A three-dimensional perception heat exchanger leakage detection method, system, device and medium

Abstract

The present disclosure relates to the technical field of heat exchanger detection, and particularly relates to a three-dimensional sensing heat exchanger leakage detection method, system, device and medium, wherein the method comprises: collecting sound wave signals, impedance and infrared image data of a heat exchanger; respectively extracting features of the sound wave signals, impedance and infrared image data and combining to obtain a combined feature vector; inputting the combined feature vector into a pre-constructed classification model to obtain a leakage type of the heat exchanger; constructing a hyperbolic equation based on space ray constraints and height plane constraints obtained from the infrared image data, range constraints obtained from the impedance and a time difference of arrival calculated from the sound wave signals; solving the hyperbolic equation to obtain three-dimensional coordinates and a leakage level; and mapping the three-dimensional coordinates, the leakage type and the leakage level to a three-dimensional model of the heat exchanger to perform an alarm, effectively detecting the leakage type and position of the heat exchanger and solving the problem of difficult operation and maintenance decision-making.

Description

Technical Field

[0001] This disclosure relates to the field of heat exchanger detection technology, and in particular to a three-dimensional sensing method, system, device and medium for heat exchanger leak detection. Background Technology

[0002] Currently, leak detection in solar thermal power plant heat exchangers mainly relies on the following methods: Macroscopic physical parameter monitoring: Leaks are identified by monitoring tank levels, pipeline pressure differences, or flow balance. Single acoustic detection: Leakage noise is monitored using acoustic emission sensors attached to the exterior of the casing. Conventional infrared detection: The external surface temperature of the equipment is captured using handheld or fixed thermal imagers.

[0003] Therefore, existing technologies struggle to detect the type and location of heat exchanger leaks, leading to difficulties in operation and maintenance decisions. Summary of the Invention

[0004] To address the aforementioned technical issues, this disclosure provides a three-dimensional sensing method, system, device, and medium for detecting leaks in heat exchangers.

[0005] This disclosure provides a three-dimensional sensing method for detecting leaks in heat exchangers, including: Acquire acoustic signals, impedance, and infrared image data of the heat exchanger; The acoustic signal, the impedance, and the infrared image data are extracted and combined to obtain a combined feature vector. The combined feature vector is input into a pre-built classification model to obtain the leakage type of the heat exchanger; A hyperbolic equation is constructed based on spatial ray constraints and height plane constraints obtained from infrared image data, impedance range constraints, and arrival time differences calculated from acoustic signals. Solving the hyperbolic equation yields the three-dimensional coordinates and leakage level; The three-dimensional coordinates, the leakage type, and the leakage level are mapped to the three-dimensional model of the heat exchanger to generate an alarm.

[0006] Further, the feature extraction of the acoustic signal, the impedance, and the infrared image data respectively includes: The acoustic signal is subjected to wavelet packet decomposition for noise reduction, and time-domain features, frequency-domain features, and Mel-frequency cepstral coefficients are extracted. Calculate the impedance change rate and extract the topological center coordinates and area of ​​the conducting region; The infrared image is corrected, the hot spot edge is extracted, and the temperature difference and temperature gradient direction histogram of the target area are calculated.

[0007] Furthermore, the construction of the classification model includes: When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is greater than the second threshold, and the temperature rise is greater than the third threshold, it is determined to be molten salt leakage; When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature drop is greater than the third threshold, it is determined to be a steam leak. When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature remains unchanged, it is determined to be external interference.

[0008] Furthermore, solving the hyperbolic equation to obtain the three-dimensional coordinates and leakage level includes: After solving the hyperbolic equation, the leakage level is determined based on the weights of the infrared image, impedance, and acoustic wave. When the acoustic wave weight is greater than the infrared image weight and the impedance weight, it is determined to be the initial stage of leakage; When the weight of the infrared image is greater than the weight of the acoustic wave and the weight of the impedance, it is determined to be in the middle stage of leakage; When the impedance weight is greater than the acoustic weight and the infrared image weight, it is determined to be in the late stage of leakage.

[0009] This disclosure also provides a three-dimensional sensing heat exchanger leak detection system for performing the three-dimensional sensing heat exchanger leak detection method, comprising: a waveguide array module, an electrode array module, and an infrared sensing module respectively installed on the heat exchanger.

[0010] Furthermore, the waveguide array module is uniformly arranged along the circumference of the heat exchanger and is arranged in layers across multiple cross sections in the axial direction.

[0011] Furthermore, the electrode array module includes an electrode mesh with at least two layers; The electrode mesh is installed on the flange face of the heat exchanger tube sheet; and / or the heat exchanger head region; And / or inside the heat exchanger insulation layer.

[0012] Furthermore, the infrared sensing module includes a top-view unit and a side-view unit; The top-view unit is installed on the top of the heat exchanger housing; The side-view unit is mounted on the side of the heat exchanger.

[0013] This disclosure also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the three-dimensional sensing heat exchanger leak detection method.

[0014] This disclosure also provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of the three-dimensional sensing heat exchanger leak detection method.

[0015] The technical solution provided in this disclosure has the following advantages compared with the prior art: By collecting acoustic signals, impedance data, and infrared image data from the heat exchanger, feature extraction and combination are performed on the acoustic signals, impedance data, and infrared image data to obtain a combined feature vector. The combined feature vector is input into a pre-built classification model to obtain the leakage type of the heat exchanger. Based on the spatial ray constraints and height plane constraints obtained from the infrared image data, the range constraints obtained from the impedance data, and the arrival time difference calculated from the acoustic signals, a hyperbolic equation is constructed. The hyperbolic equation is solved to obtain the three-dimensional coordinates and leakage level. The three-dimensional coordinates, leakage type, and leakage level are mapped onto the three-dimensional model of the heat exchanger to trigger an alarm, effectively detecting the leakage type and location of the heat exchanger and solving the problem of difficult operation and maintenance decision-making. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0017] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of a three-dimensional sensing heat exchanger leak detection method provided in an embodiment of this disclosure; Figure 2 This is a schematic diagram illustrating feature extraction provided in an embodiment of this disclosure. Detailed Implementation

[0019] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0020] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0021] Figure 1This is a schematic diagram of a three-dimensional sensing heat exchanger leak detection method provided in an embodiment of this disclosure; as shown. Figure 1 As shown, this disclosure provides a three-dimensional sensing method for detecting leaks in heat exchangers, including: Step S1: Acquire acoustic signals, impedance, and infrared image data of the heat exchanger; In this embodiment, high-frequency acoustic signals generated in the early stages of leakage are collected by an array of high-temperature resistant acoustic sensors deployed at different locations on the heat exchanger. Changes in electrical characteristics caused by the leakage are collected by an impedance detection unit located at a key structural part of the heat exchanger. Simultaneously, top-view and side-view infrared imaging devices are used to acquire temperature distribution images of the heat exchanger surface and internal channels. The reason for simultaneously collecting these three types of data is that under the operating conditions of high-temperature, high-noise, and structurally complex heat exchangers, single sensing methods often suffer from insufficient information dimensions, high false alarm rates, or insensitivity to early, minor leaks. For example, acoustic signals are easily interfered with by mechanical vibrations, infrared images may have location ambiguity under obstruction or thermal inertia conditions, and impedance detection is difficult to respond to early leaks that have not yet contacted the electrodes. By simultaneously collecting acoustic, impedance, and infrared data at the same time reference, the acoustic, electrical, and thermal characteristics of the leak can be covered from a physical mechanism perspective, providing a complete information source for subsequent multimodal fusion. On the one hand, it significantly improves the system's perception completeness of leakage events at different stages and in different media. On the other hand, it provides a data foundation that is consistent in time and complementary in information for subsequent feature extraction, classification and identification and spatial positioning, fundamentally avoiding missed detection or misjudgment caused by the failure of a single sensor.

[0022] Step S2: Extract and combine features from the acoustic signal, impedance, and infrared image data respectively to obtain a combined feature vector; In this embodiment, time-domain, frequency-domain, and time-frequency-domain analyses are performed on the acoustic signal to extract features reflecting the leakage and ejection characteristics, such as high-frequency energy distribution, statistical characteristics, and Mel-frequency cepstral coefficients. Impedance data is calculated to determine its rate of change over time, the amplitude of changes in the real and imaginary parts, and the spatial distribution characteristics of trigger nodes. For infrared image data, image correction and target detection algorithms are used to extract features such as temperature anomaly amplitude, hot spot morphology, and temperature gradient direction within the region of interest. Feature extraction and combination are crucial because the original multimodal data exhibits significant differences in scale, sampling frequency, and physical dimensions. Directly using these data for model judgment is not only computationally complex but also prone to introducing redundant information and noise interference. By transforming each modal data into discriminative feature representations and fusing them at the feature level, key information can be preserved while reducing data dimensionality, improving the stability and generalization ability of subsequent classification and localization algorithms. Transforming complex, multi-source, and heterogeneous raw monitoring data into unified, structured, and computable feature vectors provides high-information-density input for the intelligent model, significantly improving the accuracy of leak type identification and laying a reliable data foundation for subsequent spatial constraint construction and localization calculations.

[0023] Step S3: Input the combined feature vector into the pre-built classification model to obtain the leakage type of the heat exchanger; In this embodiment, the combined feature vector is input into a pre-built and trained classification model to determine the current state of the heat exchanger and output the corresponding leakage type. The classification model is trained based on the inherent correlation between multimodal features, and can comprehensively integrate acoustic features, impedance change features, and thermal imaging features to distinguish abnormal states under different physical mechanisms, such as molten salt leaks, water or steam leaks, and non-leakage external interference. In actual industrial environments, not all abnormal signals originate from actual leak events. Failure to distinguish leakage types can easily lead to false alarms or inappropriate emergency response measures. By introducing a data-driven classification model, the statistical regularities between historical samples and multimodal features can be used to accurately identify different abnormal patterns. Clearly determining the type of leaking medium before leak location and alarm not only significantly reduces the false alarm rate caused by factors such as mechanical vibration and environmental noise, but also provides a targeted basis for subsequent leak level assessment and maintenance decisions. For example, different maintenance priorities and safety handling strategies can be adopted for different media, thereby improving the overall intelligence level and practical application value of the system.

[0024] Step S4: Based on the infrared image data, spatial ray constraints and height plane constraints are obtained, impedance range constraints are obtained, and the arrival time difference is calculated from the acoustic signal to construct a hyperbola equation. In this embodiment, spatial ray constraints are formed by back-projecting the hot spot centroid location detected in the top-view infrared image through the camera imaging model; corresponding height plane constraints are obtained by mapping the image rows or columns where the abnormal temperature band is located in the side-view infrared image; spatial range constraints for the possible existence of the leak point are constructed based on the electrode grid level and region range triggered in the impedance detection unit; and corresponding hyperbolic constraint equations are established based on the arrival time difference of the leakage acoustic signals received by different sensors in the acoustic array. By constructing the above multiple constraints, it is difficult to achieve high-precision three-dimensional positioning within complex structures using a single sensing mode. For example, infrared images themselves lack depth information, impedance detection can only provide the contact area range, and acoustic positioning is easily affected by propagation paths and noise. By combining spatial constraints obtained under different physical mechanisms, the feasible solution space of the leak point can be significantly compressed geometrically. This forms a multi-constraint, complementary verification spatial positioning condition system, so that the location of the leak point no longer depends on a single sensing result, but is uniquely determined at the intersection of multi-source information, thereby significantly improving the reliability and accuracy of three-dimensional positioning.

[0025] Step S5: Solve the hyperbola equation to obtain the three-dimensional coordinates and leakage level; In this embodiment, the three-dimensional coordinates of the leak point are solved based on the constructed hyperbolic constraint equation and other spatial constraints, and the leak level is determined accordingly. Specifically, by uniformly modeling the deviations corresponding to each constraint, the degree of deviation between the candidate leak point and each constraint is calculated. Considering the reliability of different constraints, each deviation is weighted, and the solution is iteratively applied with the goal of minimizing the overall residual, ultimately obtaining the three-dimensional coordinates of the leak point that satisfy multiple constraints. Simultaneously, based on the residual distribution of each constraint, the intensity of the acoustic signal, and the duration of the anomaly during the solution process, the scale and severity of the leak are comprehensively evaluated, and the corresponding leak level is output. The leak location problem is a spatial parameter estimation problem under multiple constraints. Simply using a single constraint or direct intersection is easily affected by noise and outliers. By introducing residual minimization and weight adjustment mechanisms, the interference of unreliable constraints on the results can be effectively suppressed. Even under complex operating conditions, the three-dimensional spatial coordinates of the leak point can be obtained stably and accurately, while simultaneously providing level information reflecting the severity of the leak, providing highly reliable quantitative results for subsequent visualization and operation and maintenance decisions.

[0026] Step S6: Map the three-dimensional coordinates, leakage type, and leakage level to the three-dimensional model of the heat exchanger and issue an alarm.

[0027] In this embodiment, by spatially registering the location coordinates with the structural model of the heat exchanger, the specific location of the leak is visually marked in the model. Simultaneously, color, icon, or text prompts are used to display the type and severity of the leaking medium, triggering corresponding alarm information. Traditional leak alarm systems often only provide simple alarm signals, failing to intuitively reflect the leak location and impact range, making it difficult for on-site maintenance personnel to quickly assess and accurately handle the situation. By mapping the detection results to a 3D model, abstract data results can be transformed into intuitive and understandable spatial information. This significantly improves the visualization level and response efficiency of leak events, enabling maintenance personnel to accurately locate the leak location, understand its nature and severity, and thus develop more efficient and safer maintenance and emergency response plans. Ultimately, this improves the safety of heat exchanger operation and the overall intelligence level of system operation and maintenance.

[0028] Figure 2 A schematic diagram of feature extraction provided in the embodiments of this disclosure; as shown Figure 2 As shown, in some possible implementations, step S2 involves feature extraction from the acoustic signal, impedance, and infrared image data, including: step S21, performing wavelet packet decomposition to reduce noise in the acoustic signal and extracting time-domain features, frequency-domain features, and Mel-frequency cepstral coefficients; step S22, calculating the impedance change rate and extracting the topological center coordinates and area of ​​the conducting region; and step S23, correcting the infrared image, extracting hot spot edges, and calculating the temperature difference value and temperature gradient direction histogram of the target region.

[0029] In this embodiment, wavelet packet decomposition is used to denoise the acoustic signal. By selecting an appropriate wavelet basis and decomposition level, the useful energy in the target frequency band (e.g., 1–20 kHz) is separated and low-frequency structural vibrations and high-frequency white noise are suppressed. Then, rich time-domain features (such as peak value, kurtosis, skewness, impulse factor / peak-to-average power ratio, etc.), frequency-domain features (such as spectral centroid, band energy ratio, bandpass energy ratio), and time-frequency features (short-time Fourier transform spectral entropy, instantaneous energy envelope) are extracted from the denoised signal. At the same time, Mel-Frequency Cepstral Coefficients (MFCCs) are calculated and extracted: the signal is first pre-emphasized, framed, and windowed, then FFT (Fast Fourier Transform) is performed and mapped to the Mel frequency scale through the Mel filter bank. Finally, the logarithm of the filter energy is taken and DCT (Discrete Cosine Transform) is performed to obtain the MFCC vector. The processing of impedance data focuses on time-series changes: first, the complex impedance of each grid cell is smoothed and baseline corrected, the impedance change rate and short-time step change are calculated and detected. When the values ​​of several adjacent nodes change abruptly or decrease (or their absolute change exceeds a set threshold), the conducting region is determined based on connectivity and the topological features of the conducting region are extracted, including the topological center coordinates of the region (X, Y coordinates obtained by spatial mapping based on the grid index), the number of triggering cells and the region area (number of cells × cell area), and the region shape features (aspect ratio, boundary perimeter). For infrared images, non-uniformity correction (NUC) and temperature calibration are performed first, then hot spot detection is performed, hot spot edges are extracted using improved edge detection, the temperature difference value ΔT of the target region and the temperature gradient direction histogram (HOG) are calculated to characterize the shape and flow direction information of the hot spot, and the pixel centroid / geometric centroid and ROI (Region of Interest) area of ​​the hot spot are recorded. Finally, the vectorized features from the three modalities are normalized in a standardized order and concatenated into a unified combined feature vector, with timestamps retained in the vector to ensure consistency with temporal constraints such as TDOA (Time Difference of Arrival). The original sensor data differ significantly in physical dimensions, sampling frequency, and noise characteristics. Directly concatenating the original data is not only difficult to train machine learning models but also prone to introducing redundancy. By extracting highly discriminative and low-dimensional features targeting the physical mechanisms of each modality, model training efficiency can be significantly improved, overfitting risk reduced, and multimodal information can be compared and fused at the same semantic level.On the one hand, it improves the sensitivity to minute early leakage signals, and on the other hand, it enhances the ability to distinguish media. At the same time, it provides stable, interpretable and computationally controllable inputs for classifiers and localization algorithms, reduces false alarm rates and improves the robustness and practicality of the overall system.

[0030] In some possible implementations, the classification model is constructed as follows: when the high-frequency energy of the sound wave is greater than a first threshold, the real part of the impedance is greater than a second threshold, and the temperature rise is greater than a third threshold, it is determined to be molten salt leakage; when the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature drop is greater than the third threshold, it is determined to be steam leakage; when the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature does not change, it is determined to be external interference.

[0031] In this embodiment, three key thresholds are defined and calibrated: acoustic high-frequency energy threshold T1, impedance real part change threshold T2 (or absolute value threshold of impedance change |ΔR|>T2), and thermal field temperature difference threshold T3. The optimal threshold range is obtained through statistical methods on the test bench or historical data. The judgment rule is as follows: if the acoustic high-frequency energy exceeds T1 and the change of the real part of the impedance exceeds T2, and the infrared display shows a local temperature rise ΔT>T3, it is judged as molten salt leakage, because molten salt is high temperature and conductive. When it leaks and sprays, it generates high-frequency mechanical noise and forms local heat release in the contact area, which leads to electrode short circuit or a sudden drop in the real part of the impedance. If the acoustic high-frequency energy exceeds T1, the impedance does not have significant conduction, but the imaginary part of the impedance or the dielectric constant drifts, it is judged as water / steam leakage. Steam / water vaporizes and absorbs heat instantly, which leads to local cold spots and is not conductive at first. If there is an acoustic abnormality but no significant change in impedance and temperature, it is judged as external interference (such as pump vibration or mechanical impact), which can trigger a low-level alarm or enter a further confirmation process. It significantly reduces the false alarm rate and enhances the ability to distinguish between different media, enabling the system to make a physically credible preliminary judgment at an early stage. It can also improve the stability of the judgment, thereby ensuring safety and reducing unnecessary human intervention in industrial scenarios.

[0032] In some possible implementations, the hyperbolic equation is solved to obtain three-dimensional coordinates and leakage levels, including: after solving the hyperbolic equation, the leakage level is determined based on the weights of infrared image, impedance, and acoustic wave; when the acoustic wave weight is greater than the infrared image weight and impedance weight, it is determined to be the initial stage of leakage; when the infrared image weight is greater than the acoustic wave weight and impedance weight, it is determined to be the middle stage of leakage; when the impedance weight is greater than the acoustic wave weight and infrared image weight, it is determined to be the late stage of leakage.

[0033] In this embodiment, a set of hyperboloid equations is established based on the sensor position and the measured time difference of arrival. Simultaneously, the imaging backprojection of the top-view infrared centroid is represented as a spatial ray, the image row number of the side view is mapped to a height plane, and the impedance grid trigger provides the depth interval. These constraints are transformed into residual terms and weighted according to their respective confidence levels. Then, an iterative nonlinear weighted least squares method is used to linearize the hyperboloid residuals and solve the normal equations in each iteration to obtain the position increment. The iteration continues until convergence or the convergence criterion is met. To enhance robustness, an iterative reweighting strategy is used for abnormal residuals, and the weights are dynamically adjusted during the iteration process to reduce the impact of noise constraints. The leakage level is determined by combining weighting proportions and residual characteristics: if the acoustic residual dominates in the final solution and the acoustic weight is significantly greater than the visual and impedance weights, it is judged as the early stage of leakage, indicating that the acoustic signal has been detected but there is no hot spot or electrical contact yet; if the visual weight and hot spot dominate and the hot spot area and temperature difference are large, it is judged as the middle stage of leakage (medium level), indicating that a considerable heat flow has been formed and there are visible traces; if the impedance weight dominates and the conductive area is large and the real part of the impedance shows a significant step drop, it is judged as the late stage of leakage or severe leakage (high level), which usually means that the medium has touched the electrode network and created a continuous conductive path. By jointly solving the constraints of different physical quantities under a unified mathematical framework, the complementary advantages of acoustic temporal resolution, visual spatial resolution, and impedance contact criteria are utilized, and the weighting mechanism solves the reliability differences of each sensor at different leakage stages. High-precision three-dimensional positioning is achieved, and graded alarms can be given at the same time as positioning, providing quantifiable priority determination and confidence assessment for operation and maintenance, thereby significantly improving on-site response efficiency and reducing unnecessary downtime or manpower costs caused by misjudgment.

[0034] In some possible implementations, data preprocessing and feature engineering are involved. Acoustic processing: Wavelet packet decomposition is performed on the original signal for noise reduction. Time-domain features (peak value, kurtosis, skewness), frequency-domain features (spectral centroid, energy ratio of the 1-20kHz band), and MFCC (Mel frequency cepstral coefficients) are extracted.

[0035] Impedance processing: Calculate the impedance change rate dZ / dt, and extract the topological center coordinates and area of ​​the conducting region.

[0036] Image processing: Non-uniformity correction (NUC) is performed on the infrared image. The edges of hot spots are extracted using the Canny (Canny Edge Detection) operator, and the maximum temperature difference ΔT within the ROI region is calculated. max And temperature gradient orientation histogram (HOG). Canny is a multi-stage edge detection algorithm used to extract clear, continuous, and noise-resistant edge information from images.

[0037] Fusion diagnostic model based on LightGBM (Light Gradient Boosting Machine, a classification model): To address the false positive problem, a classification model based on LightGBM was constructed.

[0038] Input: Combined feature vector V=[E acoustic ,S kewness ,R impedance ,ΔT ir ]; In the formula, E acoustic The high-frequency energy of the sound wave signal; S kewness This refers to the skewness characteristic in statistics; R impedance This represents the real part of the impedance.

[0039] Decision logic (training objective): Molten salt leakage: High-frequency sound waves with high energy + a sharp drop in the real part of impedance R + a significant increase in local temperature T (>5℃), because molten salt is a high-temperature fluid and is conductive.

[0040] Water / steam leakage: High-frequency sound waves with high energy + impedance with no change in magnitude (or only the imaginary part X changes) + local temperature T decreases (cold spots appear). The local temperature drops due to the instantaneous vaporization of water absorbing heat, and it is not conductive at the beginning.

[0041] External interference: Only the sound wave is abnormal, and there is no change in impedance and temperature. Since it is judged to be pump vibration or external impact, no alarm is triggered.

[0042] Three-dimensional spatial location calculation (geometric intersection method): used to calculate the leak point (X,Y,Z).

[0043] Visual ray constraint: The hot spot centroid (u1, v1) is detected by the top-view infrared camera. Combined with the camera intrinsic parameter matrix, the spatial ray L1 is obtained by back projection.

[0044] Depth plane constraint: The image row number where the abnormal temperature rise zone is detected by the side-looking infrared camera is mapped to the physical height plane Z. ir Meanwhile, the impedance array-triggered grid hierarchy provides a coarse depth range Z. imp .

[0045] Acoustic TDOA correction: If vision is obstructed, an acoustic array is used. The time difference Δτ between the arrival times of the sound waves at each sensor is calculated based on the cross-correlation function. ij Construct a system of hyperbolic equations: ; In the formula, c is the equivalent sound velocity of the sound wave in the tube bundle medium; The time it takes for the sound wave signal to reach the i-th sensor; The time it takes for the acoustic signal to reach the j-th sensor is denoted as (X, Y, Z); (X, Y, Z) represents the three-dimensional spatial coordinates of the leak point. , , Let (xj, yj, zj) be the known three-dimensional coordinates of the i-th sensor; , , (j) represents the known three-dimensional coordinates of the j-th sensor.

[0046] Fusion solution: The above equations are solved using the weighted least squares (WLS) method, with the weights w dynamically adjusted according to the signal-to-noise ratio of each sensor.

[0047] Initial stage of leakage (no liquid accumulation): acoustic weighting w ac maximum.

[0048] Mid-stage leakage (hot spot formation): Infrared weighting w ir maximum.

[0049] Late stage of leakage (contact electrode): Impedance weight w imp It becomes a decisive constraint.

[0050] Digital twin visualization: The calculated coordinates (X,Y,Z) are mapped onto the 3D CAD model of the heat exchanger, and the location of the leak point, the type of leaking medium, and the diffusion trend are displayed intuitively on the large screen in the control room, assisting maintenance personnel in formulating emergency repair plans.

[0051] This disclosure also provides a three-dimensional sensing heat exchanger leak detection system for performing a three-dimensional sensing heat exchanger leak detection method, comprising: a waveguide array module, an electrode array module, and an infrared sensing module respectively installed on the heat exchanger.

[0052] The waveguide array modules are uniformly arranged along the circumference of the heat exchanger and layered in multiple cross sections along the axial direction.

[0053] The electrode array module includes an electrode mesh with at least two layers; the electrode mesh is installed on the flange face of the heat exchanger tube sheet; and / or the heat exchanger head area; and / or inside the heat exchanger insulation layer.

[0054] The infrared sensing module includes a top-view unit and a side-view unit; the top-view unit is installed on the top of the heat exchanger shell; the side-view unit is installed on the side of the heat exchanger.

[0055] In some possible implementations, the detailed structure and layout of the hardware device are as follows: This device consists of three major sensing subsystems (acoustic wave, impedance, and thermal imaging) and a data fusion center.

[0056] High-temperature resistant broadband acoustic sensing subsystem (auditory module): used to capture high-frequency jet noise (1kHz-20kHz) generated in the early stage of leakage.

[0057] Waveguide Array: Due to the high surface temperature of the heat exchanger (300-560℃), it cannot be directly attached to the sensor. Therefore, an acoustic waveguide is used. Material and Dimensions: Nickel-based alloy or stainless steel is selected to withstand high-temperature corrosion. The waveguide length L is set at 500mm, and the diameter D is 12mm (meeting a heat dissipation length-to-diameter ratio of L / D≥15), ensuring that the end temperature drops below 80℃.

[0058] Coupling method: The front end of the waveguide is designed as a conical sound-absorbing head, which penetrates the heat exchanger shell and insulation layer through full penetration welding, and directly contacts the internal fluid domain or tube sheet flange surface to reduce acoustic impedance mismatch.

[0059] Vibration damping design: The outer wall of the waveguide is wrapped with ceramic fiber cotton and flexible rubber sleeve, and a flexible seal is used where it passes through the outer skin to block the lateral transmission of environmental vibration (such as pump vibration) along the shell to the waveguide.

[0060] Sensor selection: Connect an industrial-grade piezoelectric ceramic acoustic wave sensor (or fiber optic grating sensor) to the low-temperature end (terminal) of the waveguide, with a sampling rate set to 51.2kHz and a sensitivity of not less than 100mV / Pa.

[0061] Arrangement strategy: The heat exchangers are evenly arranged along the circumference of the heat exchanger and layered at different axial sections (such as tube sheet and baffle plate) to form a three-dimensional spatial array, which provides a basis for TDOA (Time Difference of Arrival) positioning.

[0062] Mesh impedance scanning subsystem (tactile module): Utilizes the differences in the electrical properties of the medium to achieve contact-based capture and rough localization of the leaking medium.

[0063] Electrode array structure: Multiple layers of high-temperature resistant flexible detection network are laid on the flange face, end cap area and inside the insulation layer of the heat exchanger tube sheet.

[0064] Mesh weaving: High-temperature resistant nickel-chromium alloy wire with a diameter of 0.5mm is used as the conductive electrode; high-temperature resistant ceramic beads or mica sheets are used as node insulation supports to construct 50mm×50mm mesh units.

[0065] Layered design: The electrode mesh is designed with at least two layers. The inner layer is in close contact with the outer wall of the equipment (detection layer), and the outer layer is grounded (shielding layer), which can both detect leakage and shield external electromagnetic interference.

[0066] Detection principle: The system is connected to a multi-channel LCR digital bridge, which polls and scans the complex impedance of each grid node at multiple frequency points from 1kHz to 100kHz. Identification of conductive molten salt: Molten salt is a good conductor (high conductivity) at high temperatures. Once the leaking liquid comes into contact with the electrode, the real part R of the internode impedance will drop sharply (from >10MΩ in an open circuit to the kΩ level instantly), forming a short circuit signal.

[0067] Identification of non-conductive media (water / oil): Water vapor or heat transfer oil is not conductive at first, but it will change the dielectric constant of the medium, causing the imaginary part of the impedance X (capacitive reactance) to drift.

[0068] Dual-view orthogonal infrared sensing subsystem (vision module): Orthogonal / complementary viewing angle arrangement, top-view unit: mounted on the top shell of the heat exchanger, with the optical axis perpendicular to the tube bundle axis (Z-axis). Through sapphire crystal and a viewing window, the tube bundle temperature field in the XY plane is observed, capturing the planar position of hot spots.

[0069] Side-viewing unit: Installed on the side of the heat exchanger (e.g., at the gap between baffles), with the optical axis parallel to the gap between the tube rows. Through the elongated zinc sulfide viewing window, the temperature gradient distribution along the Z-axis is observed, capturing longitudinal heat marks formed by dripping or spraying leaking fluid.

[0070] Air-cooled purging device: Each viewing window base is equipped with an annular jet nozzle, through which dry nitrogen (or compressed air) is introduced. Its main functions are: firstly, to cool the viewing window glass and prevent thermal stress cracking; and secondly, to form an air curtain to prevent oil vapors and salt spray from condensing on the inner surface of the viewing window and obstructing the view.

[0071] System connection and signal flow, front-end acquisition: All analog signals (sound waves, impedance) are conditioned by a high-temperature resistant preamplifier and then transmitted to a multi-channel synchronous data acquisition card.

[0072] Time synchronization: A hardware trigger line is introduced to ensure that the acoustic waves, impedance, and infrared images have a unified nanosecond-level time reference, which is a prerequisite for subsequent data fusion.

[0073] Transmission layer: The pre-processed digital signal is transmitted to the central workstation (data fusion center) via industrial Ethernet or fiber optic cable.

[0074] This disclosure also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of a three-dimensional sensing heat exchanger leak detection method.

[0075] This disclosure also provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of a three-dimensional sensing heat exchanger leak detection method.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0077] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A three-dimensional sensing method for detecting leaks in heat exchangers, characterized in that, include: Acquire acoustic signals, impedance, and infrared image data of the heat exchanger; The acoustic signal, the impedance, and the infrared image data are extracted and combined to obtain a combined feature vector. The combined feature vector is input into a pre-built classification model to obtain the leakage type of the heat exchanger; A hyperbolic equation is constructed based on spatial ray constraints and height plane constraints obtained from infrared image data, impedance range constraints, and arrival time differences calculated from acoustic signals. Solving the hyperbolic equation yields the three-dimensional coordinates and leakage level; The three-dimensional coordinates, the leakage type, and the leakage level are mapped to the three-dimensional model of the heat exchanger to generate an alarm.

2. The three-dimensional sensing heat exchanger leakage detection method according to claim 1, characterized in that, The feature extraction of the acoustic signal, the impedance, and the infrared image data respectively includes: The acoustic signal is subjected to wavelet packet decomposition for noise reduction, and time-domain features, frequency-domain features, and Mel-frequency cepstral coefficients are extracted. Calculate the impedance change rate and extract the topological center coordinates and area of ​​the conducting region; The infrared image is corrected, the hot spot edge is extracted, and the temperature difference and temperature gradient direction histogram of the target area are calculated.

3. The three-dimensional sensing heat exchanger leakage detection method according to claim 1, characterized in that, The construction of the classification model includes: When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is greater than the second threshold, and the temperature rise is greater than the third threshold, it is determined to be molten salt leakage; When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature drop is greater than the third threshold, it is determined to be a steam leak. When the high-frequency energy of the sound wave is greater than the first threshold, the real part of the impedance is less than the second threshold, and the temperature remains unchanged, it is determined to be external interference.

4. The three-dimensional sensing heat exchanger leakage detection method according to claim 1, characterized in that, Solving the hyperbolic equation to obtain the three-dimensional coordinates and leakage level includes: After solving the hyperbolic equation, the leakage level is determined based on the weights of the infrared image, impedance, and acoustic wave. When the acoustic wave weight is greater than the infrared image weight and the impedance weight, it is determined to be the initial stage of leakage; When the weight of the infrared image is greater than the weight of the acoustic wave and the weight of the impedance, it is determined to be in the middle stage of leakage; When the impedance weight is greater than the acoustic weight and the infrared image weight, it is determined to be in the late stage of leakage.

5. A three-dimensional sensing heat exchanger leak detection system, characterized in that, The method for performing the three-dimensional sensing heat exchanger leak detection method as described in any one of claims 1 to 4 includes: a waveguide array module, an electrode array module, and an infrared sensing module respectively installed on the heat exchanger.

6. The three-dimensional sensing heat exchanger leak detection system according to claim 5, characterized in that, The waveguide array module is uniformly arranged along the circumference of the heat exchanger and is arranged in layers at multiple cross sections in the axial direction.

7. The three-dimensional sensing heat exchanger leak detection system according to claim 5, characterized in that, The electrode array module includes an electrode mesh with at least two layers. The electrode mesh is installed on the flange face of the heat exchanger tube sheet; and / or the heat exchanger head area; And / or inside the heat exchanger insulation layer.

8. The three-dimensional sensing heat exchanger leak detection system according to claim 5, characterized in that, The infrared sensing module includes a top-view unit and a side-view unit; The top-view unit is installed on the top of the heat exchanger housing; The side-view unit is mounted on the side of the heat exchanger.

9. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the three-dimensional sensing heat exchanger leak detection method as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the three-dimensional sensing heat exchanger leak detection method as described in any one of claims 1 to 4.

Images