Method and system for diagnosing abnormal state of fan coil

Abstract

The invention discloses a fan coil abnormal state diagnosis method and system, and the method comprises the steps: obtaining a fan current waveform and a coil vibration signal, and building a heat exchange efficiency reference based on a motor load change characteristic and a heat exchange tube vibration mode; identifying an efficiency sudden drop area, positioning a heat exchange dead zone expansion range, generating heat transfer coefficient jump data, and determining a heat exchange abnormal position table; obtaining temperature difference of supply water and return water and opening change of a water valve, and forming waterway unbalance characteristics through coupling analysis of water flow pulse characteristics and flow state conversion points; identifying an airflow field offset based on the heat exchange abnormal position table and the waterway imbalance feature, determining a diagnosis focusing area and generating a layered detection sequence; carrying out active and passive damage classification, extracting a damage development rate, and generating a damage superposition coefficient; and a fault coupling amplification area is identified, damage synergistic effect data is generated through chain propagation analysis, comprehensive diagnosis of the abnormal state of the fan coil is completed, and accurate identification of multiple abnormalities and accurate positioning of fault root causes are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring and diagnosis of heating, ventilation and air-conditioning equipment, and in particular to a method and system for diagnosing abnormal conditions of fan coil units. Background Art

[0002] As a crucial component of air conditioning systems, fan coil units (FCUs) are directly impacted by their operational stability, impacting both indoor environmental quality and overall system performance. In practice, these units face the combined effects of multiple degradation factors during long-term operation.

[0003] Traditional equipment condition monitoring technologies primarily rely on threshold judgments based on single parameters such as temperature, pressure, and flow, lacking a deep understanding of equipment operating mechanisms. Existing diagnostic methods often fail to accurately identify fault type and severity when faced with multiple fault interactions, and even more so, struggle to establish correlations between faults. Furthermore, current technologies often rely on a post-event maintenance model, lacking the ability to predict equipment condition trends and enabling proactive maintenance management. This leads to high maintenance costs and prolonged downtime. Summary of the Invention

[0004] The present invention provides a method and system for diagnosing abnormal conditions of fan coil units. The method establishes an equipment operation benchmark through multi-source information fusion and deep correlation analysis, identifies the distribution of abnormal areas and analyzes the system imbalance pattern, determines key monitoring areas and classifies damage types, and constructs a fault propagation model to achieve accurate diagnosis and comprehensive evaluation of abnormal conditions of fan coil units.

[0005] A first aspect of the present invention provides a method for diagnosing abnormal conditions of a fan coil unit, comprising the following steps: Obtaining a fan current waveform and a coil vibration signal, extracting a motor load variation characteristic from the current waveform, extracting a heat exchange tube vibration pattern based on the coil vibration signal, and establishing a heat exchange efficiency benchmark based on the motor load variation characteristic and the heat exchange tube vibration pattern; identifying an efficiency sudden drop region based on the heat exchange efficiency benchmark, locating a heat exchange dead zone expansion range according to the efficiency sudden drop region, generating heat transfer coefficient jump data based on the heat exchange dead zone expansion range, and determining a heat exchange abnormality location table using the heat transfer coefficient jump data; Obtaining the supply and return water temperature difference and the change in water valve opening, identifying the water flow pulse characteristics based on the supply and return water temperature difference, capturing the flow pattern transition point based on the change in water valve opening, and coupling the water flow pulse characteristics with the flow pattern transition point to form a waterway imbalance characteristic; performing coupling correlation based on the heat exchange abnormality position table and the water channel imbalance characteristics to identify the airflow field offset, determining a diagnostic focus area based on the airflow field offset, and generating a hierarchical detection sequence using the diagnostic focus area; performing active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data, performing motion trajectory analysis on the active damage data to extract damage development rate, and generating a damage superposition coefficient based on the damage development rate and the passive damage data; Based on the damage superposition coefficient and the heat exchange abnormality position table, the fault coupling amplification area is identified, and a chain analysis is performed on the fault coupling amplification area to generate damage synergistic effect data, thereby completing the abnormal state diagnosis of the fan coil unit.

[0006] A second aspect of the present invention provides a fan coil unit abnormal state diagnosis system, comprising: a signal acquisition module configured to obtain a fan current waveform and a coil vibration signal, extract a motor load variation characteristic from the current waveform, extract a heat exchange tube vibration pattern based on the coil vibration signal, and establish a heat exchange efficiency benchmark based on the motor load variation characteristic and the heat exchange tube vibration pattern; an abnormality locating module, configured to identify an efficiency sudden drop area based on the heat exchange efficiency benchmark, locate an expansion range of a heat exchange dead zone according to the efficiency sudden drop area, generate heat transfer coefficient jump data based on the expansion range of the heat exchange dead zone, and determine a heat exchange abnormality location table using the heat transfer coefficient jump data; A waterway analysis module is configured to obtain a supply / return water temperature difference and a change in water valve opening, identify water flow pulse characteristics based on the supply / return water temperature difference, capture flow pattern transition points based on the change in water valve opening, and couple the water flow pulse characteristics with the flow pattern transition points for analysis to generate waterway imbalance characteristics; a coupled diagnosis module, configured to identify an airflow field offset by performing coupled correlation based on the heat exchange abnormality position table and the water channel imbalance characteristic, determine a diagnostic focus area based on the airflow field offset, and generate a hierarchical detection sequence using the diagnostic focus area; a damage analysis module, configured to perform active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data, perform motion trajectory analysis on the active damage data to extract a damage development rate, and generate a damage superposition coefficient based on the damage development rate and the passive damage data; The collaborative diagnosis module is used to identify the fault coupling amplification area based on the damage superposition coefficient and the heat exchange abnormality position table, perform chain analysis on the fault coupling amplification area to generate damage collaborative effect data, and complete the abnormal state diagnosis of the fan coil unit.

[0007] The beneficial effects of the present invention are reflected in the following aspects: First, by collecting fan current waveforms and coil vibration signals to establish a heat exchange efficiency benchmark, combined with efficiency drop zone identification and heat exchange dead zone expansion location, it can accurately detect problems such as localized fouling in the heat exchanger, fan blade contamination, and motor performance degradation. Compared with traditional methods that rely solely on temperature differential monitoring, this multi-signal fusion approach can detect equipment performance degradation trends in advance and accurately locate the specific faulty component and affected area, avoiding the continuous deterioration of equipment efficiency and increased energy consumption caused by missed detection. Second, by acquiring waterway imbalance characteristics based on the supply and return water temperature difference and water valve opening changes, combined with airflow field offset identification to determine the diagnostic focus area, it can detect complex problems such as uneven flow distribution, pipe blockage, and airflow short circuits in the waterway system. By analyzing the phase of water flow pulse characteristics and flow transition points, the specific location and severity of hydraulic imbalance can be identified. At the same time, by locating the center of the airflow field vortex, the detection range is narrowed, avoiding the workload of comprehensive disassembly and inspection, allowing maintenance personnel to quickly identify the problem area. Finally, through active and passive damage classification and damage development rate analysis, combined with chain propagation analysis of the fault coupling amplification region, it is possible to distinguish between active internal equipment faults and passive damage caused by external environmental influences, and predict fault propagation paths and development trends. By identifying propagation characteristics such as air volume attenuation gradient, condensate retention, and fin fouling rate, the probability and time point of a single fault spreading to other system components can be determined in advance, effectively preventing minor faults from evolving into systemic failures and reducing the risk of sudden equipment downtime.

[0008] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The accompanying drawings herein illustrate specific examples of the technical solutions described in the present invention, and together with the specific implementation methods constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0010] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0011] Figure 1 The present invention is a flowchart of a method for diagnosing abnormal conditions of a fan coil unit.

[0012] Figure 2 The present invention is a structural block diagram of a fan coil abnormal state diagnosis system. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0014] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0015] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0016] The technical solutions of the embodiments of this application are introduced below.

[0017] like Figure 1 As shown, an embodiment of the present invention provides a method for diagnosing abnormal conditions of a fan coil unit, comprising the following steps S110 to S160: Step S110, obtaining the fan current waveform and the coil vibration signal, extracting the motor load variation characteristics through the current waveform, extracting the heat exchange tube vibration mode based on the coil vibration signal, and establishing a heat exchange efficiency benchmark based on the motor load variation characteristics and the heat exchange tube vibration mode.

[0018] Specifically, the fan current waveform and coil vibration signal are acquired. A high-precision current sensor, CSLA2CD, with a measurement range of 0-50A, an accuracy level of 0.5, and a sampling frequency of 5000Hz, is installed at the fan motor input. This sensor collects three-phase current waveform data in real time, forming a complete fan current waveform. Current acquisition covers operating conditions such as the fan starting current, rated current, and variable frequency regulation current. The data recording format is [timestamp, phase identifier, instantaneous current value, effective value, power factor]. A PT100 temperature sensor is installed in the water inlet and outlet pipes, with a measurement range of 0-100°C, an accuracy of ±0.1°C, and a sampling frequency of 10Hz. A turbine flowmeter is installed to measure water flow, with a range of 0.5-10m³ / h, an accuracy level of 1.0, and a sampling frequency of 1Hz. Additional temperature sensors are deployed in key heat transfer areas of the heat exchanger to form a multi-point temperature monitoring network. ADXL354 accelerometers with a sensitivity of 100 mV / g, a frequency response range of 0.5-1600 Hz, and a sampling frequency of 2000 Hz were installed at the inlet and outlet water pipes, refrigerant pipes, and fin joints of the coil heat exchanger to acquire coil vibration signals. Vibration signal collection points included key locations such as U-bends, straight pipe sections, support brackets, and fin roots. X, Y, and Z axis vibration acceleration was recorded at each location. Sensor numbers were arranged from V01 to V08, and the data format was [timestamp, sensor number, X-axis acceleration, Y-axis acceleration, Z-axis acceleration, temperature compensation value].

[0019] Motor load variation characteristics are extracted from current waveforms. Three-phase current waveform data is used to calculate load characteristic parameters such as the motor's instantaneous power, average power, and power fluctuation amplitude. Instantaneous power is obtained by summing the three-phase powers: P_inst = U_A × I_A + U_B × I_B + U_C × I_C, where P_inst is the instantaneous power (W), U_A, U_B, and U_C are the three-phase voltages (V), and I_A, I_B, and I_C are the three-phase instantaneous currents (A). Average power is calculated using a sliding window with a window length of 1 second and a sliding step of 0.1 seconds. The load change rate is defined as the ratio of the power change to the time interval and reflects the dynamic response characteristics of the motor load. Current harmonic analysis uses fast Fourier transforms (FFTs) to extract characteristic harmonic components, such as the 5th, 7th, and 11th harmonics. Periods where the harmonic content exceeds 5% are marked as abnormal motor conditions. The load fluctuation coefficient is calculated as the ratio of the power standard deviation to the average value. A fluctuation coefficient of less than 5% indicates a stable load, 5%-15% indicates mild fluctuation, and greater than 15% indicates severe fluctuation. The starting surge characteristic is identified by detecting sudden current changes. The starting current is typically 4-7 times the rated current and lasts 2-5 seconds.

[0020] The vibration modes of the heat exchange tubes are extracted based on the coil vibration signals. The coil vibration signals are used to analyze the vibration frequency, amplitude, and other modal characteristics of the heat exchange tubes. Vibration signal preprocessing includes low-pass filtering to remove high-frequency noise, with a cutoff frequency set at 800 Hz and a 4th-order Butterworth filter. Power spectrum analysis is used to identify the dominant frequencies of heat exchange tube vibration. The dominant frequencies are typically distributed within the 10-200 Hz range, encompassing three frequency bands: fluid-induced vibration (10-50 Hz), structural natural frequency (80-150 Hz), and high-frequency vibration (150-200 Hz). Vibration modal analysis identifies the primary bending and torsional modes, distinguishing between normal and abnormal vibration modes. Vibration intensity is expressed as root mean square acceleration (RMS), reflecting the overall energy level of the vibration. Temperature correlation analysis establishes how vibration characteristics change with heat exchange temperature. As temperature increases, the vibration frequency typically shifts toward higher frequencies by 2-5 Hz. Vibration modes are classified into three types: steady-state vibration, intermittent vibration, and impact vibration, based on the temporal variation of the vibration amplitude.

[0021] In some embodiments, establishing a heat exchange efficiency benchmark based on the motor load change characteristics and the heat exchange tube vibration mode includes: generating a load fluctuation timing diagram based on the motor load change characteristics; fusing the heat exchange tube vibration mode and the load fluctuation timing diagram to form a vibration-load coupling curve; extracting a stable working segment from the vibration-load coupling curve; and setting the efficiency value of the stable working segment as the heat exchange efficiency benchmark.

[0022] A load fluctuation time series diagram is generated based on the motor load variation characteristics. The time series diagram is constructed using parameters such as average power, power fluctuation amplitude, and load change rate from the motor load variation characteristics. The horizontal axis of the time series diagram represents time, with a time resolution set to 0.1 seconds, and the vertical axis represents normalized load power, with the motor rated power as the normalized benchmark. The load fluctuation curve is smoothed and filtered with a filter window length of 5 seconds to eliminate short-term noise interference. The fluctuation envelope extracts the upper and lower envelopes of the load fluctuation, and the envelope width reflects load stability. Key event points are marked in the time series diagram, including the startup moment, stable operation phase, load sudden change point, and shutdown moment. The load gradient calculates the slope of the load change at adjacent time points, and regions where the gradient exceeds a threshold are marked as transition segments. Periodicity analysis identifies the periodic characteristics of the load fluctuation, and the period length is typically related to the refrigeration cycle. Data during periods of harmonic anomalies are removed from the time series diagram to ensure the reliability of the benchmark.

[0023] The vibration mode of the heat exchange tube and the load fluctuation time series are combined to form a vibration-load coupling curve. Based on the temporal correspondence between the vibration characteristics and the load characteristics, the vibration intensity and load power are compared and analyzed on the same time axis. The coupling curve uses a two-dimensional coordinate system, with the horizontal axis representing the normalized load power and the vertical axis representing the normalized vibration intensity. The density of data points reflects the frequency of occurrence of the operating condition, and areas with high density indicate common operating conditions. The coupling relationship analysis uses correlation calculation, and the correlation coefficient r indicates the degree of correlation between vibration and load. Linear fitting identifies the main trend line of vibration-load. The fitting equation is in the form of y=ax+b, where a is the coupling coefficient and b is the basic vibration level. Data points corresponding to abnormal vibration modes are marked as outliers and excluded from the coupling analysis. The coupling curve partitioning divides the entire load range into low-load, medium-load, and high-load zones, each with different vibration characteristics.

[0024] Stable operating segments are extracted from the vibration-load coupling curve. A stable operating segment is defined as the operating range in which both vibration and load remain relatively stable. Stability criteria include a load fluctuation coefficient less than 8%, a vibration intensity change less than 10%, and a coupling relationship deviation less than 15%. The stable segment identification algorithm uses a sliding window analysis with a window length set to 60 seconds. Data within the window that meets the stability criteria is marked as a candidate stable segment. A continuity check ensures that the duration of a stable segment is greater than 5 minutes; brief periods of stability are not considered valid stable segments. Stable segment quality is quantified using a stability index, which comprehensively considers factors such as fluctuation amplitude, duration, and consistency. Multi-stable segment processing identifies multiple stable segments within the operating cycle and selects the optimal stable segment based on stability ranking. Stable segment boundaries are determined through gradient change detection, with gradient mutation points serving as the start and end boundaries of the stable segment. Environmental factor correction is based on the results of temperature correlation analysis, establishing a correction for the impact of temperature on the stable segment.

[0025] The efficiency value during the stable operating period is set as the heat transfer efficiency benchmark. Based on the average load power and heat transfer data during the stable operating period, the heat transfer efficiency η = Q / P, where η is the heat transfer efficiency (dimensionless), Q is the heat transfer (W), and P is the motor power (W). The heat transfer is calculated using temperature sensor and flow meter data: Q = m × c × ΔT, where m is the water flow rate (kg / s), c is the specific heat capacity of water (J / kg·K), and ΔT is the inlet and outlet water temperature difference (K). The benchmark efficiency is the weighted average of the efficiencies during the stable operating period, with weights allocated based on operating time. The confidence interval of the efficiency benchmark is determined through statistical analysis, and the 95% confidence interval reflects the reliability range of the benchmark. Environmental standardization is based on temperature correlation analysis, correcting the benchmark efficiency to standard environmental conditions: an ambient temperature of 25°C and a relative humidity of 60%. The correction formula is η_corrected = η × [1 + α × (T - 25)], where α is the temperature correction factor. The benchmark validity is tested through repeatability testing. The benchmark efficiency is considered valid if the deviation is less than 3% over multiple consecutive days. The benchmark update mechanism regularly updates the efficiency benchmark based on accumulated operating data, with an update cycle set at 30 days. The multi-condition benchmark considers efficiency differences under different load conditions and establishes a segmented efficiency benchmark library. Ultimately, a heat exchange efficiency benchmark was established.

[0026] Step S120: identifying an efficiency drop region based on a heat exchange efficiency benchmark, locating a heat exchange dead zone expansion range based on the efficiency drop region, generating heat transfer coefficient jump data based on the heat exchange dead zone expansion range, and determining a heat exchange abnormality location table using the heat transfer coefficient jump data.

[0027] Specifically, efficiency drop regions are identified based on a heat exchange efficiency benchmark. The relative deviation method is used: the deviation rate = (η_baseline - η_real) / η_baseline × 100%, where η_baseline is the baseline efficiency (dimensionless) and η_real is the real-time efficiency (dimensionless). Regions with a deviation rate greater than 15% are marked as efficiency drop regions, those with a deviation rate between 5% and 15% are marked as efficiency degradation regions, and those with a deviation rate less than 5% are marked as normal efficiency regions. Spatial positioning is based on the geometric coordinate system of the coil heat exchanger, dividing the heat exchanger into a grid, with each grid corresponding to an efficiency monitoring point. Efficiency monitoring utilizes an installed temperature sensor monitoring network and turbine flowmeter. The algorithm for identifying drop regions uses connected domain analysis to merge adjacent drop grids into continuous regions. Temperature compensation data from the ADXL354 vibration sensor is combined to correct temperature measurement deviations at each monitoring point. Area statistics show that individual drop regions vary widely in area, with most often being elliptical or irregular polygonal in shape. Temporal persistence analysis requires that a drop phenomenon persist for a sufficient period of time to be considered a valid drop region.

[0028] In some embodiments, locating the expansion range of the heat exchange dead zone based on the efficiency drop area includes: dividing the efficiency drop area into a central area and an edge area; emitting a detection path from the central area to the edge area; recording the positions of efficiency recovery points on the detection path to form a recovery point set; and connecting the farthest points in the recovery point set to form the expansion range of the heat exchange dead zone.

[0029] The efficiency drop region is segmented into a central region and an edge region. Based on the geometry and deviation rate distribution of the efficiency drop region, a centroid segmentation method is used to determine the region center. The central region is defined as the area surrounding the maximum deviation rate, which is typically greater than 25%. The edge region is the remainder of the drop region excluding the central region, and the deviation rate in the edge region ranges from 15% to 25%. The segmentation algorithm uses a gradient descent method to find the point with the maximum deviation rate as the center point. The center point coordinates (x_center, y_center) correspond to the physical location of the heat exchanger. The radius of the central region is dynamically adjusted based on the total area of ​​the drop region. The edge region boundary is determined by the iso-deviation rate line, with the 15% deviation rate contour line serving as the outer boundary of the edge region. The segmentation result includes geometric parameters such as the central region coordinates, radius, and the sequence of edge region boundary points. The segmentation quality is assessed by the uniformity of the efficiency distribution within the region. Segmentation is considered valid if the standard deviation of the efficiency within the central region meets the set threshold.

[0030] Detection paths are launched from the central area to the peripheral areas. A radial layout is used, with straight paths launched from the center of gravity of the central area in all directions toward the peripheral areas. The angular spacing of the paths is determined based on the complexity of the area, forming a uniform distribution of detection rays. The length of each detection path extends an appropriate distance beyond the outer boundary of the peripheral area to ensure coverage of the possible expansion of the dead zone. Sampling points along the path utilize an existing network of temperature sensors, with the sampling point spacing consistent with the sensor layout. Detection paths are numbered and named in angular order. Path weights are assigned based on the efficiency gradient in that direction: paths with large gradients receive higher weights, while paths with small gradients receive lower weights. Multi-path parallel detection can identify the primary and secondary directions of dead zone expansion. Static monitoring of the detection process is performed using a fixed sensor network.

[0031] The positions of the efficiency recovery points on the detection path are recorded to form a recovery point set. The recovery point is defined as the position along the detection path where the efficiency begins to recover to above 85% of the baseline efficiency. A sliding window detection is used, and the window length is a number of sampling points. When the efficiency in the window rises continuously and the efficiency at the end point is greater than the threshold, it is marked as a recovery point. The threshold is set to 85% of the baseline efficiency to ensure that the recovery point represents a substantial improvement in heat exchange performance. There may be multiple recovery points on each detection path, and the recovery point farthest from the center area is selected as the representative recovery point of the path. The recovery point coordinates are expressed in the heat exchanger coordinate system, and the accuracy reaches the sensor layout accuracy. The recovery point detection adopts a multi-point comparison method. When the efficiency values ​​of multiple consecutive sampling points show an increasing trend and the last point exceeds the threshold, it is confirmed as a valid recovery point. At the same time, the efficiency recovery rate and recovery amplitude at the recovery point are recorded. The recovery point set contains representative recovery points of all paths, forming a recovery point ring around the dead zone.

[0032] The farthest points in the recovery point set are connected to form the expansion range of the heat exchange dead zone. The furthest point pair algorithm is used to identify the pair of points with the greatest mutual distance in the recovery point set, which serve as the endpoints of the principal axis of the expansion range. Multi-directional expansion, based on the principal axis, searches for the farthest points in the perpendicular and oblique directions to form the key nodes of the expansion boundary. A convex hull generation method is used based on key point constraints, taking all recovery points as input to ensure that the farthest key point is contained within the convex hull boundary. The convex hull calculation uses the Graham scan algorithm, starting with the farthest point pair as the starting edge and connecting the peripheral recovery points in polar angle order to generate a minimum bounding polygon enclosing all recovery points. The convex hull boundary is the expansion range boundary of the heat exchange dead zone, and the area within the boundary is the dead zone's impact area. Boundary smoothing uses spline curve fitting to eliminate jagged edges caused by sampling errors. The area of ​​the expansion range is calculated using the polygon area formula to reflect the impact of the dead zone. A boundary buffer is added with an appropriate width around the perimeter of the expansion range to account for boundary uncertainty. Expansion direction analysis identifies the primary direction of dead zone expansion, with expansion generally being more pronounced along the direction of fluid flow. The boundary node coordinates are stored in counter-clockwise order to form a closed polygon outline.

[0033] Heat transfer coefficient jump data is generated based on the dead zone expansion range. Using the boundary information of the dead zone expansion range, the variation of the heat transfer coefficient near the boundary is calculated. The heat transfer coefficient is calculated using the local convective heat transfer coefficient h = q / (ΔT), where h is the convective heat transfer coefficient (W / m²·K), q is the heat flux (W / m²), and ΔT is the temperature difference between the wall and the fluid (K). The heat flux q is calculated using the temperature gradient of adjacent temperature sensors and the known thermal conductivity of the material, while ΔT is directly measured using the temperature sensor. Jump detection uses the difference analysis of the heat transfer coefficients on both sides of the boundary. The jump amplitude is |h_inside - h_outside| / h_outside × 100%, where h_inside is the heat transfer coefficient within the dead zone and h_outside is the heat transfer coefficient in the normal area. Locations with jump amplitudes greater than 40% are marked as significant jump points, those between 20% and 40% are marked as moderate jump points, and those less than 20% are marked as minor jump points. The spatial resolution is set to the temperature sensor spacing, and the jump detection points are evenly distributed along the expansion range boundary. Time series analysis of the jump data tracks the evolution of the heat transfer coefficient jump and identifies the development trend of the jump.

[0034] In some embodiments, the use of the heat transfer coefficient jump data to determine the heat exchange anomaly position table includes: performing heat flow blockage identification on the heat transfer coefficient jump data to generate a streamline termination area; performing expansion impact assessment based on the streamline termination area to form an expansion coefficient; generating a continuous expansion field for the expansion coefficient through spatial interpolation; and performing boundary extraction based on the continuous expansion field to generate a heat exchange anomaly position table.

[0035] Heat flow blockage identification is performed on the heat transfer coefficient jump data to generate streamline termination areas. The blocking judgment standard is the area where the jump amplitude is greater than 50% and the jump direction of adjacent detection points is consistent. Streamline tracking uses the numerical flow field analysis method to track the propagation path of the heat streamline starting from the heat exchanger inlet. Streamline termination judgment is based on the sudden change of the heat transfer coefficient gradient. When the absolute value of the gradient is greater than the set threshold, the streamline is considered to terminate at this position. Termination area identification clusters and merges adjacent streamline termination points, and the cluster radius is set to an appropriate distance. The geometric characteristics of the streamline termination area include parameters such as center coordinates, influence radius, and termination strength. The termination strength is defined as the average value of the heat transfer coefficient jump amplitude in the area, reflecting the severity of the blocking degree. Multi-scale analysis identifies streamline termination areas of different scales. Large-scale termination areas correspond to major heat exchange obstacles, and small-scale termination areas correspond to local heat exchange defects.

[0036] The expansion effect of streamline terminations is evaluated to generate an expansion coefficient. The extent and range of each streamline termination's impact on the surrounding heat transfer performance are analyzed. The extent of impact is calculated using a distance decay function: the impact intensity I(r) = I0 × exp(-r / r0), where I(r) is the dimensionless impact intensity at distance r, I0 is the dimensionless maximum impact intensity of the termination zone, and r0 is the impact decay constant. The expansion coefficient is defined as the product of the termination zone's impact intensity and distance, reflecting the contribution of that region to the dead zone expansion. Weighting is determined based on the termination intensity and impact radius of the termination zone, with regions with higher termination intensity and larger impact radius receiving higher weights. The combined effect of multiple termination zones is calculated using the linear superposition principle, with the total expansion coefficient equal to the weighted sum of the expansion coefficients of each termination zone. Directional analysis accounts for the asymmetric effect of fluid flow direction on expansion; the downstream expansion coefficient is generally greater than the upstream expansion coefficient. Time evolution analysis tracks the temporal trend of the expansion coefficient and identifies phases of expansion acceleration or deceleration.

[0037] A continuous expansion field is generated by spatial interpolation of the expansion coefficient. The Kriging interpolation method is used to expand the discrete expansion coefficient into a continuous spatial field distribution. The interpolation grid resolution is appropriately encrypted based on the sensor layout grid to cover the entire heat exchanger area. The interpolation weight is determined based on the spatial distance and data correlation. Data points with close distance and high correlation have higher weights. The boundary condition is set to zero expansion coefficient at the heat exchanger boundary to ensure the physical rationality of the field distribution. The interpolation accuracy is evaluated by cross-validation. The interpolation is considered valid if the average relative error verified by the leave-one-out method is less than the set threshold. The gradient calculation of the continuous expansion field identifies the areas with the most drastic changes in expansion intensity. Areas with large gradients usually correspond to the boundaries of heat exchange anomalies. Field smoothing uses Gaussian filtering to eliminate high-frequency noise in the interpolation process. Multi-level analysis generates expansion fields of different resolutions. Coarse resolution is used for overall trend analysis, and fine resolution is used for local feature identification.

[0038] Boundary extraction is performed based on the continuous expansion field to generate a table of heat transfer anomaly locations. Boundaries are extracted based on contour analysis of the continuous expansion field, and contour lines with expansion coefficients equal to a threshold are selected as the anomaly region boundaries. The threshold is determined using an adaptive method, taking an appropriate proportion of the maximum value of the expansion field as the boundary threshold. The contour tracing algorithm uses the marching cubes method to generate smooth, closed boundary curves. Boundary simplification removes short boundary branches, retaining the main anomaly region boundaries. Anomaly location classification categorizes the extracted boundary regions into major, minor, and local anomalies based on area size. Location encoding uses a combination of the region center coordinates and the anomaly type in the format of "type code-X coordinate-Y coordinate". The anomaly location table contains fields such as location code, boundary coordinates, area, anomaly severity, and discovery time. The table is sorted based on anomaly severity and area size, with locations with high anomaly severity and large area receiving the highest priority. A dynamic update mechanism regularly updates the anomaly location table based on new heat transfer coefficient data, with an update cycle of 24 hours.

[0039] Step S130, obtain the supply and return water temperature difference and the change in water valve opening, identify the water flow pulse characteristics through the supply and return water temperature difference, capture the flow state transition point according to the change in water valve opening, and couple the water flow pulse characteristics with the flow state transition point to form the waterway imbalance characteristics.

[0040] Specifically, the supply and return water temperature difference and valve opening changes are acquired. Using a PT100 supply and return water temperature sensor with a measurement accuracy of ±0.1°C and a sampling frequency of 10Hz, the supply and return water temperature difference ΔT is calculated as T_supply - T_return, where T_supply is the supply water temperature (°C) and T_return is the return water temperature (°C). The temperature difference data is recorded in the format [timestamp, supply water temperature, return water temperature, temperature difference]. Valve opening is monitored using an angular displacement sensor with a measurement range of 0-90°, a resolution of 0.1°, and a linearity of ±0.5%. The opening rate of change is calculated as the ratio of the opening difference between adjacent moments to the time interval, reflecting the speed of valve adjustment. Opening data includes information such as valve number, opening angle, rate of change, adjustment direction, and actuation time. Multi-point synchronous data acquisition covers key valve positions, including the main supply valve, branch regulating valves, and balancing valves. A ring buffer is used for data storage, maintaining the last hour of historical data for analysis and processing.

[0041] Identify water flow pulse characteristics using the supply and return water temperature difference. Time series data of the supply and return water temperature difference is used to analyze the pulsation patterns and periodicity of the water flow. Pulse detection uses a peak recognition algorithm, identifying valid pulses as temperature fluctuations greater than 0.5°C and lasting 3-10 seconds. Pulse frequency measures the number of pulses occurring per unit time. Under normal operation, the pulse frequency is 2-5 times / minute, and under abnormal conditions, it can reach over 10 times / minute. Pulse amplitude is defined as the maximum temperature difference variation in a single pulse; its amplitude reflects the strength of the water flow disturbance. Pulse waveform analysis includes temporal characteristics of three stages: rise time, peak duration, and fall time. These waveform characteristics are used to verify the accuracy of pulse identification. Periodicity analysis uses the autocorrelation function to identify the pulse repetition period. The period length is typically related to the system's thermal inertia and flow regulation cycle. Pulse synchronization analyzes the temporal relationship between temperature difference pulses at multiple measuring points. Synchronous pulses indicate systemic disturbances, while asynchronous pulses indicate local disturbances. Synchronous information is used to classify pulse sequences. Filtering removes high-frequency noise and low-frequency drift, preserving valid pulse information.

[0042] Flow regime transition points are captured based on changes in valve opening. Based on valve position change data, this technology identifies moments when the flow regime in the water system significantly changes. Flow regime transition criteria include an opening change greater than 10°, a change rate greater than 5° / second, and coordinated operation of multiple valves. Transition point classifications include startup transition (system shutdown to operation), regulation transition (flow change due to load regulation), switching transition (operating mode switch), and fault transition (passive regulation due to an abnormal condition). Time accuracy reaches the second level, ensuring accurate location of transition points. Transition intensity is assessed by calculating the weighted sum of opening changes, with main valves receiving a higher weight and branch valves receiving a lower weight. Transition direction identification includes flow increase (increase in opening), flow decrease (decrease in opening), and mixed (partial increase in opening and partial decrease in opening). Transition direction information is used in the symbol design for transition moment markers. Durability analysis requires that the transition state be maintained for at least 30 seconds. Brief fluctuations in opening are not considered valid transitions. Durability information is used to screen transition point validity. Impact range analysis identifies the pipeline branches and heat exchange equipment affected by the transition point. Impact range information is used to calculate the weight of the transition intensity.

[0043] In some embodiments, the coupling analysis of the water flow pulse characteristics and the flow state transition point to form a waterway imbalance characteristic includes: constructing a pulse time axis based on the water flow pulse characteristics; mapping the flow state transition point to the pulse time axis to form a transition moment mark; dividing the pulse time axis into a pre-conversion segment and a post-conversion segment with the transition moment mark as the boundary; and comparing the phase distribution characteristics of the pre-conversion segment and the post-conversion segment to form a waterway imbalance characteristic.

[0044] A pulse timeline is constructed based on the water flow pulse characteristics. A standardized timeline coordinate system is constructed using parameters such as pulse time, pulse amplitude, duration, and periodicity from the water flow pulse characteristics. The timeline is zeroed at system startup, with a time resolution of 1 second and a time span covering the entire operating cycle. Pulse markers are annotated on the timeline with the pulse time as the horizontal axis and the pulse amplitude as the vertical axis. Pulse density is calculated based on the results of periodicity analysis to determine the number of pulses within a unit time window. The window length is set based on the identified period. Timeline normalization aligns pulse data from different measurement points to the same time base, eliminating sampling time differences. Pulse sequences are sorted by pulse time to form ordered timeline markers. Pulse intensity normalization normalizes pulse amplitudes to the range of 0-1, with the maximum pulse amplitude during the observation period as the normalization reference. Phase information is calculated as the phase angle of each pulse relative to the reference period: φ = 2π × (t_pulsemodT) / T, where φ is the phase angle (radians), t_pulse is the pulse time (seconds), and T is the reference period (seconds).

[0045] Flow transition points are mapped to the pulse timeline to form transition moment markers. Based on the transition moment, transition classification, and transition intensity of the flow transition point, transition marker symbols are added to the pulse timeline. The mapping algorithm uses a time matching method to directly correspond the time coordinates of the transition point to the corresponding position on the pulse timeline. Marker symbols use different colors and shapes to distinguish transition categories: a green triangle for start-up transitions, a blue circle for adjustment transitions, a yellow square for switching transitions, and a red diamond for fault transitions. The size of the marker reflects the transition intensity: larger marker symbols for greater transition intensity and smaller marker symbols for less intensity. Time alignment accuracy reaches 1 second, ensuring accurate correspondence between transition markers and pulse events. Overlap processing: When the transition moment coincides with the pulse moment, a composite marker is used to display both information. The marker sequence forms a chain of transition events in chronological order, which is used to analyze the temporal patterns of transitions.

[0046] The pulse timeline is segmented into pre-transition and post-transition segments using transition markers as demarcations. Each transition marker serves as a demarcation point, dividing the continuous pulse timeline into multiple time segments. The pre-transition segment is defined as the time interval from the previous transition point to the current transition point, and the post-transition segment is defined as the time interval from the current transition point to the next transition point. Segment length statistics display the average length of the pre-transition and post-transition segments. Head-end processing provides special treatment for the start and end points of the timeline: the segment before the start point is set to an empty segment, and the segment after the end point continues until the end of the observation period. Segment coding uses the format "S + transition point number + preceding and following identifiers," such as S01B for the segment before the first transition point, and S01A for the segment after the first transition point. Intra-segment pulse statistics calculate statistical parameters such as the number of pulses and average amplitude within each time segment. The segmentation results in a time segment list, each of which contains information such as the start and end times, included pulses, and statistical parameters.

[0047] Exemplarily, the comparison of the phase distribution characteristics of the front section and the back section of the conversion to form a waterway imbalance characteristic includes: converting the pulse sequence of the front section of the conversion into a phase sequence; staggering and superimposing the pulse sequence of the back section of the conversion onto the phase sequence to form a phase difference diagram; extracting the phase jump accumulation from the phase difference diagram; and forming the waterway imbalance characteristic based on the distribution unevenness of the phase jump accumulation.

[0048] The pulse sequence from the pre-conversion stage is converted into a phase sequence. The phase angle of each pulse is calculated using the pulse timing and period information contained in the pre-conversion stage. The phase sequence is arranged in chronological order, with the sequence length equal to the number of pulses in the pre-conversion stage. Phase calculation uses the formula φ = 2π × (t_pulsemodT) / T, where T is the pulse period. Phase continuity processing ensures the continuity of the phase sequence by expanding the phase angle when it crosses the 2π boundary. Sequence smoothing uses a moving average filter with a filter window length of 3 pulses to eliminate phase measurement errors of individual pulses. Phase trend analysis identifies the overall trend of the phase sequence, and the trend line is fitted using the least squares method. Sequence normalization adjusts the mean of the phase sequence to zero and the standard deviation to 1 to facilitate comparison with the post-conversion stage. Sequence interpolation uses linear interpolation to generate a phase sequence with equal intervals for uneven pulse spacing.

[0049] The post-conversion pulse sequence is offset and superimposed onto the phase sequence to form a phase difference plot. Using a sliding matching method, the post-conversion phase sequence is superimposed with the pre-conversion phase sequence at varying time offsets. The offset range is set to ±50% of the sequence length, with an offset step of one pulse interval. The phase difference is calculated as Δφ = φ_after - φ_before, where Δφ is the phase difference (radians), φ_after is the post-conversion phase angle (radians), and φ_before is the pre-conversion phase angle (radians). The phase difference plot has the offset as the horizontal axis and the phase difference as the vertical axis, with color indicating the magnitude of the phase difference. Superposition weights are assigned based on pulse amplitude, with pulses with larger amplitudes receiving higher weights and pulses with smaller amplitudes receiving lower weights. The optimal matching position is determined by minimizing the mean squared error of the phase difference. The offset position with the smallest error is the optimal superposition position. The significance of the difference is assessed by the standard deviation of the phase difference; a large standard deviation indicates a significant difference between the pre-conversion and post-conversion phases.

[0050] Phase jump accumulation is extracted from the phase difference map. A phase jump is defined as a sudden change in the phase difference between adjacent locations, with a jump threshold set to π / 6 radians. Phase jumps are detected using a gradient analysis method. The first-order derivative of the phase difference is calculated using a sliding window with a window length of three sampling points and a sliding step of one sampling point. The gradient calculation uses a central difference scheme to improve numerical accuracy, while also setting upper and lower gradient limits to prevent noise interference. Jump direction identification includes positive jumps (increase in phase difference) and negative jumps (decrease in phase difference). Direction determination is verified by verifying the sign consistency of the gradients at three consecutive points to ensure reliable jump identification. Jump amplitude measures the phase difference change at each jump point. The amplitude reflects the severity of the jump. The amplitude calculation uses the difference between the local maximum values ​​before and after the jump to avoid the influence of transient fluctuations. Cumulative value calculation uses an integral method: Cumulative value C = Σ|Δφ_jump|, where C is the cumulative phase jump (in radians), Δφ_jump is the phase difference change (in radians) at each jump point, and Σ represents the summation over all jump points. Spatial distribution analysis used a two-dimensional kernel density estimation method to identify clustering patterns of transition points. A Gaussian kernel function was used, and the bandwidth was optimized through cross-validation. Areas of high clustering indicated systematic variation. Temporal correlation analysis established a cross-correlation function between the transition occurrence time and the transition moment. The peak position of the correlation coefficient was used to determine the optimal time delay. A correlation coefficient greater than 0.7 indicated a strong correlation.

[0051] Waterway imbalance characteristics are determined based on the uneven distribution of phase jump cumulative values. The Gini coefficient method is used to quantify the degree of uneven distribution of phase jump cumulative values. The calculation formula is G = (2Σi × C_i) / (n × ΣC_i) - (n+1) / n, where G is the dimensionless Gini coefficient, C_i is the cumulative value (radians) of the i-th interval, n is the total number of intervals, and i is the interval number. The Gini coefficient ranges from 0 to 1, with higher values ​​indicating more uneven distribution and more severe waterway imbalance. Interval partitioning evenly divides the phase difference image into 20 intervals based on the offset, and the sum of the cumulative values ​​is calculated for each interval. The unevenness grading method categorizes the Gini coefficient into three levels: mild imbalance (G < 0.3), moderate imbalance (0.3 ≤ G < 0.6), and severe imbalance (G ≥ 0.6). Imbalance pattern recognition analyzes the shape characteristics of the cumulative distribution to identify typical patterns such as local imbalance, overall imbalance, and cyclical imbalance. Stability assessment compares imbalance characteristics over multiple consecutive time windows to assess the stability of the imbalance state. Imbalance direction analysis identifies the location and direction of major imbalances, providing guidance for regulatory measures.

[0052] Step S140 : coupling and correlating the heat exchange abnormality position table and the water channel imbalance characteristics to identify the airflow field offset, determining a diagnosis focus area based on the airflow field offset, and generating a hierarchical detection sequence using the diagnosis focus area.

[0053] Specifically, a coupling correlation analysis is performed based on the heat exchange anomaly location table and the waterway imbalance signature to identify airflow offsets. The coupling weight is determined based on the anomaly level in the heat exchange anomaly location table and the Gini coefficient in the waterway imbalance signature. The weight formula is W = α × A_level + β × G, where W is the coupling weight, A_level is the anomaly level, and G is the Gini coefficient. α and β are weight coefficients set to 0.6 and 0.4, respectively. The offset is calculated based on the difference between the airflow disturbance intensity and the normal state: V_offset = |V_disturbed - V_normal|, where V_offset is the airflow offset, V_disturbed is the wind speed in the disturbance state inferred from vibration sensor data changes and temperature fluctuations, and V_normal is the wind speed in the normal state. Spatial correlation analysis identifies the spatial correspondence between anomaly locations in the heat exchange anomaly location table and inferred airflow offsets. Significant airflow offsets are often observed around anomaly locations. Temporal correlation analysis compares the temporal consistency between the onset of the imbalance in the waterway imbalance signature and the change in the airflow offset. A time difference of less than 5 minutes is considered a strong correlation. The strength of the association is determined by a comprehensive evaluation of spatial distance and time difference. A close distance and a small time difference indicate a strong association. The impact range analysis determines the radius of influence of each anomaly location on the surrounding airflow field.

[0054] In some embodiments, determining the diagnostic focus area based on the airflow field offset includes: converting the airflow field offset into an offset vector field; finding a vortex center point in the offset vector field; performing regional growth using the vortex center point as a seed to form a preliminary focus area; and optimizing the boundaries of the preliminary focus area to form a diagnostic focus area.

[0055] The airflow field offset is converted into an offset vector field. A gridded vector field distribution is established within the monitoring area using the airflow field offset V_offset (scalar value) and the offset direction angle θ determined by combining the main vibration direction of the vibration sensor and the temperature gradient direction. The vector field grid resolution is set to 0.5m×0.5m, covering the entire monitoring area. The vector component calculation includes the X-direction component V_x=V_offset×cos(θ) and the Y-direction component V_y=V_offset×sin(θ), where θ is the offset direction angle. Spatial interpolation uses the inverse distance weighted interpolation method, with the maximum influence radius set to 3 meters. Measurement points beyond the range do not participate in the interpolation, and the discrete offset data is expanded into a continuous vector field distribution. A two-dimensional Gaussian filter is used for vector field smoothing, and the filter parameters are adaptively adjusted according to the grid density to eliminate local noise and discontinuity, ensuring the rationality of the field distribution. The boundary condition is set so that the vector value gradually becomes zero at the boundary of the monitoring area. Vector field quality assessment includes continuity check and physical rationality verification, divergence Reflects the convergence and divergence characteristics of airflow, curl Reflects the rotation characteristics of airflow.

[0056] Find the vortex center in the offset vector field. Based on the curl distribution characteristics of the offset vector field, the curl calculation uses the second-order accuracy format of the finite difference method, and the grid boundary is processed by one-sided difference. The curl value , where ω is the curl value (1 / s), Represents partial derivative operation. The candidate point of the vortex center is determined through multi-step screening. First, the curl threshold is screened, then the local extreme value test is performed, and finally the connectivity analysis is performed. The point whose absolute value of curl is greater than the threshold 0.5 / s and is a local extreme value is marked as a candidate point. Streamline tracing verifies the authenticity of the candidate point. The streamline integral adopts the fourth-order Runge-Kutta method. The integral step size is adaptively adjusted with a step size range of 0.01-0.1 meters. The streamline distribution is drawn based on the offset vector field. The streamline tracing starts from 8 directions around the candidate point at the same time. The tracing length is set to 2 times the influence radius of the candidate point. The streamlines around the real vortex center are closed rings or spirals. The center point accuracy optimization adopts the centroid iteration algorithm. The candidate point is used as the initial value, and the centroid is calculated in the local high curl area. The iterative convergence criterion is that the change in the centroid position is less than 0.05 meters for two consecutive times. Vortex intensity is assessed by calculating circulation, where Γ = ∮V·dl, where Γ is the circulation value (m² / s) and V is the vector in the offset vector field. The circulation integral uses the rectangular integration rule, with a circular path centered on the vortex center. The radius of the integral path is automatically determined based on the curl intensity distribution, with the distance where the curl decays to 50% of its maximum value being used as the integration radius. The vortex radius is determined using multiple criteria: curl decay, velocity decay, and streamline curvature. The weighted average of these three criteria is used as the final radius. The vortex radius is determined by the distance where the curl decays to 50% of its peak value. The radius reflects the vortex's impact range. Vortex classification uses a dual criterion based on curl sign and streamline direction. Clockwise vortices (negative curl) correspond to downdraft patterns, while counterclockwise vortices (positive curl) correspond to updraft patterns. Different vortex types correspond to different airflow anomalies.

[0057] Using the vortex center as a seed, region growing is performed to form a preliminary focal area. Using the coordinates of the vortex center as the starting point for growth, an eight-neighborhood expansion strategy is used to simultaneously expand the surrounding areas, forming a continuous focal region. Region growing employs a breadth-first search strategy to ensure a systematic expansion process. Region growing is based on vector similarity. Adjacent grid points are included in the growing region if their vector direction difference is less than 45° and their relative amplitude difference is less than 30%. Similarity is determined by triple validation of direction, amplitude, and continuity. The growing radius is limited to 1.5 times the vortex radius, controlled by Euclidean distance checks to prevent overexpansion. Multi-seed competition uses a distance-proportional allocation method. Overlapping regions are assigned based on their distance to each center. Using the distance-based decision principle, grid points are assigned to the nearest vortex center. Termination criteria for growth include reaching the radius limit, failing similarity, or encountering a boundary. Regional connectivity is checked using a connected component labeling algorithm to ensure internal spatial continuity. Isolated grid points are excluded from the focal region. Growth quality is assessed using the vector direction standard deviation and amplitude coefficient of variation. Regions with good uniformity exhibit superior focusing effects. The preliminary focus areas are numbered and categorized according to their importance.

[0058] The initial focus area is optimized to form a diagnostic focus region. Based on the geometric features of the initial focus area, boundary optimization is divided into three stages: preprocessing, shape regularization, and postprocessing. Irregular boundaries are corrected to regular shapes to facilitate subsequent diagnostic operations. Morphological closing operations are used for boundary smoothing. Structuring elements are adaptively selected based on boundary complexity to eliminate jagged protrusions and small depressions. Morphological processing methods are employed. The convex hull is calculated using the Graham scan algorithm. Minimum bounding polygons are generated in polar angle order. This minimum bounding polygon enclosing the initial focus area serves as the basis for regularization. Ellipse fitting uses the least squares method, taking into account the weight distribution of boundary points, to approximate the focus area shape and obtain a regular elliptical boundary. A 0.5-meter buffer zone is added to the fitted boundary to ensure complete diagnostic coverage. Area constraints are implemented through threshold checking and proportional scaling, with undersized areas appropriately expanded and oversized areas appropriately contracted. Shape regularity is assessed using a comprehensive scoring of geometric indicators such as circularity and compactness to ensure that the optimized area is suitable for detection. Overlapping is handled using a region merging algorithm, with the merging strategy determined based on the proportion of overlapping areas. Regions are merged when significant overlap occurs. The boundary accuracy reaches 0.1 meter, meeting the diagnostic positioning requirements.

[0059] A hierarchical inspection sequence is generated using diagnostic focus areas. A hierarchical inspection strategy is established based on the importance of the diagnostic focus areas and the vortex intensity characteristics. The inspection hierarchy consists of three levels: core inspection layer, key inspection layer, and routine inspection layer. The core inspection layer targets critical focus areas with vortex intensities greater than 2.0 m² / s and an area larger than 10 m², with an inspection density of 4 inspection points per square meter. The key inspection layer targets medium focus areas with vortex intensities between 1.0 and 2.0 m² / s, with an inspection density of 2 inspection points per square meter. The routine inspection layer covers general focus areas with vortex intensities less than 1.0 m² / s, with an inspection density of 1 inspection point per square meter. Inspection points are arranged in a hexagonal pattern to ensure uniform coverage. The inspection sequence is prioritized according to the focus areas, with critical areas prioritized. Inspection parameters include environmental parameters such as temperature, humidity, wind speed, and pressure, generating comprehensive diagnostic data. Time scheduling considers inspection efficiency, with the total inspection time limited to 2 hours. Path optimization is based on the spatial distribution of the focus areas and employs a traveling salesman problem algorithm to minimize the distance and time traveled by the inspection equipment. The detection accuracy requirements are ±0.1℃ for the core layer, ±0.2℃ for the key layer, and ±0.5℃ for the regular layer.

[0060] Step S150: Active and passive injuries are classified based on the hierarchical detection sequence to generate active injury data and passive injury data, motion trajectory analysis is performed on the active injury data to extract the injury development rate, and the injury superposition coefficient is generated based on the injury development rate and the passive injury data.

[0061] In some embodiments, the active and passive damage classification based on the layered detection sequence to generate active damage data and passive damage data includes: locating and identifying the damage source of the layered detection sequence to generate endogenous damage and exogenous damage; using the endogenous damage to evaluate the impact of the exogenous damage to form a damage association matrix; generating a dominant damage vector through eigenvalue decomposition of the damage association matrix; and implementing damage classification based on the dominant damage vector to generate active damage data and passive damage data.

[0062] Damage source location and identification are performed on a layered detection sequence, generating endogenous and exogenous damage. Based on detection parameters such as temperature, humidity, wind speed, and pressure in the layered detection sequence, anomaly propagation path analysis is used to track parameter gradient changes from the detection point with the highest anomaly intensity toward the surrounding area. Criteria for identifying endogenous damage include characteristics such as the anomaly source being located within the air conditioning system equipment, the anomaly parameter being directly related to the equipment's operating status, and the anomaly intensity varying with equipment load. Criteria for identifying exogenous damage include characteristics such as the anomaly source originating from the external environment, the anomaly parameter being related to changes in external conditions, and the anomaly intensity being unaffected by equipment adjustments. Damage intensity is quantified using the weighted average of parameter deviations: intensity I = Σ(wi × |Pi - Pref|) / Σ(wi), where I is the damage intensity, wi is the weight of the i-th parameter, Pi is the measured parameter value, and Pref is the reference value. Spatial clustering merges adjacent anomaly detection points into damage regions, based on the spatial distribution characteristics of the layered detection sequence. Temporal feature analysis analyzes the onset, duration, and development trends of damage based on the time series of the layered detection sequence. Endogenous damage usually manifests as intermittent and controllable characteristics, while exogenous damage manifests as continuous and uncontrollable characteristics.

[0063] The impact of endogenous damage on exogenous damage is evaluated to form a damage correlation matrix. Based on the location, intensity and time characteristics of endogenous damage, its impact on exogenous damage is evaluated. The spatial impact is calculated using the distance attenuation function, and the influence function is I_spatial=exp(-d / d_0), where I_spatial is the spatial impact, d is the distance between the two damage centers, and d_0 is the impact attenuation constant, which is 3 meters. The temporal impact is based on the temporal characteristic analysis results of endogenous and exogenous damage. Damages that occur simultaneously or with a time difference of less than 10 minutes have a high temporal impact. The intensity impact is calculated by the correlation between the intensity of endogenous damage and the intensity of exogenous damage. Damages with consistent intensity change trends have a high intensity impact. The comprehensive impact I_total = α × I_spatial + β × I_temporal + γ × I_intensity, where I_total is the comprehensive impact, I_temporal is the temporal impact, and I_intensity is the intensity impact. α, β, and γ are weight coefficients, taking them as 0.4, 0.3, and 0.3, respectively. The damage association matrix M is an n×m matrix, where n is the number of intrinsic damages and m is the number of exogenous damages. The matrix element M_ij represents the influence of the i-th intrinsic damage on the j-th exogenous damage.

[0064] The damage correlation matrix is ​​subjected to eigenvalue decomposition to generate the dominant damage vector. Based on the numerical distribution characteristics of the damage correlation matrix M, matrix preprocessing is first performed, including data centering and normalization, to eliminate dimensionality effects. Using the singular value decomposition (SVD) method, the damage correlation matrix M is decomposed into M=UΣV^T, where U is the left singular vector matrix, Σ is the singular value diagonal matrix, and V is the right singular vector matrix. The SVD decomposition employs a bidiagonalization algorithm, first converting the matrix to bidiagonal form via a Householder transformation, and then solving for the singular values ​​using a QR iteration. Numerical stability is ensured by checking the condition number. When the matrix condition number exceeds a set threshold, a truncated SVD method is used to improve numerical stability. The dominant damage vector corresponds to the singular vector with the largest singular value and reflects the dominant mode in the damaged system. The dominant vector is determined by analyzing the energy contribution ratio of the singular values, calculating the proportion of each singular value to the total energy. The size of the singular value reflects the importance of the corresponding mode. The cumulative contribution ratio criterion is used to determine the number of singular values ​​to be retained, typically retaining the first few singular values ​​with a cumulative contribution ratio of 90%. The elements of the dominant vector represent the contribution of each damage to the dominant mode. Larger values ​​indicate a significant role for that damage in the dominant mode. The sub-dominant vector corresponds to the second-largest singular value and reflects the secondary mode of the damaged system. Sub-dominant modes are used to verify the reliability and integrity of the dominant mode. The mode contribution rate is calculated using the variance explained ratio to assess the ability of each mode to explain the variation in the original data. This contribution rate calculation considers orthogonality constraints between modes.

[0065] Damage classification is performed based on the dominant damage vector to generate active and passive damage data. Damage classification is determined based on the numerical distribution and mode contribution rate of each element in the dominant damage vector. Active damage is determined when the dominant vector element value is greater than 0.3 and the corresponding damage exhibits controllable and responsive characteristics. Passive damage is determined when the absolute value of the dominant vector element value is less than 0.3 or the corresponding damage exhibits uncontrollable and lagging characteristics. The classification threshold of 0.3 is dynamically adjusted based on the mode contribution rate, with a stricter threshold applied to dominant modes with high contribution rates. Active damage data includes information such as damage location, damage type, control parameters, response time, and adjustable range. Passive damage data includes information such as damage location, damage source, impact level, duration, and adaptation strategy. Cross-validation verifies classification results using secondary dominant vectors to improve classification accuracy. Boundary damage processing uses a comprehensive assessment of damage vector elements with values ​​close to the threshold, combining temporal characteristics and spatial clustering results.

[0066] Active damage data is analyzed for motion trajectory analysis to extract damage growth rates. Based on the damage location, time information, and control parameter changes in the active damage data, a damage trajectory is constructed in space and time. Trajectory construction utilizes time series analysis, connecting the position coordinates of the same damage at different times to form a motion trajectory. Position accuracy reaches 0.1 meters, and time accuracy reaches the minute level, ensuring accurate trajectory analysis. Trajectory smoothing utilizes a Kalman filter to eliminate the impact of position measurement noise on the trajectory. Rate calculation utilizes the central difference method: rate v = (s_{i+1} - s_{i-1}) / (2Δt), where v is the damage growth rate, s_i is the position coordinate at the i-th moment, and Δt is the time interval. The rate direction, combined with analysis of control parameter changes in the active damage data, reflects the primary direction of damage growth. Trajectory pattern recognition includes typical patterns such as linear, radial, and spiral growth. The pattern type is used to calculate and modify the superposition coefficient. The expansion range is calculated using the trajectory envelope area, with the area serving as a spatial factor in the superposition coefficient calculation. Statistical analysis of the growth rate calculates statistical parameters such as the average rate, maximum rate, and rate variance.

[0067] A damage superposition coefficient is generated based on damage growth rate and passive damage data. The damage superposition effect is calculated using statistical parameters of the damage growth rate and information such as location, intensity, and duration from the passive damage data. Spatial overlap analyzes the degree of overlap between the active damage extension range and the passive damage location. The overlap area ratio, O, is calculated as A_overlap / A_total, where O represents the overlap, A_overlap represents the overlap area, and A_total represents the total affected area. The rate influence factor is based on the average damage growth rate: F = 1 + k × v_avg, where F represents the rate influence factor, k represents the rate coefficient, set to 0.1 min / m, and v_avg represents the average growth rate. Intensity superposition considers the influence of active damage growth and passive damage data. The superposed intensity, I_superposed, is calculated as √(I_active² + I_passive²), where I_superposed represents the superimposed intensity, I_active represents the active damage intensity, and I_passive represents the passive damage intensity. Temporal synchronization analysis uses the development time of active damage and the duration of passive damage data. Damages with a time difference of less than 5 minutes are superimposed. Trajectory pattern matching analyzes the degree of match between the active damage expansion pattern and the passive damage spatial distribution pattern. A higher degree of match indicates a stronger superposition effect. The superposition coefficient C = O × F × (I_superposed / I_max), where C is the damage superposition coefficient and I_max is the maximum damage intensity of the system. The coefficient grading system classifies the superposition coefficient into three levels: mild superposition (C < 0.3), moderate superposition (0.3 ≤ C < 0.7), and severe superposition (C ≥ 0.7).

[0068] Step S160 , identifying the fault coupling amplification area based on the damage superposition coefficient and the heat exchange abnormality position table, performing chain analysis on the fault coupling amplification area to generate damage synergy effect data, and completing the abnormal state diagnosis of the fan coil unit.

[0069] Specifically, fault coupling amplification regions are identified based on the damage superposition coefficient and the heat exchange anomaly location table. Parameters such as the superposition level, overlap, and rate impact factor in the damage superposition coefficient generated in S150 are combined with data such as the anomaly location coordinates, anomaly severity, and area in the heat exchange anomaly location table established in S120 to locate key areas where multiple faults couple and generate an amplification effect. Coupling criteria include a damage superposition coefficient greater than 0.5, an anomaly area greater than 8 square meters, and an anomaly severity exceeding moderate. Spatial coupling analysis identifies anomaly locations based on their spatial proximity; anomaly locations within a distance of less than 3 meters exhibit strong spatial coupling. Temporal coupling analysis examines the consistency of anomaly occurrence times; anomaly events with a time difference of less than 15 minutes exhibit temporal coupling. Amplification effects are quantified using the multiplier effect formula M=C×(1+A_level / A_max), where M is the amplification factor, C is the damage superposition coefficient, A_level is the anomaly severity level, and A_max is the maximum anomaly severity. Region boundaries are delineated by lines of equal amplification factors. Contiguous regions with an amplification factor greater than 1.5 are defined as fault coupling amplification regions. The impact intensity classification divides the amplification region into three levels: strong coupling (M>2.0), medium coupling (1.5≤M≤2.0), and weak coupling (1.2≤M<1.5). Regional characteristics include parameters such as center coordinates, boundary range, amplification factor, coupling type, and impacting equipment. The data structure for a fault coupling amplification region is [region number, center coordinates, boundary coordinates, amplification factor, coupling strength, and primary fault type].

[0070] In some embodiments, the chain analysis of the fault coupling amplification area to generate damage synergy effect data includes: determining an analysis path based on identifying chain propagation characteristics of the fault coupling amplification area, the chain propagation characteristics including air volume attenuation gradient, condensate retention degree and fin fouling rate; tracing the fault transmission process along the analysis path to form a propagation chain diagram; extracting the amplification coefficient of each chain link from the propagation chain diagram; and arranging the amplification coefficients in propagation order to generate damage synergy effect data.

[0071] The analysis path is determined by identifying chain propagation characteristics based on the fault coupling amplification region. The spatial distribution and coupling strength of the fault coupling amplification region are used to infer three chain propagation characteristics. The airflow attenuation gradient is inferred from changes in vibration sensor data. Areas with abnormal vibration intensity correspond to areas of severe airflow attenuation. The gradient value is estimated by the ratio of the vibration intensity change to the distance interval. Areas with a gradient greater than 0.51 / s are marked as areas of rapid airflow attenuation, 0.2-0.51 / s as areas of moderate attenuation, and less than 0.21 / s as areas of slow attenuation. The degree of condensate retention is inferred from localized temperature anomalies detected by the temperature sensor. Areas with uneven temperature distribution are often associated with condensate retention. The retention index is calculated based on the relative percentage of temperature deviation. A retention index greater than 30% indicates severe attenuation, 15%-30% indicates moderate attenuation, and less than 15% indicates mild attenuation. The fin fouling rate is inferred from the load variation trend of the current sensor. A continuous increase in motor load reflects increased fin resistance. The fouling rate is calculated by the ratio of the load increase to the time interval. A scaling rate greater than 50 Pa / day is considered rapid scaling, 20-50 Pa / day is considered moderate scaling, and less than 20 Pa / day is considered slow scaling. The analysis path is determined based on the impact intensity classification results, with the main propagation path being from the strong coupling zone to the weak coupling zone. Path weights are assigned based on the combined impact of the three propagation characteristics, with weight coefficients of 0.4 for air volume attenuation, 0.3 for condensate retention, and 0.3 for fin scaling.

[0072] The fault propagation process is traced along the analysis path to form a propagation chain diagram. Based on the spatial distribution and propagation direction of the analysis path, propagation nodes are set every 2 meters or at locations where the fault characteristics change significantly. Each node records the fault status and propagation parameters at its current location. The propagation intensity decay uses an exponential decay model: propagation intensity I(x) = I_0 × exp(-αx), where I(x) is the propagation intensity at distance x, I_0 is the initial propagation intensity (set based on the amplification factor M of the impact intensity grading), α is the attenuation coefficient, and x is the propagation distance. The inter-node connection relationship is established based on the causal relationship of fault propagation, with the fault status of upstream nodes influencing the fault development of downstream nodes. The propagation delay time considers the time required for the fault to propagate from one node to the next and is used to determine the diagnostic time window. The propagation chain diagram uses a directed graph structure, with nodes representing the fault status and arrows indicating the propagation direction and intensity. Branching processing: When the fault propagation forks, multiple parallel propagation chains are established. Converging processing: When multiple propagation chains converge at a node, the combined fault intensity at that node is calculated using the superposition effect.

[0073] The amplification factor of each link is extracted from the propagation chain diagram. Based on the node fault intensity distribution of the propagation chain diagram, the amplification factor is defined as the amplification factor of each link relative to the fault intensity of the upstream node. The direct amplification factor is calculated as the ratio of the fault intensity of the current node to the fault intensity of the upstream node. The cumulative amplification factor considers the total amplification effect from the starting node to the current node and is used to assess the overall amplification degree of fault propagation. The local amplification factor reflects the fault amplification capacity of the link itself, excluding the influence of propagation attenuation. The environmental correction factor is based on temperature sensor data and accounts for the impact of temperature changes on the amplification factor. The corrected amplification factor is obtained by multiplying the direct amplification factor by the environmental correction factor. Abnormal amplification detection identifies links with abnormally high amplification factors, which typically correspond to critical fault nodes or amplification mechanisms. Amplification factor distribution analysis statistically analyzes the distribution of amplification factors throughout the propagation chain and identifies the nodes that contribute most to the amplification effect. Link importance assessment is a comprehensive evaluation of amplification factors and node locations. Highly important links are used to identify key maintenance locations.

[0074] Amplification coefficients are arranged in propagation order to generate damage synergy effect data. Based on the amplification coefficients of each link and the spatiotemporal order of the propagation chain diagram, the timestamp of the fault arrival at each node is used, accounting for the impact of propagation delay. Spatial sorting is performed based on the distance from the fault source to the affected terminal. Synergy effect quantification is combined with path weights for a weighted calculation: the synergy coefficient S_synergy = ∏(wi × A_i), where S_synergy is the synergy coefficient, A_i is the amplification coefficient of the i-th link, and w_i is the corresponding path weight. Effect classification categorizes synergy effects into three types: enhancement (S_synergy > 1.5), maintenance (0.8 ≤ S_synergy ≤ 1.5), and attenuation (S_synergy < 0.8). Critical links are identified by detecting mutation points in the amplification coefficients. This information is used to provide key maintenance recommendations in diagnostic reports. Synergy pattern analysis identifies different synergy effect patterns, including linear, nonlinear, and threshold synergy. Effect intensity is assessed through a comprehensive evaluation of the synergy coefficient and impact range. Data integrity checking ensures that the synergy effect data covers the complete fault propagation process.

[0075] Based on in-depth analysis of damage synergy effect data, a comprehensive assessment of abnormal conditions and fault types in fan coil units is conducted. Using information such as the synergy coefficient, effect classification, and critical links in the damage synergy effect data, the diagnostic classification categorizes abnormal conditions into four levels: minor, moderate, severe, and critical. Minor abnormalities, corresponding to a synergy coefficient of 1.0-1.3, impact local performance but not overall operation. Moderate abnormalities, corresponding to a synergy coefficient of 1.3-1.8, require immediate maintenance but do not temporarily impact normal operation. Severe abnormalities, corresponding to a synergy coefficient of 1.8-2.5, impact system operation and require immediate attention. Critical abnormalities, corresponding to a synergy coefficient greater than 2.5, pose a risk of system failure and require emergency shutdown and maintenance. Fault location is based on critical link identification results, accurately pinpointing specific equipment components, including fan blades, motor bearings, heat exchanger fins, condensate trays, and control valves. Root cause analysis identifies the initial cause and triggering mechanism of the fault by tracing back through the propagation chain diagram. Maintenance recommendations provide specific maintenance measures and schedules based on the link importance assessment results and the abnormality level. Through accurate abnormal state diagnosis, it is possible to accurately identify the key factors affecting the energy efficiency of the system, avoid the blindness and excessive maintenance of traditional empirical maintenance, and achieve targeted problem solving. The diagnostic system can promptly detect hidden energy consumption factors such as decreased heat exchange efficiency, water channel imbalance, and uneven airflow distribution, and formulate targeted energy-saving plans such as deep cleaning of heat exchangers, rebalancing of water channels, and precise adjustment of air volume, effectively reducing system operating energy consumption and extending the service life of equipment. The prognostic assessment combines the cumulative amplification factor and propagation delay time to predict the remaining service life and performance degradation of the system under the current fault development trend. Report generation automatically generates a diagnostic report containing abnormal status, fault location, cause analysis, maintenance recommendations, etc. Through systematic analysis of damage synergy effects and multi-dimensional comprehensive evaluation, accurate and reliable fan coil abnormal state diagnosis was finally completed.

[0076] In order to implement a fan coil abnormal state diagnosis method corresponding to the above method embodiment, to achieve the corresponding functions and technical effects. Figure 2 , Figure 2 The following is a block diagram of a fan coil abnormality diagnosis system 200 provided in an embodiment of the present application. For ease of explanation, only the parts related to this embodiment are shown. The fan coil abnormality diagnosis system 200 provided in an embodiment of the present application includes: Signal acquisition module 201, configured to obtain a fan current waveform and a coil vibration signal, extract a motor load variation characteristic from the current waveform, extract a heat exchange tube vibration pattern based on the coil vibration signal, and establish a heat exchange efficiency benchmark based on the motor load variation characteristic and the heat exchange tube vibration pattern; an abnormality locating module 202 for identifying an efficiency sudden drop region based on the heat exchange efficiency benchmark, locating a heat exchange dead zone expansion range based on the efficiency sudden drop region, generating heat transfer coefficient jump data based on the heat exchange dead zone expansion range, and determining a heat exchange abnormality location table using the heat transfer coefficient jump data; The waterway analysis module 203 is configured to obtain the supply and return water temperature difference and the change in water valve opening, identify water flow pulse characteristics based on the supply and return water temperature difference, capture flow pattern transition points based on the change in water valve opening, and couple the water flow pulse characteristics with the flow pattern transition points to generate waterway imbalance characteristics. A coupled diagnosis module 204 is configured to perform coupled correlation to identify an airflow field offset based on the heat exchange anomaly location table and the waterway imbalance characteristics, determine a diagnostic focus area based on the airflow field offset, and generate a hierarchical detection sequence using the diagnostic focus area; a damage analysis module 205 configured to perform active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data, perform motion trajectory analysis on the active damage data to extract damage development rate, and generate a damage superposition coefficient based on the damage development rate and the passive damage data; The collaborative diagnosis module 206 is used to identify the fault coupling amplification area based on the damage superposition coefficient and the heat exchange abnormality position table, perform chain analysis on the fault coupling amplification area to generate damage collaborative effect data, and complete the abnormal state diagnosis of the fan coil unit.

[0077] The fan coil unit abnormality diagnosis system 200 described above can implement a fan coil unit abnormality diagnosis method according to the aforementioned method embodiment. The optional options in the aforementioned method embodiment also apply to this embodiment and are not further described here. The remaining contents of the present application embodiment can be referenced to the contents of the aforementioned method embodiment and are not further described in this embodiment.

[0078] The purpose of the above embodiments is to exemplify and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosed content of the present invention, and it does not limit the scope of protection of the present invention.

[0079] The above embodiments are not exhaustive and may include many other embodiments not listed above. Any replacements and improvements made without violating the concept of the present invention are within the scope of protection of the present invention.

Claims

1. A method for diagnosing abnormal conditions of a fan coil unit, characterized in that: include: Obtaining a fan current waveform and a coil vibration signal, extracting a motor load variation characteristic from the current waveform, extracting a heat exchange tube vibration pattern based on the coil vibration signal, and establishing a heat exchange efficiency benchmark based on the motor load variation characteristic and the heat exchange tube vibration pattern; identifying an efficiency sudden drop region based on the heat exchange efficiency benchmark, locating a heat exchange dead zone expansion range according to the efficiency sudden drop region, generating heat transfer coefficient jump data based on the heat exchange dead zone expansion range, and determining a heat exchange abnormality location table using the heat transfer coefficient jump data; Obtaining the supply and return water temperature difference and the change in water valve opening, identifying the water flow pulse characteristics based on the supply and return water temperature difference, capturing the flow pattern transition point based on the change in water valve opening, and coupling the water flow pulse characteristics with the flow pattern transition point to form a waterway imbalance characteristic; performing coupling correlation based on the heat exchange abnormality position table and the water channel imbalance characteristics to identify the airflow field offset, determining a diagnostic focus area based on the airflow field offset, and generating a hierarchical detection sequence using the diagnostic focus area; performing active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data, performing motion trajectory analysis on the active damage data to extract damage development rate, and generating a damage superposition coefficient based on the damage development rate and the passive damage data; Based on the damage superposition coefficient and the heat exchange abnormality position table, the fault coupling amplification area is identified, and a chain analysis is performed on the fault coupling amplification area to generate damage synergistic effect data, thereby completing the abnormal state diagnosis of the fan coil unit.

2. The method according to claim 1, characterized in that The establishing of a heat exchange efficiency benchmark according to the motor load variation characteristics and the heat exchange tube vibration mode includes: generating a load fluctuation timing diagram based on the motor load variation characteristics; Combining the heat exchange tube vibration mode and the load fluctuation time sequence diagram to form a vibration-load coupling curve; extracting a stable working segment from the vibration-load coupling curve; The efficiency value of the stable working section is set as the heat exchange efficiency benchmark.

3. The method according to claim 1, characterized in that The step of locating the expansion range of the heat exchange dead zone according to the efficiency sudden drop area includes: dividing the efficiency drop region into a central region and a peripheral region; transmitting a detection path from the central region to the edge region; Recording the locations of efficiency recovery points on the detection path to form a recovery point set; The farthest points in the recovery point set are connected to form a heat exchange dead zone expansion range.

4. The method according to claim 1, wherein The method of determining a heat exchange abnormality position table by using the heat transfer coefficient jump data includes: Performing heat flow blocking identification on the heat transfer coefficient jump data to generate a streamline termination area; Performing an expansion impact assessment based on the streamline termination area to form an expansion coefficient; generating a continuous expansion field by spatial interpolation of the expansion coefficients; Boundary extraction is performed based on the continuous expansion field to generate a heat exchange abnormality position table.

5. The method according to claim 1, wherein The coupling analysis of the water flow pulse characteristics and the flow state transition point to form the waterway imbalance characteristics includes: Constructing a pulse time axis based on the water flow pulse characteristics; Mapping the flow state transition point to the pulse time axis to form a transition moment mark; Dividing the pulse time axis into a pre-conversion segment and a post-conversion segment with the conversion moment mark as a boundary; The phase distribution characteristics of the front section and the back section of the conversion are compared to form a waterway imbalance characteristic.

6. The method according to claim 1, wherein The determining of the diagnosis focus area based on the airflow field offset includes: Converting the airflow field offset into an offset vector field; Finding a vortex center point in the offset vector field; Taking the vortex center point as a seed, regional growth is performed to form a preliminary focusing area; Boundary optimization is performed on the preliminary focusing area to form a diagnostic focusing region.

7. The method according to claim 1, characterized in that The active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data includes: Performing damage source location identification on the layered detection sequence to generate endogenous damage and exogenous damage; Using the endogenous damage to evaluate the impact of the exogenous damage to form a damage correlation matrix; generating a dominant damage vector by performing eigenvalue decomposition on the damage correlation matrix; Damage classification is performed according to the dominant damage vector to generate active damage data and passive damage data.

8. The method according to claim 1, characterized in that The performing chain analysis on the fault coupling amplification region to generate damage synergy effect data includes: Determining an analysis path based on identifying chain propagation characteristics of the fault coupling amplification region, wherein the chain propagation characteristics include air volume attenuation gradient, condensate retention degree, and fin fouling rate; Tracing the fault transmission process along the analysis path to form a propagation chain diagram; Extracting the amplification factor of each chain link from the propagation chain diagram; The amplification factors are arranged in propagation order to generate damage synergy effect data.

9. The method according to claim 5, characterized in that The forming of a waterway imbalance feature by comparing the phase distribution characteristics of the pre-conversion section and the post-conversion section includes: Converting the pulse sequence of the front-end conversion section into a phase sequence; Staggering and superimposing the pulse sequence of the post-conversion segment onto the phase sequence to form a phase difference map; Extracting the phase jump cumulative amount from the phase difference map; A waterway imbalance feature is formed based on the uneven distribution of the phase jump accumulation amount.

10. A fan coil unit abnormal state diagnosis system, characterized in that: include: a signal acquisition module configured to obtain a fan current waveform and a coil vibration signal, extract a motor load variation characteristic from the current waveform, extract a heat exchange tube vibration pattern based on the coil vibration signal, and establish a heat exchange efficiency benchmark based on the motor load variation characteristic and the heat exchange tube vibration pattern; an abnormality locating module, configured to identify an efficiency sudden drop area based on the heat exchange efficiency benchmark, locate an expansion range of a heat exchange dead zone according to the efficiency sudden drop area, generate heat transfer coefficient jump data based on the expansion range of the heat exchange dead zone, and determine a heat exchange abnormality location table using the heat transfer coefficient jump data; A waterway analysis module is configured to obtain a supply / return water temperature difference and a change in water valve opening, identify water flow pulse characteristics based on the supply / return water temperature difference, capture flow pattern transition points based on the change in water valve opening, and couple the water flow pulse characteristics with the flow pattern transition points for analysis to generate waterway imbalance characteristics; a coupled diagnosis module, configured to identify an airflow field offset by performing coupled correlation based on the heat exchange abnormality position table and the water channel imbalance characteristic, determine a diagnostic focus area based on the airflow field offset, and generate a hierarchical detection sequence using the diagnostic focus area; a damage analysis module, configured to perform active and passive damage classification based on the hierarchical detection sequence to generate active damage data and passive damage data, perform motion trajectory analysis on the active damage data to extract a damage development rate, and generate a damage superposition coefficient based on the damage development rate and the passive damage data; The collaborative diagnosis module is used to identify the fault coupling amplification area based on the damage superposition coefficient and the heat exchange abnormality position table, perform chain analysis on the fault coupling amplification area to generate damage collaborative effect data, and complete the abnormal state diagnosis of the fan coil unit.

Images