Rapid diagnosis method for leakage fault of heat exchanger

Abstract

The invention discloses a rapid diagnosis method for a leakage fault of a heat exchanger, and relates to the technical field of air energy heat pumps. Comprising the steps of S1, arranging basic parameter sensors on fluid channels such as a refrigerant side, a water side and a heat exchange channel of a heat exchanger, arranging acoustics and vibration sensors in a shell, arranging environment parameter sensors on the outer side to construct a monitoring network, and S2, extracting time sequence, frequency and physical mechanism correlation characteristics after data preprocessing, and constructing a multi-dimensional characteristic system. S3, analyzing sample distribution in real time, calculating an index deviating from a normal data cluster, and triggering early warning and activating diagnosis when the standard is reached; and inputting the feature data into the model, outputting a fault category, confidence and a basis, and performing dynamic threshold verification in combination with a working condition. And S4, constructing a positioning equation set based on the ultrasonic signal, correcting the propagation velocity through the flow velocity and the flow direction of the refrigerant, and solving the coordinates of the leakage point. And S5, establishing a coordinate mapping reference according to the layout of the cooling fins and the condenser pipes, converting coordinates into actual physical positions, and feeding back the actual physical positions to maintenance personnel in a multi-dimensional manner.

Description

Technical Field

[0001] This invention belongs to the field of air source heat exchanger technology, and particularly relates to a rapid diagnosis method for heat exchanger leakage faults. Background Technology

[0002] As the core energy exchange component of an air source heat pump system, the sealing performance of the air source heat exchanger directly affects the system's operating efficiency, energy consumption level, and safety. As a key energy-saving device, the evaporator and condenser of an air source heat pump operate in a high-temperature, high-humidity, and corrosive environment for extended periods, making them highly susceptible to refrigerant micro-leakage due to material fatigue or manufacturing defects. Initially, such leaks often manifest as slow changes in pressure gradient, delayed temperature response, vibration spectrum disturbances, and distortions in flow velocity and direction. Traditional fault diagnosis methods struggle to detect leaks in their early stages, and by the time the leak worsens and is discovered, it can cause significant losses.

[0003] like Figure 3 As shown, during the installation of large tubular or plate heat exchangers, due to the large overall size of the equipment, even if a minor leak occurs inside the heat exchanger, it is difficult for staff to accurately locate the leak and disassemble and repair it. This requires a significant amount of time for overall inspection and internal disassembly, greatly reducing maintenance efficiency.

[0004] Traditional pressure detection methods require shutdown and pressure holding for 1-2 hours, resulting in delayed response and inability to detect micro-leakage (leakage rate <0.1L / h). Mainstream threshold alarm methods only trigger analysis when parameters exceed limits, missing the optimal recovery window. Single pressure sensors are susceptible to environmental interference, and existing intelligent diagnostic models do not fully integrate key parameters such as condenser flow velocity and direction, and heat sink temperature field, limiting their ability to identify weak leak signals and ensuring ineffective positioning accuracy for pipe sections and pipe groups. The mapping deviation between leak point coordinates and actual physical location is large, lacking consideration of equipment structure and temperature field characteristics, making it difficult for maintenance personnel to quickly and intuitively determine the leak location. Leak prevention operations are disconnected from diagnostic results, failing to develop targeted intervention strategies based on leak location, flow velocity and direction, and temperature field distribution, resulting in limited leak prevention effectiveness. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a rapid diagnostic method for heat exchanger leakage faults, thus solving the problems in the background technology.

[0006] This invention provides the following technical solution: A rapid diagnostic method for heat exchanger leakage faults includes the following steps: S1: Basic parameter sensors are installed in the fluid channels of the heat exchanger, including the refrigerant side inlet and outlet, the water side inlet and outlet, and the heat exchange channel. An acoustic sensor array and vibration sensor are installed inside the heat exchanger shell, and environmental parameter sensors are installed outside the equipment to establish a monitoring network. Fluid parameters, acoustic signals, vibration parameters and environmental parameters are collected synchronously to form a standardized dataset with timestamps and sensor identifiers. S2: Process the collected data, extract time-series features, frequency features, and physical mechanism correlation features from the processed data, and construct a multi-dimensional feature system; S3: Real-time analysis of data sample distribution characteristics, calculation of the correlation index of sample deviation from normal data clusters, and triggering an early warning and activating the diagnostic process when the index of multiple consecutive monitoring windows is lower than the set threshold; inputting the extracted feature data into the diagnostic model for processing, outputting fault category, diagnostic confidence and key diagnostic basis, and verifying the diagnostic results with a dynamic threshold constructed in combination with operating condition parameters; S4: Construct a set of positioning equations using ultrasonic signal data, correct the ultrasonic propagation speed by adjusting the flow rate and direction of the refrigerant in the condenser tube, solve the set of equations, and obtain the coordinates of the leak point.

[0007] S5: Establish a coordinate mapping benchmark based on the layout of the heat exchanger fins and condenser tubes, convert the coordinates of the leak point into the actual physical location, and provide feedback to maintenance personnel in a multi-dimensional way so that maintenance personnel can intuitively judge the location of the leak.

[0008] Preferably, the basic parameter sensors mentioned in step S1 include a pressure sensor, a temperature sensor, a flow sensor, and a pressure transmitter; the acoustic sensor is an ultrasonic sensor array, which is arranged on the front, back, left, and right sides of the heat exchanger shell to form a fully enclosed monitoring system, and the vibration sensor is arranged at the four corners of the shell.

[0009] Preferably, in step S2, the data processing includes: a) standardizing the format of the collected raw data, converting it into a triplet structure of timestamp, sensor ID, and parameter value, and storing it in a standardized format; validating the data, removing data with missing timestamps or parameter values ​​exceeding the sensor's range; supplementing short-term interrupted data using cubic spline interpolation; and marking long-term invalid data and triggering alarms; b) dividing the standardized data into data blocks according to the time series, performing cluster analysis using a clustering algorithm to distinguish between normal data clusters, slightly abnormal clusters, and severely abnormal clusters; automatically locating cluster inflection points through curvature analysis to dynamically update clustering parameters to adapt to dynamic changes in operating conditions; and replacing abnormal data points with statistical characteristic values ​​of normal data clusters; c) subsequently performing noise reduction and normalization processing on the data.

[0010] Preferably, the flow velocity sensor is an electromagnetic flow velocity sensor, and the flow direction sensor is installed in the same direction as the condenser tube axis; the flow velocity sensor and the flow direction sensor are arranged at the inlet and outlet of the condenser tube, key bends and tube branches, and synchronously record the pipe section diameter, curvature and flow direction angle parameters at the installation position; the flow velocity and flow direction correction is combined with the pipe section structural parameters to correct the measurement error caused by the sensor installation deviation.

[0011] Preferably, in step S2, time-domain features, frequency-domain features, physical driving features, and flow velocity and direction correlation features are extracted from the processed data to generate feature data samples. Feature weights are dynamically adjusted through an attention mechanism to enhance the characterization of key fault features. The distribution characteristics of the data samples are analyzed. When multiple consecutive monitoring window indices are lower than a set threshold, the diagnostic process is triggered.

[0012] Preferably, the temperature sensor includes an infrared temperature sensor array and point temperature sensors. The infrared temperature sensors are arranged in a grid pattern on the surface of the heat sink, and the point temperature sensors are installed at intervals on the condenser tube sections. The data features extracted by the temperature sensor include the amplitude of local temperature anomalies on the heat sink, the abrupt change rate of the temperature gradient, and the axial temperature deviation of the tube section.

[0013] Preferably, in step S3, the warning level is divided according to the fault type and confidence level, and corresponding audible and visual alarms and information push warning actions are triggered for different warning levels. The pushed information includes the estimated leakage rate, location coordinates, and diagnostic basis.

[0014] Preferably, the procedure also includes the step of: after the leak point is determined, performing sealing protection, operation process control or leak classification operation based on the diagnostic results, flow rate and direction and temperature field data.

[0015] Preferably, in step S3, the fault categories include no leakage, micro-leakage (≥0.03L / h), moderate leakage (1.0-5L / h), and severe leakage (>5L / h); the method is applicable to plate heat exchangers and shell-and-tube heat exchangers in air source heat pump systems and is directly compatible with existing heat pump systems.

[0016] Preferably, in step S4, the positioning method corrects the ultrasonic propagation speed by adjusting the flow velocity and direction of the condenser tube. The corrected speed c' satisfies: c'=c×(1+v×cosθ / c)), where v is the fluid velocity in the tube, θ is the angle between the flow direction and the ultrasonic propagation direction, and c is the ultrasonic propagation speed under standard conditions of 340m / s.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a rapid diagnostic method for heat exchanger leaks. This technical solution integrates temperature difference characteristics and multiple physical field parameters to reduce location errors and accurately determine the location of leaks in all scenarios, including micro-leaks, seal damage, and other issues, thus resolving the problem of misjudgment of adjacent pipe sections. It also integrates acoustic, pressure, flow velocity, and temperature difference data for cross-validation, reducing the risk of misjudgment based on a single parameter anomaly and improving diagnostic stability under complex operating conditions. An infrared temperature sensor visualizes abnormal temperature difference areas, and combined with the physical feature mapping of the heat exchanger fins, the field terminal directly outputs data such as the core temperature difference area, physical location description, and audio-visual guidance, quickly locating the leak and shortening the leak location time. For annular temperature difference anomalies in tube sheet seal leaks and sudden temperature drops in damaged heat exchangers, it directly outputs specific information such as the seal segment number and heat exchanger grid coordinates, saving maintenance time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the leak location method of the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the principle and flow of the intuitive observation method of the present invention.

[0021] Figure 3 It is an accompanying diagram in the background art. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 a part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0024] Example: like Figure 1-2 As shown, a rapid diagnostic method for heat exchanger leakage faults includes the following steps: S1: Basic parameter sensors are installed in the fluid channels of the heat exchanger, including the refrigerant-side inlet and outlet, the water-side inlet and outlet, and the heat exchange channel. An acoustic sensor array and vibration sensors are installed inside the heat exchanger shell, and environmental parameter sensors are installed outside the equipment to establish a monitoring network. Fluid parameters, acoustic signals, vibration parameters, and environmental parameters are collected synchronously to form a standardized dataset with timestamps and sensor identifiers. Three pressure sensors, two temperature sensors, and one mass flow sensor are installed on the refrigerant side (inlet, outlet, shell side) and the water side (inlet, outlet, heat exchange channel), respectively. High-precision pressure transmitters (0.1%FS) are added to the gas and liquid pipes, and a mass flow meter is installed on the main refrigerant loop. Four sets of ultrasonic sensor arrays (six sensors per set, center frequency 500kHz, detection distance 0-50cm, sensor spacing 8cm) are installed on the front, back, left, and right sides of the heat exchanger shell to form a fully enclosed acoustic monitoring network. Triaxial MEMS accelerometers are installed at the four corners of the heat exchanger shell to monitor vibration parameters. Electromagnetic velocity and flow direction sensors are installed at the inlet and outlet of the heat exchanger condenser tubes, key bends, and tube branches. The sensors are installed in the same direction as the condenser tube axis. The structural parameters of the tube section, such as diameter and curvature, at the sensor installation locations are recorded simultaneously to provide a basis for data correction. All sensors use a 100Hz sampling frequency (accelerometer sampling frequency 5kHz), are connected to the edge computing unit via industrial Ethernet, and achieve nanosecond-level synchronization using the IEEE 1588v2 precision time protocol. A differentiable time alignment layer (DTA) compensates for transmission delay differences, ensuring data transmission latency ≤10ms and that flow velocity and flow direction data are consistent with other monitoring data in time and space.

[0025] S2: Process the collected data, extract time-series features, frequency features, and physical mechanism correlation features from the processed data, and construct a multi-dimensional feature system; S3: Real-time analysis of data sample distribution characteristics, calculation of the correlation index of sample deviation from normal data clusters, and triggering an early warning and activating the diagnostic process when the index of multiple consecutive monitoring windows is lower than the set threshold; inputting the extracted feature data into the diagnostic model for processing, outputting fault category, diagnostic confidence and key diagnostic basis, and verifying the diagnostic results with a dynamic threshold constructed in combination with operating condition parameters; S4: Construct a set of positioning equations using ultrasonic signal data, correct the ultrasonic propagation speed by adjusting the flow rate and direction of the refrigerant in the condenser tube, solve the set of equations, and obtain the coordinates of the leak point.

[0026] The specific method for locating the leak signal is as follows: taking the first group of sensor arrays as a reference, extract the initial peak times t1, t2, t3, and t4 when the leakage signal arrives at each group of sensor arrays, and calculate the time differences Δt12=t2-t1, Δt13=t3-t1, Δt14=t4-t1, and Δt23=t3-t2. Establish a set of nonlinear positioning equations: Based on the principle of spherical wave propagation, with the coordinates of each sensor array as the center of a sphere and the ultrasonic wave propagation distance (c × time difference) as the radius, construct the following set of nonlinear equations: |(x-x1) 2 +(y-y1) 2 +(z-z1) 2 -(x-x2) 2 +(y-y2) 2 +(z-z2) 2 |=c 2 ×Δt12 2 ; |(x-x1) 2 +(y-y1) 2 +(z-z1) 2 -(x-x3) 2 +(y-y3) 2 +(z-z3) 2 |=c 2 ×Δt13 2 ; |(x-x1) 2 +(y-y1) 2 +(z-z1) 2 -(x-x4) 2 +(y-y4) 2 +(z-z4) 2 |=c 2 ×Δt14 2 ; |(x-x2) 2 +(y-y2) 2 +(z-z2) 2 -(x-x3) 2 +(y-y3) 2 +(z-z3) 2 |=c 2 ×Δt23 2 ; Where (x, y, z) are the unknown coordinates of the leak point, and c is the calibrated ultrasonic wave propagation speed; The equations were solved using a particle swarm optimization algorithm (50 iterations, 100 particles), with a positioning error of ≤ ±1.5 cm.

[0027] S5: Establish a coordinate mapping benchmark based on the layout of the heat exchanger fins and condenser tubes, convert the coordinates of the leak point into the actual physical location, and provide feedback to maintenance personnel in a multi-dimensional way so that maintenance personnel can intuitively judge the location of the leak.

[0028] To enable maintenance personnel to more intuitively pinpoint leak locations, as minute leaks are difficult to detect within multi-layered heat exchangers, a more intuitive observation method has been established. The specific steps are as follows: First, using the tube sheet center and condenser tube arrangement baseline marked on the heat exchanger's factory drawings as the physical origin, simultaneously calibrate the digital coordinate system against the actual equipment structure. Use a laser rangefinder to measure key parameters such as tube spacing and tube length to correct physical deviations. Set at least three permanent calibration marks (e.g., laser-etched points) on the heat exchanger tube sheet and shell edge. Regularly (quarterly) remeasure the mark coordinates using a high-precision total station to update the coordinate system deviation compensation value, avoiding reference drift caused by equipment installation tilt or long-term operational deformation. After installing sensors such as flow rate, pressure, and ultrasonic sensors, record their actual coordinates using a laser locator and compare them with the preset installation positions to correct coordinate deviations and ensure the spatial position accuracy corresponding to the sensor data. Step 2: Establish a condenser tube group database in two dimensions (column and row). Accurately input the center Y-axis / X-axis coordinates, tube spacing, tube segment diameter, and tube extension length from the tube sheet for each column / row of condenser tubes. Differentiate between different tube segment types such as elbows, straight pipes, and branch pipes, and mark the coordinate interval boundaries of each tube segment. Mark the Z-axis coordinates of key nodes such as the tube sheet connection end, elbow inflection point, and support point of each condenser tube in the digital model, and clarify the division thresholds for the tube beginning, middle, and end. Step 3: Establish a three-dimensional physical coordinate system based on the heat exchanger's geometric reference point. Affix reference labels with coordinate scales to the shell and pre-store the equipment structural parameter library. Coordinate unification conversion: Convert the three-dimensional coordinates of the leak point to the actual coordinates on site, eliminating equipment installation deviations. Tube segment and tube group positioning: Determine the tube column by combining the Y-axis coordinate, judge the tube segment position by the Z-axis coordinate, and lock the interval by the X-axis coordinate. Feature location association: Mark the intuitive features around the leak point (such as the distance to the reference label, bolt holes, and the positional relationship of the support ribs).

[0029] The basic parameter sensors mentioned in step S1 include a pressure sensor, a temperature sensor, a flow sensor, and a pressure transmitter; the acoustic sensor is an ultrasonic sensor array, which is arranged on the front, back, left, and right sides of the heat exchanger shell to form a fully enclosed monitoring system, and the vibration sensor is arranged at the four corners of the shell.

[0030] In step S2, the data processing includes: a) unifying the format of the collected raw data, converting it into a triplet structure of timestamp, sensor ID, and parameter value, and storing it in a standardized format; validating the data, removing data with missing timestamps or parameter values ​​exceeding the sensor's range; supplementing short-term interrupted data using cubic spline interpolation; and marking long-term invalid data and triggering alarms; b) dividing the standardized data into data blocks according to the time series, performing cluster analysis using a clustering algorithm to distinguish between normal data clusters, slightly abnormal clusters, and severely abnormal clusters; automatically locating cluster inflection points through curvature analysis to dynamically update clustering parameters to adapt to dynamic changes in operating conditions; and replacing abnormal data points with statistical characteristic values ​​of normal data clusters; c) subsequently performing noise reduction and normalization processing on the data.

[0031] Specifically, multiple sources of raw data are simultaneously collected, including pressure (P), temperature (T), flow rate (Q), ultrasonic signal (U), vibration signal (A), environmental parameters (Ta, RH), flow velocity (v), and flow direction angle (θ). The data format is standardized and unified. The standardized pressure, flow rate, temperature, and flow velocity data are divided into data blocks according to time series continuity, determined based on the original sampling frequency, ensuring each block contains 100 consecutive data points. This process generates 100 independent data blocks. For each data block, a random sampling method is used to randomly select three non-repeating data points from the 100 data points as initial cluster centers C1, C2, and C3. The resolution of each data point within the data block is calculated to three... The Euclidean distance between the initial cluster centers is used to assign each data point to the cluster to which the nearest cluster center belongs, generating three initial clusters: Cluster_1 (corresponding to C1), Cluster_2 (corresponding to C2), and Cluster_3 (corresponding to C3). The mean of all data points within each of the three clusters is calculated, and these mean values ​​are used as the new cluster centers C1', C2', and C3'. The average Euclidean distance between the old and new cluster centers is calculated as: a = (∣C1′−C1∣ + ∣C2′−C2∣ + ∣C3′−C3∣) / 3. If a ≤ 0.001 or the number of iterations reaches 50, the process stops. If the above conditions are not met, the above steps are repeated with the new cluster centers until the data converges successfully. The variance of the data within each of the three converged clusters is calculated. Clusters with a variance ≤ 0.05 are considered normal data clusters, and clusters with a variance > 0.05 are considered abnormal clusters. The larger the variance, the higher the degree of abnormality. The mean μ of outlier data clusters is taken, and all points within the outlier clusters are replaced with the mean μ. After processing all data, the influence of outliers is eliminated to improve the accuracy of the output data. For vibration signals and flow direction data, filters are used to remove outliers. For the processed pressure, temperature, and flow data, a Kalman filter model (state transition matrix A=1, control matrix B=0, process noise variance 0.001, observation noise calibrated according to sensor accuracy) is constructed for iterative filtering, and the noise-reduced stationary time series is output. The ultrasonic signal is decomposed using a combination of Kalman filtering and wavelet thresholding. The wavelet basis function is db4, with 3 decomposition layers. The threshold is λ = σ × √(2lnN), where σ is the noise standard deviation and N is the data length. The vibration signal is decomposed using wavelet packet decomposition to remove low-frequency interference. The flow velocity and direction data are processed using an adaptive filtering algorithm to reduce noise interference caused by fluid disturbance. The Min-Max normalization method is used to map all preprocessed data to the [0,1] interval. The formula is: X_norm = (X - X_min) / (X_max - X_min), where X_min and X_max are the historical extreme values ​​of each parameter. The window length is set to T = 60 seconds and the step size is S = 10 seconds. The normalized data is divided into sliding windows, and each window generates the original input matrix for subsequent feature extraction.

[0032] The flow velocity sensor is an electromagnetic flow velocity sensor, and the flow direction sensor is installed in the same direction as the axis of the condenser tube. The flow velocity sensor and the flow direction sensor are arranged at the inlet and outlet of the condenser tube, key bends and tube branches, and synchronously record the pipe section diameter, curvature and flow direction angle parameters at the installation position. The flow velocity and flow direction correction is combined with the pipe section structural parameters to correct the measurement error caused by the sensor installation deviation.

[0033] In step S2, time-domain features, frequency-domain features, physical driving features, and flow velocity and direction correlation features are extracted from the processed data to generate feature data samples. Feature weights are dynamically adjusted through an attention mechanism to enhance the characterization of key fault features. The distribution characteristics of the data samples are analyzed. When multiple consecutive monitoring window indices are lower than the set threshold, the diagnostic process is triggered.

[0034] The core features extracted include: pressure change rate ΔP / Δt, peak pressure P_peak, pressure standard deviation P_std; flow deviation ΔQ, flow fluctuation coefficient Cv, cumulative flow deviation ΣΔQ; temperature gradient ΔT / ΔL, temperature difference ΔT_in-out; vibration signal peak value A_peak, vibration RMS value A_rms; ultrasonic signal peak value U_peak, ultrasonic RMS value U_rms, pulse width τ; Fast Fourier Transform (FFT) is performed on the ultrasonic and vibration signals to extract the characteristic frequency f0 corresponding to the peak power spectral density (PSD), PSD integral value (10-30kHz band), and frequency domain entropy H; high-frequency energy ratio of the vibration signal (HFEP, the proportion of energy in the 1000-5000Hz band to the total energy); phase change deviation PDI (calculated based on the REFPROP database lookup table); and cycle efficiency decay rate CEDR (characterizing the heat transfer rate per unit pressure ratio). The following characteristics are considered: (Descending trend), pressure-temperature coupling coefficient P-T_couple, flow-temperature synergy index Q-T_index, vibration-acoustic correlation feature A-U_correlate; Flow velocity characteristics: pipe segment velocity gradient (ratio of inlet / outlet velocity difference to pipe length), velocity fluctuation coefficient (ratio of velocity standard deviation to mean), velocity attenuation rate in leakage-sensitive areas (proportion of velocity difference between high-risk leakage areas and upstream normal areas); Flow direction characteristics: flow direction offset angle (angle between actual flow direction and pipe segment design axis), flow direction stability index (variance of flow direction angle change per unit time); Coupling characteristics: velocity-pressure coupling coefficient (product of velocity change rate and pressure change rate), flow direction-temperature synergy index (correlation between flow direction offset angle and pipe segment temperature gradient); Physical driving characteristics: fluid momentum loss rate based on velocity and flow direction (calculated by combining pipe segment cross-sectional area and fluid density), leakage-induced velocity distortion rate (degree of deviation between velocity distribution under leakage conditions and normal conditions).

[0035] After extracting effective features, calculate the Local Density Disturbance Index (LDDI): Define the proportion of each sample belonging to the main cluster (normal cluster) in its ε neighborhood as LDDI, and the formula is LDDI = (number of main cluster samples in the neighborhood / total number of samples in the neighborhood).

[0036] Anomaly warning determination: When LDDI < 0.65 for 3 consecutive windows, the system is determined to enter the "anomaly warning state" and the second-stage deep diagnosis process is activated; if LDDI ≥ threshold, monitoring continues.

[0037] By combining the features extracted above with a dual attention mechanism model and then classifying and filtering them, leakage fault classification is obtained. The specific process is as follows: Input layer: Receives a 100×24-dimensional feature matrix (time step 100, number of features 24).

[0038] Physical consistency embedding layer: Injects priors of thermodynamic conservation and fluid momentum conservation (based on flow velocity, flow direction and pipe segment structure parameters), integrates physical driving features, flow velocity and flow direction features with network hidden states, and improves the physical rationality of the model.

[0039] CNN feature extraction layer: 3 convolutional layers (convolutional kernel sizes 3×3, 5×5, and 7×7, with 32, 64, and 128 kernels respectively), using ReLU activation function, followed by a max pooling layer (pooling kernel 2×2) after each layer to extract spatial correlations of features.

[0040] Spatiotemporal Co-attention Layer: Includes temporal dimension attention and channel dimension attention, which performs weight allocation on the feature map output by CNN and the input sequence of LSTM, and automatically focuses on sensitive time segments and key sensor channels.

[0041] LSTM temporal modeling layer: 2 hidden layers (256 neurons each), dropout rate set to 0.2, to capture the dynamic changes of the feature time series.

[0042] Output layer: Fully connected layer (64 neurons) + Softmax activation function, outputting four types of diagnostic results: "no leakage", "micro-leakage (≥0.03L / h)", "moderate leakage (1.0-5L / h)" and "severe leakage (>5L / h)". It also outputs confidence scores and key evidence paths (Top-2 time windows + Top-2 sensor channels + Top-3 core features).

[0043] The temperature sensor includes an infrared temperature sensor array and point temperature sensors. The infrared temperature sensors are arranged in a grid pattern on the surface of the heat sink, and the point temperature sensors are installed at intervals on the condenser tube sections. The data features extracted by the temperature sensor include the amplitude of local temperature anomalies on the heat sink, the abrupt change rate of the temperature gradient, and the axial temperature deviation of the tube section.

[0044] In step S3, the dynamic thresholds under different operating conditions are as follows: Low flow rate condition: Dynamic pressure threshold P = P1 × (1 + k1 × T + k2 × Q + k3) × 0.9, Dynamic ultrasonic threshold U = U1 × (1 + k6 × f0) × 1.1 (k1 = 0.002 / ℃, k2 = 0.001 / kW, k3 = 0.05, k6 = 0.01 / kHz); P1 is the original pressure threshold, U1 is the original ultrasonic threshold, and f0 is the ultrasonic characteristic frequency. High flow rate condition: Dynamic pressure change rate threshold = original threshold × 1.2, Dynamic flow rate fluctuation coefficient threshold = original threshold × 0.8; Flow direction deviation condition: When the flow direction deviation angle > 10°, the flow direction correction factor is activated (the threshold is linearly adjusted according to the deviation angle, with a maximum correction range of ±15%). The above process, combined with ambient temperature, workload, flow rate level, and flow direction stability, generates dynamic thresholds to verify diagnostic results, reduce misjudgments, and improve accuracy.

[0045] In step S3, the warning level is divided according to the fault type and confidence level. Corresponding audible and visual alarms and information push warning actions are triggered for different warning levels. The pushed information includes the estimated leakage rate, location coordinates, and diagnostic basis.

[0046] It also includes the following steps: after the leak point is determined, based on the diagnostic results, flow rate and direction and temperature field data, to perform sealing protection, operation process control or leak classification operations.

[0047] In step S3, the fault categories include no leakage, micro-leakage (≥0.03L / h), moderate leakage (1.0-5L / h), and severe leakage (>5L / h); the method is applicable to plate heat exchangers and shell-and-tube heat exchangers in air source heat pump systems and is directly compatible with existing heat pump systems.

[0048] In step S4, the positioning method corrects the ultrasonic propagation speed by adjusting the flow velocity and direction of the condenser tube. The corrected speed c' satisfies: c'=c×(1+v×cosθ / c)), where v is the fluid velocity in the tube, θ is the angle between the flow direction and the ultrasonic propagation direction, and c is the ultrasonic propagation speed under standard conditions of 340m / s.

[0049] To further improve the accuracy of leak location, a quantitative method based on cumulative temperature difference and relative temperature difference changes is used to capture subtle temperature anomalies that are easily missed by traditional methods. Combined with the characteristic of slight flow velocity attenuation, this enables precise location of micro-leaks. The specific steps for differential-assisted leak location determination are as follows: I. Establishment of Temperature Difference Reference and Initial Temperature Calibration Initial temperature acquisition of heat exchanger fins: After the heat exchanger is put into operation for the first time or after maintenance is completed, it is run stably for 2 hours under rated conditions. The initial temperature T0 of each heat exchanger fin is collected by the infrared temperature sensor array on the surface of the heat sink in a 5cm×5cm grid. The data is recorded as the initial temperature mapping table of heat exchanger fin number and stored in the system database.

[0050] Temperature difference threshold calibration: Based on the heat exchanger material (such as aluminum, copper), heat exchange area and normal operation heat conduction law, calculate the normal temperature difference range (usually ≤2℃) of adjacent heat exchangers and different areas of the same heat exchanger, and set the leakage judgment temperature difference threshold ΔT1=3℃ (local abnormal temperature difference) and ΔT2=5℃ (temperature difference of the core leakage area).

[0051] Temperature field benchmark model construction: The initial temperature data is used to generate a continuous temperature field benchmark model of the heat exchanger plate, and the normal temperature distribution characteristics of different regions (tube sheet end, middle of tube section, and edge region) are clarified, which serves as a benchmark for subsequent leakage detection.

[0052] II. Real-time Temperature Difference Data Acquisition and Preprocessing Dynamic temperature difference monitoring: During the operation of the heat exchanger, the infrared temperature sensor array collects the surface temperature data T of the heat exchanger plate at a frequency of 10Hz in real time, and synchronously records the collection timestamp and the corresponding heat exchanger plate number to ensure that the data is synchronized with the leak location coordinates in time and space.

[0053] Temperature difference calculation: For each heat exchanger and grid area, calculate the absolute temperature difference between the real-time temperature and the initial temperature ΔT=|Treal-time-T0|, and at the same time calculate the relative temperature difference between adjacent heat exchangers (horizontal and vertical) ΔTnei=|Treal-time_iTreal-time_j|.

[0054] Temperature difference data noise reduction: Gaussian filtering algorithm is used to smooth the temperature difference data, eliminating instantaneous temperature difference changes (single change > 8℃ and duration < 1 second) caused by environmental interference and sensor error, while retaining the continuous temperature difference anomalies related to leakage.

[0055] III. Initial screening of leakage areas based on temperature difference Abnormal temperature difference area identification: Traverse all heat exchanger grid data, filter out areas that meet ΔT≥ΔT1 (3℃) and last for ≥30 seconds, and mark them as abnormal temperature difference areas; further filter out areas that ΔT≥ΔT2 (5℃), and mark them as high suspected leakage core areas.

[0056] Temperature gradient analysis: Calculate the temperature gradient (rate of change of temperature per unit distance) in the abnormal temperature difference area. The temperature gradient caused by leakage has a distribution characteristic of being high in the core area and gradually decreasing outward. Based on this, uniform temperature anomalies (such as changes in ambient temperature or overall decrease in heat exchange efficiency) can be ruled out.

[0057] Multi-parameter correlation initial screening: Combining flow velocity and direction data (flow velocity attenuation rate in the leakage area ≥5%) and pressure data (pressure mutation ≥0.05MPa), cross-validation is performed on areas with abnormal temperature differences, retaining the overlapping areas of "abnormal temperature difference + abnormal flow velocity / pressure", and narrowing down the candidate range of leakage locations.

[0058] IV. Precise Leak Location Identification Through the Integration of Temperature Difference and Coordinates Coordinate-Heat Exchanger Mapping Association: Map the initial location coordinates (x, y, z) of the leak point to the physical coordinate system of the heat exchanger, determine the target heat exchanger number and specific grid position corresponding to the coordinates, and extract the temperature difference data of the position and the surrounding 3×3 grid.

[0059] Temperature difference characteristics match the leak location: If the target location exhibits the characteristic of "ΔT≥ΔT2 (5℃) + highest temperature difference in the central region and gradient decay towards the periphery", then the location is determined to be the core point of the leak. If the target location ΔT ≥ ΔT1 (3℃) but does not reach ΔT2, and considering that the relative temperature difference between adjacent grids ΔT ≥ 2℃ and the flow direction data shows that the area is downstream of the fluid, the leak point is determined to be located 1-2cm upstream of the target location. If a continuous temperature difference abnormal zone appears in the edge area of ​​the heat exchanger plate, combined with the tube section coordinates (Z axis near the tube sheet end), it can be determined that the leak point is located in the sealing area where the tube sheet and the heat exchanger plate are connected.

[0060] Compare the spatial overlap between the abnormal temperature difference area and the peak area of ​​the ultrasonic signal. If the overlap rate is ≥80%, the leak location can be confirmed. Based on the coordinate range of the flow velocity and direction distortion area, if the deviation between the abnormal temperature difference area and the distortion area is ≤1cm, the original judgment is maintained; if the deviation is >1cm, the midpoint between the core temperature difference area and the center of flow velocity distortion is taken as the corrected leak location.

[0061] V. Temperature Drift Correction: The initial temperature T0 of the heat exchanger is re-collected every quarter to update the temperature field reference model and correct temperature drift errors caused by equipment aging and dust accumulation, ensuring the accuracy of temperature difference calculation.

[0062] Adaptive threshold adjustment based on operating conditions: The temperature difference threshold is dynamically adjusted according to the real-time operating load (20%-100%) and ambient temperature (-10℃-40℃): ΔT1 and ΔT2 are reduced by 10% under low load conditions and increased by 10% under high load conditions to avoid misjudgment caused by operating condition fluctuations.

[0063] VI. Micro-leakage scenario: The temperature difference caused by micro-leakage is relatively weak. Extend the data acquisition time window to 60 seconds and use the cumulative temperature difference integral (ΔT cumulative = ΣΔT × Δt) to assist in the judgment. If the cumulative temperature difference integral is ≥10℃・s and is accompanied by a slight decrease in local flow velocity, the leak location can be located.

[0064] Tube sheet seal leakage scenario: The temperature difference of this type of leakage is often "ring-shaped" (around the edge of the tube sheet). The key is to calculate the radial temperature difference gradient in the connection area between the tube sheet and the heat exchange fins. When the gradient value is >0.5℃ / cm, it is judged as a seal leakage. Combined with the tube section coordinates, the seal failure point can be accurately located.

[0065] Heat exchanger fin damage and leakage scenario: Heat exchanger fin damage will cause a sudden drop in local temperature, and the abnormal temperature difference area completely coincides with the location of the damage. High-resolution data from an infrared temperature sensor array can directly locate the coordinates of the damage point with an error ≤ ±0.5cm. Other technical solutions not described in detail in this invention are all existing technologies in the field and will not be elaborated here.

[0066] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can be modified and varied in various ways. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid diagnostic method for heat exchanger leakage faults, characterized in that, Includes the following steps: S1: Basic parameter sensors are installed in the fluid channels of the heat exchanger, including the refrigerant side inlet and outlet, the water side inlet and outlet, and the heat exchange channel. An acoustic sensor array and vibration sensor are installed inside the heat exchanger shell, and environmental parameter sensors are installed outside the equipment to establish a monitoring network. Fluid parameters, acoustic signals, vibration parameters and environmental parameters are collected synchronously to form a standardized dataset with timestamps and sensor identifiers. S2: Process the collected data, extract time-series features, frequency features, and physical mechanism correlation features from the processed data, and construct a multi-dimensional feature system; S3: Real-time analysis of data sample distribution characteristics, calculation of the correlation index of sample deviation from normal data clusters, and triggering an early warning and activating the diagnostic process when the index of multiple consecutive monitoring windows is lower than the set threshold; inputting the extracted feature data into the diagnostic model for processing, outputting fault category, diagnostic confidence and key diagnostic basis, and verifying the diagnostic results with a dynamic threshold constructed in combination with operating condition parameters; S4: Construct a set of positioning equations using ultrasonic signal data, correct the ultrasonic propagation speed by adjusting the flow rate and direction of the refrigerant in the condenser tube, solve the set of equations, and obtain the coordinates of the leak point; S5: Establish a coordinate mapping benchmark based on the layout of the heat exchanger fins and condenser tubes, convert the coordinates of the leak point into the actual physical location, and provide feedback to maintenance personnel in a multi-dimensional way so that maintenance personnel can intuitively judge the location of the leak.

2. The rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, The basic parameter sensors mentioned in step S1 include a pressure sensor, a temperature sensor, a flow sensor, and a pressure transmitter; the acoustic sensor is an ultrasonic sensor array, which is arranged on the front, back, left, and right sides of the heat exchanger shell to form a fully enclosed monitoring system, and the vibration sensor is arranged at the four corners of the shell.

3. The rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, In step S2, the data processing includes: a) unifying the format of the collected raw data, converting it into a triplet structure of timestamp, sensor ID, and parameter value, and storing it in a standardized format; validating the data, removing data with missing timestamps or parameter values ​​exceeding the sensor's range; supplementing short-term interrupted data using cubic spline interpolation; and marking long-term invalid data and triggering alarms; b) dividing the standardized data into data blocks according to the time series, performing cluster analysis using a clustering algorithm to distinguish between normal data clusters, slightly abnormal clusters, and severely abnormal clusters; automatically locating cluster inflection points through curvature analysis to dynamically update clustering parameters to adapt to dynamic changes in operating conditions; and replacing abnormal data points with statistical characteristic values ​​of normal data clusters; c) subsequently performing noise reduction and normalization processing on the data.

4. A rapid diagnosis method for heat exchanger leakage faults according to claim 2, characterized in that, The flow velocity sensor is an electromagnetic flow velocity sensor, and the flow direction sensor is installed in the same direction as the axis of the condenser tube. The flow velocity sensor and the flow direction sensor are arranged at the inlet and outlet of the condenser tube, key bends and tube branches, and synchronously record the pipe section diameter, curvature and flow direction angle parameters at the installation position. The flow velocity and flow direction correction is combined with the pipe section structural parameters to correct the measurement error caused by the sensor installation deviation.

5. A rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, In step S2, time-domain features, frequency-domain features, physical driving features, and flow velocity and direction correlation features are extracted from the processed data to generate feature data samples. Feature weights are dynamically adjusted through an attention mechanism to enhance the characterization of key fault features. The distribution characteristics of the data samples are analyzed. When multiple consecutive monitoring window indices are lower than the set threshold, the diagnostic process is triggered.

6. A rapid diagnosis method for heat exchanger leakage faults according to claim 2, characterized in that, The temperature sensor includes an infrared temperature sensor array and point temperature sensors. The infrared temperature sensors are arranged in a grid pattern on the surface of the heat sink, and the point temperature sensors are installed at intervals on the condenser tube sections. The data features extracted by the temperature sensor include the amplitude of local temperature anomalies on the heat sink, the abrupt change rate of the temperature gradient, and the axial temperature deviation of the tube section.

7. A rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, In step S3, the warning level is divided according to the fault type and confidence level. Corresponding audible and visual alarms and information push warning actions are triggered for different warning levels. The pushed information includes the estimated leakage rate, location coordinates, and diagnostic basis.

8. A rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, It also includes the following steps: after the leak point is determined, based on the diagnostic results, flow rate and direction and temperature field data, to perform sealing protection, operation process control or leak classification operations.

9. A rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, In step S3, the fault categories include no leakage, minor leakage, moderate leakage, and severe leakage; this method is applicable to plate heat exchangers and shell-and-tube heat exchangers in air source heat pump systems and is directly compatible with existing heat pump systems.

10. A rapid diagnosis method for heat exchanger leakage faults according to claim 1, characterized in that, In step S4, the positioning method corrects the ultrasonic propagation speed by adjusting the flow velocity and direction of the condenser tube. The corrected speed c' satisfies: c'=c×(1+v×cosθ / c)), where v is the fluid velocity in the tube, θ is the angle between the flow direction and the ultrasonic propagation direction, and c is the ultrasonic propagation speed under standard conditions of 340m / s.

Images