Method and system for detecting abnormal sound of heading machine through analysis of acoustic signal

Abstract

The application provides a tunneling machine abnormal sound detection method and system for sound wave signal analysis, and relates to the technical field of tunneling machine detection. The method comprises the following steps: continuously monitoring the sound of the tunneling machine by a distributed piezoelectric ceramic sound wave probe; pre-processing and analyzing the original array sound wave data according to a sound wave preprocessing mechanism; calling an energy focusing estimation strategy to estimate and analyze the reflection matrix; forming a working sound visual diagram of the tunneling machine according to the wave velocity distribution information, obtaining the abnormal sound source position, predicting the abnormal type of the abnormal sound source position, and performing early warning processing on the abnormal sound source position. The technical problem of low accuracy and poor real-time performance of abnormal sound positioning and identification in the prior art is solved by relying on manual and simple sensor monitoring. The technical effects of improving the accuracy and real-time performance of abnormal sound positioning and identification of the tunneling machine and ensuring the operation safety and fault early warning capability of the tunneling machine are achieved.

Description

Technical Field

[0001] This invention relates to the field of tunneling machine testing technology, specifically to a method and system for detecting abnormal noises in tunneling machines using acoustic signal analysis. Background Technology

[0002] In the modern coal mining and underground engineering construction fields, tunneling machines are key construction equipment, and their operating status directly affects project progress and safety. Currently, the detection of abnormal noises from tunneling machines mainly relies on manual inspections and traditional, simple monitoring methods. Manual inspections are easily affected by subjective factors, as different personnel have varying abilities to perceive and judge abnormal noises. Furthermore, in complex and harsh construction environments, manual inspections struggle to accurately capture abnormal noise signals. Traditional, simple monitoring methods generally suffer from limitations in sensor deployment, insufficient spatial coverage, significant noise interference, and inadequate signal processing capabilities. This results in the inability to promptly and accurately identify and locate abnormal noises, making it difficult to ensure construction safety and fault early warning capabilities.

[0003] In existing technologies, abnormal noise detection of tunneling machines relies on manual labor and simple sensor monitoring, which has technical problems such as low accuracy in abnormal noise location and identification and poor real-time performance. Summary of the Invention

[0004] This application provides a method and system for detecting abnormal noises in tunneling machines based on acoustic signal analysis. This method addresses the technical problems in existing technologies where abnormal noise detection in tunneling machines relies on manual labor and simple sensor monitoring, resulting in low accuracy and poor real-time performance in noise location and identification.

[0005] In view of the above problems, this application provides a method and system for detecting abnormal noises in tunneling machines by analyzing acoustic signals.

[0006] The first aspect of this application provides a method for detecting abnormal noises during tunneling machine operation using acoustic signal analysis. The method includes: dynamically and continuously monitoring the tunneling machine's sound using a distributed piezoelectric ceramic acoustic probe to obtain raw array acoustic data; preprocessing and analyzing the raw array acoustic data according to an acoustic preprocessing mechanism to obtain a reflection matrix; estimating and analyzing the reflection matrix using an energy focusing estimation strategy to obtain wave velocity distribution information; forming a working sound visualization of the tunneling machine based on the wave velocity distribution information, and identifying the location of abnormal noise sources in the working sound visualization; predicting the anomaly type of the abnormal noise source location, and performing early warning processing for the anomaly type of the abnormal noise source location.

[0007] Preferably, the method further includes: the distributed piezoelectric ceramic acoustic probes are deployed on the cutterhead surface of the tunneling machine based on a radial survey line pre-plan, and are completed by a rotation completion pre-plan; wherein, the radial survey line pre-plan refers to setting 6 radial survey lines with an included angle of 60° on the cutterhead surface, with M probes deployed on each survey line, where M is an integer greater than or equal to 3 and less than or equal to 5; wherein, the rotation completion pre-plan refers to the cutterhead surface being automatically relocked after being rotated clockwise by 60°, 120° or 180°, wherein the coupling state between the distributed piezoelectric ceramic acoustic probes and the cutterhead surface remains unchanged.

[0008] Preferably, the original array acoustic data is preprocessed and analyzed according to the acoustic preprocessing mechanism to obtain a reflection matrix, including: preprocessing the original array acoustic data according to the primary preprocessing scheme in the acoustic preprocessing mechanism to obtain first-level acoustic data; preprocessing the first-level acoustic data according to the secondary preprocessing scheme in the acoustic preprocessing mechanism to obtain second-level acoustic data; and establishing the reflection matrix based on the second-level acoustic data; wherein, the primary preprocessing scheme includes preprocessing steps such as direct wave cutoff, bandpass filtering, FK filtering, energy equalization, and gain compensation.

[0009] Preferably, the primary acoustic data is preprocessed according to the secondary preprocessing scheme in the acoustic preprocessing mechanism to obtain secondary acoustic data, including: activating a prediction filter to preprocess the primary acoustic data according to the secondary preprocessing scheme to obtain the secondary acoustic data; wherein, before activating the prediction filter, the following steps are taken: calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe according to the plane wave deconstruction strategy, wherein the first acoustic probe refers to any one of the distributed piezoelectric ceramic acoustic probes; calculating the first tilt angle variance based on the first local tilt angle, and using the first tilt angle variance as the first structural complexity index of the first part, wherein the first part refers to the position of the cutterhead surface monitored by the first acoustic probe; matching the first filter factor length corresponding to the first structural complexity index, and constructing the prediction filter based on the first filter factor length.

[0010] Preferably, calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe according to the plane wave deconstruction strategy includes: performing a Fourier transform on the first acoustic signal to obtain a first frequency domain signal; calculating the first local tilt angle based on the first frequency domain signal, wherein calculating the first local tilt angle includes: obtaining the first adjacent probe of the first acoustic probe; comparing the first acoustic probe with the first adjacent probe to obtain a first comparison parameter, wherein the first comparison parameter includes a first phase delay and a first probe spacing; obtaining the first signal frequency of the first frequency domain signal, and calculating the first local tilt angle according to the plane wave deconstruction strategy, combined with the first phase delay and the first probe spacing.

[0011] Preferably, the method further includes: the first structural complexity index is proportional to the length of the first filter factor.

[0012] Preferably, the energy focusing estimation strategy is invoked to estimate and analyze the reflection matrix to obtain wave velocity distribution information, including: performing Radon transform on the reflection matrix and extracting the energy focusing width at the slope k=-1 of the main diagonal; and analyzing the energy focusing width according to the energy focusing estimation strategy to obtain the wave velocity distribution information.

[0013] Preferably, forming the working sound visualization of the tunneling machine based on the wave velocity distribution information includes: calculating the first pseudospectral energy intensity of the first acoustic signal and forming a pseudospectral energy function; using the azimuth angle corresponding to the maximum value of the pseudospectral energy function as a weighting coefficient to perform weighted correction on the Green function to obtain a target Green function; processing the target Green function using a time reversal operator to form the working sound visualization, wherein forming the working sound visualization includes: reversing the target Green function on the time axis and convolving the reversed target Green function with the first acoustic signal to obtain a first convolution result; integrating the first convolution result to obtain the working sound visualization.

[0014] Preferably, predicting the anomaly type of the abnormal noise source location includes: collecting real-time operating status information of the abnormal noise source location; traversing the vectorized real-time operating status information in the historical operation database to obtain the most similar historical operating status information; extracting the historical anomaly type from the most similar historical record corresponding to the most similar historical operating status information, and using it as the anomaly type.

[0015] A second aspect of this application provides a tunneling machine noise detection system based on acoustic signal analysis. The system includes: an acoustic monitoring module for dynamically and continuously monitoring the tunneling machine's sound using a distributed piezoelectric ceramic acoustic probe to obtain raw array acoustic data; a data preprocessing and analysis module for preprocessing and analyzing the raw array acoustic data according to an acoustic preprocessing mechanism to obtain a reflection matrix; an energy focusing estimation module for retrieving an energy focusing estimation strategy to estimate and analyze the reflection matrix to obtain wave velocity distribution information; a noise source acquisition module for forming a working sound visualization of the tunneling machine based on the wave velocity distribution information and acquiring the location of the noise source in the working sound visualization; and an anomaly prediction and early warning module for predicting the anomaly type of the noise source location and performing early warning processing for the anomaly type.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] The method provided in this application embodiment performs dynamic and continuous sound monitoring of a tunneling machine using a distributed piezoelectric ceramic acoustic wave probe to obtain raw array acoustic wave data. The raw array acoustic wave data is preprocessed and analyzed according to an acoustic wave preprocessing mechanism to obtain a reflection matrix. An energy focusing estimation strategy is then used to estimate and analyze the reflection matrix to obtain wave velocity distribution information. Based on the wave velocity distribution information, a working sound visualization of the tunneling machine is formed, and the location of abnormal noise sources in the working sound visualization is obtained. The abnormality type of the abnormal noise source location is predicted, and early warning processing for the abnormality type is performed on the abnormal noise source location. This achieves the technical effect of improving the accuracy and real-time performance of abnormal noise location and identification of the tunneling machine, ensuring the safe operation of the tunneling machine and its fault early warning capability. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the method for detecting abnormal noises during tunneling machine operation based on acoustic signal analysis provided in this application.

[0019] Figure 2 A schematic diagram of the structure of the tunneling machine noise detection system for acoustic signal analysis provided in this application.

[0020] Figure labeling: 10 sound wave monitoring module, 20 data preprocessing and analysis module, 30 energy focusing estimation module, 40 abnormal sound source acquisition module, and 50 anomaly prediction and early warning module. Detailed Implementation

[0021] This application provides a method and system for detecting abnormal noises in tunnel boring machines (TBMs) based on acoustic signal analysis. This addresses the technical problems of existing TBM noise detection methods, which rely on manual labor and simple sensor monitoring, resulting in low accuracy and poor real-time performance in noise location and identification. The method improves the accuracy and real-time performance of abnormal noise location and identification in TBMs, thereby enhancing construction safety and fault early warning capabilities.

[0022] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.

[0023] Example 1, as Figure 1 As shown, this application provides a method for detecting abnormal noises during tunneling machine operation based on acoustic signal analysis. The method includes:

[0024] Dynamic and continuous sound monitoring of the tunneling machine is performed using distributed piezoelectric ceramic acoustic probes to obtain raw array acoustic data.

[0025] Specifically, piezoelectric ceramic acoustic wave probes are sensitive elements that utilize the piezoelectric effect to convert the mechanical vibration signals of sound waves into measurable electrical signals. They feature high sensitivity, fast response speed, and adaptability to harsh environments, enabling them to capture subtle structural sound changes in high-noise and high-vibration environments. By employing a distributed layout, multiple piezoelectric ceramic acoustic wave probes are installed on the cutterhead surface of a tunneling machine. This distributed layout allows for the capture of sound signals generated during the cutterhead's operation from different positions and angles, avoiding monitoring blind spots and ensuring the completeness and representativeness of the acquired sound information.

[0026] During the operation of the tunneling machine, the distributed piezoelectric ceramic acoustic probes are activated to track the changes in the working status of the tunneling machine in real time and continuously monitor the sound, acquiring multiple real-time acoustic data. The acoustic data from different probes are arranged and combined according to the spatial distribution order and time order of the probe array to form the original array acoustic data, which can completely reflect the sound field changes at different positions of the cutterhead throughout the construction process, improving the comprehensiveness and reliability of the data.

[0027] Furthermore, the method also includes: the distributed piezoelectric ceramic acoustic probes are deployed on the cutterhead surface of the tunneling machine based on a radial survey line pre-plan, and are completed by a rotation completion pre-plan; wherein, the radial survey line pre-plan refers to setting 6 radial survey lines with an included angle of 60° on the cutterhead surface, with M probes deployed on each survey line, where M is an integer greater than or equal to 3 and less than or equal to 5; wherein, the rotation completion pre-plan refers to the cutterhead surface automatically relocking after being rotated clockwise by 60°, 120° or 180°, wherein the coupling state between the distributed piezoelectric ceramic acoustic probes and the cutterhead surface remains unchanged.

[0028] Specifically, the distributed piezoelectric ceramic acoustic wave probes are deployed using a radial survey line layout on the cutterhead surface of the tunnel boring machine (TBM), supplemented by a rotational completion plan. The radial survey line layout forms the basic framework for the entire probe deployment. On the TBM cutterhead surface, with the cutterhead center as the reference, six radially distributed survey lines are set, with the angle between adjacent survey lines precisely set at 60°. Each radial survey line contains M piezoelectric ceramic acoustic wave probes, where M is an integer greater than or equal to 3 and less than or equal to 5. This uniform radial layout creates a uniform spatial sampling grid without increasing the number of probes excessively, providing comprehensive coverage of the cutterhead surface and ensuring the capture of sound signals from different angles during cutterhead operation. To further improve the comprehensiveness and accuracy of acoustic monitoring, a rotational completion scheme is introduced. This scheme involves rotating the cutterhead surface clockwise by 60°, 120°, or 180° after the initial radial survey line layout. Each rotation automatically relocks the cutterhead, allowing the probe array to repeatedly cover the monitoring area from multiple spatial orientations while maintaining the coupling state between the probe and the cutterhead surface. This coupling state refers to the tight acoustic and mechanical contact between the probe and the measured surface, ensuring stable signal acquisition and accurate reflection of structural acoustic characteristics. In this way, areas not initially covered by the initial radial survey lines can be monitored by probes at new positions, achieving comprehensive acoustic monitoring of the cutterhead surface. For example, after rotating the cutterhead 60° clockwise, the area previously between two survey lines is now covered by a new survey line, allowing the probe to collect the acoustic signal from that area, further enriching the monitoring data. By combining radial survey line pre-planning and rotational completion pre-planning, the sound signals of the tunnel boring machine cutterhead during operation can be collected comprehensively and meticulously, obtaining comprehensive and accurate raw array acoustic wave data, effectively supporting the precise location and type identification of abnormal noise sources.

[0029] The original array acoustic data is preprocessed and analyzed according to the acoustic preprocessing mechanism to obtain the reflection matrix.

[0030] Specifically, a pre-set acoustic preprocessing mechanism is used to suppress noise, remove interference, and enhance features of the raw array acoustic data collected by monitoring. The acoustic preprocessing mechanism includes a series of signal processing techniques, such as direct wave cutoff, filtering, energy equalization, and gain compensation. Combined with targeted algorithms such as predictive filtering, multiple feature data that can reflect the propagation and reflection characteristics of sound waves in the tunneling machine structure are extracted. The multiple feature data are then integrated according to the spatial and temporal correspondence of the probe array to form a reflection matrix, providing effective data support for the localization of abnormal sound sources.

[0031] Furthermore, the original array acoustic data is preprocessed and analyzed according to the acoustic preprocessing mechanism to obtain a reflection matrix, including: preprocessing the original array acoustic data according to the primary preprocessing scheme in the acoustic preprocessing mechanism to obtain first-level acoustic data; preprocessing the first-level acoustic data according to the secondary preprocessing scheme in the acoustic preprocessing mechanism to obtain second-level acoustic data; and establishing the reflection matrix based on the second-level acoustic data; wherein, the primary preprocessing scheme includes preprocessing steps such as direct wave cutoff, bandpass filtering, FK filtering, energy equalization, and gain compensation.

[0032] Specifically, the raw array acoustic data is preprocessed and analyzed according to the acoustic preprocessing mechanism. First, a primary preprocessing scheme is executed: for the raw array acoustic data acquired by the distributed piezoelectric ceramic acoustic probes, operations such as direct wave cutoff, bandpass filtering, FK filtering, energy equalization, and gain compensation are performed sequentially. Direct wave cutoff removes the direct propagation signal between the probe and the sound source through time window interception to reduce its interference with reflection characteristics. Bandpass filtering suppresses low-frequency and high-frequency noise by limiting the frequency range, retaining only the effective frequency band related to structural features. FK filtering, or frequency-wavenumber filtering, selectively retains wavefield components in specific propagation directions using wavenumber domain characteristics, thereby suppressing irrelevant wave energy. Energy equalization adjusts the signal energy of different probes and time slots to make the overall data amplitude distribution more consistent, preventing certain strong energy channels from masking weak reflection signals. Gain compensation corrects the amplitude of signals with long propagation distances or numerous reflections that cause energy attenuation. After these processes, first-level acoustic data with suppressed interference and preserved main structural reflection characteristics is obtained.

[0033] The system then proceeds to a secondary preprocessing stage. In this stage, a predictive filter is used to further eliminate residual noise and non-stationary interference. The parameters of the predictive filter are dynamically determined based on a plane wave deconstruction strategy. The filter factor length is matched by calculating the local tilt angle and its variance, allowing the filter to adaptively adjust in monitoring areas with varying structural complexity. This secondary processing yields secondary acoustic data with higher signal-to-noise ratio and spatial resolution. A reflection matrix is ​​then established based on this secondary acoustic data. The reflection matrix is ​​a two-dimensional data array that structurally stores the cross-correlation or reflection coefficient information between channels according to the spatial layout and time sequence of the probe array. It comprehensively describes the propagation and reflection characteristics of sound waves within the structure. Preprocessing the original array acoustic data using the primary preprocessing scheme and further optimizing the data through the secondary preprocessing scheme improves data quality, thereby enhancing the accuracy and reliability of abnormal sound source detection.

[0034] Furthermore, the primary acoustic data is preprocessed according to the secondary preprocessing scheme in the acoustic preprocessing mechanism to obtain secondary acoustic data, including: activating a prediction filter to preprocess the primary acoustic data according to the secondary preprocessing scheme to obtain the secondary acoustic data; wherein, before activating the prediction filter, the following steps are taken: calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe according to the plane wave deconstruction strategy, wherein the first acoustic probe refers to any one of the distributed piezoelectric ceramic acoustic probes; calculating the first tilt angle variance based on the first local tilt angle, and using the first tilt angle variance as the first structural complexity index of the first part, wherein the first part refers to the position of the cutterhead surface monitored by the first acoustic probe; matching the first filter factor length corresponding to the first structural complexity index, and constructing the prediction filter based on the first filter factor length.

[0035] The first structural complexity index is proportional to the length of the first filter factor.

[0036] Specifically, after preprocessing the original array acoustic data according to the primary preprocessing scheme in the acoustic preprocessing mechanism to obtain the first-level acoustic data, in order to further suppress residual acoustic signal noise and enhance the resolution of structural reflection signals, a predictive filter is activated to preprocess the first-level acoustic data according to the secondary preprocessing scheme in the acoustic preprocessing mechanism. The predictive filter is a digital filter that adaptively models and suppresses interference based on local signal features; its filtering performance depends on the set filter factor length. Before activating the predictive filter, it is necessary to construct it. The specific construction process is as follows: Any probe is selected from the distributed piezoelectric ceramic acoustic probes as the first acoustic probe, and the signal it receives is the first acoustic signal. The first acoustic signal is analyzed and calculated using a plane wave deconstruction strategy to obtain the first local tilt angle. The first local tilt angle reflects the propagation direction of the acoustic wave at a specific location and time. After obtaining the first local tilt angle, the variance of the first tilt angle is further calculated by statistical analysis within a certain spatiotemporal sampling range. Centered on the monitoring location of the first acoustic probe, a continuous segment of sampling points in the time series is selected. Combined with the tilt angle values ​​of adjacent measuring points in space, the average of multiple tilt angle samples is first calculated. Then, the squared deviation of each sample from the average is calculated and averaged to obtain the first local tilt angle variance. The first local tilt angle variance is a quantitative description of the degree of local tilt angle fluctuation, reflecting the degree of fluctuation in the direction of sound wave propagation with time and space. The cutterhead surface position corresponding to the first tilt angle variance, i.e., the first part monitored by the first acoustic probe, is used as the first structural complexity index of the first part. The larger the first structural complexity index, the more drastic the change in the sound field at that part. Based on experimental data and expert experience, a pre-defined proportionality coefficient or mapping function maps the structural complexity index obtained from the variance of the first tilt angle to the corresponding filter factor length range. The first structural complexity index is proportional to the first filter factor length. For example, the minimum and maximum filter factor length ranges are specified during filter design, with the lowest complexity corresponding to the minimum length and the highest complexity corresponding to the maximum length. Intermediate values ​​are obtained by linear interpolation or by calculating specific length parameters using a piecewise function. After matching the first filter factor length, it is input into the filter design process as the order parameter of the prediction filter. The type of prediction filter is selected, such as a linear prediction filter based on an autoregressive model. Then, using first-order acoustic data as training input, the filter coefficients are iteratively solved using adaptive algorithms such as minimum mean square error or least squares. This allows the prediction filter to predict the current sampling point using historical sampling points and effectively separate and suppress components in the prediction residual that are unrelated to the structural reflection characteristics, thus forming an adaptively optimized prediction filter.By filtering the primary acoustic data using a predictive filter, secondary acoustic data with a higher signal-to-noise ratio and clearer waveform characteristics are obtained, thereby improving the accuracy and reliability of detecting abnormal noises during tunneling machine operation.

[0037] Furthermore, calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe according to the plane wave deconstruction strategy includes: performing a Fourier transform on the first acoustic signal to obtain a first frequency domain signal; calculating the first local tilt angle based on the first frequency domain signal, wherein calculating the first local tilt angle includes: obtaining the first adjacent probe of the first acoustic probe; comparing the first acoustic probe with the first adjacent probe to obtain a first comparison parameter, wherein the first comparison parameter includes a first phase delay and a first probe spacing; obtaining the first signal frequency of the first frequency domain signal, and calculating the first local tilt angle according to the plane wave deconstruction strategy, combined with the first phase delay and the first probe spacing.

[0038] Specifically, in calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe based on the plane wave deconstruction strategy, the first acoustic signal is first subjected to a Fourier transform to obtain a first frequency domain signal. The Fourier transform is a mathematical tool that can convert a time-domain signal into a frequency-domain signal; through this transformation, the first frequency domain signal is obtained. Then, the first adjacent probes of the first acoustic probe are acquired. The first adjacent probes refer to the piezoelectric ceramic acoustic probes that are spatially adjacent to the first acoustic probe. Acoustic signal data within the same time window are extracted from the signals received by the first acoustic probe and the first adjacent probe, and compared and analyzed to obtain first comparison parameters. The first comparison parameters include a first phase delay and a first probe spacing. Specifically, the acoustic data signals of the first acoustic probe and the first adjacent probe are analyzed using a cross-correlation function. The position offset corresponding to the peak value of the function can be used to calculate the first phase delay. That is, the first phase delay refers to the difference in phase of the same signal received by the two probes, reflecting the time sequence of the acoustic waves arriving at different probes. At the same time, with the help of the coordinate information of the probe layout, the physical distance between the first acoustic probe and the first adjacent probe is directly measured or obtained through preset probe installation parameters to obtain the first probe spacing.

[0039] Spectral analysis is performed on the first frequency domain signal, iterating through the amplitude corresponding to each frequency point. The frequency corresponding to the peak amplitude in the first frequency domain signal is extracted as the first signal frequency. This frequency is then substituted into the calculation formula of the plane wave deconstruction strategy. Combined with the first phase delay and the first probe spacing, the first local tilt angle is calculated. The calculation formula is as follows: Where σ is the tilt angle, Let f be the phase delay, f be the signal frequency, and Δx be the probe spacing. The first phase delay, the first signal frequency, and the first probe spacing are then substituted into the formula to calculate the first local tilt angle. By converting the signal to the frequency domain using Fourier transform, the frequency and phase information of the signal can be extracted more accurately. By acquiring adjacent probes and calculating comparison parameters, and utilizing the different propagation characteristics of sound waves at different locations, the local tilt angle is finally calculated, further improving the accuracy and reliability of abnormal noise detection in tunneling machines.

[0040] The energy focusing estimation strategy is invoked to estimate and analyze the reflection matrix, thereby obtaining wave velocity distribution information.

[0041] Furthermore, the energy focusing estimation strategy is invoked to estimate and analyze the reflection matrix to obtain wave velocity distribution information, including: performing Radon transform on the reflection matrix and extracting the energy focusing width at the slope k=-1 of the main diagonal; and analyzing the energy focusing width according to the energy focusing estimation strategy to obtain the wave velocity distribution information.

[0042] Specifically, firstly, a Radon transform is performed on the reflection matrix. The Radon transform is an integral transform method that projects a two-dimensional signal onto a parameterized linear space. It maps the energy distribution of sound waves at different tilt angles to the slope domain, which is used to identify the direction and velocity of sound wave propagation. After completing the Radon transform, the energy distribution curve along the main diagonal with a slope of k=-1 is located in the slope domain of the transform result. The maximum energy point on this curve is identified as the peak position. Centered on this peak, the curve is then searched for points where the energy attenuates to a specific threshold, such as 50% of the peak energy. The interval between these two threshold points is calculated. This length is the energy focusing width at the main diagonal slope k=-1. The energy focusing width reflects the degree of energy concentration in this specific direction; the narrower the width, the more concentrated the energy. Subsequently, the extracted energy focusing width is analyzed according to the energy focusing estimation strategy. This strategy is based on the principles of sound wave propagation and energy distribution. For example, different wave velocities affect the propagation and reflection of sound waves in a medium, leading to differences in energy distribution in the reflection matrix. When the medium wave velocity is high, the energy at a specific slope will exhibit a narrow and concentrated peak in the Radon domain, while at lower wave velocities, the energy peak will be wider, and the peak position may shift. Therefore, an energy focusing estimation strategy can be established by deriving the sound wave propagation formula or through experimental calibration, using the peak width and peak position characteristics corresponding to different wave velocity ranges to establish a width-wave velocity mapping relationship. The energy focusing width is then input into the energy focusing estimation strategy for analysis to obtain the corresponding wave velocity distribution information. Wave velocity distribution information refers to the distribution of sound wave propagation speed at different locations, enabling subsequent sound source visualization imaging and precise anomaly location based on a real wave velocity model, thereby improving imaging accuracy and location reliability, and enhancing anomaly detection accuracy.

[0043] Based on the wave velocity distribution information, a working sound visualization of the tunneling machine is generated, and the location of abnormal noise sources in the working sound visualization is obtained.

[0044] Specifically, the acquired wave velocity distribution information is analyzed, and the wave velocity data corresponding to different locations is transformed into visual elements to construct a working sound visualization chart of the tunneling machine. This chart visually presents the sound propagation characteristics in various areas. The working sound visualization chart uses different colors or brightness levels to intuitively display the acoustic energy intensity distribution at each location, thereby directly and accurately identifying and locating areas of abnormal energy concentration, i.e., the source of abnormal noise. By transforming the wave velocity distribution into a visualization and locating the source of abnormal noise, potential faults or abnormal operating conditions of the tunneling machine can be detected in a timely manner, providing accurate basis for equipment maintenance and fault diagnosis, and ensuring the safe and efficient operation of the tunneling machine.

[0045] Furthermore, forming the working sound visualization of the tunneling machine based on the wave velocity distribution information includes: calculating the first pseudospectral energy intensity of the first acoustic signal based on the wave velocity distribution information, and forming a pseudospectral energy function; using the azimuth angle corresponding to the maximum value of the pseudospectral energy function as a weighting coefficient to perform weighted correction on the Green function to obtain a target Green function; processing the target Green function using a time reversal operator to form the working sound visualization, wherein forming the working sound visualization includes: reversing the target Green function on the time axis, and convolving the reversed target Green function with the first acoustic signal to obtain a first convolution result; integrating the first convolution result to obtain the working sound visualization.

[0046] Specifically, wave velocity distribution information reflects the propagation characteristics of sound waves at different locations and directions. Based on the wave velocity distribution information, pseudospectral analysis is performed on the first sound wave signal received by the first sound wave probe. Spatial array signal processing methods, such as MUSIC or Capon algorithms, are used to analyze and process the first sound wave signal, extracting its frequency and azimuth information, and calculating the first pseudospectral energy intensity. This pseudospectral energy intensity can obtain high-resolution wave source azimuth characteristics under low signal-to-noise ratio conditions. The energy intensities corresponding to each azimuth angle are combined to form a pseudospectral energy function, which describes the pseudospectral energy distribution of the signal at different azimuth angles.

[0047] The azimuth angle corresponding to the maximum value of the pseudospectral energy function is used as a weighting coefficient to correct the Green's function. The Green's function is a mathematical function describing the response at any point in a field under point source excitation. In the sound wave propagation problem, the Green's function reflects the sound field distribution generated by the sound wave in space. Since sound wave propagation in the actual environment is affected by various factors, the propagation characteristics differ in different directions, thus requiring correction. The target Green's function obtained by weighting the Green's function with the azimuth angle corresponding to the maximum value of the pseudospectral energy function can more accurately reflect the actual sound wave propagation situation.

[0048] The target Green's function is processed using a time-reversal operator. Time reversal is a signal processing method that reverses the propagation process of a wave in time. Its principle is to play the received signal backwards, allowing the wave to refocus at the original sound source location during its reverse propagation. Specifically, the target Green's function is first reversed on the time axis to obtain a time-reversed Green's function, which is then convolved with the first sound wave signal. The convolution result reflects the convergence of the sound wave in time and is denoted as the first convolution result. Convolution is a mathematical operation; by performing a sliding multiplication and integration of the time-reversed Green's function and the first sound wave signal in time, the two are effectively fused. Through this fusion, key information related to the sound source location and characteristics in the first sound wave signal is extracted. Because the time-reversed Green's function carries prior knowledge of the sound wave propagation path and sound source characteristics, its combination with the first sound wave signal highlights components closely related to the location and characteristics of the abnormal sound source. Finally, the first convolution result is arranged according to spatial location and time sequence to form a spatiotemporal distribution matrix. Then, the convolution value corresponding to each spatial location in the matrix is ​​numerically integrated along the time dimension. The integration can be performed using numerical methods such as summation or trapezoidal rule to gather the energy information of that location over the entire time period. The integrated results of all spatial locations are mapped to the corresponding spatial coordinate system and visualized through color gradients or brightness changes, thus generating a complete two-dimensional or three-dimensional working sound visualization. For example, 30 distributed piezoelectric ceramic probes are distributed on the surface of the tunnel boring machine cutterhead. After the previous steps, the first convolution result is obtained. The first convolution result is a two-dimensional matrix with probe positions as rows and time sampling points as columns. For example, the matrix size is 30×1000, with 30 positions and 1000 convolution values ​​for each time sampling point. For each row, that is, the 1000 convolution values ​​of each probe position are integrated by summation to obtain the total energy value of that location over the entire observation period. For example, the integration result for the first probe is 85, the second probe is 210, the third probe is 45, and so on. These values ​​represent the degree of acoustic energy concentration at each location. Based on the actual installation coordinates of each probe on the cutter head, the integrated energy value is mapped to the spatial position of the cutter head, for example, using two-dimensional coordinates (x, y) to mark the probe position. Visualization rendering is then performed, selecting colors such as blue for low energy and red for high energy, converting the integrated energy value of each probe into color fills. If a more continuous image is needed, spatial interpolation can be performed on the energy values ​​to generate a smooth cutter head acoustic wave distribution map. The rendered image becomes the working sound visualization map, where the area with the most concentrated color corresponds to the location of the strongest energy focus, which is also the possible source of abnormal noise.

[0049] The working sound visualization visually displays the acoustic energy distribution at various spatial locations through changes in color or brightness, enabling direct identification of areas with abnormal energy concentration, i.e., the source of abnormal noise. By calculating the pseudo-spectral energy intensity and constructing the pseudo-spectral energy function, the energy characteristics of the first sound wave signal can be accurately grasped. Weighted correction of the Green's function improves the accuracy of the sound wave propagation model. The working sound visualization visual is generated by using a time-reversal operator, transforming the abstract sound wave signal into an intuitive image. This facilitates accurate and direct analysis of abnormal noises in the tunneling machine's working sound, timely detection of potential faults, and provides strong support for the safe operation and maintenance of the tunneling machine.

[0050] The abnormality type of the abnormal noise source is predicted, and an early warning process for the abnormality type is performed on the abnormal noise source.

[0051] Specifically, after constructing the working sound visualization and locating the source of the abnormal noise, real-time operating status information of the source location is collected. This information is then combined with historical data to determine the anomaly type of the source location. For example, by comparing and analyzing the currently collected real-time operating status information with historical data, the anomaly type of the current abnormal noise source location is predicted and identified. Different levels of warning signals are set according to the severity of the anomaly type; for example, a level 1 warning indicates a minor anomaly, a level 2 warning indicates a moderate anomaly, and a level 3 warning indicates a severe anomaly. After obtaining the predicted type, the warning processing mechanism is triggered, automatically generating warning information based on the danger level of the anomaly type. This information can be displayed on the operating terminal or trigger an audible and visual alarm, and, if necessary, trigger automatic shutdown or maintenance prompts. By combining the sound source location results with historical experience data, the process from detecting abnormal noise to determining the anomaly type and then to proactively issuing warnings is realized. This improves the accuracy and response speed of anomaly diagnosis during tunneling machine operation, thereby reducing the risk of tunneling machine failure and ensuring equipment operation safety.

[0052] Furthermore, predicting the anomaly type of the abnormal noise source location includes: collecting real-time operating status information of the abnormal noise source location; traversing the vectorized real-time operating status information in the historical operation database to obtain the most similar historical operating status information; extracting the historical anomaly type from the most similar historical record corresponding to the most similar historical operating status information, and using it as the anomaly type.

[0053] Specifically, real-time operational status information is collected for the located abnormal sound source. This information includes not only acoustic features but also operating parameters of the tunneling machine collected by various high-precision sensors installed on the machine, such as rotational speed, torque, thrust, and vibration acceleration, ensuring that the data comprehensively reflects the operational status of the abnormal sound source. The collected data is then vectorized, transforming features of different types and dimensions into a unified feature vector using normalization or standardization methods. This vector is then traversed in a pre-built historical operation database, which stores a large amount of operational status information and corresponding anomaly records from past tunneling machine operations. Using a specific similarity algorithm, such as cosine similarity, the similarity between the vectorized real-time operational status information and each record in the historical operation database is calculated. This yields the historical operational status information most similar to the current feature vector. The historical anomaly type corresponding to this most similar historical operational status information is extracted from the most similar historical records and directly used as the predicted type for the current abnormal sound source. By combining historical experience with real-time monitoring, the determination of anomaly types is based on a large number of existing cases and feature matching. This not only enables the rapid and accurate prediction of the anomaly type at the source of abnormal noise, but also provides a direct basis for decision-making in subsequent early warning and maintenance. This allows for rapid response and efficient handling, ensuring the safe and stable operation of the tunneling machine and improving production efficiency and equipment lifespan.

[0054] Example 2, based on the same inventive concept as the method for detecting abnormal noises in tunneling machines using acoustic signal analysis in the aforementioned examples, such as... Figure 2 As shown, this application provides a tunneling machine operating noise detection system based on acoustic signal analysis, wherein the system includes:

[0055] The acoustic monitoring module 10 is used to dynamically and continuously monitor the tunneling machine using distributed piezoelectric ceramic acoustic probes to obtain raw array acoustic data. The data preprocessing and analysis module 20 is used to preprocess and analyze the raw array acoustic data according to an acoustic preprocessing mechanism to obtain a reflection matrix. The energy focusing estimation module 30 is used to retrieve an energy focusing estimation strategy to estimate and analyze the reflection matrix to obtain wave velocity distribution information. The abnormal noise source acquisition module 40 is used to generate a working sound visualization of the tunneling machine based on the wave velocity distribution information and to acquire the location of abnormal noise sources in the working sound visualization. The anomaly prediction and early warning module 50 is used to predict the anomaly type of the abnormal noise source location and to perform early warning processing for the anomaly type.

[0056] Furthermore, the acoustic monitoring module 10 in the tunneling machine noise detection system for acoustic signal analysis is also used for: the distributed piezoelectric ceramic acoustic probes are deployed on the cutterhead surface of the tunneling machine based on a radial survey line pre-plan, and are completed by a rotation completion pre-plan; wherein, the radial survey line pre-plan refers to setting 6 radial survey lines with an included angle of 60° on the cutterhead surface, with M probes deployed on each survey line, where M is an integer greater than or equal to 3 and less than or equal to 5; wherein, the rotation completion pre-plan refers to automatically relocking the cutterhead surface after rotating it clockwise by 60°, 120° or 180°, wherein the coupling state between the distributed piezoelectric ceramic acoustic probes and the cutterhead surface remains unchanged.

[0057] Furthermore, the data preprocessing and analysis module 20 in the tunneling machine noise detection system for acoustic signal analysis is also used for: preprocessing the original array acoustic data according to the primary preprocessing scheme in the acoustic preprocessing mechanism to obtain first-level acoustic data; preprocessing the first-level acoustic data according to the secondary preprocessing scheme in the acoustic preprocessing mechanism to obtain second-level acoustic data; and establishing the reflection matrix based on the second-level acoustic data; wherein the primary preprocessing scheme includes preprocessing steps such as direct wave cutoff, bandpass filtering, FK filtering, energy equalization, and gain compensation.

[0058] Furthermore, the data preprocessing and analysis module 20 in the tunneling machine noise detection system for acoustic signal analysis is also used to: activate a prediction filter to preprocess the primary acoustic data according to the secondary preprocessing scheme to obtain the secondary acoustic data; wherein, before activating the prediction filter, the following steps are taken: calculating the first local tilt angle of the first acoustic signal received by the first acoustic probe according to the plane wave deconstruction strategy, wherein the first acoustic probe refers to any one of the distributed piezoelectric ceramic acoustic probes; calculating the first tilt angle variance based on the first local tilt angle, and using the first tilt angle variance as the first structural complexity index of the first part, wherein the first part refers to the cutterhead surface position monitored by the first acoustic probe; matching the first filter factor length corresponding to the first structural complexity index, and constructing the prediction filter based on the first filter factor length.

[0059] Furthermore, the data preprocessing and analysis module 20 in the tunneling machine noise detection system for acoustic signal analysis is also used for: performing a Fourier transform on the first acoustic signal to obtain a first frequency domain signal; calculating the first local tilt angle based on the first frequency domain signal, wherein calculating the first local tilt angle includes: acquiring the first adjacent probe of the first acoustic probe; comparing the first acoustic probe with the first adjacent probe to obtain a first comparison parameter, wherein the first comparison parameter includes a first phase delay and a first probe spacing; acquiring the first signal frequency of the first frequency domain signal, and calculating the first local tilt angle based on the plane wave deconstruction strategy, combined with the first phase delay and the first probe spacing.

[0060] Furthermore, the data preprocessing and analysis module 20 in the tunneling machine noise detection system for acoustic signal analysis is also used to: make the first structural complexity index proportional to the length of the first filter factor.

[0061] Furthermore, the energy focusing estimation module 30 in the tunneling machine working noise detection system for acoustic signal analysis is also used to: perform Radon transform on the reflection matrix and extract the energy focusing width at the slope k=-1 of the main diagonal; analyze the energy focusing width according to the energy focusing estimation strategy to obtain the wave velocity distribution information.

[0062] Furthermore, the abnormal noise source acquisition module 40 in the tunneling machine abnormal noise detection system for acoustic signal analysis is also used for: calculating the first pseudo-spectral energy intensity of the first acoustic signal based on the wave velocity distribution information, and forming a pseudo-spectral energy function; using the azimuth angle corresponding to the maximum value of the pseudo-spectral energy function as a weighting coefficient to perform weighted correction on the Green function to obtain a target Green function; processing the target Green function using a time reversal operator to form the working sound visualization, wherein forming the working sound visualization includes: reversing the target Green function on the time axis, and convolving the reversed target Green function with the first acoustic signal to obtain a first convolution result; integrating the first convolution result to obtain the working sound visualization.

[0063] Furthermore, the abnormal prediction and early warning module 50 in the tunneling machine abnormal noise detection system based on acoustic signal analysis is also used to: collect real-time operating status information of the abnormal noise source location; traverse the vectorized real-time operating status information in the historical operation database to obtain the most similar historical operating status information; extract the historical abnormal type from the most similar historical record corresponding to the most similar historical operating status information, and use it as the abnormal type.

[0064] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0065] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting abnormal sound of a boring machine in operation by analyzing a sound wave signal, characterized by, The method comprises the steps of: obtaining original array acoustic wave data by performing dynamic and continuous sound monitoring on the roadheader through a distributed piezoelectric ceramic acoustic wave probe; preprocessing and analyzing the original array acoustic wave data according to an acoustic wave preprocessing mechanism to obtain a reflection matrix; obtaining wave velocity distribution information by performing estimation and analysis on the reflection matrix through an energy focusing estimation strategy; forming a working sound visual graph of the roadheader according to the wave velocity distribution information, and obtaining an abnormal sound source position in the working sound visual graph; predicting an abnormal type of the abnormal sound source position, and performing early warning processing on the abnormal sound source position according to the abnormal type; performing estimation and analysis on the reflection matrix through the energy focusing estimation strategy to obtain wave velocity distribution information, comprising: performing Radon transformation on the reflection matrix, and extracting the energy focusing width at k=-1 of the main diagonal line slope; analyzing the energy focusing width according to the energy focusing estimation strategy to obtain the wave velocity distribution information; establishing an energy focusing estimation strategy of the width-wave velocity mapping relationship through the peak width and the peak position characteristics corresponding to different wave velocity ranges.

2. The method according to claim 1, wherein the method is a method for detecting abnormal sound of a boring machine, and the sound signal is a sound signal of the boring machine. The distributed piezoelectric ceramic acoustic wave probe is arranged on the cutter head surface of the roadheader based on a radial survey line plan, and is completed through a rotation completion plan; wherein the radial survey line plan refers to arranging six radial survey lines with an included angle of 60° on the cutter head surface, and arranging M probes on each survey line, M being an integer greater than or equal to 3 and less than or equal to 5; wherein the rotation completion plan refers to automatically relocking the cutter head surface after rotating 60°, 120° or 180° clockwise, wherein the coupling state of the distributed piezoelectric ceramic acoustic wave probe and the cutter head surface remains unchanged.

3. The method according to claim 1, wherein the method is characterized by: The method comprises the steps of: preprocessing and analyzing the original array acoustic wave data according to an acoustic wave preprocessing mechanism to obtain a reflection matrix, comprising: preprocessing the original array acoustic wave data according to a primary preprocessing scheme in the acoustic wave preprocessing mechanism to obtain first-level acoustic wave data; preprocessing the first-level acoustic wave data according to a secondary preprocessing scheme in the acoustic wave preprocessing mechanism to obtain second-level acoustic wave data; establishing the reflection matrix based on the second-level acoustic wave data; 4. The method according to claim 3, wherein the method is characterized by: wherein the primary preprocessing scheme comprises preprocessing steps of direct wave cutting, band-pass filtering, FK filtering, energy equalization and gain compensation. The method comprises the steps of: preprocessing the original array acoustic wave data according to an acoustic wave preprocessing mechanism to obtain a reflection matrix, comprising: preprocessing the original array acoustic wave data according to a primary preprocessing scheme in the acoustic wave preprocessing mechanism to obtain first-level acoustic wave data; preprocessing the first-level acoustic wave data according to a secondary preprocessing scheme in the acoustic wave preprocessing mechanism to obtain second-level acoustic wave data; wherein, before activating the prediction filter, the method comprises the steps of: calculating a first local inclination of a first acoustic wave signal received by a first acoustic wave probe according to a plane wave deconstruction strategy, wherein the first acoustic wave probe refers to any one of the distributed piezoelectric ceramic acoustic wave probes; calculating a first inclination variance based on the first local inclination, and taking the first inclination variance as a first structural complexity index of a first position, wherein the first position refers to a cutter head surface position monitored by the first acoustic wave probe. The first filter factor length corresponding to the first structural complexity index is matched, and the prediction filter is constructed based on the first filter factor length.

5. The method according to claim 4, wherein the method is characterized by: The first local inclination of the first acoustic wave signal received by the first acoustic wave probe is calculated according to a plane wave deconstruction strategy, including: The first acoustic wave signal is subjected to Fourier transform to obtain a first frequency domain signal; The first local inclination is calculated according to the first frequency domain signal, wherein the calculation of the first local inclination includes: A first adjacent probe of the first acoustic wave probe is acquired; A first contrast parameter is obtained by comparing the first acoustic wave probe and the first adjacent probe, wherein the first contrast parameter includes a first phase delay and a first probe spacing; A first signal frequency of the first frequency domain signal is acquired, and the first local inclination is calculated according to the plane wave deconstruction strategy, in combination with the first phase delay and the first probe spacing.

6. The method according to claim 4, wherein the method is characterized by: The first structural complexity index is directly proportional to the first filter factor length.

7. The method according to claim 4, wherein the method is characterized by: The working sound visual map of the tunneling machine is formed according to the wave velocity distribution information, including: The first pseudo-spectrum energy intensity of the first acoustic wave signal is calculated according to the wave velocity distribution information, and a pseudo-spectrum energy function is formed; The azimuth angle corresponding to the maximum value of the pseudo-spectrum energy function is taken as a weight coefficient to perform weighted correction on the Green function, to obtain a target Green function; The target Green function is processed by using a time reversal operator to form the working sound visual map, wherein the formation of the working sound visual map includes: The target Green function is inverted on the time axis, and the inverted target Green function is convolved with the first acoustic wave signal to obtain a first convolution result; The working sound visual map is obtained by integrating the first convolution result.

8. The method according to claim 7, wherein the method is a method for detecting abnormal sound of a boring machine, and the sound signal is a sound signal of the boring machine. The abnormal type of the abnormal sound source part is predicted, including: Real-time running state information of the abnormal sound source part is collected; The vectorized real-time running state information is traversed in a historical operation database to obtain most similar historical running state information; The historical abnormal type in the most similar historical record corresponding to the most similar historical running state information is extracted and taken as the abnormal type.

9. A boring machine abnormal sound detection system for acoustic signal analysis, characterized in that Steps for implementing the method of any one of claims 1 to 8, including: An acoustic wave monitoring module (10) is configured to perform dynamic and continuous sound monitoring on a tunneling machine by using a distributed piezoelectric ceramic acoustic wave probe to obtain original array acoustic wave data; A data preprocessing analysis module (20) is configured to perform preprocessing analysis on the original array acoustic wave data according to an acoustic wave preprocessing mechanism to obtain a reflection matrix; An energy focusing estimation module (30) is configured to call an energy focusing estimation strategy to perform estimation analysis on the reflection matrix to obtain wave velocity distribution information; An abnormal sound source acquisition module (40) is configured to form a working sound visual map of the tunneling machine according to the wave velocity distribution information, and acquire an abnormal sound source part in the working sound visual map; An abnormality prediction and early warning module (50) is configured to predict an abnormal type of the abnormal sound source part, and perform early warning processing of the abnormal type on the abnormal sound source part.

Images