A crack prediction method and device based on multi-attribute fusion technology

Abstract

This invention discloses a fracture prediction method and apparatus based on multi-attribute fusion technology. The method includes: S1. acquiring pre-stack data to be identified; S2. processing the pre-stack data to obtain a high signal-to-noise ratio (SNR) pre-stack seismic data volume; S3. performing high-resolution and low-resolution target processing on the high SNR pre-stack seismic data volume; S4. reprocessing the processing results of S3 using a third-generation coherence volume algorithm, maximum likelihood analysis, and a positive flexure prediction method, and predicting large and small fractures and flexure zones based on the reprocessing results; S5. constructing a data volume fusion model using a data volume fusion method; S6. quantitatively simulating the small fracture prediction results and flexure zone prediction results obtained by the maximum likelihood method, and inputting the quantitative simulation results into the model, which outputs the prediction results of the multi-scale fracture data volume. This method solves the problem of low accuracy in predicting the distribution patterns of oil and gas-bearing fractures.

Description

Technical Field

[0001] This invention relates to the field of petroleum geophysical exploration, and in particular to a fracture prediction method and apparatus based on multi-attribute fusion technology. Background Technology

[0002] The study of natural fractures in rocks began abroad as early as the 1920s. Due to the complexity of the natural fracturing process and its distribution, although theories of "rock mechanics," and later "rock mass mechanics" and "fracture mechanics" were established long ago, our understanding of the fracturing process in deep sedimentary basins and the prediction of fracture distribution remain quite limited. Compared to the study of porous reservoirs, long-standing limitations in understanding and technology have slowed the progress of reservoir fracture research. Facing the global challenge of exploring and evaluating low-permeability and fractured reservoirs—a challenge that remains largely unresolved in petroleum geology—both domestically and internationally, research is still in the exploratory and developmental stage. Accurate methods for predicting the distribution patterns of oil and gas-bearing fractures have not yet been truly developed; in other words, existing methods have low accuracy in predicting the distribution patterns of oil and gas-bearing fractures. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the low accuracy of the prediction of the distribution law of oil and gas fractures in the prior art, and to provide a fracture prediction method and device based on multi-attribute fusion technology.

[0004] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0005] A crack prediction method based on multi-attribute fusion technology includes the following steps:

[0006] S1. Obtain the pre-stack data to be identified;

[0007] S2. The pre-stack data to be identified is processed using a construction-guided filtering algorithm to obtain a high signal-to-noise ratio pre-stack seismic data volume. When S / N>5, the pre-stack seismic data volume with a high signal-to-noise ratio is defined as a pre-stack data volume with a high signal-to-noise ratio, where S is the signal and N is the noise.

[0008] S3. Perform high-resolution target processing on the high signal-to-noise ratio pre-stack seismic data volume to obtain a high-resolution target processed data volume, and perform low-resolution target processing to obtain a low-resolution target processed data volume. Define frequencies higher than the dominant frequency of the original seismic data as high frequencies and frequencies lower than the dominant frequency of the original seismic data as low frequencies.

[0009] S4. The high-resolution target processing data volume is processed using the third-generation coherence volume algorithm, the maximum likelihood analysis method, and the positive flexure prediction method. The processing results are the prediction results of different methods for small cracks and flexure zones. Small cracks are cracks that need to be processed in their seismic data volume to be identified.

[0010] S5. The third-generation coherence volume algorithm is used to process the low-resolution target processing data volume. The processing result is the prediction result for large cracks. Large cracks are cracks that can be identified without processing their seismic data volume.

[0011] S6. The results calculated by the third-generation coherent volume algorithm in steps S4 and S5 are fused into a single data volume using the data volume fusion method to form a data volume fusion model.

[0012] S7. Quantitatively simulate the prediction results of small cracks and flexural zones predicted by the maximum likelihood method, and input the quantitative simulation results into the data volume fusion model. The data volume fusion model outputs the prediction results of crack data volume.

[0013] Preferably, in step S2, a construction-guided filtering algorithm is used to process the pre-stack data to be identified, obtaining a high signal-to-noise ratio pre-stack seismic data volume, including the following steps:

[0014] S21. Conduct signal-to-noise ratio analysis of the raw seismic data volume and test the structural steering filter parameters;

[0015] S22. Establish the optimal filter parameters by adjusting the tested parameters of the guided filter;

[0016] S23. The pre-stack data to be identified is processed using the optimal filtering parameters and the constructed steerable filtering algorithm to obtain a pre-stack seismic data volume with high signal-to-noise ratio.

[0017] Preferably, in step S21, the guided filtering parameters include the data scale factor, the diffusion factor, and the number of iterations.

[0018] Preferably, in step S3, high-resolution target processing and low-resolution target processing are performed on the high signal-to-noise ratio pre-stack seismic data volume, including the following steps:

[0019] S31. Perform dominant frequency spectrum analysis on the target layer of the high signal-to-noise ratio pre-stack seismic data volume to identify the effective frequency band and the cutoff frequencies of high and low frequencies;

[0020] S32. Utilize WFI data volume target processing technology and employ Ricker wavelet or Yu's wavelet to perform frequency extension or frequency reduction processing on the spectrum of high signal-to-noise ratio pre-stack seismic data volume.

[0021] Preferably, in step S4, the third-generation coherent volume algorithm uses a 3*3*3 grid for calculation; in step S5, the third-generation coherent volume algorithm uses a 9*9*9 grid for calculation.

[0022] Preferably, in step S6, the parameters in the data volume fusion model include data volume fusion threshold and data volume fusion weight. By setting different data volume fusion threshold and data volume fusion weight values, crack prediction data volumes of different scales are obtained. The data volume fusion model fuses crack prediction data volumes of different scales and outputs the prediction results of crack data volumes.

[0023] Preferably, in step S7, a quantitative simulation method for cracks and flexural zones is used to quantitatively simulate the prediction results of small cracks and flexural zones predicted by the maximum likelihood method.

[0024] A crack prediction device based on multi-attribute fusion technology includes at least one processor and at least one memory communicatively connected to the processor. The memory stores instructions that are executed by the at least one processor to enable the at least one processor to perform any step of the method.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] This invention processes pre-stack data by constructing a guided filtering algorithm, uses the third-generation coherence volume algorithm, the maximum likelihood analysis method, and constructs a positive flexure prediction method to predict cracks and flexure zones at different scales, and fuses the prediction results through a data volume fusion method to form a data volume fusion model, thereby realizing the prediction of cracks at different scales. It directly starts from the data to carry out crack prediction of different scales and sizes, thus improving the prediction accuracy of cracks at different scales. Attached Figure Description

[0027] Figure 1 This is a flowchart of a crack prediction method based on multi-attribute fusion technology according to the present invention;

[0028] Figure 2 The images show the original seismic profile with low signal-to-noise ratio (left) and the profile with complex domain nonlinear anisotropic diffusion filtering (right).

[0029] Figure 3 This is a comparison chart before and after diffusion filtering;

[0030] Figure 4 A comparison of time slices before and after 2.7s filtering in the target layer;

[0031] Figure 5 Here is a flowchart of the generalized S-transform;

[0032] Figure 6Before resolution processing of the inline1476 line through the well, the cross-section image was processed to improve the resolution.

[0033] Figure 7 A cross-sectional view after resolution enhancement for the inline1476 line passing through the well;

[0034] Figure 8 It is a comparison diagram of high-resolution coherence and high-resolution frequency-division small-scale coherence;

[0035] Figure 9 Three-dimensional visualization feature maps of crack fusion bodies at different scales (left) and crack-amplitude fusion bodies (right);

[0036] Figure 10 A comprehensive interpretation diagram of crack fusion at different scales. Detailed Implementation

[0037] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0038] Example

[0039] A crack prediction method based on multi-attribute fusion technology, the flowchart of which is as follows: Figure 1 As shown, it includes the following steps:

[0040] S1. Obtain the pre-stack data to be identified;

[0041] S2. A construction-guided filtering algorithm is used to process the pre-stack data to be identified, resulting in a pre-stack seismic data volume with a high signal-to-noise ratio;

[0042] S3. Perform high-resolution target processing on the high signal-to-noise ratio pre-stack seismic data volume to obtain a high-resolution target-processed data volume, and perform low-resolution target processing to obtain a low-resolution target-processed data volume.

[0043] S4. The high-resolution target processing data volume is processed using the third-generation coherent volume algorithm, the maximum likelihood analysis method, and the positive flexure prediction method. The processing results are the prediction results of different methods for small cracks and flexure zones.

[0044] S5. The low-resolution target processing data volume is processed using the third-generation coherent volume algorithm, and the processing result is the prediction result for large cracks.

[0045] S6. The results calculated by the third-generation coherent volume algorithm in steps S4 and S5 are fused into a single data volume using the data volume fusion method to form a data volume fusion model.

[0046] S7. Quantitatively simulate the prediction results of small cracks and flexural zones predicted by the maximum likelihood method, and input the quantitative simulation results into the data volume fusion model. The data volume fusion model outputs the prediction results of crack data volume.

[0047] In step S2, the high signal-to-noise ratio seismic data volume is defined as follows: the signal-to-noise ratio is a key factor in the evaluation of seismic data quality, usually evaluated by the S / N ratio (S refers to signal and N refers to noise). Seismic data with S / N>5 are usually called high signal-to-noise ratio seismic data. The signal refers to the electronic signal from outside the equipment that needs to be processed by the equipment. The noise refers to the irregular additional signal (or information) generated after passing through the equipment that does not exist in the original signal, and this kind of signal does not change with the change of the original signal.

[0048] In step S2, the process of using the constructed guided filter algorithm to process the pre-stack data to be identified is as follows:

[0049] S21. Conduct signal-to-noise ratio analysis of the original seismic data volume and test the structural steering filter parameters, including the data scale factor, diffusion factor, and number of iterations.

[0050] S22. Establish the optimal filter parameters by adjusting the tested parameters of the guided filter;

[0051] S23. The pre-stack data to be identified is processed using the optimal filtering parameters and the constructed steerable filtering algorithm to obtain a pre-stack seismic data volume with high signal-to-noise ratio.

[0052] Seismic images contain numerous edges formed by phase axes, rich in geological information. However, seismic data from reservoir fracture zones often exhibits lower quality. Therefore, noise suppression necessitates preserving or even enhancing all edge information to provide reliable data for geological interpretation in seismic exploration. In recent years, diffusion filtering techniques based on partial differential equations have gained increasing attention and are widely used in image denoising and enhancement. These techniques preserve or even enhance textured structures such as edges and lines while denoising. Their development has progressed from linear to nonlinear, from isotropic to anisotropic diffusion, and from the real domain to the complex domain. By incorporating a Shock filter into the linear anisotropic diffusion term in the complex domain, diffusion filtering is extended to nonlinearity, overcoming the noise sensitivity of the Shock filter and the poor edge preservation ability of diffusion filtering. This aims to suppress seismic signal noise, improve the signal-to-noise ratio, and reduce the ambiguity of reservoir fracture predictions.

[0053] Figure 2 This is a low signal-to-noise ratio original seismic profile and a complex domain nonlinear anisotropic diffusion-filtered profile. Figure 2As can be seen, the original seismic profile has a discontinuous phase axis (left side), making it difficult to describe the fine structural features (elliptical position). Using complex domain nonlinear anisotropic diffusion filtering can effectively suppress noise, enhance the continuity of the phase axis of the profile, and enhance edge information, thereby improving the signal-to-noise ratio of the data (elliptical position), which is beneficial for the fine interpretation of the structure (right side).

[0054] Figure 3 This is a comparison chart before and after diffusion filtering, from... Figure 3 As can be seen, the signal-to-noise ratio of some phases has been improved. Figure 4 The comparison of time slices before and after 2.7s filtering of the target layer shows that the random noise is suppressed to a certain extent after complex domain nonlinear anisotropic diffusion filtering, and the signal-to-noise ratio of the seismic data is slightly improved.

[0055] In step S3, high frequency and low frequency are defined as follows: resolution is a major evaluation parameter of the longitudinal resolution capability of seismic data. Usually, the dominant frequency of the original seismic data is used as a benchmark (e.g., 30 Hz). Data volumes with a dominant frequency lower than that of the original data (e.g., reduced to 20 Hz) are low-resolution target processed data volumes, while data volumes with a dominant frequency higher than that of the original data (e.g., increased to 40 Hz) are high-resolution target processed data volumes.

[0056] In step S3, high-resolution target processing is performed on the high signal-to-noise ratio pre-stack seismic data volume to obtain a high-resolution target-processed data volume, and low-resolution target processing is performed to obtain a low-resolution target-processed data volume. The specific processing steps are as follows:

[0057] S31. Perform dominant frequency spectrum analysis on the target layer of the high signal-to-noise ratio pre-stack seismic data volume to identify the effective frequency band and the cutoff frequencies of high and low frequencies;

[0058] S32. Utilize WFI data volume target processing technology and employ Ricker wavelet or Yu's wavelet to perform frequency extension or frequency reduction processing on the spectrum of high signal-to-noise ratio pre-stack seismic data volume.

[0059] Improving the resolution of seismic data is one of the fundamental goals of seismic data processing. Only by improving the resolution of seismic data can the requirements of reservoir research and oil reservoir characterization be met. In seismic exploration, seismic signals are non-stationary signals, and the amplitude and frequency of wavelets have different characteristics at different propagation times. At the same time, data from different frequency bands also have different amplitude characteristics, which traditional Fourier transform cannot fully analyze. Only by converting seismic data to the time-frequency domain through time-frequency analysis can we understand the characteristics of seismic data from multiple aspects and angles, ultimately achieving the goal of improving the resolution of seismic data. The high-resolution seismic data processing technology in this embodiment is based on the generalized S-transform and, taking into account the characteristics of the amplitude and phase of seismic records changing with time, space, and frequency, performs amplitude compensation and phase correction of seismic data in three-dimensional (time, space, and frequency) space.

[0060] According to the generalized S-transform theory, seismic records can be converted into a high-resolution time-frequency plane distribution, which not only possesses multi-scale focusing capabilities but also directly relates to the Fourier spectrum, preserving the absolute phase of the frequencies. It also avoids the time window width problem of the Short-Time Fourier Transform (STFT) and the scale width problem of the Wavelet Transform (WT). Therefore, performing amplitude spectrum compensation and phase spectrum correction of seismic data in the generalized S-domain can better improve data resolution. The specific steps are as follows: Figure 5 As shown:

[0061] ① Perform a generalized S-transform on each seismic signal to obtain its time-frequency distribution;

[0062] ② After fitting the time-frequency amplitude spectrum of the earthquake at a certain moment and obtaining the wavelet amplitude spectrum, the time-frequency amplitude spectrum at a certain moment can be compensated and corrected.

[0063] ③ Perform three-dimensional (time, space and frequency) component phase correction for each frequency component record;

[0064] ④ The generalized S-transform coefficients of the frequency corresponding to that moment are weighted by factors after amplitude spectrum compensation and phase spectrum correction;

[0065] ⑤ Reconstruct the compensation and correction results for each frequency within all time ranges back into the seismic record to complete the amplitude spectrum compensation and phase spectrum correction in the S-domain.

[0066] Figure 6 and Figure 7 These are comparison images of the cross-section before and after resolution enhancement processing of the inline 1476 line through the well. Figure 6 and Figure 7As can be seen, the profile resolution was significantly improved, the phase characteristics were well preserved, the reflection characteristics of the stratigraphic interface wave groups remained unchanged, and the interlayer information was abundant. From the comparison of the target layer spectrum, this high-resolution processing effectively broadened the bandwidth, increasing the bandwidth of the original pure wave data from 5-50 Hz to 5-70 Hz, and the dominant frequency from 22 Hz to 32 Hz. Conventional high-resolution processing often tends to reduce the signal-to-noise ratio and fidelity of seismic data. The stability of the seismic phase axis is improved, and the changes in seismic profile intensity, frequency, and phase relative relationships are well preserved. From the comparison analysis of the coherence attributes in the red circle area (within the blue box on the coherence plane) before and after processing, the low coherence revealed by the processed coherence attributes is less affected by complex phase amplitude, and the predicted fracture has less ambiguity. From the comparison of the original seismic record and the well-side calibration results of the high-resolution processed record, the correlation between the high-resolution processed results and the well-side traces is significantly improved, and the phase and amplitude characteristics correspond well with the synthetic record, indicating that this resolution-enhanced processing has high fidelity.

[0067] In step S4, the definitions of large and small cracks are as follows: The scale of cracks is a relative concept. Large cracks (fractures) refer to the fracture system that can be directly identified in the seismic data volume, usually with a width of 10 meters and a length of hundreds of meters; small cracks are cracks that are indirectly identified after special processing of the seismic data volume, and their scale is usually in the meter range.

[0068] In step S4, the third-generation coherence volume algorithm calculates using a small grid, setting a small grid of (△X, △Y, △Z), such as a 3*3*3 grid, to predict the distribution of relatively small fracture data volumes. In step S5, the third-generation coherence volume algorithm calculates using a large grid, for example, setting a large grid of (△X, △Y, △Z), such as a 9*9*9 grid, to predict the distribution of larger fracture data volumes. Different grid calculations represent different prediction methods for fractures of different scales. The grid size setting varies for data volumes with different dominant frequencies, and there is no unified quantitative standard. For data volumes with a dominant frequency of 30 Hz, a 9*9*9 grid is typically used for predicting larger fractures, while a 3*3*3 grid is typically used for predicting small-scale fractures. Through repeated experiments, the third-generation (C3) coherence volume algorithm in the EPOS3.0 interpretation system was selected, and it is believed that the third-generation coherence volume algorithm has significant advantages in terms of resolution and noise resistance, making it more suitable for reservoir fracture prediction research. The third-generation coherence technique essentially reveals the lateral heterogeneity of underground geological bodies from the perspective of the continuity of waveform characteristics between traces. These heterogeneities are usually related to abrupt changes in the properties of the geological body, such as lithological abrupt changes and fracture development zones. It is currently the most commonly used geometric seismic attribute for predicting the "chaotic reflection" characteristics of fracture zones.

[0069] Frequency-division coherence is used to calculate the degree of variation of seismic waveforms within different grid ranges. Large-scale frequency-division coherence can usually identify the distribution characteristics and patterns of large faults and clarify the distribution patterns of major fault zones, and is usually used to identify large faults. Small-scale frequency-division coherence is usually used to identify detailed features of faults and the patterns of fault combinations, and is of great significance for identifying fault-associated fractures and small fault systems.

[0070] Figure 8 This is a comparison chart of high-resolution coherence and high-resolution frequency-division small-scale coherence, from... Figure 8 The results show that high-resolution coherence can identify the planar extension trend of complex fractures and provide a clear understanding of the distribution patterns of large fractures; while small-scale frequency-division coherence can identify the detailed composition of large fractures, which are composed of multiple smaller fractures. Combining the two not only provides a clear understanding of the distribution of the strike of large fractures but also allows for the identification of the combination characteristics of smaller fractures.

[0071] In step S6, the parameters in the data volume fusion model include the data volume fusion threshold and the data volume fusion weight. By setting different data volume fusion thresholds and weights, fracture prediction data volumes at different scales are obtained. The data volume fusion model fuses these fracture prediction data volumes at different scales and outputs the prediction results for the fracture data volumes. Data volume fusion is a color fusion technique. By setting the transparency of the fused data volumes, the lower limit of the transparency is used to ensure that the fracture anomaly can be visually identified as a composite geological feature. This lower limit is the threshold of the fused volume, and there is no unified standard. Typically, for the third-generation coherent algorithm, the threshold for fracture interruption is set to 0.3. Fracture data with a value less than 0.3 are retained and participate in data fusion.

[0072] Pixel imaging fracture volume sculpting utilizes the changes in the continuity of longitudinal and transverse seismic waveforms to identify the distribution of small fracture systems. Specifically, it employs multiple methods to analyze the longitudinal and transverse continuity of seismic reflection wave groups, and enhances the energy of data in highly discontinuous sections. After energy enhancement, specific attribute values ​​are assigned to the data volume based on threshold differences to achieve fracture volume sculpting. Finally, the sculpted fracture volume is embedded into the seismic data to achieve three-dimensional fracture volume sculpting. Figure 9 As shown. Based on the extraction of properties along the layer slices of the fracture fusion body, the analysis of fracture properties at different scales in the main target layer is realized, clearly showing the planar distribution of the fracture system at different scales, such as... Figure 10 As shown.

[0073] In step S7, a quantitative simulation method for cracks and flexure zones is used to quantitatively simulate the small crack prediction results and flexure zone prediction results predicted by the maximum likelihood method. The specific process of the quantitative simulation method for cracks and flexure zones is as follows: for the crack and flexure zone prediction data volume, a threshold value for crack and flexure anomalies is set, the anomaly area inside the threshold value is quantified (e.g., defined as 1), and the non-anomaly area outside the threshold value is defined as 0, thereby realizing the quantitative simulation of the crack and flexure zone prediction data volume.

[0074] A crack prediction device based on multi-attribute fusion technology uses a Core i7-12700 processor and a Samsung 980PRO1T solid-state drive for memory.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A crack prediction method based on multi-attribute fusion technology, characterized in that, Includes the following steps: S1. Obtain the pre-stack data to be identified; S2. The pre-stack data to be identified is processed using a construction-guided filtering algorithm to obtain a high signal-to-noise ratio pre-stack seismic data volume. When S / N>5, the pre-stack seismic data volume with a high signal-to-noise ratio is defined as a pre-stack data volume with a high signal-to-noise ratio, where S is the signal and N is the noise. S3. Perform high-resolution target processing on the high signal-to-noise ratio pre-stack seismic data volume to obtain a high-resolution target processed data volume, and perform low-resolution target processing to obtain a low-resolution target processed data volume. Define frequencies higher than the dominant frequency of the original seismic data as high frequencies and frequencies lower than the dominant frequency of the original seismic data as low frequencies. S4. The high-resolution target processing data volume is processed using the third-generation coherence volume algorithm, the maximum likelihood analysis method, and the positive flexure prediction method. The processing results are the prediction results of different methods for small cracks and flexure zones. Small cracks are cracks that need to be processed in their seismic data volume to be identified. S5. The third-generation coherence volume algorithm is used to process the low-resolution target processing data volume. The processing result is the prediction result for large cracks. Large cracks are cracks that can be identified without processing their seismic data volume. S6. The results calculated by the third-generation coherent volume algorithm in steps S4 and S5 are fused into a single data volume using the data volume fusion method to form a data volume fusion model. S7. Quantitatively simulate the prediction results of small cracks and flexural zones predicted by the maximum likelihood method, and input the quantitative simulation results into the data volume fusion model. The data volume fusion model outputs the prediction results of crack data volume.

2. The crack prediction method based on multi-attribute fusion technology according to claim 1, characterized in that, In step S2, a construction-guided filtering algorithm is used to process the pre-stack data to be identified, resulting in a high signal-to-noise ratio pre-stack seismic data volume, including the following steps: S21. Conduct signal-to-noise ratio analysis of the raw seismic data volume and test the structural steering filter parameters; S22. Establish the optimal filter parameters by adjusting the tested parameters of the guided filter; S23. The pre-stack data to be identified is processed using the optimal filtering parameters and the constructed steerable filtering algorithm to obtain a pre-stack seismic data volume with high signal-to-noise ratio.

3. The crack prediction method based on multi-attribute fusion technology according to claim 2, characterized in that, In step S21, the guided filtering parameters include the data scale factor, the diffusion factor, and the number of iterations.

4. The crack prediction method based on multi-attribute fusion technology according to claim 1, characterized in that, In step S3, high-resolution target processing and low-resolution target processing are performed on the high signal-to-noise ratio pre-stack seismic data volume, including the following steps: S31. Perform dominant frequency spectrum analysis on the target layer of the high signal-to-noise ratio pre-stack seismic data volume to identify the effective frequency band and the cutoff frequencies of high and low frequencies; S32. Utilize WFI data volume target processing technology and employ Ricker wavelet or Yu's wavelet to perform frequency extension or frequency reduction processing on the spectrum of high signal-to-noise ratio pre-stack seismic data volume.

5. The crack prediction method based on multi-attribute fusion technology according to claim 1, characterized in that, In step S4, the third-generation coherent volume algorithm uses a 3*3*3 grid for calculation; in step S5, the third-generation coherent volume algorithm uses a 9*9*9 grid for calculation.

6. The crack prediction method based on multi-attribute fusion technology according to claim 1, characterized in that, In step S6, the parameters in the data volume fusion model include the data volume fusion threshold and the data volume fusion weight. By setting different data volume fusion thresholds and data volume fusion weights, crack prediction data volumes of different scales are obtained. The data volume fusion model fuses crack prediction data volumes of different scales and outputs the prediction results of crack data volumes.

7. The crack prediction method based on multi-attribute fusion technology according to claim 1, characterized in that, In step S7, a quantitative simulation method for cracks and flexural zones is used to quantitatively simulate the prediction results of small cracks and flexural zones predicted by the maximum likelihood method.

8. A crack prediction device based on multi-attribute fusion technology, characterized in that, The method includes at least one processor and at least one memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor to enable the at least one processor to perform any step of the method according to any one of claims 1-7.

Images