; in all the multiple experimental data, different rock types led to differences in the experimental process, and thus Tn and There will be differences.

6. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 1, characterized in that: During training, the input on the left side is a CT image, an acoustic emission cumulative distribution map, and an ultrasound velocity map that do not contain piezoelectric ceramic sensor artifacts; the output on the right side is a fused image; the loss function is the structural similarity measure and mutual information of the three modes.

7. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 1, characterized in that: In the step of "Training the detection network using the training dataset", the dataset is divided into a training set and a test set in a ratio of 8:

2. The data fusion and interpretation module adopts the stochastic gradient descent optimization method, sets a dynamic learning rate with an initial value of 0.0001, and reduces it by half every 50 iterations. The batch size is set to 40, and the number of iterations is 200. The training of the data fusion interpretation module is performed on the GPU image processing unit.

8. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 1, characterized in that, The training dataset is updated when the following conditions are met: The first CT image after the rock fractured in this experiment was selected. The CT image The first CT image after rock fracturing in any experiment in the dataset. Compare and calculate the image similarity values ​​of the two images. When this value is greater than 10%.

9. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 2, characterized in that: The indoor hydraulic fracturing experimental setup includes a pressure vessel, a loading system, an acoustic emission counting and waveform acquisition system, and a CT monitoring system.

10. The multiphysics monitoring data fusion and interpretation method based on artificial intelligence according to claim 2, characterized in that: During the pressurization of rock samples using different loading strategies, active source ultrasonic data were collected at set times. During active source ultrasonic monitoring, some piezoelectric ceramic sensors acted as transmitting probes to excite ultrasonic signals, while the remaining piezoelectric ceramic sensors acted as receiving probes to receive ultrasonic signals. During the pressurization of rock samples using different loading strategies, the piezoelectric ceramic sensor is used as both a transmitting and receiving probe during active source ultrasonic acquisition, and at other times it serves as a receiving probe to receive acoustic emission signals generated by changes in the rock sample.

Images