Multi-node cooperative electromagnetic ultrasonic online monitoring method and system

Abstract

The application discloses a kind of multi-node coordination's electromagnetic ultrasonic online monitoring method and system, and the application relates to the technical field of online monitoring, to multiple electromagnetic ultrasonic probe introduction orthogonal code demodulation, so that multiple electromagnetic ultrasonic probe detects the detection of the same to-be-detected object complementary interference, the detection data output by each monitoring node is collected, and the echo timestamp of the to-be-detected object is extracted, the clock drift of each monitoring node is compensated along the echo timestamp in reverse, to trigger the space-time alignment of each detection data, and the corresponding multi-node cooperative detection system is constructed in combination with attention mechanism, improve the accuracy of multi-node cooperative detection system.According to the dynamic identification of multiple electromagnetic ultrasonic online monitoring content, the corresponding suspected defect data is determined, the suspected defect data is loaded into the online monitoring finite element module, and the corresponding online monitoring defect route is determined in combination with the corresponding multi-physical coupling relationship, to improve the accuracy of online monitoring defect route.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology, and in particular to a multi-node collaborative electromagnetic ultrasonic online monitoring method and system. Background Technology

[0002] Electromagnetic ultrasonic testing technology is widely used in the online monitoring of key equipment such as industrial pipelines and pressure vessels due to its advantages such as non-contact operation, no need for coupling agent, and suitability for high-temperature and high-speed environments. As industrial equipment develops towards larger and more complex sizes, the monitoring range of a single probe is limited and it is difficult to cover the entire area to be tested. Therefore, the use of multi-probe collaborative monitoring has gradually become the mainstream technical approach.

[0003] Existing multidimensional monitoring networks typically employ time-division excitation or frequency-division multiplexing (FDM) techniques to avoid signal interference when multiple probes operate simultaneously. While time-division excitation avoids interference, it reduces detection efficiency and makes it difficult to capture transient defect signals. FDM, on the other hand, is limited by bandwidth resources. When the number of nodes is large, bandwidth allocation becomes difficult and intermodulation interference is easily generated. Furthermore, it lacks unified control over the spatiotemporal dimensions of multi-node detection data. The local clocks of each node often drift, and the objects under inspection (such as pipe components) undergo spatial displacement due to thermal expansion and contraction, resulting in misalignment of the data collected by each node in terms of timestamps and spatial locations. This affects the accuracy of the multi-node collaborative detection system and makes it impossible to achieve adaptive adjustment of the monitoring parameters of multiple monitoring nodes. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a multi-node collaborative electromagnetic ultrasonic online monitoring method and system.

[0005] This invention provides a multi-node collaborative online electromagnetic ultrasound monitoring method, comprising:

[0006] The current position of the object to be detected is marked, and a multi-dimensional monitoring network is constructed by combining the distribution positions of multiple electromagnetic ultrasonic probes. Orthogonal codes are introduced to demodulate multiple electromagnetic ultrasonic probes simultaneously, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Multiple monitoring nodes are marked in the multi-dimensional monitoring network.

[0007] Collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data. Combine the attention mechanism to build the corresponding multi-node collaborative detection system.

[0008] In this multi-node collaborative detection system, the electromagnetic impedance change and ultrasonic guided wave transmission loss of two adjacent monitoring nodes are fused by tensor in the feature layer, and multiple cross-space collaborative detection layers are output. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions.

[0009] Based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, the corresponding suspected defect data is determined. The suspected defect data is loaded into the online monitoring finite element module, and the corresponding online monitoring defect route is determined in combination with the corresponding multi-physical coupling relationship. This triggers the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

[0010] This invention provides a multi-node collaborative electromagnetic ultrasound online monitoring system, which is applied to the aforementioned multi-node collaborative electromagnetic ultrasound online monitoring method; the multi-node collaborative electromagnetic ultrasound online monitoring system includes:

[0011] The multidimensional monitoring module is used to mark the current position of the object to be detected. It constructs a multidimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes. It synchronously introduces orthogonal code demodulation to multiple electromagnetic ultrasonic probes, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Multiple monitoring nodes are marked in the multidimensional monitoring network.

[0012] The multi-node collaborative detection system module is used to collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data, and build the corresponding multi-node collaborative detection system in combination with the attention mechanism.

[0013] The electromagnetic ultrasonic online monitoring module is used in this multi-node collaborative detection system to perform tensor fusion of the electromagnetic impedance changes and ultrasonic guided wave transmission loss of two adjacent monitoring nodes in the feature layer, and output multiple corresponding cross-space collaborative detection layers. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions.

[0014] The adaptive adjustment module is used to determine the corresponding suspected defect data based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, load the suspected defect data into the online monitoring finite element module, and determine the corresponding online monitoring defect route in combination with the corresponding multi-physics coupling relationship, and trigger the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] (1) Mark the current position of the object to be tested, construct a multi-dimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes, introduce orthogonal code demodulation to multiple electromagnetic ultrasonic probes simultaneously, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive, mark multiple monitoring nodes in the multi-dimensional monitoring network; collect the detection data output by each monitoring node, extract the echo timestamp of the object to be tested, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data, and construct the corresponding multi-node collaborative detection system by combining the attention mechanism. The multi-dimensional monitoring network is introduced to further control the detection data output by each monitoring node and improve the accuracy of the multi-node collaborative detection system.

[0017] (2) In this multi-node collaborative detection system, the electromagnetic impedance change and ultrasonic guided wave transmission loss of two adjacent monitoring nodes are fused by tensor in the feature layer, and multiple cross-space collaborative detection layers are output. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions. Based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, the corresponding suspected defect data is determined. The suspected defect data is loaded into the online monitoring finite element module, and the corresponding online monitoring defect route is determined in combination with the corresponding multi-physical coupling relationship. The multiple cross-space collaborative detection layers are controlled, and the suspected defect data and the online monitoring finite element module are fully considered, which improves the accuracy of the online monitoring defect route and triggers the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the multi-node collaborative electromagnetic ultrasonic online monitoring method in this embodiment of the invention.

[0019] Figure 2 This is a schematic diagram of the structure of a multi-node collaborative electromagnetic ultrasonic online monitoring system in an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the structure of a computer system suitable for implementing the electronic devices of the present application. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1 to 2 A multi-node collaborative online electromagnetic ultrasound monitoring method is proposed and applied to online monitoring scenarios. The multi-node collaborative online electromagnetic ultrasound monitoring method includes:

[0023] Step S11: Mark the current position of the object to be detected, construct a multi-dimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes, and simultaneously introduce orthogonal code demodulation to multiple electromagnetic ultrasonic probes so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Mark multiple monitoring nodes in the multi-dimensional monitoring network.

[0024] Step S12: Collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data, and build the corresponding multi-node collaborative detection system in combination with the attention mechanism.

[0025] Step S13: In this multi-node collaborative detection system, the electromagnetic impedance change and ultrasonic guided wave transmission loss of two adjacent monitoring nodes are fused by tensor in the feature layer, and multiple cross-space collaborative detection layers are output. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions.

[0026] Step S14: Based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, determine the corresponding suspected defect data, load the suspected defect data into the online monitoring finite element module, and determine the corresponding online monitoring defect route in combination with the corresponding multi-physical coupling relationship, and trigger the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

[0027] In step S11, the specific steps are as follows:

[0028] S111: Acquire the current position of the object to be tested, dynamically mark the global coordinate system and local morphological features of the object to be tested, output the corresponding topological constraint information, trigger the spatial deployment of multiple electromagnetic ultrasonic probes relative to the same object to be tested along the topological constraint information, and construct a multi-dimensional monitoring network with full-view coverage capability.

[0029] S112: In the multidimensional monitoring network, each electromagnetic ultrasonic probe is assigned a correlated orthogonal coding sequence, so that the ultrasonic electromagnetic fields emitted within the same time window remain orthogonal and interference-free in physical space. Zero crosstalk complementary detection of the same object to be detected is performed along multiple detection channels. Simultaneously, the demodulated electromagnetic ultrasonic signal stream after orthogonal coding sequence is mapped to the multidimensional monitoring network, dynamically generating multiple monitoring nodes with independent spatial attributes and timestamp references.

[0030] In the embodiments of this application, the current position of the object to be tested is collected, the global coordinate system and local morphological features of the object to be tested are dynamically marked, and the corresponding topological constraint information is output. Multiple electromagnetic ultrasonic probes are deployed in space relative to the same object to be tested along the topological constraint information, and a multi-dimensional monitoring network with full-view coverage capability is constructed, thus introducing a multi-dimensional monitoring network with full-view coverage capability.

[0031] At this point, an absolute spatial reference for the monitoring system is introduced; high-precision positioning sensing units (such as laser trackers or industrial vision positioning modules) are used to collect the actual physical position of the object to be detected at the monitoring site, and the sensor measurement coordinates are converted into the global reference coordinates of the system through a coordinate transformation matrix, thereby realizing the "existence establishment" of the object to be detected in the digital monitoring space.

[0032] After establishing the global location, the local geometric features and physical properties of the object to be inspected are extracted in depth. Using structured light scanning or point cloud data processing, local morphological features such as curvature changes, wall thickness distribution, and weld orientation are identified. Based on these features, the accessibility and energy attenuation characteristics of the acoustic propagation path are calculated to generate topological constraint information. This information specifies the feasible domain for probe deployment, the limit of the sound beam incident angle, and the safe distance to avoid edge diffraction interference.

[0033] Based on topological constraints, an optimization approach (such as genetic or multi-objective particle swarm optimization) is used to plan the spatial deployment of multiple electromagnetic ultrasound probes (EMAT). This process involves solving for the optimal spatial position and attitude angle of the probe array to maximize the detection coverage area and minimize the detection blind zone. A multi-degree-of-freedom robotic arm or adaptive flexible fixture is triggered to precisely install the probes in the designated positions, achieving a hard connection in physical space.

[0034] After the physical deployment of the probes is completed, the spatial coordinates, detection direction, and frequency response characteristics of each probe are mapped to the virtual monitoring space to construct a multi-dimensional monitoring network. Through the network calibration process, the communication links and acoustic paths between each node are verified. The so-called "full-view coverage" means that the network has the ability to perform slice-like observation of the object to be tested from different spatial dimensions (such as circumferential, axial, and radial), ensuring that defects in any direction can be captured by at least one monitoring node.

[0035] Specifically, the object to be inspected is a pipeline assembly. At the monitoring site of the pipeline assembly, the laser tracker emits a laser beam to scan the flange end face and key weld positions of the pipeline assembly. The system obtains its absolute coordinates (X,Y,Z) in the workshop coordinate system. Due to the slight displacement of the pipeline assembly caused by thermal expansion and contraction or fluid impact, the system captures its current spatial attitude in real time, establishes a rigid body kinematic model, and ensures that the coordinates of the probes deployed later strictly coincide with the actual physical position of the pipeline.

[0036] The system focuses on scanning the geometric abrupt change region at the connection between the "bend" and "straight" pipe sections; local morphological feature analysis identifies that the radius of curvature RR of the pipe's outer wall gradually changes from infinity in the straight pipe section to a specific value in the bend; topological constraint information clearly indicates that the probes must be arranged at a specific arc along the pipe axis, and the incident angle of the sound beam must compensate for the curvature change of the pipe surface to ensure that the ultrasonic guided wave can effectively penetrate the pipe wall, rather than being scattered on the surface; at the same time, the weld reinforcement area is marked as a "probe-undesertible area", forming a spatial topological constraint boundary.

[0037] On the pipe assembly, the system, based on topological constraints, controls the robotic arm to deploy high-frequency surface wave EMAT probes on the outer arc side of the bend section to monitor surface cracks; and low-frequency guided wave EMAT probes on the inner arc side and straight pipe sections to monitor internal corrosion and wall thinning. The probe array is distributed in a "spiral staggered" pattern to ensure that the sound beam coverage of each two adjacent probes has a 10%-15% overlap area. This deployment method strictly follows the curvature adaptability and weld avoidance principles in the topological constraints, ensuring that the probes fit tightly against the pipe surface and achieve high-fidelity signal coupling.

[0038] Nodes 1-4 are distributed circumferentially, providing a "transverse view" of the pipe's circumferential cross-section; nodes 5-8 are distributed axially, providing a "side view" of the pipe's longitudinal direction. This network not only covers the main structure of the pipe but also eliminates the acoustic shadow area inside the bend section through a special array layout. The system completes a self-test, confirming that all probes can transmit and receive signals, and that the acoustic path between adjacent probes can cover the entire critical monitoring area of ​​the pipe. This completes the construction of the multi-dimensional monitoring network, laying the physical groundwork for the orthogonal coding access in the subsequent step S112.

[0039] Furthermore, in the multidimensional monitoring network, each electromagnetic ultrasonic probe is assigned a correlated orthogonal coding sequence, so that the ultrasonic electromagnetic fields emitted within the same time window remain orthogonal and interference-free in physical space. Zero-crosstalk complementary detection of the same object to be detected is performed along multiple detection channels. Simultaneously, the electromagnetic ultrasonic signal stream demodulated by the orthogonal coding sequence is mapped to the multidimensional monitoring network, dynamically generating multiple monitoring nodes with independent spatial attributes and timestamp references.

[0040] At this point, based on the principle of Code Division Multiple Access (CDMA), the system selects coded sequences whose cross-correlation function values ​​approach zero from orthogonal codebooks (such as Walsh matrices or Gold sequence sets); these orthogonal coded sequences are used as carrier modulation signals and loaded into the excitation circuits of each probe respectively. By adjusting the chip rate and length of the coded sequences, their spectral characteristics are matched with the center frequency of the electromagnetic ultrasonic probe, thereby endowing each probe signal with orthogonal properties in the coding domain without changing the hardware transmission power.

[0041] Under unified GPS timing or high-precision crystal oscillator clock synchronization, all electromagnetic ultrasonic probes are triggered to simultaneously emit ultrasonic pulses within the same time window. Since the excitation signal has been loaded with orthogonal coding, although the electromagnetic field excited by each probe and the ultrasonic guided wave field propagate superimposed on each other in the medium in physical space, at the signal demodulation level, the autocorrelation characteristics (sharp autocorrelation peak) and cross-correlation characteristics (cross-correlation value approaches zero) of the orthogonal coding are used to make the composite wave field maintain an "orthogonal non-interference state" in a statistical sense. That is, in the echo signal received by a certain probe, the interference signals emitted by other probes are regarded as broadband white noise and filtered out after correlation demodulation.

[0042] By performing convolution operations or correlation detection on the acquired mixed signals using a matched filter bank, the echo signals corresponding to each probe are separated, achieving "zero crosstalk" detection. Here, "complementary detection" refers to the fact that the detection areas of different probes overlap (complement) in physical space. Through orthogonal demodulation technology, the multi-source signals in the overlapping area are separated, thereby achieving multi-view, multi-dimensional imaging of the same object to be inspected (such as defects or stress concentration areas). This not only utilizes the spatial coverage advantage of multiple probes but also avoids signal confusion.

[0043] The system combines the global coordinates (X,Y,Z) obtained in step S111 with the current timestamp information to dynamically generate monitoring node metadata for each frame of signal data. This node not only contains waveform data, but also binds its independent spatial attributes (position coordinates, beam pointing) and temporal attributes (transmission time, flight time). In this way, the original physical signal is transformed into a digital twin monitoring node with clear physical meaning, providing a standardized data interface for subsequent spatiotemporal alignment and collaborative detection.

[0044] Specifically, at the monitoring site of the pipeline assembly, the system has confirmed the stable attitude of the pipeline under the rigid body kinematics model. At this time, the multi-dimensional monitoring network control center sends independent Walsh orthogonal codes to the eight electromagnetic ultrasonic probes deployed on the flange end face and the bend section of the pipeline. For example, the probe deployed on the flange end face is loaded with the code sequence numbered W1, while the probe monitoring the critical weld is loaded with the W2 sequence. Although these probes are physically close to each other and receive the weak stress wave signals generated by the thermal expansion and contraction of the pipeline, the excitation signals are orthogonalized in the frequency domain to ensure that the source of the signal can be accurately distinguished through the coding features.

[0045] When fluid impact causes pipe vibration, the system needs to collect data at high frequency. At time T0, eight probes simultaneously emit pulses loaded with orthogonal codes. Ultrasonic guided waves propagate along the pipe wall to detect microcracks caused by thermal expansion and contraction. In physical space, sound waves from the straight pipe section and the sound waves from the curved pipe section are superimposed. However, based on the waveform design of orthogonal coding, the matched filter at the receiving end can accurately extract the echo energy of the target probe, while suppressing the sound wave interference from other probes to the background noise level (e.g., signal-to-noise ratio better than -60dB), thus achieving "soft isolation" of the physical field in complex vibration environments.

[0046] At the critical weld of the pipeline, the detection blind zones of two adjacent probes overlap. The system processes the signal streams from these two probes in parallel. Although both probes detect defects at the weld location simultaneously, the demodulation module outputs two independent detection results: one reflects the surface wave characteristics of the upper surface of the weld, and the other reflects the transverse wave characteristics inside the weld. These two signals are strictly aligned on the time axis and do not contain each other's echo interference. This zero-crosstalk complementary detection enables the system to distinguish between surface oxide scale peeling and internal crack propagation from the microstructure, greatly improving the confidence of the monitoring data.

[0047] On the digital twin interface of the pipeline assembly, the system generates a series of monitoring nodes with timestamp references in real time. For example, for a region where pipe wall thinning is detected, the system generates a monitoring node with the following attributes: "Time: Tnow, Location: (Xpipe, Ypipe, Zpipe), Probe ID: #3, Feature: Wall thickness loss 5%". Because a rigid body kinematic model has been established, even if the pipeline undergoes millimeter-level displacement due to thermal expansion in the workshop coordinate system, these monitoring nodes can be accurately anchored to the corresponding position of the actual physical structure through inverse coordinate transformation, ensuring accurate mapping between monitoring data and physical entities.

[0048] In step S12, the specific steps are as follows:

[0049] S121: Real-time monitoring of multiple monitoring nodes, synchronous collection of detection data output by each monitoring node, extraction of the echo timestamp reflected by the object under test through envelope detection and adaptive threshold, reverse compensation of the echo timestamp, and reverse correction of the local clock error of each monitoring node using the echo timestamp as the absolute time anchor point, thereby triggering spatiotemporal alignment of all detection data in the absolute spatiotemporal coordinate system to output the spatiotemporally aligned dataset;

[0050] S122: Input the dataset into the corresponding attention space and introduce a time decay factor into the attention space to autonomously focus on the key waveform segments containing defect features between different monitoring nodes. Based on the fusion of multiple key waveform segments, a multi-node collaborative detection system with global feature perception and local anomaly capture capabilities is constructed.

[0051] In the embodiments of this application, multiple monitoring nodes are monitored in real time, and the detection data output by each monitoring node is collected synchronously. The echo timestamp reflected by the object to be detected is extracted by envelope detection and adaptive threshold. The echo timestamp is compensated in reverse, and the local clock error of each monitoring node is corrected in reverse using the echo timestamp as the absolute time anchor point. This triggers the spatiotemporal alignment of all detection data in the absolute spatiotemporal coordinate system to output the spatiotemporally aligned dataset, thus introducing the spatiotemporally aligned dataset.

[0052] At this time, the system monitors all nodes in the multidimensional monitoring network in parallel through a high-speed data acquisition card (DAQ) to acquire electromagnetic ultrasonic echo signals in real time; it extracts the instantaneous amplitude envelope of the signal using Hilbert transform or envelope detection to eliminate high-frequency carrier interference; and it introduces an adaptive threshold method (such as a floating threshold or a signal-to-noise ratio adaptive threshold) to accurately identify the start and end times of the reflected echo of the object under test (such as bottom surface echo or reference reflector echo) in the background of dynamic noise, thereby extracting a high-precision echo timestamp.

[0053] Because there are slight deviations in the local crystal oscillator frequencies of each monitoring node, long-term operation will cause timestamp drift. The system uses the theoretical acoustic time calculated based on the known geometry of the object under test (such as standard wall thickness) as a reference, and compares the extracted actual echo timestamp with the theoretical time. The time deviation Δt is calculated, which is the clock drift. Through reverse compensation, the drift is converted into a correction coefficient, and the local clock of each monitoring node is reversed to make its time reference converge to the absolute time reference of the system.

[0054] Combining the rigid body kinematics model and global coordinates established in step S111, the time-domain waveform data of each monitoring node is mapped to a unified spatial coordinate system. This process uses the time-space transformation relationship (distance = speed of sound × time / 2) to convert the sampling points on the time axis into depth or position coordinates on the spatial axis. This triggers the alignment of all detection data in the absolute spatiotemporal coordinate system, enabling data collected by different nodes at different times to be compared and fused in the same spatial dimension.

[0055] The multi-channel detection data, after spatiotemporal alignment, is standardized and packaged. The output includes a multidimensional dataset containing three-dimensional spatial coordinates (X,Y,Z), a calibrated timestamp Tcorrected, and the corresponding amplitude value A. This dataset not only reflects the time-frequency characteristics of the signal but also carries precise physical spatial attributes, providing a high-quality data foundation for constructing the attention mechanism and multi-node collaborative detection system in the subsequent step S12.

[0056] Specifically, the system collects ultrasonic signals from the pipeline under different operating conditions in real time. Given that the pipeline wall thickness changes at the micrometer level due to thermal expansion and contraction, and there is fluid impact noise on site, the system performs envelope detection on the collected signals to capture the "bottom wave" signal reflected back from the inner (or outer) wall of the pipeline. Through an adaptive threshold method, the system successfully filters out low-frequency interference caused by fluid impact and accurately locks the arrival time of the bottom wave signal, Tarrival, which reflects the combined propagation characteristics of the current pipeline wall thickness and sound velocity.

[0057] For the pipeline assembly, the system has established a rigid body kinematic model and obtained its absolute coordinates (X,Y,Z), from which the theoretical propagation time Ttheory of the ultrasonic guided wave under standard wall thickness can be calculated. For example, the arrival time of the bottom wave recorded by the local clock of a certain node is Tlocal, but the time is too fast due to crystal aging. The system calculates the deviation between the actual time Tarrival and Ttheory and finds that there is a positive drift of 0.5μs at this node. The system then performs reverse compensation (subtracting the drift) on all subsequent timestamps of this node, eliminating the measurement error caused by the inconsistency of the hardware clock and ensuring the synchronization of the time reference of each node.

[0058] When the pipeline assembly undergoes a slight displacement due to thermal expansion, although the relative position of the probe and the pipeline changes, the system performs spatial calculations based on the absolute time anchor point. The system maps the corrected time axis of each node to the "propagation distance axis" along the pipe wall. Due to the use of a unified time anchor point (calibrated bottom wave time), even if the pipeline undergoes overall displacement, the echo signal from the weld defect can be accurately located in the spatial coordinate system. The system aligns the detection data of all nodes to the absolute geometry of the pipeline, eliminating the waveform position ambiguity caused by the thermal expansion and contraction of the pipeline, and realizing the "spatial overlap" and "time synchronization" of multi-node data.

[0059] The system outputs a standardized monitoring dataset for pipeline components. This dataset clearly records the echo energy distribution of various parts of the pipeline (such as flange end faces, bends, and welds) at different times. For example, the dataset marks a microcrack located at coordinates (Xdefect, Ydefect, Zdefect), and this location information has been freed from the influence of clock drift and thermal expansion displacement. This spatiotemporally aligned dataset is transmitted to the data processing center in real time, providing reliable data support for the subsequent construction of a high-precision multi-node collaborative detection system.

[0060] Furthermore, the dataset is input into the corresponding attention space, and a time decay factor is introduced into this attention space to autonomously focus on key waveform segments containing defect features between different monitoring nodes. Based on the fusion of multiple key waveform segments, a multi-node collaborative detection system with global feature perception and local anomaly capture capabilities is constructed. This system takes into account the overall consideration of multiple key waveform segments, ensuring the accuracy of the multi-node collaborative detection system with global feature perception and local anomaly capture capabilities. At the same time, a multi-dimensional monitoring network is introduced to further control the detection data output by each monitoring node, thereby improving the accuracy of the multi-node collaborative detection system.

[0061] At this point, the system receives the spatiotemporally aligned dataset output in step S121 and constructs a multidimensional attention space. Within this space, the original electromagnetic ultrasound signal waveform is no longer just a time series, but is transformed into a feature tensor. The system initializes the weight matrix and performs linear transformation and nonlinear activation on the dataset, providing a computational basis for subsequent feature focusing.

[0062] In the attention space, this factor acts as a dynamic weighting coefficient on the time axis of the signal. According to the propagation characteristics of ultrasonic signals, the signal energy usually decays exponentially with time (i.e., the sound wave increases in depth inside the material). Based on this physical law, the time decay factor performs nonlinear compensation or suppression on the signal in the later part of the time axis to prevent deep noise from being misjudged as a defect, while ensuring that shallow micro-defects are not missed.

[0063] The system calculates the correlation score of each monitoring node signal in the attention space, and autonomously calculates the attention weight distribution of each time segment by combining the time decay factor. Through Softmax normalization, the region with high weight is identified as the key waveform segment containing defect features. The system performs "soft focusing" on these segments to suppress background noise in non-critical areas and extract a feature subset with high discriminative power.

[0064] By combining local anomaly features (such as crack echoes at a single node) with global background features (such as the overall wall thickness variation trend of the pipeline) through tensor splicing or feature aggregation operations, a multi-node collaborative detection system is constructed. This system not only has the ability to keenly capture local minor defects, but also can judge the impact of defects on the overall structural integrity from a global perspective, and output detection results with high robustness.

[0065] Specifically, the system receives a spatiotemporally aligned dataset. Although the pipeline undergoes slight displacement due to thermal expansion and contraction, the signals in the dataset have achieved spatial overlap under the correction of the rigid body kinematics model. At this point, the system projects these waveform signals from the flange end face and key weld locations into the attention space. In this space, the weak crack echo signals that were originally submerged in noise are given higher feature dimension weights, while the stable bottom wave signals serve as background benchmarks, preparing for subsequent feature selection.

[0066] Since the flow noise generated by fluid impact is usually random and distributed in the middle and later part of the signal, the system introduces a time decay factor in the attention space and applies an exponential decay weight to the sound path region of the pipe wall thickness. When the ultrasonic guided wave passes through the pipe wall to detect internal corrosion, the time decay factor automatically adjusts the gain weight, effectively suppressing the clutter interference reflected from the bottom of the pipe, while retaining the weak scattered wave signal caused by the coarsening of grains in the heat-affected zone in the weld area, thus achieving signal-to-noise ratio optimization for potential defect signals.

[0067] In the monitoring of critical welds in pipeline components, the sound beams of multiple monitoring nodes cover the weld and its heat-affected zone. When a node captures the intermittent echo generated by the lack of fusion inside the weld, the attention mechanism detects that the waveform segment has significant differences from the standard bottom wave characteristics (such as waveform broadening and phase reversal). The system then autonomously increases the attention weight of this time segment and marks it as a "critical waveform segment". At the same time, for the uniform deformation signal generated by the pipeline due to thermal expansion and contraction, the attention weight is automatically reduced because it conforms to the preset normal pattern. This achieves accurate capture of abnormal defect signals and builds a multi-node collaborative detection system for pipeline components.

[0068] The multi-node collaborative detection system integrates global wall thickness monitoring data (reflecting the overall corrosion thinning trend) from straight pipe section nodes with local crack feature data from bend pipe section nodes. For example, the system determines that although the overall pipe wall thickness is within the normal range (global feature), there are microcracks (local anomalies) caused by local stress concentration at the weld. The spatial location of the crack is confirmed through cross-verification of multi-node data. This collaborative detection system overcomes the one-sidedness of single-point monitoring and realizes comprehensive and three-dimensional monitoring of the health status of pipeline components.

[0069] In step S13, the specific steps are as follows:

[0070] S131: Real-time monitoring of the multi-node collaborative detection system, extracting the electromagnetic impedance change of two adjacent monitoring nodes under the action of electromagnetic field and the ultrasonic guided wave transmission loss on the sound field propagation path, mapping the electromagnetic impedance change and ultrasonic guided wave transmission loss to a high-dimensional feature space for fusion. This fusion not only preserves their respective physical gradient directions, but also mines the nonlinear response characteristics of electromagnetic-acoustic coupling at the defect edge through cross tensor product, and constructs multiple cross-space collaborative detection layers with cross-space physical correlation attributes along each nonlinear response characteristic.

[0071] S132: Multiple cross-space collaborative detection layers are loaded into a pre-constructed monitoring gradient architecture with monitoring constraints. This monitoring gradient architecture uses the physical continuity of adjacent monitoring nodes as a priori constraints and dynamically suppresses operating noise through gradient backpropagation, thereby triggering the dynamic output of electromagnetic ultrasonic online monitoring content of each monitoring node under different complex operating conditions, so as to realize the adaptiveness of electromagnetic ultrasonic online monitoring content of each monitoring node.

[0072] In the embodiments of this application, the multi-node collaborative detection system is monitored in real time. The electromagnetic impedance changes of two adjacent monitoring nodes under the action of electromagnetic field and the ultrasonic guided wave transmission loss on the sound field propagation path are extracted. The electromagnetic impedance changes and ultrasonic guided wave transmission loss are mapped to a high-dimensional feature space for fusion. This fusion not only preserves their respective physical gradient directions, but also mines the nonlinear response characteristics of electromagnetic-acoustic coupling at the defect edge through cross tensor product. Multiple cross-space collaborative detection layers with cross-space physical correlation attributes are constructed along each nonlinear response characteristic. This takes into account the overall consideration of each nonlinear response characteristic and ensures the accuracy of multiple cross-space collaborative detection layers with cross-space physical correlation attributes.

[0073] At this time, the system monitors two adjacent monitoring nodes in the multi-node collaborative detection system in real time and extracts the characteristic parameters of their electromagnetic and acoustic domains respectively. For the electromagnetic field, the impedance spectrum change (Z(t)) of the electromagnetic ultrasonic probe is collected in real time. Through vector network analysis or real-time impedance analyzer, the changes in resistance and reactance components are calculated. This change reflects the change in the electromagnetic coupling state between the probe coil and the surface of the tested material (such as eddy current path distortion caused by cracks). For the acoustic field, the transmission loss (TL) of the ultrasonic guided wave on the propagation path between the two nodes is calculated, that is, the logarithmic difference between the transmitted signal energy and the received signal energy. This index directly reflects the degree of scattering and absorption of sound wave energy inside the material.

[0074] The extracted one-dimensional electromagnetic impedance change data and one-dimensional ultrasonic transmission loss data are mapped to a unified high-dimensional feature space. Using the embedding layer or manifold learning method in deep neural networks, the low-dimensional physical scalar is mapped to a high-dimensional feature vector. In this process, a loss function is specially designed to preserve their respective physical gradient directions. That is, in the feature space, the gradient direction of impedance change is consistent with the direction of material surface conductivity change, and the gradient direction of transmission loss is consistent with the direction of material internal attenuation coefficient, ensuring that the physical essence is not lost during the fusion process.

[0075] By using cross tensor product operations, an outer product operation is performed on the electromagnetic feature tensor and the acoustic feature tensor to generate a higher-order fused tensor. This fusion method goes beyond simple weighted summation and can capture the nonlinear coupling relationship between the two physical fields. By calculating the principal components of the tensor or using convolution kernel scanning, the nonlinear response characteristics of electromagnetic-acoustic coupling at the defect edge can be extracted. For example, when a defect simultaneously affects the surface eddy current distribution (electromagnetic domain) and the propagation of internal stress waves (acoustic field), its cross term will exhibit significant heterogeneity, which is strong evidence for the existence of the defect.

[0076] Based on the mined nonlinear response features, a cross-space collaborative detection layer is constructed. Each detection layer represents a specific defect mode or physical state (such as "surface crack layer", "internal corrosion layer", "stress concentration layer"). These detection layers do not exist in isolation, but are interconnected in the spatial dimension through feature indexes, forming a multi-level three-dimensional description of the monitoring area. The system superimposes these detection layers and outputs a comprehensive feature map containing cross-space physical correlation attributes, providing input for the monitoring gradient architecture in the subsequent step S13.

[0077] Specifically, the system selects two adjacent monitoring nodes covering the critical weld area (the first node is located in the straight pipe section, and the second node is located in the bend pipe section). When the fluid impact inside the pipe causes stress concentration in the pipe wall, the physical state between the first node and the second node changes. The system collects the electromagnetic impedance data of the first node in real time and finds that its real part (resistance component) has a small jump, indicating that there is a sudden change in magnetic permeability caused by oxide scale peeling or cracks on the pipe surface. At the same time, the transmission loss of the ultrasonic guided wave transmitted from the first node to the second node is calculated, and it is found that the transmission coefficient has decreased by 1.5dB, indicating that there is a scattering attenuation caused by defects in the sound wave propagation path.

[0078] The system maps the electromagnetic impedance change vector VZ and the guided wave transmission loss vector VTL at the weld to the same Hilbert feature space. In this space, the overall geometric deformation of the pipeline caused by thermal expansion and contraction is characterized by a low-frequency, gradually changing gradient, while local crack defects are characterized by a high-frequency, abrupt gradient. By preserving this gradient feature, the system successfully distinguishes the different trajectories of the overall thermal expansion of the pipeline (which leads to uniform and minute changes in impedance and transmittance) and weld cracks (which lead to abrupt changes in impedance and a sharp increase in transmittance) in the feature space, thus avoiding misjudgment.

[0079] At critical weld locations in the pipeline assembly, the system performs tensor fusion of electromagnetic and acoustic characteristics. Assuming an incomplete fusion defect exists within the weld, simply observing the electromagnetic impedance is considered signal fluctuation, and simply observing the transmission loss is considered poor coupling. However, through cross tensor product calculation, the system finds that the "electromagnetic-acoustic cross eigenvalue" in this region is significantly increased. This is because the edge of the defect not only blocks the propagation path of the ultrasonic guided wave (leading to a nonlinear increase in transmission loss), but also disrupts the electromagnetic field distribution at the tip of the defect (leading to drastic impedance changes). This nonlinear abrupt change in the cross term accurately marks the edge contour and depth information of the defect.

[0080] The system constructs a multi-layered, cross-space collaborative detection layer on the pipeline assembly. The first layer focuses on the electromagnetic anomaly area on the pipeline surface, marking the abrupt change in potential distribution at the flange end face. The second layer focuses on the acoustic shadow area inside the pipe wall, marking the wall thickness reduction trend of the bend section. The third layer is the nonlinear response layer of the newly discovered weld defects, which precisely covers the location of the weld defects in spatial coordinates (X,Y,Z). The system outputs these detection layers to form an "electromagnetic-acoustic composite hologram," which clearly shows the health status of the pipeline assembly under complex operating conditions. It not only points out the location of the defects but also reveals the physical nature of the defects under the dual effects of electromagnetic and mechanical forces.

[0081] Furthermore, multiple cross-space collaborative detection layers are loaded into a pre-constructed monitoring gradient architecture with monitoring constraints. This monitoring gradient architecture uses the physical continuity of adjacent monitoring nodes as a priori constraints and dynamically suppresses operating noise through gradient backpropagation, thereby triggering the dynamic output of electromagnetic ultrasonic online monitoring content of each monitoring node under different complex operating conditions. This achieves the self-adaptation of electromagnetic ultrasonic online monitoring content of each monitoring node and introduces the self-adaptation of electromagnetic ultrasonic online monitoring content of each monitoring node.

[0082] At this point, the system pre-constructs a monitoring gradient architecture, which is essentially a deep neural network or optimization solver constrained by physical laws. Its core lies in introducing "physical continuity" as a priori constraint: that is, for two adjacent monitoring nodes, the physical quantities they detect (such as wall thickness, stress, and temperature field) should be continuously and gradually changing in space, without any abrupt changes without physical basis. This constraint is transformed into a regularization term or a transmission rule between network layers to guide the flow of data processing.

[0083] The multiple cross-space collaborative detection layers (containing electromagnetic-acoustic fusion features) generated in step S131 are used as input data and loaded into the monitoring gradient architecture. The architecture calculates the predicted output of each monitoring node through forward propagation and calculates the "constraint deviation" between the output result and the physical continuity constraint. This deviation is defined as a gradient signal, which is used to quantify the degree to which the current monitoring content deviates from the physical law.

[0084] Using gradient backpropagation, the system adjusts the feature weights or threshold parameters in the monitoring gradient architecture in reverse according to the constraint deviation. For high-frequency abrupt signals that violate physical continuity (usually corresponding to operating noise, such as fluid impact or electromagnetic interference), the architecture dynamically reduces their weights through a negative feedback mechanism to achieve noise suppression. Meanwhile, for slowly varying signals that conform to physical laws (usually corresponding to real defects or structural changes), their weights are retained or enhanced. This process is carried out online in real time without the need for offline training.

[0085] After filtering and reconstruction using a gradient architecture, the system outputs the final online electromagnetic ultrasound monitoring content. This content is no longer a simple accumulation of the original signals, but a standardized monitoring result that has undergone physical constraint verification and noise suppression. The output content can automatically adjust its confidence level and presentation format (such as B-scan images, C-scan imaging, or defect parameter lists) according to the current working conditions (such as high temperature, high pressure, vibration), thus achieving the adaptability of the monitoring system.

[0086] Specifically, the system establishes a monitoring gradient architecture based on a rigid body kinematics model. Considering that the distribution of surface temperature and stress fields of pipe components is spatially continuous during thermal expansion and contraction, the architecture is set such that the rate of change of electromagnetic properties detected by two adjacent monitoring nodes (such as nodes located at the junction of straight and curved pipe sections) must conform to Gaussian distribution or smooth curve constraints. This constraint ensures that the system automatically eliminates abnormal signals that violate the laws of spatial continuity when processing data, laying a mathematical foundation for subsequent noise suppression.

[0087] When the pipeline is subjected to fluid impact and generates vibration noise, the system inputs a cross-space collaborative detection layer containing noise characteristics into the architecture. At this time, some nodes output abrupt signals due to poor vibration coupling. When the architecture calculates the gradient, it finds that these abrupt signals cause the characteristic gradient between adjacent nodes to have extreme values, which violates the prior of "physical continuity" (i.e., the pipeline wall thickness cannot change abruptly at the centimeter level within a millimeter-level distance). This violation is transformed into a high-gradient error signal, providing a target for the next step of correction.

[0088] When fluid impact causes the entire pipeline to shake, all nodes will receive low-frequency vibration interference signals. Since these signals are spatially consistent and do not conform to the characteristics of local defects, the architecture automatically adjusts the filtering parameters of each node through gradient backpropagation, which greatly suppresses the ultrasonic guided wave phase jitter noise caused by vibration. Conversely, if there is a real crack at the weld, although the nonlinear response characteristics it produces are weak, they are fixed in spatial location and conform to the law of local stress concentration. The architecture will retain this characteristic through gradient positive feedback, thereby achieving "locking" of the real defect in the background of strong noise.

[0089] The system implements adaptive monitoring output on pipeline components. During the "cold" start-up phase of the pipeline, the system automatically reduces its sensitivity to minute displacement signals and outputs high-precision wall thickness reference data. During the "hot" operation phase, in the face of geometric position changes caused by thermal expansion and contraction of the pipeline, the monitoring gradient architecture adjusts the monitoring threshold in real time based on the rigid body kinematics model. The output monitoring content automatically eliminates false defect signals caused by thermal expansion and only retains and highlights the potential crack propagation trend at key welds. This ensures that the monitoring reports obtained by maintenance personnel are always accurate and reliable, unaffected by the complex operating conditions of the pipeline.

[0090] In step S14, the specific steps are as follows:

[0091] S141: Acquire multiple online electromagnetic ultrasound monitoring contents, dynamically identify multiple online electromagnetic ultrasound monitoring contents, and determine the corresponding suspected defect data by combining the anomaly probability distribution mechanism. Inject the suspected defect data into the online monitoring finite element module in real time, thereby reconstructing the local geometry and boundary conditions near the defect through the virtual space of the online monitoring finite element module, and solving the sound field scattering and eddy current distortion response of the defect under different excitation parameters by combining the multi-physics coupling relationship, thereby depicting the online monitoring defect path that reflects the defect morphology;

[0092] S142: Following the solution results of this multi-physical coupling relationship, the optimization guide is directly triggered and reversely controlled to adaptively adjust the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network, forming a complete closed loop from physical perception - data collaboration - digital twin verification - parameter self-evolution. The monitoring parameters of multiple monitoring nodes cover excitation frequency, bias magnetic field strength, probe array aperture and sampling rate.

[0093] In the embodiments of this application, multiple online electromagnetic ultrasound monitoring contents are acquired, and these contents are dynamically identified. A corresponding suspected defect data is determined by combining an anomaly probability distribution mechanism. This suspected defect data is then injected into the online monitoring finite element module in real time. The local geometry and boundary conditions near the defect are reconstructed through the virtual space of the online monitoring finite element module. Furthermore, by combining multi-physics coupling relationships, the acoustic field scattering and eddy current distortion responses of the defect under different excitation parameters are solved in a forward manner. This process depicts the online monitoring defect path reflecting the defect morphology, incorporating the overall consideration of the anomaly probability distribution mechanism and ensuring the accuracy of the corresponding suspected defect data.

[0094] At this point, the system implements adaptive monitoring output on the pipeline components. During the "cold" start-up phase of the pipeline, the system automatically reduces its sensitivity to minute displacement signals and outputs high-precision wall thickness reference data. During the "hot" operation phase, in the face of geometric position changes in the pipeline caused by thermal expansion and contraction, the monitoring gradient architecture adjusts the monitoring threshold in real time based on the rigid body kinematics model. The output monitoring content automatically eliminates false defect signals caused by thermal expansion and only retains and highlights the potential crack propagation trend at key welds. This ensures that the monitoring reports obtained by maintenance personnel are always accurate and reliable, unaffected by the complex operating conditions of the pipeline.

[0095] The system injects the extracted suspected defect data into the online monitoring finite element module of the lightweight or reduced-order model (ROM) in real time. Based on the global coordinate system and local morphological characteristics of the A-type test object, the module reconstructs the local geometry near the defect in virtual space. Based on the characteristic parameters of the suspected defect data (such as echo delay time and phase reversal information), the initial geometric parameters of the defect (such as depth and orientation) are initially set, and boundary conditions (such as temperature load and pressure constraint) consistent with the actual working conditions are applied.

[0096] In the finite element environment, Maxwell's equations and elastic dynamics equations are solved by combining the electromagnetic-ultrasonic multiphysics coupling relationship. The system simulates the scattering behavior of the sound field at the defect and the distortion response of the eddy current field at the defect edge under different excitation parameters (such as different frequencies and different pulse widths) of the electromagnetic ultrasonic probe. The echo signal received by the virtual sensor is calculated, its amplitude, phase and mode conversion characteristics are analyzed, and it is compared with suspected defect data collected in the physical world.

[0097] In the virtual pipeline weld model, the system simulated the excitation process of the electromagnetic ultrasonic probe; the solver calculated the accumulation effect of eddy currents at the tip of the virtual crack, as well as the reflected waves and mode-converted waves generated after the ultrasonic guided wave encounters the crack surface; the system adjusted the excitation frequency and performed multiple forward solutions, and found that when the frequency was set to a specific value, the characteristics of the echo signal generated by the virtual simulation (waveform phase flip angle, amplitude ratio) were highly consistent with the suspected defect signal collected on site. This result not only verified the authenticity of the defect, but also revealed the specific path of eddy current distortion caused by the crack.

[0098] Specifically, the system analyzes monitoring data from the flange end face and key weld locations in real time. Although the pipeline undergoes overall displacement due to thermal expansion and contraction, the system has eliminated background noise through a rigid body kinematics model. At a certain moment, the system detects a sudden increase in ultrasonic guided wave transmission loss at the key weld and an abnormal jump in electromagnetic impedance. The anomaly probability distribution mechanism calculates that the anomaly confidence level in this area reaches 92%, far exceeding the set alarm threshold of 85%. The system then locks onto this area, extracts the time-domain waveform, spectral characteristics, and spatial coordinates of the weld, marks it as suspected crack defect data, and prepares to inject it into the simulation module.

[0099] For the marked suspected weld crack data, the online monitoring finite element module quickly generates a local "weld-crack" refined mesh model. Based on the current absolute coordinates and thermal expansion state of the pipeline, the module applies a temperature field of 425℃ and an internal pressure boundary condition of 12MPa consistent with the field conditions in the virtual model. At the same time, based on the ultrasonic echo time difference, a virtual crack surface with a depth of about 2mm is initialized at the weld fusion line, thus completing the geometric reconstruction of the physical defect in digital space.

[0100] The system simulated the excitation process of an electromagnetic ultrasonic probe; the solver calculated the accumulation effect of eddy currents at the tip of a virtual crack, as well as the reflected waves and mode-converted waves generated when the ultrasonic guided wave encounters the crack surface; the system adjusted the excitation frequency and performed multiple forward solutions, finding that when the frequency was set to a specific value, the characteristics of the echo signal generated by the virtual simulation (waveform phase flip angle, amplitude ratio) were highly consistent with the suspected defect signal collected on site. This result not only verified the authenticity of the defect, but also revealed the specific path of eddy current distortion caused by the crack.

[0101] The system outputs the online monitoring defect path of weld cracks in pipeline components. This path clearly shows the three-dimensional morphology of the crack along the weld fusion line and marks the location of the deepest point of the crack tip at coordinates (Xtip, Ytip, Ztip). Simulation results show that the strongest point of acoustic field scattering at the crack edge is not at the geometric center, but at the lower end of the crack. Based on this, the system generates a defect path map, indicating to maintenance personnel that the crack is very likely to extend into the heat-affected zone during the next maintenance cycle, and suggests that the online monitoring system focus on tracking the guided wave attenuation in this area, completing a closed-loop process from data identification to qualitative and quantitative analysis of defects.

[0102] Furthermore, by using the optimization guidance based on the solution results of this multi-physical coupling relationship, the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network are directly triggered and reversely controlled to adaptively adjust, forming a complete closed loop from physical perception to data collaboration, digital twin verification, and parameter self-evolution. The monitoring parameters of multiple monitoring nodes cover excitation frequency, bias magnetic field strength, probe array aperture, and sampling rate. The monitoring parameters of multiple monitoring nodes are introduced, and at the same time, multiple cross-space collaborative detection layers are controlled, fully considering suspected defect data and online monitoring finite element modules, improving the accuracy of online defect detection routes, and triggering the adaptive adjustment of monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

[0103] At this point, the system performs in-depth analysis of the multi-physics coupling solution results output by the online monitoring finite element module in step S141. By constructing objective functions (such as maximizing the defect echo signal-to-noise ratio and minimizing signal attenuation), it guides the optimization in the parameter space. It analyzes the sensitivity and resolution of the virtual sensor received signals under different parameter settings (such as different frequencies and different magnetic field strengths), determines the optimal combination of excitation and reception parameters for the current defect morphology, and generates a "parameter adjustment instruction set".

[0104] The system sends the generated parameter adjustment instruction set back to the control terminal of the multi-dimensional monitoring network via a high-speed industrial bus or fiber optic network. This process uses a low-latency closed-loop control protocol to ensure that the instructions can reach the hardware driver layer of each monitoring node without loss and in real time. At the same time, the system performs a safety check to prevent the adjusted parameters from exceeding the safe operating range of the hardware or affecting the normal operation of other nodes.

[0105] Each monitoring node coordinates adjustments to key monitoring parameters according to instructions: Excitation frequency adjustment: The signal generator adjusts the carrier frequency to match the characteristic detection frequency of the defect; Bias magnetic field strength adjustment: The adjustable constant current source changes the current of the input electromagnet or pulse magnetization coil to optimize the transduction efficiency; Probe array aperture adjustment: The number of effective working elements of the probe array is controlled by an electronic switch matrix to change the directivity and focusing depth of the sound beam; Sampling rate adjustment: The high-speed acquisition card adjusts the sampling frequency according to the new signal bandwidth to ensure no signal aliasing and optimal data volume.

[0106] The system confirms the effectiveness of the closed loop through a continuous feedback mechanism; the adjusted node immediately begins a new round of data acquisition, inputting the updated monitoring data back into the collaborative detection system; the system compares the signal-to-noise ratio and feature clarity before and after the adjustment to verify the effect of parameter optimization; if the effect meets expectations, the current parameters are locked; if the operating conditions change again, the optimization process is retried; thus, a complete closed loop is formed from physical perception (data acquisition), data collaboration (feature fusion), digital twin verification (simulation optimization) to parameter self-evolution (reverse control), ensuring that the monitoring system is always in the best working state.

[0107] Specifically, in the digital twin model of the pipeline assembly, simulation results show that for the microcracks identified at the weld, when the excitation frequency is adjusted to a specific high-frequency band (such as 2.5MHz), the wavelength of the ultrasonic guided wave resonates with the crack depth, and the echo amplitude is increased by 40% compared to the current frequency. At the same time, the simulation shows that appropriately increasing the bias magnetic field strength can effectively suppress magnetic domain noise in the high-temperature environment of the pipeline. Based on this, the system generates optimization guidance instructions, instructing the physical probe to adjust towards "high frequency" and "strong magnetic field" to accurately capture the crack signal.

[0108] The system packages the optimized commands such as "excitation frequency 2.5MHz, bias current 1.5A" and sends them to the monitoring and control cabinet at the pipeline component site through the control network. After receiving the reverse control command, the control cabinet immediately verifies the validity of the command and confirms that the parameter combination is within the rated operating range of the probe. Then, it triggers each hardware module (signal generator, power amplifier, bias power supply) to enter the parameter reconfiguration state, completing the command transmission from digital space to physical space.

[0109] At the pipeline site, coordinated adjustments were made to the monitoring nodes for weld cracks: the signal generator increased the ultrasonic pulse center frequency from 1MHz to 2.5MHz, improving the resolution of small cracks; the bias power supply output current was increased to 1.5A, enhancing the transduction efficiency of the probe in high-temperature environments; the array aperture control module activated more probe elements around the weld, locking the sound beam focus point at the simulated crack tip position; and the acquisition card simultaneously increased the sampling rate to 50MHz to record the detailed features of high-frequency echoes with high fidelity. This series of adjustments required no manual intervention, achieving "targeted" monitoring of specific defects.

[0110] After the adjustment was completed, the pipeline component monitoring system immediately acquired a set of clear, high signal-to-noise ratio crack echo signals. System verification showed that the signal-to-noise ratio improved from 6dB before the adjustment to 18dB, successfully capturing the diffraction waves from the previously submerged crack tip. The system locked this state and entered the next monitoring cycle. When the pipeline is displaced again due to fluid impact or the operating conditions change due to temperature changes, this closed-loop mechanism will automatically repeat the above process, continuously and dynamically optimizing the monitoring parameters, realizing intelligent and adaptive online monitoring of the entire life cycle of the pipeline component.

[0111] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a multi-node collaborative electromagnetic ultrasound online monitoring system according to an embodiment of the present invention; the multi-node collaborative electromagnetic ultrasound online monitoring system is applied to the above-described multi-node collaborative electromagnetic ultrasound online monitoring method; the multi-node collaborative electromagnetic ultrasound online monitoring system includes:

[0112] The multidimensional monitoring module 21 is used to mark the current position of the object to be detected, and to construct a multidimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes. Orthogonal codes are introduced to demodulate multiple electromagnetic ultrasonic probes simultaneously, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Multiple monitoring nodes are marked in the multidimensional monitoring network.

[0113] The multi-node collaborative detection system module 22 is used to collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, compensate the clock drift of each monitoring node in reverse along the echo timestamp, trigger the spatiotemporal alignment of each detection data, and construct the corresponding multi-node collaborative detection system in combination with the attention mechanism.

[0114] The electromagnetic ultrasonic online monitoring module 23 is used in the multi-node collaborative detection system to perform tensor fusion of the electromagnetic impedance change and ultrasonic guided wave transmission loss of two adjacent monitoring nodes in the feature layer, and output multiple cross-space collaborative detection layers. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions.

[0115] The adaptive adjustment module 24 is used to determine the corresponding suspected defect data based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, load the suspected defect data into the online monitoring finite element module, and determine the corresponding online monitoring defect route in combination with the corresponding multi-physical coupling relationship, and trigger the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

[0116] HYPERLINK "javascript:;" Figure 3 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown; it should be noted that... Figure 3 The computer system of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0117] For example, HYPERLINK "javascript:;" Figure 3 As shown, the computer system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage portion 308 into a random access memory (RAM) 303, such as performing the methods described in the above embodiments; various programs and data required for system operation are also stored in the RAM 303; the CPU 301, ROM 302 and RAM 303 are interconnected with each other via a bus 304; an input / output (I / O) interface 305 is also connected to the bus 304.

[0118] The following components are connected to I / O interface 305: input section 306 including keyboard, mouse, etc.; output section 307 including cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; storage section 308 including hard disk, etc.; and communication section 309 including network interface card such as LAN (Local Area Network), modem, etc., which performs communication processing via a network such as the Internet; drive 310 is also connected to I / O interface 305 as needed; removable media 311, such as disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from it can be installed into storage section 308 as needed.

[0119] In particular, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs; for example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts; in such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311; when the computer program is executed by central processing unit (CPU) 301, it performs various functions defined in the system of this application.

[0120] It should be noted that although multiple modules are mentioned in the detailed description above, this division is not mandatory; in fact, according to the embodiments of this disclosure, the features and functions of two or more modules or described above can be embodied in one module; conversely, the features and functions of one module described above can be further divided into multiple modules to be embodied.

[0121] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein; this application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein; the specification and embodiments are to be considered exemplary only.

[0122] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-node collaborative online electromagnetic ultrasound monitoring method, characterized in that, include: The current position of the object to be detected is marked, and a multi-dimensional monitoring network is constructed by combining the distribution positions of multiple electromagnetic ultrasonic probes. Orthogonal codes are introduced to demodulate multiple electromagnetic ultrasonic probes simultaneously, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Multiple monitoring nodes are marked in the multi-dimensional monitoring network. Collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data. Combine the attention mechanism to build the corresponding multi-node collaborative detection system. In this multi-node collaborative detection system, the electromagnetic impedance change and ultrasonic guided wave transmission loss of two adjacent monitoring nodes are fused by tensor in the feature layer, and multiple cross-space collaborative detection layers are output. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions. Based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, corresponding suspected defect data is determined. This suspected defect data is loaded into the online monitoring finite element module, and the corresponding online monitoring defect route is determined in combination with the corresponding multi-physics coupling relationship. This triggers the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network. The process includes: acquiring multiple electromagnetic ultrasonic online monitoring contents, dynamically identifying multiple electromagnetic ultrasonic online monitoring contents, determining the corresponding suspected defect data in combination with the anomaly probability distribution mechanism, and injecting the suspected defect data into the online monitoring finite element module in real time. This allows the local geometry and boundary conditions near the defect to be reconstructed in the virtual space of the online monitoring finite element module. In combination with the multi-physics coupling relationship, the acoustic field scattering and eddy current distortion response of the defect under different excitation parameters are solved in a forward manner, thereby depicting the online monitoring defect route that reflects the defect morphology. The optimization is guided by the solution results of this multi-physical coupling relationship, which directly triggers and reverses the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network, forming a complete closed loop from physical perception to data collaboration, digital twin verification, and parameter self-evolution. The monitoring parameters of multiple monitoring nodes cover excitation frequency, bias magnetic field strength, probe array aperture, and sampling rate.

2. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 1, characterized in that, The current position of the object to be detected is marked, and a multi-dimensional monitoring network is constructed by combining the distribution positions of multiple electromagnetic ultrasonic probes. Orthogonal code demodulation is introduced into multiple electromagnetic ultrasonic probes simultaneously, so that the detection of the same object by multiple electromagnetic ultrasonic probes is mutually complementary and does not interfere with each other. Multiple monitoring nodes are marked in the multi-dimensional monitoring network, including: The system acquires the current position of the object under test, dynamically marks the global coordinate system and local morphological features of the object under test, outputs the corresponding topological constraint information, triggers the spatial deployment of multiple electromagnetic ultrasonic probes relative to the same object under test along the topological constraint information, and constructs a multi-dimensional monitoring network with full-view coverage capability.

3. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 2, characterized in that, The method of marking the current position of the object to be detected, constructing a multi-dimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes, and simultaneously introducing orthogonal code demodulation to multiple electromagnetic ultrasonic probes to ensure that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and does not interfere with each other, and marking multiple monitoring nodes in the multi-dimensional monitoring network, also includes: In the multidimensional monitoring network, each electromagnetic ultrasonic probe is assigned a correlated orthogonal coding sequence, so that the ultrasonic electromagnetic fields emitted within the same time window remain orthogonal and interference-free in physical space. Zero crosstalk complementary detection of the same object to be detected is performed along multiple detection channels. Simultaneously, the demodulated electromagnetic ultrasonic signal stream after orthogonal coding sequence is mapped to the multidimensional monitoring network, dynamically generating multiple monitoring nodes with independent spatial attributes and timestamp references.

4. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 1, characterized in that, The process involves collecting detection data output from each monitoring node, extracting the echo timestamp of the object to be detected, compensating for clock drift of each monitoring node along the echo timestamp to trigger spatiotemporal alignment of the detection data, and constructing a corresponding multi-node collaborative detection system using an attention mechanism, including: Multiple monitoring nodes are monitored in real time, and the detection data output by each monitoring node is collected synchronously. The echo timestamp reflected by the object under test is extracted by envelope detection and adaptive threshold. The echo timestamp is compensated in reverse, and the local clock error of each monitoring node is corrected in reverse using the echo timestamp as the absolute time anchor point. This triggers the spatiotemporal alignment of all detection data in the absolute spatiotemporal coordinate system to output the spatiotemporally aligned dataset.

5. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 4, characterized in that, The process of collecting detection data output from each monitoring node, extracting the echo timestamp of the object to be detected, compensating for clock drift of each monitoring node in reverse along the echo timestamp to trigger spatiotemporal alignment of each detection data, and constructing a corresponding multi-node collaborative detection system by combining an attention mechanism, also includes: The dataset is input into the corresponding attention space, and a time decay factor is introduced into the attention space to autonomously focus on the key waveform segments containing defect features between different monitoring nodes. Based on the fusion of multiple key waveform segments, a multi-node collaborative detection system with global feature perception and local anomaly capture capabilities is constructed.

6. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 1, characterized in that, In this multi-node collaborative detection system, the electromagnetic impedance changes and ultrasonic guided wave transmission losses of two adjacent monitoring nodes are tensor-fused at the feature layer, and multiple corresponding cross-space collaborative detection layers are output. These multiple cross-space collaborative detection layers are loaded onto a monitoring gradient architecture with monitoring constraints to dynamically output the online electromagnetic ultrasonic monitoring content of each monitoring node under different operating conditions, including: The multi-node collaborative detection system is monitored in real time. The electromagnetic impedance changes of two adjacent monitoring nodes under the action of electromagnetic field and the ultrasonic guided wave transmission loss on the sound field propagation path are extracted. The electromagnetic impedance changes and ultrasonic guided wave transmission loss are mapped to a high-dimensional feature space for fusion. This fusion not only preserves their respective physical gradient directions, but also mines the nonlinear response characteristics of electromagnetic-acoustic coupling at the defect edge through cross tensor product. Multiple cross-space collaborative detection layers with cross-space physical correlation attributes are constructed along each nonlinear response characteristic.

7. The multi-node collaborative electromagnetic ultrasonic online monitoring method according to claim 6, characterized in that, In this multi-node collaborative detection system, the electromagnetic impedance changes and ultrasonic guided wave transmission losses of two adjacent monitoring nodes are tensor-fused at the feature layer, and multiple corresponding cross-space collaborative detection layers are output. These multiple cross-space collaborative detection layers are then loaded onto a monitoring gradient architecture with monitoring constraints to dynamically output the online electromagnetic ultrasonic monitoring content of each monitoring node under different operating conditions. The system also includes: Multiple cross-space collaborative detection layers are loaded into a pre-constructed monitoring gradient architecture with monitoring constraints. This monitoring gradient architecture uses the physical continuity of adjacent monitoring nodes as a priori constraint and dynamically suppresses operating noise through gradient backpropagation, thereby triggering the dynamic output of electromagnetic ultrasonic online monitoring content of each monitoring node under different complex operating conditions, so as to realize the self-adaptation of electromagnetic ultrasonic online monitoring content of each monitoring node.

8. A multi-node collaborative electromagnetic ultrasonic online monitoring system, characterized in that, The multi-node collaborative electromagnetic ultrasound online monitoring system is applied to the multi-node collaborative electromagnetic ultrasound online monitoring method as described in any one of claims 1-7; the multi-node collaborative electromagnetic ultrasound online monitoring system includes: The multidimensional monitoring module is used to mark the current position of the object to be detected. It constructs a multidimensional monitoring network by combining the distribution positions of multiple electromagnetic ultrasonic probes. It synchronously introduces orthogonal code demodulation to multiple electromagnetic ultrasonic probes, so that the detection of the same object by multiple electromagnetic ultrasonic probes is complementary and mutually exclusive. Multiple monitoring nodes are marked in the multidimensional monitoring network. The multi-node collaborative detection system module is used to collect the detection data output by each monitoring node, extract the echo timestamp of the object to be detected, and compensate the clock drift of each monitoring node in reverse along the echo timestamp to trigger the spatiotemporal alignment of each detection data, and build the corresponding multi-node collaborative detection system in combination with the attention mechanism. The electromagnetic ultrasonic online monitoring module is used in this multi-node collaborative detection system to perform tensor fusion of the electromagnetic impedance changes and ultrasonic guided wave transmission loss of two adjacent monitoring nodes in the feature layer, and output multiple corresponding cross-space collaborative detection layers. The multiple cross-space collaborative detection layers are loaded onto the monitoring gradient architecture with monitoring constraints to dynamically output the electromagnetic ultrasonic online monitoring content of each monitoring node under different working conditions. The adaptive adjustment module is used to determine the corresponding suspected defect data based on the dynamic identification of multiple electromagnetic ultrasonic online monitoring contents, load the suspected defect data into the online monitoring finite element module, and determine the corresponding online monitoring defect route in combination with the corresponding multi-physics coupling relationship, and trigger the adaptive adjustment of the monitoring parameters of multiple monitoring nodes in the multi-dimensional monitoring network.

Images