A tunnel boring machine array acoustic wave processing method and system based on predictive filtering

Abstract

The application provides a tunnel boring machine array acoustic wave processing method and system based on predictive filtering, which pre-processes the acquired tunnel boring machine array acoustic wave; calculates the structural complexity of the pre-processed data by using plane wave decomposition; extracts the data difference of different space-time positions by using the structural complexity, selects different filtering factors for different space-time positions; constructs an autoregressive prediction model based on the filtering factors; solves the autoregressive prediction model coefficients in different space-time positions by using the least square method, and realizes structure complexity guided denoising. The application uses the combined noise removal strategy of conventional data processing and structure complexity guided predictive filtering method, realizes time-space differentiated noise removal, improves the signal-to-noise ratio of the data, and lays a foundation for fine imaging of abnormal bodies.

Description

Technical Field

[0001] This invention belongs to the field of noise processing technology, and relates to a method and system for acoustic wave processing of tunnel boring machine arrays based on predictive filtering. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, most tunnel noise removal methods are based on the assumption of stable data. The suppression of unstable noise is ineffective, leaving residual noise that fails to meet the data requirements for precise detection during tunnel boring machine (TBM) construction. Furthermore, the noise components generated by mechanical vibrations at different locations are complex and intertwined, leading to a lack of clarity regarding the noise patterns within TBMs. In particular, the lack of methods for removing unstable noise with strong temporal and spatial differences is a key challenge hindering the full identification of effective array acoustic signals. Therefore, it is urgent to conduct noise characteristic analysis of tunnel data and propose an adaptive parameter selection method based on the differences in local data characteristics, using different filtering parameters for different temporal and spatial locations, to achieve high signal-to-noise ratio data acquisition for TBM array acoustic waves. Summary of the Invention

[0004] To address the aforementioned problems, this invention proposes a method and system for acoustic processing of tunnel boring machine arrays based on predictive filtering. This invention utilizes a combined noise removal strategy of conventional data processing and predictive filtering guided by structural complexity to achieve spatiotemporal noise removal, improve the signal-to-noise ratio of the data, and lay the foundation for fine imaging of anomalies.

[0005] According to some embodiments, the present invention adopts the following technical solution:

[0006] A method for acoustic wave processing of a tunnel boring machine array based on predictive filtering includes the following steps:

[0007] Preprocess the acquired acoustic waves from the tunnel boring machine array;

[0008] The structural complexity of the preprocessed data is calculated using plane wave deconstruction.

[0009] By leveraging structural complexity, data differences at different spatiotemporal locations are extracted, and different filtering factors are selected for different spatiotemporal locations.

[0010] Based on the filtering factor, an autoregressive prediction model is constructed.

[0011] By using the least squares method, the coefficients of the autoregressive prediction model in different spatiotemporal locations are solved to achieve denoising guided by structural complexity.

[0012] As an alternative implementation, the preprocessing includes:

[0013] Remove direct waves;

[0014] Enhance signal strength at long distances and deep locations using equalization and gain methods;

[0015] FK filtering is used to remove signals in the data whose apparent velocity interference exceeds a predetermined value;

[0016] Bandpass filtering is used to remove random noise.

[0017] As an alternative implementation, the specific process of calculating the structural complexity of the preprocessed data using plane wave decomposition includes estimating the tilt angle of the preprocessed data using adjacent acoustic channel data.

[0018] As an alternative implementation method, the specific process of selecting different filter factors for different spatiotemporal locations includes selecting different filter factors according to the different data complexity. The higher the data complexity, the longer the filter factor is used.

[0019] As an alternative implementation method, the calculation method for the filtering factor of two-dimensional data includes:

[0020] Spatiotemporal windowing is performed on the two-dimensional data, and an expression for the adaptive factor within a certain window is constructed based on the size of the window in the time direction and the size of the offset direction.

[0021] The distribution of the adaptive factor is averaged and grouped.

[0022] The size of the filter factor is determined based on the adaptive factor, and the size of the filter factor is positively correlated with the adaptive factor.

[0023] As a further limitation, the expression for the adaptive factor is:

[0024]

[0025] Where Cg represents the adaptive factor of the g-th sliding window, σ ij The tilt angles of the i-th sampling point and the j-th probe are calculated using the plane wave deconstruction method. The window size is M in the time direction and N in the offset direction. S is the average tilt angle within the sliding window.

[0026] As an alternative implementation method, the method for calculating the filtering factor of three-dimensional data includes first using the three-dimensional plane wave deconstruction method to calculate the local dip angle of the data along the main survey line direction and the connecting survey line direction, and then establishing an adaptive factor along the main survey line direction and the connecting survey line direction.

[0027] The adaptive factors are grouped along the main survey line direction and the connecting survey line direction;

[0028] The statistical characteristics of the data within the sliding window are re-statistically analyzed to obtain different filtering factors.

[0029] As a further constraint, the adaptive factor along the main survey line direction and the connecting survey line direction is:

[0030]

[0031]

[0032] in The data is locally tilted within a certain sliding three-dimensional window, and its lengths in the sampling time direction, the main survey line direction, and the connecting survey line direction are g, h, and k, respectively. and These are the average local tilt angles of the sliding window, and the active window size for the 3D data is w1×w2×w3.

[0033] As an alternative implementation method, the specific process of constructing an autoregressive prediction model based on the filtering factor includes:

[0034] Based on the characteristics of plane waves in acoustic wave data, plane waves in the frequency domain are expressed.

[0035] L channels of acoustic wave data are selected for prediction, and the prediction process is the process of constructing an autoregressive prediction model for time series.

[0036] A tunnel boring machine array acoustic wave processing system based on predictive filtering includes:

[0037] The preprocessing module is configured to preprocess the acquired acoustic waves from the tunnel boring machine array;

[0038] The structural complexity calculation module is configured to use plane wave deconstruction to calculate the structural complexity of the preprocessed data.

[0039] The filter factor selection module is configured to extract data differences at different spatiotemporal locations using structural complexity and select different filter factors for different spatiotemporal locations.

[0040] The model building module is configured to build an autoregressive prediction model based on the filtering factor;

[0041] The denoising module is configured to use the least squares method to solve the coefficients of the autoregressive prediction model, thereby achieving denoising guided by structural complexity.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] This invention obtains the tilt angle information of the data through the plane wave deconstruction method. Based on this, it constructs the relationship between the local features of the data and the tilt angle using the structural complexity evaluation method, and obtains the adaptive filtering factor of the data at different locations. This provides filtering parameters for the removal of least squares noise, and the final denoised data is cleaner. There is no noise residue on the difference profile and the local similarity profile, the signal leakage phenomenon is weak, and the noise removal effect is good.

[0044] This invention has excellent noise removal capabilities and weak signal leakage. Compared with traditional predictive filtering methods, it improves the signal-to-noise ratio and data quality. It can suppress noise while reducing signal leakage and obtain array acoustic data with a high signal-to-noise ratio. Attached Figure Description

[0045] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0046] Figure 1 This is a flowchart illustrating the present invention;

[0047] Figure 2 This is a flowchart illustrating a typical embodiment of the present invention;

[0048] Figure 3 This is the statistical analysis and probability distribution of the filter factors in the comparative example of this invention;

[0049] Figure 4 This is a comparison chart of the denoising results using the SCG-PF method;

[0050] Figure 5 This is a comparison chart of the spectral results of different denoising methods when x = 2.5m;

[0051] Figure 6 It is simple three-dimensional acoustic wave data;

[0052] Figure 7 This is a comparison chart of the denoising results using traditional methods;

[0053] Figure 8 It is a three-dimensional simple data filtering factor. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0057] This invention proposes a joint noise removal strategy combining conventional data processing and structural complexity-guided predictive filtering (SCG-PF) methods, focusing on noise removal from non-steady-state data with strong temporal and spatial differences.

[0058] This invention uses tilt angle information as a bridge to construct the relationship between data features and structural complexity, proposes an adaptive factor calculation method guided by local structural complexity, and constructs a denoising model based on adaptive filtering factors to achieve non-steady-state noise removal from tunnel boring machines. This method, starting from noise characteristics, achieves spatiotemporally differentiated noise removal, improves the signal-to-noise ratio of the data, and lays the foundation for fine imaging of anomalies.

[0059] Specifically, such as Figure 1 As shown, reflected waves are mainly used as effective information. Direct waves, regular interference noise with continuous in-phase axes, and random interference noise with discontinuous in-phase axes are the information that needs to be removed. First, conventional data processing is used: the strong amplitude of the direct wave reduces the recognition ability of other data and needs to be removed first; when the data contains velocity-induced in-phase axis interference, for example, when there is a strong in-phase axis with an apparent velocity of approximately 340 m / s in the array acoustic wave data, it indicates that the data contains alarm, knocking, or other information. FK filtering is used to remove signals with obvious apparent velocity interference from the data; to compensate for the influence of signal diffusion effects, equalization and gain methods are used to enhance the signal strength of distant and deep signals. After removing data with specific apparent velocities, frequency domain removal methods are needed. It can be found that the frequency band of some noise is not within the frequency band excited by the array acoustic wave probe; therefore, preliminary bandpass filtering can be used to remove some random noise.

[0060] Then, the focus is on removing non-steady-state random noise with strong spatiotemporal differences: ① The structural complexity of the data is calculated by deconstructing the data using plane waves; ② Considering the strong spatiotemporal differences of the noise, a method is proposed to extract the data differences at different spatiotemporal locations using structural complexity, and different filtering factors are selected for different spatiotemporal locations; ③ The Levinson matrix of the non-steady-state data is constructed, and the auto-regressive (AR) prediction model for forward and backward prediction is calculated; ④ Based on this, the coefficients of the AR model are solved using the least squares method to achieve SCG-PF denoising guided by structural complexity.

[0061] The specific steps will be described in detail below.

[0062] First, there is a structural complexity analysis based on tilt angle characteristics.

[0063] Arrayed acoustic wave data exhibits local continuity along the same phase axis and local plane wave characteristics with no energy attenuation. Therefore, the tilt angle of the data can be estimated using data from adjacent acoustic wave channels. Assuming the acoustic signal is x(t), and the profile formed by multiple probes is X = {x1(t), x2(t), ..., xn(t)}, where t is the sampling time and n is the number of acoustic wave probes, the acoustic signal satisfies the following relationship:

[0064]

[0065] Its general solution in the frequency domain can be expressed as:

[0066] U(x)=U(0)e iωσx

[0067] Where U is the frequency domain field value obtained by Fourier transform of the plane wave field u, and σ is the slope of the acoustic signal. Therefore, the field value at any position can be obtained by applying a certain phase delay to the field value U(0) of the first channel.

[0068] Similarly, the slope can be used as a bridge to obtain the relationship between any two acoustic wave data:

[0069] U(x)=U(x-1)e iωσ

[0070] The above equation can be expressed by introducing the Z-transform as follows:

[0071] S(Zx)=1-e iωσ Zx

[0072] Based on this, the optimization problem of minimizing the residuals of adjacent acoustic wave data can be solved to obtain:

[0073] Q'(σ0)Δσx+C(σ)x≈0

[0074] Where Δσ is the change in tilt angle, and the initial tilt angle is σ0. The slope can be estimated by solving the plane wave prediction operator C using the Gauss-Newton or quasi-Newton method.

[0075] The tilt angle variation of data can characterize the structural complexity of the data. However, how to extract the spatiotemporal differences in the observed data and obtain adaptive filter factors for different spatiotemporal locations is the key research topic of this section. Especially for predictive filtering methods, the reasonable selection of filter factors plays a crucial role in obtaining data with high signal-to-noise ratio. Therefore, before performing noise removal, it is necessary to study the influence of filter factors on noise removal and how to obtain adaptive filter factors based on the spatiotemporal differences of the data.

[0076] In the process of denoising acoustic data using predictive filtering, the choice of filter factor directly affects the denoising effect. Based on the data structure complexity, the spatiotemporal differences of the data are extracted. Then, different filter factor lengths are selected using these spatiotemporal differences to remove random noise from non-steady-state data.

[0077] Experimental research revealed that the length of the filter factor significantly impacts the data denoising effect. A shorter filter factor results in stronger denoising capability but at the cost of losing effective signal; a longer filter factor, while having less denoising capability, retains more effective signal and minimizes signal leakage. Therefore, adaptively selecting the filter factor for different spatiotemporal locations is crucial for improving the signal-to-noise ratio. In the data processing, this paper found that for structurally complex data, a longer filter factor can retain effective signal and avoid signal leakage, while for structurally simple data, a shorter filter factor can remove more noise.

[0078] This invention proposes an adaptive factor calculation method based on statistical data patterns, bridging the gap between tilt angle and filter factor, and constructing an adaptive filter factor selection method for different spatiotemporal locations, ultimately achieving noise removal from non-steady-state data with strong spatiotemporal variations. The specific implementation method is as follows:

[0079] To achieve different filter factors at different locations, this paper performs spatiotemporal windowing on the data. For two-dimensional data, the window size is M in the time direction and N in the offset direction. The adaptive factor within a certain window can then be expressed as:

[0080]

[0081] Where Cg represents the adaptive factor of the g-th sliding window, σ ijThe tilt angles of the i-th sampling point and the j-th probe are calculated using the plane wave deconstruction method. The window size limits the length of the adaptive filter factor. Therefore, the window size is larger than the maximum adaptive filter factor length. Here, S is the average tilt angle within the sliding window. Therefore, once each Cg within the sliding window is calculated, the local adaptive factor for all windows of that data can be obtained. The distribution of Cg is divided into 10 groups with an interval of:

[0082]

[0083] To avoid the problem of boundary data not being able to be grouped, ε is a very small constant, its size being one ten-thousandth of the difference between the maximum and minimum Cg. Once the adaptive factor Cg is determined, the length of the filter factor for any window will be adaptively selected according to the magnitude of Cg. The adaptive factor is essentially the variance of the tilt angle data, representing the degree of dispersion of the tilt angle feature. This feature remains unchanged under the same data but different phase delays, thus solving the requirement that the evaluation parameters do not change when the data is the same but the tilt angle is different. For example, when the adaptive factor is large, it means that the tilt angles of the data within the window are dispersed, indicating that the data is more complex. In this case, a larger filter factor is selected to avoid signal leakage. When the adaptive factor is small, it means that the tilt angle values ​​of the data within the window are clustered, indicating that the data is more simple. In this case, a smaller filter factor is selected to improve the noise reduction level.

[0084] Based on the statistical characteristics of the tilt angle, an adaptive filtering factor acquisition method is proposed as follows:

[0085] p = (i-1) + 2, min(C g )+(i-1)×dC g <C g <min(C g )+i×dC g ,

[0086] i = 1, 2, ..., 10

[0087] For three-dimensional acoustic data, besides the time orientation being consistent with two-dimensional data, the offset direction includes both the main survey line direction and the connecting survey line direction (In-line and Cross-line). Therefore, the active window size for the data is W1×W2×W3. In calculating the three-dimensional adaptation factor, the local dip angles along the In-line and Cross-line directions are first calculated using a three-dimensional plane wave deconstruction method. The adaptation factors calculated along the In-line and Cross-line directions are:

[0088]

[0089]

[0090] in The data is locally tilted within a sliding 3D window, with lengths g, h, and k in the sampling time direction, in-line direction, and cross-line direction, respectively. The obtained adaptive factor... and These represent the acoustic data along the in-line and cross-line directions, respectively. and These are the average values ​​of the local tilt angles of the sliding window. Similar to the two-dimensional case, this paper also divides the adaptive factors into 10 groups in both the In-line and Cross-line directions. To avoid situations where boundary values ​​cannot be grouped, this paper also adds a small threshold εx and εy to the difference between the formulas in the two directions. Therefore, the interval between each group of adaptive factors in the In-line and Cross-line directions is:

[0091]

[0092]

[0093] After grouping the adaptive factors, the statistical characteristics of the data within the sliding window can be re-statistically analyzed, thereby obtaining different filter factors. Unlike two-dimensional data, three-dimensional data requires obtaining filter factors along the in-line and cross-line directions, where wi and wj are the grouping indices in the two directions, each with a size of [1, 2, ..., 10].

[0094]

[0095]

[0096] In summary, by introducing an adaptive factor, we proposed the spatiotemporal variation information of the data, built a bridge between structural complexity and filtering parameters (filtering factors), and realized the selection of adaptive filtering factors based on spatiotemporal location differences.

[0097] An AR model is constructed using filter factors and least squares constraints are solved to achieve adaptive noise removal by selecting different filter factors at different spatiotemporal locations. This removes noise while preserving as much signal as possible to obtain high signal-to-noise ratio data.

[0098] We begin by introducing a noise removal method for steady-state data, based on the plane wave characteristics of acoustic data. A plane wave in the frequency domain can be represented as:

[0099] U F (x+1)=e 2fπσ UF (x).

[0100] Among them, U F (x) is the Fourier transform of the original data, and f is a frequency slice. Therefore, L channels of acoustic data can be used for prediction. The length of L represents the length of the filter factor. This allows for noise removal. For example, L-1 acoustic data points can be used to predict L channels of acoustic data; therefore, the prediction process can be:

[0101]

[0102] in This prediction process is the construction process of the AR model for the time series. Of course, this only applies to the process from x, x-1, up to x+L. Compared to acoustic data, this process involves both forward prediction from smaller to larger measurement point numbers and reverse prediction from larger to smaller measurement point numbers. Therefore, an AR model that includes both forward and reverse prediction can be represented as:

[0103]

[0104] By processing the frequency points after the Fourier transform in sequence, noise removal of steady-state characteristic data can be achieved.

[0105] For non-stationary data with strong spatiotemporal differences, the data tilt angle σ changes with the acoustic wave reception offset. Therefore, it is necessary to perform non-stationary noise removal work. For data containing non-stationary noise, the AR model for predictive filtering is expressed as:

[0106]

[0107] Comparing AR models for steady-state and non-steady-state data reveals that the tilt angle of non-steady-state data is related to the offset x of the acoustic data, while under steady-state conditions, the tilt angle change is independent of the data offset. For three-dimensional data, the AR model can be expressed as:

[0108]

[0109] In the text, Lx and Ly represent the filter factor lengths along the in-line and cross-line directions, respectively. Noise removal can be achieved by solving the above equation.

[0110] In summary, this paper obtains the tilt information of the data through the plane wave deconstruction method. Based on this, the relationship between the local features of the data and the tilt angle is constructed using the structural complexity assessment method, and adaptive filtering factors of the data at different locations are obtained, providing filtering parameters for the removal of least squares noise.

[0111] Based on the constructed autoregressive AR model, this paper utilizes the least squares method to achieve efficient and steady-state AR model solving. The above equation can be simplified to matrix form as follows:

[0112] U = Fy

[0113] Where U is The matrix representation of y is The matrix form, F is U F The matrix form of (x+xk, y+yk). Therefore, bidirectional prediction can be represented as:

[0114]

[0115] The filter factor has a length of p, and the equation consists of two parts. From the left side of the equation, we can see that the first part is from p+1 to position X, and the second part is from 1 to Xp. This corresponds to the forward prediction from small offset to large offset and the reverse prediction from large offset to small offset, which correspond to data at different spatiotemporal locations.

[0116] The process of solving for y is equivalent to the inverse problem of solving y given UF. This predictive noise filtering method can be achieved by constructing a problem of minimizing the error:

[0117]

[0118] For SCG-PF filtering, the above formula is expressed as:

[0119]

[0120] By solving the least squares problem of minimizing the above equation by differentiating it, we have:

[0121] F T U = F T Fy

[0122] Matrix form F T F is in the left Blitz form, which can be solved using the Levinson recursive method. To improve the steady-state performance of the equations for solving array acoustic data, a small perturbation variable ε3 is added. Therefore, the solution obtained is:

[0123]

[0124] Where I is the identity matrix, at this time The filtered data can be obtained by multiplying the predicted parameters by the original matrix F.

[0125]

[0126] As a typical embodiment, its data processing flow is as follows: Figure 2 As shown.

[0127] The removal of 3D data is included; for 2D data, only the time and offset dimensions need to be considered. Therefore, the SCG-PF method for denoising includes the following steps:

[0128] (I) Select the time-space window size for the data;

[0129] (II) Calculate the local slope of the traversing sliding window using the plane wave deconstruction method;

[0130] (III) Calculate the adaptive factor based on the statistical characteristics of the slope within the window;

[0131] (IV) Divide all adaptive factors in the windows into 10 groups according to their numerical values. The length of the filter factor in each group is [3, 12].

[0132] (V) The least squares method is used to solve the autoregressive model of data in different time-space windows to achieve noise removal.

[0133] To verify the effectiveness of the proposed SCG-PF method for removing noise from tunnel boring machines with strong spatiotemporal variations, a comparative analysis was conducted between traditional predictive filtering methods, singular spectrum decomposition, energy information constraints, and similarity information constraints, and a predictive filtering method guided by structural complexity.

[0134] Two-dimensional acoustic wave data: For acoustic wave data with simple structures, the tilt angle is first evaluated using plane wave deconstruction. Based on this, an adaptive factor is calculated and a filtering factor is obtained using a time-space windowing strategy according to the statistical characteristics of the local tilt angle of the data (e.g., ...). Figure 3 As shown in Figure a, the probability distribution of different filter factors is as follows: Figure 3 As shown in b.

[0135] It can be seen that the filter factors exhibit a cross-shaped distribution in the profile, consistent with the shape of the data phase axis. Furthermore, it can be observed that the filter factors are longer in the middle part of the phase axis, and the filter factors are smallest (p=3) in the region without a phase axis distribution, which is consistent with the distribution of data structure complexity. In addition, the pie chart of filter factor statistics shows that most filter factors are short (p=3), accounting for 70%. Filter factors 3-5 are defined as short filter factors, 6-8 as medium filter factors, and 9-12 as long filter factors. Short filter factors account for nearly three-quarters, while medium and long filter factors account for only one-quarter. This aligns with the requirement of using shorter filter factors for simple data structures. Based on this, the data was denoised using Multichannel Singular Spectrum Analysis (MSSA) and the SCG-PF method proposed in this study. The results are as follows... Figure 4 As shown.

[0136] The SSA method yielded an SNR of 6.63 dB, while the SCG-PF method yielded an SNR of 6.74 dB. The comparison shows that both SSA and SCG-PF methods demonstrate good denoising performance for simple structured data, particularly in differential profiles. Figure 4 Signals (c) and (d) show almost no leakage. Furthermore, this paper compares the amplitude spectra of the denoised data with offset x = 2.5 m after filtering with filter factor length p = 3 and p = 12, using the MSSA method and the SCG-PF method. The results are as follows: Figure 5 As shown in ae.

[0137] from Figure 5 It can be seen that the SSA and SCG-PF methods effectively remove most of the noise, especially around the 4kHz frequency (arrow position in the figure). The two methods result in minimal data frequency loss within the main frequency band and can effectively reduce signal leakage.

[0138] 2) Three-dimensional acoustic data: For three-dimensional acoustic data, this paper constructs three sets of linear in-phase axis data that are staggered vertically. From top to bottom, they consist of convolutions of Ricker wavelets with dominant frequencies of 4kHz, 3kHz, and 4kHz, with a reflection coefficient of 1. The sampling interval is 2e-5s, and the number of sampling points in the Time, In-line, and Cross-line directions are 300, 58, and 60, respectively. The selected sliding window size is 30×30×30. After adding noise to the noiseless data, the signal-to-noise ratio and peak signal-to-noise ratio are 2.82dB and 18.85dB, respectively. Figure 6 As shown.

[0139] Predictive filtering (PF) and multichannel singular spectrum analysis (MSSA) are used to analyze noisy data. Figure 6 The denoised profile obtained after processing is as follows: Figure 7 As shown, to further analyze the signal leakage phenomenon of different denoising methods, this paper introduces a local similarity method. This method calculates the local similarity of the denoised profile and the noise-removed profile and compares the difference between the two. When the local similarity at a certain location is strong, it indicates that there is signal leakage in that area. Conversely, if the local similarity result on the profile is close to 0, it indicates that there is no signal leakage and the denoising effect is good.

[0140] For this noisy data, the tilt angle information is calculated using a plane wave deconstruction method based on a 30×30×30 window, such as... Figure 8 As shown, for two intersecting in-phase axes, it can be observed that in the in-line direction, as the offset increases, the data phase exhibits a positive delay, resulting in a positive tilt angle value; conversely, in the cross-line direction, as the offset increases, the data phase exhibits a negative delay, resulting in a negative tilt angle value. This characteristic also applies to the cross-line direction. Based on this, this paper utilizes an adaptive factor to evaluate tilt angle information and uses statistical laws to match filter factors at different locations according to the spatiotemporal differences in the data.

[0141] It can be observed that the filter factor is longer in areas with complex data structures (bottom) and shorter in areas with simple structures (top). This better adapts to changes in data structure, which helps avoid signal leakage and improves the signal-to-noise ratio. Using the filter factor to denoise noisy data, it is clear that the denoised data is cleaner. No noise residue is observed in the difference profile and local similarity profile, signal leakage is weaker, and the noise removal effect is better.

[0142] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0143] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0144] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0145] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0146] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0147] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for processing acoustic waves in a tunnel boring machine array based on predictive filtering, characterized by, The method comprises the following steps: The tunnel boring machine array acoustic wave obtained is preprocessed; The structural complexity of the preprocessed data is calculated by using plane wave decomposition; Data differences at different space-time positions are extracted by using the structural complexity, and different filtering factors are selected for different space-time positions, and the filtering factor calculation method of two-dimensional data comprises: Time-space window processing is performed on the two-dimensional data, and an expression of an adaptive factor in a certain window is constructed according to the size of the window in the time direction and the size of the offset distance direction; The distribution of the adaptive factor is evenly grouped; The size of the filtering factor is determined according to the adaptive factor, and the size of the filtering factor is positively correlated with the adaptive factor; An autoregressive prediction model is constructed based on the filtering factor; The autoregressive prediction model coefficients in different space-time positions are solved by using the least square method, and the structural complexity guided denoising is realized.

2. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as claimed in claim 1, characterized in that, The preprocessing comprises: Removing direct waves; Enhancing the signal strength of far and deep paths by using equalization and gain methods; Removing signals with a viewing speed interference exceeding a predetermined value in the data by using F-K filtering; Removing random noise by using band-pass filtering.

3. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as claimed in claim 1, characterized in that, The specific process of calculating the structural complexity of the preprocessed data by using plane wave decomposition comprises estimating the dip angle of the preprocessed data by using adjacent acoustic wave channel data.

4. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as defined in claim 1, characterized in that, The specific process of selecting different filtering factors for different space-time positions comprises selecting different filtering factors according to the different data complexity, and the higher the data complexity, the longer the filtering factor used.

5. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as defined in claim 1, wherein, The expression of the adaptive factor is: wherein, Cg denotes the adaptive factor for the g th sliding window, is the dip angle of the i th sample and j th receiver calculated by the plane wave decomposition method, the window size is M in the time direction and N in the offset direction, S is the average dip angle in the sliding window.

6. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as defined in claim 1, wherein, The filtering factor calculation method of three-dimensional data comprises first calculating the local dip angle of the data along the main line direction and the contact line direction by using the three-dimensional plane wave decomposition method, and establishing the adaptive factor along the main line direction and the contact line direction; The adaptive factors along the main line direction and the contact line direction are grouped; The statistical characteristics of the data in the sliding window are re-counted to obtain different filtering factors.

7. A method of processing acoustic waves in a tunnel boring machine array based on predictive filtering as claimed in claim 6, characterized in that, The adaptive factors along the main line direction and the contact line direction are: wherein is the data of local dip angle in a certain sliding stereoscopic window, the length of which in the sampling time direction, the main survey line direction and the liaison survey line direction is g, h and k , and is the average value of the local dip angle of the sliding window respectively, and the active window size of the three-dimensional data is w1×w2×w3 .

8. A method of processing array acoustic waves for a tunnel boring machine based on predictive filtering as defined in claim 1, wherein, The specific process of constructing the autoregressive prediction model based on the filtering factor comprises: Based on the characteristics of the plane wave of the acoustic wave data, the frequency domain plane wave is expressed. Selecting L The prediction process is the autoregressive prediction model construction process of time series.

9. A tunnel boring machine array acoustic wave processing system based on predictive filtering, characterized by, The method comprises: A preprocessing module configured to preprocess the tunnel boring machine array acoustic wave obtained; A structural complexity calculation module configured to calculate the structural complexity of the preprocessed data by using plane wave decomposition; A filtering factor selection module configured to extract data differences at different space-time positions by using the structural complexity, and select different filtering factors for different space-time positions, and the filtering factor calculation method of two-dimensional data comprises: Time-space window processing is performed on the two-dimensional data, and an expression of an adaptive factor in a certain window is constructed according to the size of the window in the time direction and the size of the offset distance direction; The distribution of the adaptive factor is evenly grouped; The size of the filtering factor is determined according to the adaptive factor, and the size of the filtering factor is positively correlated with the adaptive factor; A model construction module configured to construct an autoregressive prediction model based on the filtering factor; The denoising module is configured to solve the autoregressive prediction model coefficient by using a least square method, and realize the structure complexity guided denoising.

Images