A method and system for fault prediction and health management of terminal equipment in a ventilation system.

Abstract

This invention provides a method and system for fault prediction and health management of terminal equipment in a ventilation system, relating to the field of condition monitoring technology. The method includes: Step 1, using a standardized operational dataset as the processing object, extracting key time-frequency features characterizing the mechanical operation of the fan and the airflow state in the duct; and constructing a three-dimensional spatial mapping ellipsoid based on the spatial distribution coordinates of three key monitoring nodes and the aeromechanical impedance characteristics of each node; Step 2, performing adaptive mesh partitioning and flow-vibration energy transfer numerical simulation on the three-dimensional spatial mapping ellipsoid along the principal axis to obtain spatial phase coupling correction values. This invention achieves accurate perception of the multi-physics coupling state of equipment, quantitative early warning of early hidden faults, and closed-loop management of predictive maintenance.

Description

Technical Field

[0001] This invention relates to the field of condition monitoring technology, and in particular to a method and system for fault prediction and health management of terminal equipment in a ventilation system. Background Technology

[0002] Taking the centralized ventilation system of the underground parking garage and supporting equipment room of a large commercial complex as an example, the centrifugal fans, ducted supply and exhaust fans, fireproof smoke exhaust valves and static pressure boxes, etc., which are densely deployed at the end, are in a humid and dusty underground environment and operate continuously under high load for 24 hours a day to ensure daily air replacement and emergency fire smoke exhaust needs. The existing operation and maintenance mode relies on fixed-period manual visual inspection and simple operating current measurement to judge the equipment status. It is difficult to collect and integrate multi-dimensional dynamic characteristics such as vibration spectrum, bearing temperature rise, air outlet air pressure and motor speed in real time. It lacks the ability to continuously and quantitatively monitor the trend of equipment performance degradation and the ability to perform flow-machine multi-physics field coupling analysis. As a result, hidden faults such as early bearing micro-wear, impeller dust adhesion imbalance and valve lubrication jamming are difficult to accurately identify and warn in advance. This can easily lead to the equipment rapidly evolving from slight degradation to sudden complete failure, causing a sharp drop in ventilation efficiency and excessive air quality in the underground space. In addition, it poses a major safety hazard that the smoke exhaust equipment cannot be started in fire emergency conditions. At the same time, passive post-event maintenance increases the system's unplanned downtime losses and additional operation and maintenance costs. Summary of the Invention

[0003] This invention provides a method and system for fault prediction and health management of terminal equipment in a ventilation system, enabling accurate perception of the multi-physical field coupling state of the equipment, quantitative early warning of early hidden faults, and closed-loop management of predictive maintenance.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0005] Firstly, a method for fault prediction and health management of terminal equipment in a ventilation system, the method comprising:

[0006] Step 1: Using the standardized operational dataset as the processing object, extract key time-frequency features that characterize the mechanical operation of the fan and the airflow state in the pipeline; based on the spatial distribution coordinates of the three key monitoring nodes and the aeromechanical impedance characteristics of each node, fit and construct a three-dimensional spatial mapping ellipsoid.

[0007] Step 2: Adaptive mesh generation and flow-vibration energy transfer numerical simulation are performed on the three-dimensional spatially mapped ellipsoid along the principal axis to obtain the spatial phase coupling correction value;

[0008] Step 3: Phase alignment and energy weight calibration of key time-frequency features are performed using spatial phase coupling correction values ​​to complete multi-dimensional feature fusion and construct a dynamic operation feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operation feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained;

[0009] Step 4: Using the real-time health index as the fitting benchmark, perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory; match and map the equipment performance degradation trajectory with the preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and deduce the fault warning level and remaining service life prediction results.

[0010] Step 5: Based on the fault warning level and the predicted remaining service life, maintenance work orders and dynamic compensation instructions for system air volume and air pressure are obtained to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

[0011] Furthermore, step 1 is preceded by:

[0012] Multi-dimensional operating parameters of the terminal equipment of the underground space ventilation system are collected in real time during continuous operation. The centrifugal fan bearing housing coupling end, the fireproof smoke exhaust valve plate shaft sealing area, and the static pressure box guide array outlet surface are used as three key monitoring nodes. The spatial topological coordinates and physical boundary attributes of each node are recorded synchronously. The multi-dimensional operating parameters are processed by spatiotemporal alignment, noise filtering and dimension normalization to obtain a standardized operating dataset.

[0013] Furthermore, using the standardized operational dataset as the processing object, key time-frequency features characterizing the mechanical operation of the fan and the airflow state in the pipeline are extracted; based on the spatial distribution coordinates of three key monitoring nodes and the aeromechanical impedance characteristics of each node, a three-dimensional spatial mapping ellipsoid is fitted and constructed, including:

[0014] An adaptive time-frequency decomposition operation is performed on the standardized operating dataset to separate the mechanical vibration frequency band and the pneumatic vibration frequency band. The instantaneous amplitude envelope, dominant frequency offset and phase difference information of each frequency band are extracted and combined to obtain a set of key time-frequency feature vectors characterizing the mechanical operation of the fan and the airflow state of the pipeline.

[0015] The key time-frequency feature vector set and the spatial distribution coordinates of the three key monitoring nodes are transformed to obtain a spatial reference vector. The aeromechanical impedance characteristics of each node are quantized into frequency domain impedance weighting coefficients. The key time-frequency feature vector set is weighted and projected using the impedance weighting coefficients to obtain a multi-source feature covariance matrix.

[0016] Using the multi-source feature covariance matrix as the processing object, eigenvalues ​​and eigenvectors are extracted as spatial topological constraints. The principal axis direction parameters and semi-axis length parameters of the ellipsoid quadratic surface equation are solved. The principal axis direction parameters and semi-axis length parameters are mapped to a three-dimensional Cartesian coordinate system, and a three-dimensional spatial mapping ellipsoid representing the boundary of the coupled operating state of the equipment is fitted and constructed.

[0017] Furthermore, adaptive mesh generation and flow-vibration energy transfer numerical simulations were performed on the three-dimensional spatially mapped ellipsoid along the principal axis to obtain spatial phase coupling correction values, including:

[0018] Along the principal axis of the three-dimensional spatially mapped ellipsoid, a variable density adaptive meshing operation is performed based on the aerodynamic-mechanical impedance gradient distribution of the nodes to obtain a three-dimensional discrete mesh topology. The airflow pulsation load and mechanical excitation load are applied to the mesh nodes as virtual excitation sources to perform a numerical simulation of flow-vibration energy transfer and obtain the set of multi-directional normal stress and shear stress components of each mesh element section under transient conditions.

[0019] Using the set of multi-directional normal stress and shear stress components as input data, a time-series stress state evolution sequence is constructed, the stress extreme point sequence is extracted, and the coordinates are mapped to a two-dimensional stress polar coordinate system to obtain the stress state geometric transformation trajectory of the periodic evolution of stress state.

[0020] Geometric feature tracing is performed on the stress state geometric transformation trajectory to extract the spatiotemporal translation vector of the trajectory center coordinates and the extreme envelope of the trajectory radius; based on the geometric mapping relationship between the spatiotemporal translation vector and the extreme envelope, the principal stress amplitude distribution field and the spatial tilt tensor of the maximum shear stress surface of each grid element are solved in reverse.

[0021] Substituting the principal stress amplitude distribution field and the spatial tilt tensor into the material fatigue constitutive relation, and combining the fluid-structure interaction boundary conditions, the mechanical vibration phase lag angle and the aerodynamic response energy attenuation coefficient are calculated; tensor synthesis and normalization operations are performed on the mechanical vibration phase lag angle and the aerodynamic response energy attenuation coefficient to obtain the spatial phase coupling correction value.

[0022] Furthermore, by using spatial phase coupling correction values ​​to perform phase alignment and energy weight calibration on key time-frequency features, multi-dimensional feature fusion is completed, and a dynamic operating feature sequence of the equipment is constructed. By performing sliding window time-series evolution analysis on the dynamic operating feature sequence, a real-time health index of the equipment's current degradation level is obtained, including:

[0023] The phase lag compensation parameter and energy attenuation weight parameter are extracted from the spatial phase coupling correction value. The phase lag compensation parameter is applied to the mechanical vibration phase angle and pneumatic pulsation phase angle of key time-frequency characteristics to achieve phase alignment between the fan operation signal and the pipeline airflow signal.

[0024] A multi-band energy allocation matrix is ​​constructed using energy attenuation weight parameters. The key time-frequency features after phase alignment are weighted and superimposed with the energy allocation matrix according to the feature frequency band to complete the multi-dimensional feature fusion and obtain the dynamic operation feature sequence of the equipment.

[0025] Set a sliding time window with a fixed time span and step interval, input the dynamic operation feature sequence of the equipment into the sliding time window in sequence, extract the statistical moments and frequency domain energy ratio of the feature vectors in each time window, calculate the distance divergence and evolution difference rate of the feature vectors of adjacent time windows, and construct the temporal evolution difference matrix.

[0026] Principal component extraction is performed on the temporal evolution difference matrix to obtain the characteristic energy decay ratio in the principal evolution direction. The characteristic energy decay ratio is then mapped to the standard health baseline to obtain the real-time health index of the current degradation level of the device through a nonlinear decay function.

[0027] Furthermore, using the real-time health index as a fitting benchmark, nonlinear trend extrapolation is performed along a continuous time axis to obtain the equipment performance degradation trajectory. This trajectory is then matched and mapped with a pre-defined fault mechanism feature library to analyze the specific early latent faults corresponding to the current degradation stage. Finally, the fault warning level and remaining service life prediction results are derived, including:

[0028] The real-time health index in the continuous time series is smoothed and denoised, and the inflection point of health status decay and the rate change node are extracted. Using the decay inflection point as the segment boundary, nonlinear trend extrapolation is performed along the continuous time axis to obtain the equipment performance degradation trajectory.

[0029] The equipment performance degradation trajectory is divided into multiple degradation feature segments. The attenuation slope, curvature change rate and frequency domain jitter features of each degradation feature segment are extracted and input into a preset fault mechanism feature library. The multidimensional spatial similarity with the standard degradation template in the library is calculated to obtain the trajectory matching similarity vector.

[0030] Extreme value retrieval and clustering are performed on the trajectory matching similarity vector to filter out the fault mechanism category corresponding to the similarity peak. Combined with the aerodynamic-mechanical impedance characteristics of the equipment, the specific early latent faults corresponding to the current deterioration stage are analyzed.

[0031] Based on the damage evolution rate of specific early latent faults and the remaining attenuation margin of equipment performance degradation trajectory, combined with the preset safety failure threshold, life attenuation integral extrapolation is performed to divide risk intervals to obtain fault warning levels, and the predicted results of the remaining service life of the equipment are output.

[0032] Furthermore, based on the fault warning level and remaining service life prediction results, maintenance work orders and dynamic compensation instructions for system airflow and pressure are obtained, completing predictive maintenance and closed-loop health management of the ventilation system's terminal equipment, including:

[0033] The fault warning level and the remaining service life prediction results are input into the preset maintenance decision matrix, and the corresponding equipment maintenance priority and intervention strategy are matched to obtain a maintenance work order that includes fault location coordinates, hidden danger type identification, expected operation window period and standardized handling process.

[0034] Based on the hazard type identifier in the maintenance work order and the current attenuation margin of the equipment performance degradation trajectory, the air volume gain threshold and air pressure balance margin required to maintain the underground space environmental standard are calculated, and the system air volume and air pressure dynamic compensation command is obtained by combining them.

[0035] The maintenance work order and the system air volume and air pressure dynamic compensation instruction are encoded with communication protocol and encapsulated with execution timing, and sent to the centralized control cabinet to synchronously trigger the fan frequency conversion speed regulation, valve opening adjustment and maintenance terminal work order push.

[0036] The system collects the feedback value of the air pressure at the end of the pipeline and the response current of the equipment after the execution of the dynamic compensation command for air volume and air pressure in real time. It then compares and verifies the deviation with the expected compensation target. If the deviation exceeds the allowable threshold, it triggers the iterative correction of the compensation parameters and reissues them. If the deviation is within the allowable threshold, it locks the current intervention parameters and updates the baseline of the equipment health status, thus completing the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

[0037] Secondly, the fault prediction and health management system for terminal equipment of the ventilation system includes:

[0038] The module is used to extract key time-frequency features that characterize the mechanical operation of the fan and the airflow state in the pipeline by using the standardized operational dataset as the processing object; and to fit and construct a three-dimensional spatial mapping ellipsoid based on the spatial distribution coordinates of three key monitoring nodes and the aeromechanical impedance characteristics of each node.

[0039] The calculation module is used to perform adaptive mesh generation and numerical simulation of flow-vibration energy transfer on a three-dimensional spatially mapped ellipsoid along the principal axis, and obtain spatial phase coupling correction values.

[0040] The analysis module is used to perform phase alignment and energy weight calibration of key time-frequency features through spatial phase coupling correction values, complete multi-dimensional feature fusion, and construct a dynamic operating feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operating feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained;

[0041] The prediction module is used to extrapolate the nonlinear trend along the continuous time axis using the real-time health index as the fitting benchmark to obtain the equipment performance degradation trajectory; the equipment performance degradation trajectory is matched and mapped with the preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and the fault warning level and remaining service life prediction results are deduced.

[0042] The compensation module is used to obtain maintenance work orders and dynamic compensation instructions for system air volume and air pressure based on the fault warning level and the prediction results of the remaining service life, so as to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

[0043] Thirdly, a computing device, comprising:

[0044] One or more processors;

[0045] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0046] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0047] The above-described solution of the present invention has at least the following beneficial effects:

[0048] A three-dimensional spatial mapping ellipsoid is constructed by fitting the spatial distribution coordinates of three key monitoring nodes with aerodynamic-mechanical impedance characteristics. Adaptive meshing and flow-vibration energy transfer numerical simulations are performed along the principal axis to calculate the spatial phase coupling correction value. This correction value is then used to perform phase alignment and energy weight calibration on key time-frequency features to achieve multi-dimensional feature fusion. Combined with sliding window time-series evolution analysis, a real-time health index is obtained, and nonlinear trend extrapolation is performed to generate equipment performance degradation trajectories. By matching and mapping with a preset fault mechanism feature library, specific early latent faults and remaining service life are analyzed. Finally, maintenance work orders and dynamic compensation of system airflow and pressure are generated simultaneously based on the early warning level and service life prediction results. The technical means of the command overcomes the technical problems of existing operation and maintenance models that rely solely on fixed-cycle manual inspections and simple current measurements, which cannot collect and integrate multi-dimensional dynamic features in real time, lack continuous quantitative monitoring of equipment performance degradation trends and multi-physical field coupling analysis capabilities, and make it difficult to accurately identify and provide early warnings for early hidden faults such as bearing micro-wear and valve body jamming, which can easily develop into sudden and complete failures. Thus, it achieves the technical effect of accurately sensing the multi-physical field coupling operation status of equipment, realizing quantitative early warning of early hidden faults and predictive maintenance closed-loop management, ensuring the daily ventilation efficiency and emergency safety of fire smoke extraction in underground spaces, and reducing unplanned downtime losses and passive maintenance costs. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating a method for fault prediction and health management of terminal equipment in a ventilation system, provided by an embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of a fault prediction and health management system for terminal equipment of a ventilation system provided in an embodiment of the present invention.

[0051] Figure 3 This is a statistical diagram showing the number of alarms and maintenance work orders for each fault warning level.

[0052] Figure 4 This is a diagram comparing the operating efficiency of equipment under predictive maintenance and reactive maintenance modes. Detailed Implementation

[0053] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art.

[0054] like Figure 1 As shown, an embodiment of the present invention proposes a method for fault prediction and health management of terminal equipment in a ventilation system, the method comprising the following steps:

[0055] Step 1: Using the standardized operational dataset as the processing object, extract key time-frequency features that characterize the mechanical operation of the fan and the airflow state in the pipeline; based on the spatial distribution coordinates of the three key monitoring nodes and the aeromechanical impedance characteristics of each node, fit and construct a three-dimensional spatial mapping ellipsoid.

[0056] Step 2: Adaptive mesh generation and flow-vibration energy transfer numerical simulation are performed on the three-dimensional spatially mapped ellipsoid along the principal axis to obtain the spatial phase coupling correction value;

[0057] Step 3: Phase alignment and energy weight calibration of key time-frequency features are performed using spatial phase coupling correction values ​​to complete multi-dimensional feature fusion and construct a dynamic operation feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operation feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained;

[0058] Step 4: Using the real-time health index as the fitting benchmark, perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory; match and map the equipment performance degradation trajectory with the preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and deduce the fault warning level and remaining service life prediction results.

[0059] Step 5: Based on the fault warning level and the predicted remaining service life, maintenance work orders and dynamic compensation instructions for system air volume and air pressure are obtained to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

[0060] In this embodiment of the invention, a standardized operational dataset is used as the processing object. Key time-frequency features are extracted and a three-dimensional spatial mapping ellipsoid is constructed. Spatial phase coupling correction values ​​are obtained through adaptive mesh partitioning and numerical simulation of flow-vibration energy transfer. Real-time health index is obtained through multi-dimensional feature fusion and sliding window time-series evolution analysis. Then, fault warning levels and remaining service life are deduced through nonlinear trend extrapolation and fault mechanism feature library matching. Finally, maintenance work orders and system air volume and pressure dynamic compensation instructions are generated to achieve closed-loop management. Therefore, this invention overcomes the technical problems of existing ventilation system terminal equipment fault prediction, such as difficulty in accurately capturing early hidden degradation characteristics of fan mechanical operation and duct airflow coupling, delayed fault warning, inaccurate prediction of remaining service life, lack of full-process closed-loop health management, and inability to achieve predictive maintenance. Thus, it achieves the technical effects of accurately identifying early hidden faults of equipment, issuing fault warnings in advance, accurately predicting remaining service life, realizing predictive maintenance and closed-loop health management of equipment, improving the operational stability and reliability of ventilation system terminal equipment, and reducing operation and maintenance costs.

[0061] In a preferred embodiment of the present invention, the method further includes the following step before step 1:

[0062] Step 01: Real-time acquisition of multi-dimensional operating parameters of the terminal equipment of the underground space ventilation system during continuous operation. Three key monitoring nodes are selected: the coupling end of the centrifugal fan bearing seat, the sealing area of ​​the fireproof smoke exhaust valve plate shaft, and the outlet surface of the static pressure box guide array. The spatial topological coordinates and physical boundary attributes of each node are recorded synchronously. The multi-dimensional operating parameters are then processed through spatiotemporal alignment, noise filtering, and dimensional normalization to obtain a standardized operating dataset. Specifically, this includes real-time acquisition and preprocessing of operating parameters for the terminal equipment of the ventilation system under all operating conditions, addressing the operating environment and maintenance pain points of centralized ventilation systems in underground parking garages and supporting equipment rooms of large commercial complexes. Real-time acquisition of multi-dimensional operating parameters of centrifugal fans, ducted supply and exhaust fans, fireproof smoke exhaust valves, and static pressure boxes during continuous operation is conducted. The acquired parameters cover equipment mechanical vibration signals, bearing seat temperature rise values, outlet air pressure and air volume pulsation values, motor operating current and speed signals, valve body shaft torque, and static pressure box outlet airflow disturbance data. The acquisition process is synchronized with the actual operating sequence of the equipment, ensuring no parameter changes under transient conditions are missed.

[0063] The centrifugal fan bearing housing coupling end, the fireproof smoke exhaust valve plate shaft sealing area, and the static pressure box guide array outlet surface were selected as three key monitoring nodes. While collecting operating parameters, the three-dimensional spatial topological coordinates of each node in the underground ventilation duct network were recorded by spatial positioning. At the same time, the physical boundary attributes such as material properties, structural dimensions, and airflow cross-sections of each node were registered. The spatial coordinate information was synchronously bound with the operating parameters to provide basic data for subsequent spatial feature mapping.

[0064] After completing the collection of multi-dimensional operating parameters and node information, the collected data is spatiotemporally aligned. Using a unified system clock as a reference, the asynchronous time-series data collected by different sensors are matched and sorted according to timestamps, so that the data of different monitoring nodes and different types of parameters at the same time are consistent in the time dimension, forming a time-aligned original data sequence.

[0065] Subsequently, noise filtering was performed on the original data sequence after spatiotemporal alignment. An adaptive filtering algorithm was used to remove random interference signals and power frequency noise caused by the damp and dusty underground environment, while retaining the effective signal components that truly reflect the operating status of the equipment. During the filtering process, the amplitude difference between the original signal and the filtered signal was compared point by point to ensure that the effective features were not overly smoothed or filtered out.

[0066] Next, the filtered data is normalized to map parameters with different dimensions, such as vibration amplitude, temperature, wind pressure, current, and rotational speed, to the same numerical range. By eliminating the differences in dimensions and magnitudes between different physical quantities, various parameters have a basis for direct fusion and comparative analysis. After spatiotemporal alignment, noise filtering, and dimensional normalization, a standardized operational dataset with regularized time sequence, noise removal, and unified numerical values ​​is finally formed.

[0067] In this embodiment of the invention, multi-dimensional operating parameters of the terminal equipment of the underground space ventilation system are collected in real time during continuous operation. The spatial topological coordinates and physical boundary attributes of each node are recorded synchronously using the coupling end of the centrifugal fan bearing seat, the sealing area of ​​the valve plate of the fireproof smoke exhaust valve, and the outlet surface of the static pressure box guide array as three key monitoring nodes. The collected data is then processed by performing spatiotemporal alignment, noise filtering, and dimensional normalization to generate a standardized operating dataset. This overcomes the technical problems of existing ventilation systems, such as data acquisition lag, single monitoring dimension, susceptibility to noise interference from the damp and dusty underground environment, and lack of unified spatiotemporal reference and dimensional standard for multi-source heterogeneous parameters. Thus, it achieves the technical effect of real-time synchronous acquisition and high-precision standardized preprocessing of multi-source operating data, eliminating the constraints of environmental interference and dimensional differences on feature analysis.

[0068] In a preferred embodiment of the present invention, step 1 above may include:

[0069] Step 1.1 involves performing an adaptive time-frequency decomposition operation on the standardized operating dataset to separate the mechanical vibration frequency band and the aerodynamic frequency band. The instantaneous amplitude envelope, dominant frequency offset, and phase difference information of each frequency band are extracted and combined to obtain a key time-frequency feature vector set characterizing the mechanical operation of the fan and the airflow state in the pipeline. Specifically, addressing the problem of mechanical degradation and airflow disturbance characteristics being difficult to separate due to the long-term high-load operation of ventilation terminal equipment in the underground parking garage and supporting equipment room of large commercial complexes in a humid and dusty environment, the standardized operating dataset, after spatiotemporal alignment, noise filtering, and dimensional normalization, is used as the processing object, and an adaptive time-frequency decomposition operation is performed on it. This decomposition process automatically distinguishes and separates the mechanical vibration frequency band, reflecting the mechanical state of the fan bearing operation and impeller rotation, and the aerodynamic frequency band, reflecting the airflow transport in the pipeline and the airflow pulsation in the plenum, based on the signal frequency distribution characteristics. This avoids mutual interference between the two types of features, preventing distortion of degradation information.

[0070] After frequency band separation, feature extraction is performed on the mechanical vibration band and the pneumatic pulse band respectively. The instantaneous amplitude envelope of each frequency band is calculated by continuous fitting of the time-series signal amplitude to characterize the real-time variation of signal energy. Next, the dominant frequency offset is calculated, and the formula for calculating the dominant frequency offset is: ,in This is the dominant frequency in the current operating band under the current conditions. This is the reference dominant frequency under healthy equipment conditions. This value can intuitively reflect the degree of frequency drift caused by equipment degradation; simultaneously, phase difference information is calculated, and the formula for calculating the phase difference is... ,in For the phase of the mechanical vibration signal, The phase of the pneumatic pulse signal is used to reflect the degree of synchronization between mechanical motion and airflow transmission. The instantaneous amplitude envelope, dominant frequency offset, and phase difference information of each frequency band are systematically integrated to obtain a set of key time-frequency feature vectors that comprehensively characterize the mechanical operation state of the fan and the airflow state of the pipeline.

[0071] Step 1.2 involves performing coordinate transformation on the key time-frequency feature vector set and the spatial distribution coordinates of the three key monitoring nodes to obtain a spatial reference vector. The aeromechanical impedance characteristics of each node are then quantified into frequency domain impedance weighting coefficients. A weighted projection operation is performed on the key time-frequency feature vector set using these impedance weighting coefficients to obtain a multi-source feature covariance matrix. Specifically, after obtaining the key time-frequency feature vector set, coordinate transformation is performed on the vector set and the spatial distribution coordinates of the three key monitoring nodes: the coupling end of the centrifugal fan bearing seat, the sealing area of ​​the fireproof smoke exhaust valve plate shaft, and the outlet surface of the static pressure box guide array. A spatial reference vector integrating feature information and spatial location information is obtained by performing linear mapping operations between the dimensional components of the feature vector and the three-dimensional spatial coordinate components of the corresponding nodes.

[0072] Simultaneously, based on the actual operating conditions of each monitoring node, the aerodynamic-mechanical impedance characteristics of the node itself are quantitatively calculated to obtain the frequency domain impedance weighting coefficient. The formula for calculating the frequency domain impedance weighting coefficient is as follows: ,in The measured value of the node aerodynamic impedance. This is the measured value of the node's mechanical impedance, and this coefficient reflects the proportion of contributions from airflow and mechanical structure at the node. The calculated frequency domain impedance weighting coefficients are used to perform a weighted projection operation on the key time-frequency feature vector set. The formula for calculating the weighted projection feature is as follows: ,in These are the original key time-frequency features. These are the features after weighted projection.

[0073] After weighted projection, a multi-source feature covariance matrix is ​​constructed to quantify the correlation characteristics among multi-source features. The formula for calculating the multi-source feature covariance matrix is ​​as follows: ,in The number of feature samples used in the calculation. The feature matrix after weighted projection. The transpose of the weighted projection feature matrix is ​​used to calculate the multi-source feature covariance matrix, which can fully reflect the coupling relationship between mechanical features, aerodynamic features and spatial location features.

[0074] Step 1.3: Using the multi-source feature covariance matrix as the processing object, extract eigenvalues ​​and eigenvectors as spatial topological constraints, solve the principal axis direction parameters and semi-axis length parameters of the ellipsoidal quadratic surface equation, map the principal axis direction parameters and semi-axis length parameters to a three-dimensional Cartesian coordinate system, and fit and construct a three-dimensional spatial mapping ellipsoid representing the boundary of the coupled operating state of the equipment. Specifically, this includes: using the calculated multi-source feature covariance matrix as the processing object, performing eigenvalue decomposition operation on it, extracting the corresponding eigenvalues ​​and eigenvectors of the matrix, using the eigenvalues ​​as parameters representing the degree of dispersion of the feature distribution, and the eigenvectors as parameters representing the direction of the feature spatial distribution. The two together constitute the three-dimensional spatial topological constraints, limiting the spatial distribution range of the coupled operating state of the equipment.

[0075] The core parameters for solving the equation of the ellipsoidal quadratic surface are derived from eigenvalues ​​and eigenvectors. The principal axis orientation parameter of the ellipsoid is directly determined by the spatial orientation of the eigenvectors, and the semi-axis length parameter is obtained by taking the square root of the eigenvalue. The formula for calculating the semi-axis length is as follows: ,in , , These are the three eigenvalues ​​obtained from the multi-source feature covariance matrix decomposition. , , These correspond to the lengths of the three semi-axis of the ellipsoid in three-dimensional space.

[0076] The principal axis direction parameters and semi-axis length parameters obtained from the solution are mapped to a standard three-dimensional Cartesian coordinate system and substituted into the quadratic surface equation of the ellipsoid for surface fitting. The fitting process uses three key monitoring nodes as spatial anchor points to constrain the spatial position and shape boundary of the ellipsoid, so that the ellipsoid completely encapsulates all the operating state information under the coupling effect of the mechanical operation of the fan and the airflow in the pipeline. Finally, a three-dimensional spatial mapping ellipsoid that can accurately characterize the flow-machine coupling operating state boundary of the terminal equipment of the ventilation system is obtained by fitting.

[0077] In this embodiment of the invention, an adaptive time-frequency decomposition is performed on a standardized operating dataset to separate the mechanical vibration and airflow pulsation frequency bands, and instantaneous amplitude envelope, dominant frequency offset, and phase difference information are extracted to construct a key time-frequency feature vector set. Combined with spatial coordinate transformation of key monitoring nodes and quantization of frequency domain impedance weighting coefficients, a weighted projection operation is performed to obtain a multi-source feature covariance matrix. Then, eigenvalues ​​and eigenvectors are extracted as spatial topological constraints to solve the ellipsoid quadratic surface equation parameters and mapped to a three-dimensional Cartesian coordinate system to fit and construct a three-dimensional spatial mapping ellipsoid. This overcomes the technical problems of existing ventilation system terminal equipment operating feature extraction being of a single dimension, difficulty in separating the cross-coupling of mechanical vibration and airflow pulsation signals, lack of a unified spatial topological association and impedance mapping mechanism for multi-source heterogeneous data, and inability to intuitively quantify and characterize the boundaries of the equipment's multi-physics coupled operating state under complex working conditions. This achieves the technical effect of accurate decoupling of multi-band dynamic features and deep fusion of spatial topology. The constructed three-dimensional spatial mapping ellipsoid accurately depicts the dynamic operating boundary and state evolution trajectory of the equipment under flow-mechanical coupling.

[0078] In a preferred embodiment of the present invention, step 2 above may include:

[0079] Step 2.1: Along the principal axis of the 3D spatially mapped ellipsoid, perform a variable-density adaptive meshing operation based on the aerodynamic-mechanical impedance gradient distribution at the nodes to obtain a 3D discrete mesh topology. Apply airflow pulsation load and mechanical excitation load as virtual excitation sources to the mesh nodes and perform numerical simulation of flow-vibration energy transfer to obtain the set of multi-directional normal stress and shear stress components of each mesh element section under transient conditions. Specifically, addressing the problem of local stress concentration and difficulty in quantifying degradation characteristics caused by the coupling effect of airflow pulsation and mechanical vibration in the ventilation terminal equipment of underground space in large commercial complexes during high-load continuous operation, a variable-density adaptive meshing operation is carried out along the principal axis of the ellipsoid, using the constructed 3D spatially mapped ellipsoid as the geometric basis. During the meshing process, the aerodynamic-mechanical impedance gradient of the three key monitoring nodes is used as the dividing basis. Regions with larger impedance gradient values ​​have denser meshing, while regions with smaller impedance gradient values ​​have sparser meshing. The formula for calculating the aerodynamic-mechanical impedance gradient is... ,in and These are the frequency domain impedance weighting coefficients for adjacent nodes. The spatial distance between adjacent nodes is used to complete the partitioning through this gradient constraint, resulting in a three-dimensional discrete mesh topology that conforms to the flow-machine coupling characteristics of the device.

[0080] After the mesh is constructed, the airflow pulsation load and mechanical excitation load generated during the actual operation of the equipment are applied as virtual excitation sources to the corresponding mesh nodes. The airflow pulsation load is taken from the measured data of wind pressure pulsation in the pipeline, and the mechanical excitation load is taken from the vibration excitation data of the fan bearing and impeller. Then, the flow-vibration energy transfer numerical simulation is performed. During the simulation, the multi-directional normal stress and shear stress components of each mesh unit section under transient operating conditions are solved by the elasticity control equation. The multi-directional normal stress includes horizontal normal stress, vertical normal stress and axial normal stress, and the shear stress includes in-plane shear stress and cross-sectional torsional shear stress. Finally, the complete set of multi-directional normal stress and shear stress components is obtained.

[0081] Step 2.2: Using the set of multi-directional normal stress and shear stress components as input data, a time-series stress state evolution sequence is constructed. The sequence of stress extremum points is extracted, and their coordinates are mapped to a two-dimensional stress polar coordinate system to obtain the geometric transformation trajectory of the periodic evolution of the stress state. Specifically, this includes: using the acquired set of multi-directional normal stress and shear stress components as basic input data, arranging them according to the continuously acquired timestamps, and constructing a time-series stress state evolution sequence that reflects the dynamic changes of the equipment stress state with operating time. The maximum and minimum values ​​of stress values ​​are retrieved periodically in the time-series stress state evolution sequence, all stress extremum points are extracted, and a stress extremum point sequence is formed. To intuitively present the periodic change law of the stress state, the coordinates of the stress extremum points in the rectangular coordinate system are transformed to the two-dimensional stress polar coordinate system. The coordinate transformation formula is: , ,in For polar coordinates, the polar radius The polar angle is in polar coordinates. These are the normal stress components in the corresponding directions. This corresponds to the plane shear stress components.

[0082] The transformed polar coordinate stress extreme points are continuously and smoothly fitted, and the fitted points at each time point are connected in sequence to form a stress state geometric transformation trajectory that can fully reflect the periodic evolution law of the stress state of the equipment under the action of flow-vibration coupling. This trajectory can intuitively reflect the stress distortion characteristics caused by hidden faults such as bearing micro-wear, impeller dust imbalance, and valve body jamming.

[0083] Step 2.3: Perform geometric feature tracking processing on the stress state geometric transformation trajectory to extract the spatiotemporal translation vector of the trajectory center coordinates and the extreme value envelope of the trajectory radius. Based on the geometric mapping relationship between the spatiotemporal translation vector and the extreme value envelope, solve inversely the principal stress amplitude distribution field and the spatial tilt tensor of the maximum shear stress surface for each grid element. Specifically, this includes: performing continuous geometric feature tracking processing on the stress state geometric transformation trajectory, calculating the geometric center coordinates of the trajectory in real time during the tracking process, recording the offset of the geometric center with time and space, forming the spatiotemporal translation vector of the trajectory center coordinates. The formula for calculating the spatiotemporal translation vector is as follows: ,in The three-dimensional coordinates of the trajectory center at the current moment. The baseline coordinates of the trajectory center are determined when the device is in good health. The radius values ​​of all points on the trajectory are traversed, the maximum and minimum values ​​of the radius are extracted, and the extreme values ​​of the radius at different times are connected in sequence to form the extreme value envelope of the trajectory radius.

[0084] Based on the geometric mapping relationship between the spatiotemporal translation vector and the extreme value envelope, the principal stress amplitude distribution field of each grid element is calculated by reverse derivation. The formula for calculating the principal stress amplitude is as follows: ,in The amplitudes of the first principal stress and the second principal stress are, This represents the bidirectional normal stress components. Simultaneously, the spatial tilt tensor of the surface of maximum shear stress is solved, ultimately yielding the complete principal stress amplitude distribution field and the spatial tilt tensor of the surface of maximum shear stress.

[0085] Step 2.4: Substitute the principal stress amplitude distribution field and the spatial tilt tensor into the material fatigue constitutive equation, and calculate the mechanical vibration phase lag angle and aerodynamic response energy attenuation coefficient based on the fluid-structure interaction boundary conditions. Perform tensor synthesis and normalization operations on the mechanical vibration phase lag angle and aerodynamic response energy attenuation coefficient to obtain the spatial phase coupling correction value. Specifically, this includes: substituting the solved principal stress amplitude distribution field and the spatial tilt tensor of the maximum shear stress surface into the material fatigue constitutive equation, and calculating the mechanical vibration phase lag angle and aerodynamic response energy attenuation coefficient based on the fluid-structure interaction boundary constraints of the underground ventilation equipment. The formula for calculating the mechanical vibration phase lag angle is as follows: ,in For the aerodynamic excitation phase angle, The phase angle of the mechanical vibration response; the formula for calculating the energy attenuation coefficient of the aerodynamic response is: ,in The input stream is the vibrational energy value. This represents the output energy value after being transmitted through the grid cells.

[0086] After obtaining the mechanical vibration phase lag angle and the aerodynamic response energy attenuation coefficient, tensor synthesis is performed on the two. The synthesis formula is as follows: ,in To synthesize tensors, For the space tilt tensor, Using the principal stress tensor, the synthesized tensor is normalized to obtain the corrected spatial phase coupling value of the multi-source characteristic phase and energy, which can effectively solve the problems of phase misalignment and energy attenuation distortion of flow-vibration signals under complex underground working conditions.

[0087] In this embodiment of the invention, a three-dimensional discrete mesh topology is constructed by performing variable-density adaptive meshing along the principal axis of a three-dimensional spatially mapped ellipsoid based on the aerodynamic-mechanical impedance gradient distribution of the nodes. Airflow pulsation loads and mechanical excitation loads are used as virtual excitation sources to perform numerical simulations of flow-vibration energy transfer to obtain a set of multi-directional stress components under transient conditions. A temporal stress state evolution sequence is then constructed and mapped to a two-dimensional stress polar coordinate system to generate a geometric transformation trajectory of the stress state. By tracking the geometric features of the trajectory, the spatiotemporal translation vector and the radius extremum envelope are extracted to inversely solve the principal stress amplitude distribution field and the spatial tilt tensor of the maximum shear stress surface. Finally, the mechanical vibration phase lag angle and aerodynamic... The technique of obtaining spatial phase coupling correction values ​​by responding to the energy attenuation coefficient and performing tensor synthesis and normalization operations overcomes the technical problems of traditional stress monitoring and analysis, which are difficult to accurately characterize the spatial distribution and periodic evolution of transient multi-directional stress under complex fluid-structure interaction conditions, cannot quantify the phase lag and energy attenuation characteristics in the process of mechanical vibration and pneumatic dynamic energy transfer, and make it difficult to accurately invert the evolution mechanism of hidden damage inside the equipment and the intensity of multi-field coupling interference. It achieves the technical effect of refined numerical analysis of the fluid-vibration multi-physics energy transfer process and geometric characterization of stress state. The obtained spatial phase coupling correction values ​​reflect the intensity of multi-field coupling interference and phase deviation of the equipment under alternating loads with high fidelity.

[0088] In a preferred embodiment of the present invention, step 3 above may include:

[0089] Step 3.1: Extract the phase lag compensation parameter and energy attenuation weight parameter from the spatial phase coupling correction value. Apply the phase lag compensation parameter to the mechanical vibration phase angle and pneumatic vibration phase angle of key time-frequency features to achieve phase alignment between the fan operation signal and the pipeline airflow signal. Specifically, addressing the problem of phase misalignment between the fan mechanical operation signal and the pipeline airflow signal under the flow-vibration coupling effect of ventilation terminal equipment in underground space, which leads to feature distortion, separate and extract the phase lag compensation parameter and energy attenuation weight parameter from the obtained spatial phase coupling correction value. The phase lag compensation parameter is used to correct the signal phase deviation, and the energy attenuation weight parameter is used for subsequent feature energy calibration. Together, they provide an accurate correction basis for multi-dimensional feature fusion.

[0090] After extraction, the phase lag compensation parameter is applied to the mechanical vibration phase angle and pneumatic pulse phase angle in the obtained key time-frequency features to achieve phase alignment of the two types of signals. The specific calculation process is as follows: The mechanical vibration phase angle compensation formula is... The formula for compensating for the phase angle of the pneumatic artery is: ,in The phase angle of the original mechanical vibration. The original pneumatic artery phase angle, These are the phase lag compensation parameters. , These are the compensated mechanical vibration phase angle and the pneumatic vibration phase angle, respectively. Through this compensation calculation, the signal phase difference caused by the lag in flow-vibration energy transfer is eliminated.

[0091] Step 3.2 involves constructing a multi-band energy allocation matrix using energy attenuation weight parameters. The key time-frequency features after phase alignment are then weighted and superimposed on the energy allocation matrix according to their respective feature bands to complete multi-dimensional feature fusion and obtain the dynamic operation feature sequence of the equipment. Specifically, after signal phase alignment, a multi-band energy allocation matrix is ​​constructed based on the extracted energy attenuation weight parameters. This matrix quantifies the energy contribution ratio of the mechanical vibration frequency band and the pneumatic vibration frequency band, adapting to the flow-mechanical coupling operation characteristics of underground ventilation equipment. The construction method of the multi-band energy allocation matrix is ​​as follows: the energy attenuation weight parameters are used as the diagonal elements of the matrix, and the off-diagonal elements are set to 0. The matrix dimension is consistent with the dimension of the key time-frequency feature vector set, i.e. ,in The energy decay weight parameters are for each feature. For multi-band energy allocation matrix, This is a diagonal matrix generation function that constructs a diagonal matrix with the parameters in parentheses as diagonal elements and all other off-diagonal elements set to 0.

[0092] The phase-aligned key time-frequency feature vector set is weighted and superimposed with the multi-band energy allocation matrix to achieve the fusion of multi-dimensional features. The weighted superposition calculation formula is as follows: ,in This is the set of key time-frequency feature vectors after phase alignment. For multi-band energy allocation matrix, This is the sequence of dynamic operating characteristics of the merged equipment.

[0093] Step 3.3: Set a sliding time window with a fixed time span and step interval. Input the dynamic operation feature sequence of the equipment into the sliding time window in sequence. Extract the statistical moments and frequency domain energy ratio of the feature vectors in each time window. Calculate the distance divergence and evolution difference rate of the feature vectors of adjacent time windows. Construct a time-series evolution difference matrix. Specifically, to achieve continuous quantitative monitoring of the equipment performance degradation trend and solve the pain point that the existing operation and maintenance mode cannot capture the gradual process of latent faults, a sliding time window with a fixed time span and step interval is set. The time span is determined according to the equipment operation cycle and the evolution rate of latent faults. The step interval is set to 1 / 5 of the time span to ensure that feature changes can be fully captured and data redundancy can be avoided.

[0094] The obtained dynamic operating feature sequence of the device is sequentially input into a sliding time window. Feature extraction is performed on the feature vectors within each time window, and the statistical moments and frequency domain energy proportions of the feature vectors are calculated. The statistical moments include the first moment (mean), the second moment (variance), and the third moment (skewness), used to characterize the distribution characteristics of the feature vectors. The frequency domain energy proportion is the ratio of the energy of each feature frequency band to the total energy, calculated using the following formula: ,in For the first Energy of each characteristic frequency band The total energy across all characteristic frequency bands. For the first The proportion of frequency domain energy in each characteristic frequency band.

[0095] After extraction, the distance divergence and evolutionary difference rate of feature vectors within two adjacent time windows are calculated. The distance divergence is calculated using Euclidean distance, and the formula is as follows: ,in For the first Within the first time window Each feature component For the first Within the first time window Each feature component Let the distance divergence be... The dimension of the feature vector within a single time window; the formula for calculating the evolutionary dissimilarity rate is... ,in This represents the baseline distance divergence between adjacent time windows under healthy device conditions. To obtain the evolutionary difference rate, the evolutionary difference rates of all adjacent time windows are arranged sequentially to construct a time-series evolutionary difference matrix. Step 3.4 involves performing principal component extraction on the time-series evolutionary difference matrix to obtain the characteristic energy decay ratio along the dominant evolutionary direction. This characteristic energy decay ratio is then mapped to the standard health baseline using a deviation mapping. A real-time health index reflecting the current degree of equipment degradation is obtained through a nonlinear decay function. Specifically, this includes performing principal component extraction on the constructed time-series evolutionary difference matrix, using a principal component analysis algorithm to remove redundant information from the matrix, retaining the dominant evolutionary direction that reflects the equipment degradation trend, and extracting the characteristic energy decay ratio along that direction. The formula for calculating the characteristic energy decay ratio is... ,in This is the characteristic reference energy of the equipment in a healthy state. The characteristic actual energy at the current moment, The characteristic energy decay ratio.

[0096] The calculated characteristic energy attenuation ratio is mapped to a preset standard health baseline. The standard health baseline is the threshold range of the characteristic energy attenuation ratio under the device's healthy state. The deviation mapping process involves calculating the deviation between the characteristic energy attenuation ratio and the baseline mean, using the following formula: ,in The mean of the standard healthy baseline, This is the deviation value.

[0097] Substituting the deviation value into the nonlinear decay function, a real-time health index representing the current degree of equipment degradation is calculated. The nonlinear decay function uses an exponential decay formula, i.e. ,in For real-time health index, The attenuation coefficient is calibrated based on the equipment material and operating conditions, with a value range of 0.8 to 1.2. The real-time health index ranges from 0 to 1. The closer to 1, the better the equipment's health status; the closer to 0, the more severe the equipment deterioration.

[0098] In this embodiment of the invention, the phase lag compensation parameter and energy attenuation weight parameter are extracted from the spatial phase coupling correction value to perform reverse compensation on the mechanical vibration phase angle and the pneumatic pulse phase angle respectively to achieve signal phase alignment. A multi-band energy allocation matrix is ​​constructed, and the key time-frequency features after phase alignment are weighted and superimposed according to the feature frequency bands to complete multi-dimensional feature fusion and generate a dynamic operating feature sequence of the equipment. Then, a sliding time window with a fixed time span and step interval is set, and the sequence is input sequentially. The statistical moments and frequency domain energy proportions of the feature vectors within each window are extracted, and the distance divergence and evolutionary difference rate of adjacent windows are calculated to construct a time-series evolutionary difference matrix. Finally, principal component extraction is performed on this matrix to obtain the feature energy attenuation ratio in the dominant evolutionary direction, and this ratio is deviated from the standard healthy baseline. The technique of mapping and calculating the real-time health index through a nonlinear attenuation function overcomes the technical problems of existing ventilation system terminal equipment's multi-source monitoring signals causing phase misalignment and energy distribution imbalance due to flow-vibration coupling interference, traditional feature fusion methods lacking frequency band weight allocation mechanisms which easily introduce cross-frequency band redundancy interference, and the difficulty in continuously capturing the micro-differences and degradation trends of equipment performance over time in static or single index evaluations, leading to inaccurate quantitative assessment of health status. It achieves high-precision phase synchronization and adaptive weighted fusion of multi-physics field monitoring signals, and accurately quantifies the transient degradation degree of equipment through sliding window time sequence difference analysis and nonlinear baseline deviation mapping. The output real-time health index provides a high-fidelity representation of the health degradation level of the equipment's current operating status.

[0099] In a preferred embodiment of the present invention, step 4 above may include:

[0100] Step 4.1 involves smoothing and denoising the real-time health index within the continuous time series, extracting the inflection points of health state decay and rate mutation nodes, and using the decay inflection points as segment boundaries to perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory. Specifically, this includes addressing the issue that underground space ventilation terminal equipment operates continuously under high load in a humid and dusty environment, resulting in slow deterioration of latent faults and fluctuations in the real-time health index due to environmental interference, making it difficult to accurately capture the equipment performance degradation trend. The aim of smoothing and denoising the real-time health index within the continuous time series is to eliminate random fluctuations caused by environmental interference and ensure that the real-time health index accurately and stably reflects the actual degradation state of the equipment. The smoothing and denoising uses a moving average filtering algorithm. The specific calculation process involves selecting a fixed number of sampling periods as the moving window, taking the real-time health index at the current moment and the previous few sampling periods within the window, calculating the arithmetic mean of the health index, and using this mean as the smoothed real-time health index at the current moment. The number of sampling periods included in the moving window is calibrated according to the actual operating cycle of the equipment, ranging from 5 to 10 sampling periods.

[0101] After smoothing and denoising, the inflection point of health status decay and the rate mutation node are further extracted. These two nodes are the core basis for dividing the equipment degradation stages. The decay inflection point is the starting point where the equipment's health status begins to decline continuously from a stable state. This is extracted by calculating the second difference of the real-time health index sequence. The calculation process is as follows: subtract twice the real-time health index of the previous moment from the current real-time health index, and add the real-time health indices of the two previous moments. When the result changes from 0 to a negative value and remains negative, the corresponding moment is the health status decay inflection point. The rate mutation node is the point where the rate of decay of the equipment health index suddenly increases. This is extracted by calculating the first difference of the real-time health index sequence. The calculation process is as follows: subtract the real-time health index of the previous moment from the current real-time health index. When the absolute value of the result suddenly increases and exceeds a preset threshold (which is three times the average of the first difference in the equipment's health status), the corresponding moment is the rate mutation node.

[0102] Using the extracted inflection point of health status decay as the segment boundary, the real-time health index of the continuous time series is divided into multiple different degradation stages, each corresponding to a process of equipment degradation. An adaptive piecewise fitting algorithm is employed to perform nonlinear trend extrapolation calculations on the real-time health index of each degradation stage along the continuous time axis. Based on the real-time health index at the initial moment of each degradation stage, the value is multiplied by the product of a negative decay coefficient (natural constant) and time, and then a health index threshold at the time of equipment failure is added. The decay coefficient is obtained by fitting the changing trend of the real-time health index within that degradation stage; the decay coefficient varies for different degradation stages, and the health index threshold at the time of equipment failure is preset to 0.1. Through the above adaptive piecewise fitting and nonlinear trend extrapolation calculations, a complete trajectory reflecting the gradual degradation of equipment performance from a healthy state to a failure state is finally obtained.

[0103] Step 4.2: Divide the equipment performance degradation trajectory into multiple degradation feature segments, extract the attenuation slope, curvature change rate, and frequency domain jitter features of each degradation feature segment, and input them into a preset fault mechanism feature library. Calculate the multidimensional spatial similarity with the standard degradation template in the library to obtain the trajectory matching similarity vector. Specifically, to accurately identify the specific fault mechanism corresponding to equipment degradation and solve the pain point that the existing operation and maintenance mode cannot distinguish between different types of hidden faults such as early bearing micro-wear, impeller dust adhesion imbalance, and valve body lubrication jamming, the obtained equipment performance degradation trajectory is divided into multiple degradation feature segments according to the previously extracted health state attenuation inflection point and rate change node. Each degradation feature segment corresponds to a specific degradation stage of the equipment, and the degradation rate and degradation features of the equipment within each segment are consistent.

[0104] For each degradation feature segment, its core degradation feature parameters are extracted, including attenuation slope, curvature change rate, and frequency domain jitter. These three features characterize the equipment degradation state from different dimensions. The attenuation slope is calculated by subtracting the real-time health index at the start of the segment from the real-time health index at the end of the segment, and then dividing by the time difference between the start and end times. The larger the absolute value of the attenuation slope, the faster the equipment degradation rate at that stage. The curvature change rate is calculated by subtracting the attenuation slope of the current degradation feature segment from the attenuation slope of the next adjacent degradation feature segment, and then dividing by the time difference between the start and end times of the next adjacent degradation feature segment. The difference is the curvature change rate of the current degradation feature segment, which is used to characterize the degree of change in the equipment degradation rate. The calculation process of the frequency domain jitter feature is as follows: first, calculate the average value of all real-time health indices within the degradation feature segment; then, subtract the average value from the real-time health index of each sampling point within the segment; square each difference obtained; and then calculate the arithmetic mean of all squared differences. This average value is the frequency domain jitter feature of the degradation feature segment, which is used to reflect the degree of fluctuation of the real-time health index at this stage. The greater the fluctuation, the more unstable the equipment degradation.

[0105] The attenuation slope, curvature change rate, and frequency domain jitter features of each extracted degradation feature segment are organized into a feature vector with a unified format. This feature vector is then input into a pre-defined fault mechanism feature library. This fault mechanism feature library is based on common latent faults in the terminal equipment of underground ventilation systems. It is formed by collecting complete degradation data of various faults from occurrence to development in advance, extracting standard degradation feature parameters, and constructing standard degradation templates. Each standard degradation template corresponds to a specific latent fault and includes the standard attenuation slope, standard curvature change rate, and standard frequency domain jitter features corresponding to that fault.

[0106] Calculate the multidimensional spatial similarity between the current deteriorated feature vector and each standard deterioration template in the fault mechanism feature library. Calculate the sum of squares of all feature parameters of the current deteriorated feature vector, and take the square root of this sum to obtain the magnitude of the current feature vector. Calculate the magnitude of the feature vector of the standard deterioration template using the same method. Divide the dot product of the two feature vectors by the product of the two magnitudes; the result is the multidimensional spatial similarity between the current deteriorated feature vector and the standard deterioration template. The closer the similarity value is to 1, the higher the matching degree; the closer it is to 0, the lower the matching degree. Arrange the multidimensional spatial similarities between the current deteriorated feature vector and all standard deterioration templates in the fault mechanism feature library in the order of the standard templates to form a trajectory matching similarity vector.

[0107] Step 4.3 involves performing extreme value retrieval and clustering on the trajectory matching similarity vector to filter out the fault mechanism category corresponding to the similarity peak. Combined with the aerodynamic-mechanical impedance characteristics of the equipment, the specific early latent fault corresponding to the current deterioration stage is analyzed. Specifically, this includes performing extreme value retrieval and clustering on the obtained trajectory matching similarity vector. The purpose is to filter out the fault mechanism category with the highest matching degree, eliminate interference terms, and ensure the accuracy of fault mechanism identification.

[0108] The specific process of extreme value retrieval is as follows: traverse all elements in the trajectory matching similarity vector, find the element with the largest value, and the standard degradation template in the fault mechanism feature library corresponding to the largest element is the fault mechanism category that best matches the current equipment degradation feature. The clustering process adopts the K-means clustering algorithm. Specifically, elements with similar values ​​in the trajectory matching similarity vector are grouped into one category, and elements with values ​​far below the average and extremely low matching degree are grouped into the interference category and removed. The clustering process further verifies the results obtained from extreme value retrieval, ensuring that the selected fault mechanism category is the type that best matches the current equipment degradation, and avoiding identification errors caused by a single extreme value.

[0109] After determining the fault mechanism category with the highest matching degree, the aerodynamic-mechanical impedance characteristics of each key monitoring node are combined to further analyze and verify the specific early latent faults corresponding to the current deterioration stage, ensuring the accuracy of fault identification and avoiding misjudgment. The specific analysis process is carried out based on the characteristics of different fault types: If the fault mechanism category obtained from extreme value retrieval and clustering is early bearing micro-wear, then the focus is on verifying the aerodynamic-mechanical impedance characteristics of the centrifugal fan bearing housing coupling end. Early bearing micro-wear will cause an increase in the mechanical impedance value of this node, thus causing an abnormal change in the frequency domain impedance weighting coefficient of this node. By comparing the current aerodynamic-mechanical impedance value of this node with the reference impedance value under the equipment's healthy state, it is confirmed whether the impedance change conforms to the characteristics of early bearing micro-wear, thus clarifying that the current fault is early bearing micro-wear; if the matched fault mechanism category is impeller dust adhesion imbalance, then the focus is on verifying the aerodynamic characteristics of the centrifugal fan impeller. Impedance characteristics were verified. Imbalanced dust adhesion on the impeller caused abnormal fluctuations in its aerodynamic impedance, and the fluctuation pattern was related to the severity of dust adhesion. By analyzing the fluctuation characteristics of the aerodynamic impedance at this node and combining it with the changing trend of the real-time health index, the specific fault type was confirmed as impeller dust adhesion imbalance, and the severity of dust adhesion was also assessed. If the matched fault mechanism category was valve body lubrication jamming, the aerodynamic-mechanical impedance characteristics of the valve plate shaft sealing area of ​​the fireproof smoke exhaust valve were used for verification. Valve body lubrication jamming caused an increase in mechanical impedance and abnormal aerodynamic impedance at this node. By comparing the difference between the reference impedance and the current impedance, the specific fault type was confirmed as valve body lubrication jamming. Through the above analysis process combining aerodynamic-mechanical impedance characteristics, the specific early latent fault corresponding to the current deterioration stage was finally accurately determined.

[0110] Step 4.4: Based on the damage evolution rate of specific early latent faults and the remaining attenuation margin of equipment performance degradation trajectory, and combined with a preset safety failure threshold, perform lifetime attenuation integral extrapolation, divide risk intervals to obtain fault warning levels, and output the predicted remaining service life of the equipment. Specifically, this includes: calculating the damage evolution rate of the specific early latent fault obtained from the analysis. The damage evolution rate is used to characterize the speed of fault development. The specific calculation process is as follows: select the characteristic energy attenuation ratio of two adjacent sampling periods, subtract the characteristic energy attenuation ratio of the previous sampling period from the characteristic energy attenuation ratio of the later sampling period to obtain the characteristic energy attenuation ratio difference, and then divide this difference by the time of the two sampling periods. The result obtained is the damage evolution rate of the latent fault. The larger the damage evolution rate, the faster the fault develops and the more rapidly the equipment deteriorates. The remaining decay margin of the equipment performance degradation trajectory is calculated. The remaining decay margin is used to characterize the remaining health space of the equipment from the current state to the failure state. The specific calculation process is to subtract the health index threshold at the time of equipment failure from the real-time health index at the current moment. The result is the remaining decay margin. The larger the remaining decay margin, the better the remaining health status of the equipment and the longer the time to failure. The smaller the remaining decay margin, the more severe the equipment deterioration and the shorter the time to failure. The health index threshold at the time of equipment failure is consistent with the preset value, which is 0.1.

[0111] Based on a preset safety failure threshold, the remaining service life of the equipment is estimated by integral degradation. The core of the integral degradation is to calculate the time required for the equipment to deteriorate from its current state to failure state based on the current damage evolution rate and the remaining degradation margin. The specific calculation process is as follows: taking the real-time health index at the current moment as the starting point of integration and the health index threshold at the time of equipment failure as the ending point of integration, the change in the real-time health index is divided by the damage evolution rate. The damage evolution rate changes with the real-time health index and is obtained by fitting the equipment performance degradation trajectory. The result of the integral calculation is the remaining service life of the equipment.

[0112] After predicting the remaining useful life, based on the remaining attenuation margin, remaining useful life, and damage evolution rate, and in conjunction with preset classification criteria, three risk zones are divided to determine the fault warning level. The classification criteria for the three risk zones and their corresponding warning levels are as follows: The first risk zone is the low-risk zone, defined as a remaining attenuation margin greater than or equal to 0.5, a remaining useful life greater than or equal to 30 days, and a damage evolution rate less than or equal to 0.01 per day. The corresponding warning level is Level 3, i.e., a routine warning, indicating that the equipment is currently only slightly degraded and the fault is developing slowly, and the equipment status can be monitored according to the routine maintenance cycle. The second risk zone is the medium-risk zone, defined as a remaining attenuation margin greater than or equal to 0.2 and less than 0.5, a remaining useful life greater than or equal to 10 days and less than 30 days, and a damage evolution rate less than or equal to 0.01 per day. The first risk zone is a high-risk zone, defined as a damage evolution rate greater than 0.01 per day and less than or equal to 0.03 per day. The second risk zone is a high-risk zone, defined as a remaining attenuation margin less than 0.2, a remaining service life less than 10 days, and a damage evolution rate greater than 0.03 per day. The third risk zone is a high-risk zone, defined as a damage evolution rate greater than 0.03 per day. The third risk zone is a high-risk zone, defined as a damage evolution rate greater than 0.03 per day. The fourth risk zone is a high-risk zone, defined as a high-risk zone. The fifth risk zone is a high-risk zone, defined as a high-risk zone. The sixth risk zone is a high-risk zone, defined as a high-risk zone. The seventh risk zone is a high-risk zone. The eighth risk zone is a high-risk zone. The eighth risk zone is a high-risk zone. The eighth risk zone is a high-risk zone. The eighth risk zone is a high-risk zone. The ninth risk zone is a high-risk zone. The eighth ...

[0113] In this embodiment of the invention, the real-time health index within a continuous time series is smoothed and denoised, and the inflection point of health state decay and the rate mutation node are extracted as segment boundaries. An adaptive segmented fitting algorithm is used to perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory. The equipment performance degradation trajectory is divided into multiple deterioration feature segments, and the decay slope, curvature change rate, and frequency domain jitter features of each segment are extracted and input into a preset fault mechanism feature library to calculate multi-dimensional spatial similarity to obtain a trajectory matching similarity vector. The fault mechanism category corresponding to the similarity peak is screened through extreme value retrieval and clustering, and the specific early latent fault is analyzed by combining the aerodynamic-mechanical impedance characteristics of the equipment. Finally, the lifetime decay integral extrapolation is performed based on the damage evolution rate of the specific early latent fault and the remaining decay margin of the equipment performance degradation trajectory, combined with a preset safety failure threshold. This technology, which obtains fault warning levels and outputs remaining service life prediction results from risk intervals, overcomes the technical problems of existing ventilation system terminal equipment health status assessment relying on a single static threshold or linear fitting model, unable to continuously and accurately depict the multi-stage nonlinear performance degradation evolution law of equipment, difficult to establish a mapping relationship between abstract trajectory change characteristics and specific early latent fault mechanisms such as bearing micro-wear or valve body jamming, and lack of a quantitative life extrapolation mechanism based on damage accumulation rate and safety margin, resulting in delayed fault warnings or frequent false alarms. It achieves high-precision reconstruction of the nonlinear trajectory of the entire process of equipment performance degradation, accurate pattern recognition and quantitative analysis of early latent fault mechanisms, and dynamic quantitative assessment of remaining service life and risk warning levels, thus improving the predictability of the state of ventilation system terminal equipment from latent degradation to the critical failure period.

[0114] In a preferred embodiment of the present invention, step 5 above may include:

[0115] Step 5.1 involves inputting the fault warning level and remaining service life prediction results into a pre-set maintenance decision matrix. This matches the corresponding equipment maintenance priorities and intervention strategies, resulting in a maintenance work order that includes fault location coordinates, hazard type identification, expected work window, and standardized handling procedures. Specifically, this includes inputting the obtained fault warning level and remaining service life prediction results into the pre-set maintenance decision matrix. This maintenance decision matrix is ​​pre-set based on the operation and maintenance needs, fault types, and equipment importance of the terminal equipment in the underground ventilation system. It pre-defines equipment maintenance priorities and intervention strategies corresponding to different fault warning levels and different remaining service lives. The matrix clearly indicates the maintenance priorities corresponding to Level 1 warning (high risk), Level 2 warning (medium risk), and Level 3 warning (low risk), as well as the intervention methods under different remaining service lives, ensuring the scientific and targeted nature of maintenance decisions.

[0116] After substituting the fault warning level and remaining service life prediction results into the maintenance decision matrix, the corresponding equipment maintenance priority and intervention strategy are automatically matched according to the matching rules within the matrix: if it is a Level 1 warning (emergency warning) with a remaining service life of less than 10 days, the highest maintenance priority is matched, and the intervention strategy is to immediately shut down for maintenance; if it is a Level 2 warning (key warning) with a remaining service life of 10 to 30 days, the medium maintenance priority is matched, and the intervention strategy is to complete the maintenance within 3 days; if it is a Level 3 warning (routine warning) with a remaining service life of 30 days or more, the ordinary maintenance priority is matched, and the intervention strategy is to complete the maintenance according to the normal operation and maintenance cycle.

[0117] After determining the maintenance priorities and intervention strategies, a complete maintenance work order is generated. All items in the work order are precisely generated based on the results of previous monitoring and analysis: fault location coordinates are taken from the spatial topological coordinates of the three key monitoring nodes determined in step 1. Based on the specific early latent faults analyzed in step 1, the coordinates of the corresponding monitoring nodes are marked to ensure maintenance personnel can quickly locate the fault; the hazard type identification directly uses the specific early latent fault names obtained from the analysis, clearly indicating the hazard type, such as early micro-wear of bearings, impeller dust adhesion imbalance, etc.; the expected operating window is based on the remaining service life prediction results and equipment operating conditions. The situation is determined to ensure that maintenance work does not affect the daily ventilation and fire smoke extraction needs of the underground space. For example, for a level-two warning with 15 days of remaining service life, the expected operation window is set to the off-peak ventilation period within 3 days. Standardized handling procedures are preset according to the corresponding hazard type identification, and specific maintenance steps are formulated for different hidden faults. For example, the standardized handling procedure for early micro-wear of bearings includes bearing cleaning, lubrication, wear detection, and precision calibration. The standardized handling procedure for impeller dust adhesion imbalance includes impeller disassembly, dust cleaning, and dynamic balance testing. This ensures that maintenance work is standardized and efficient, and avoids recurrence of faults due to inadequate maintenance.

[0118] Step 5.2: Based on the hazard type identifier in the maintenance work order and the current attenuation margin of the equipment performance degradation trajectory, calculate the airflow gain threshold and air pressure balance margin required to maintain the underground space environmental standard, and combine them to obtain the system airflow and air pressure dynamic compensation command. Specifically, based on the hazard type identifier in the maintenance work order and the current attenuation margin of the equipment performance degradation trajectory, first clarify the impact law of different hazard types on airflow and air pressure: early micro-wear of bearings will cause fan speed fluctuations, which will reduce the airflow at the outlet and make the air pressure unstable; impeller dust adhesion imbalance will cause the fan airflow to decrease and the pipeline air pressure distribution to be uneven; valve body lubrication jamming will cause the pipeline airflow resistance to increase, the air pressure to rise and the airflow to decrease. Based on these influencing patterns and the current attenuation margin of equipment performance degradation trajectories, the required airflow gain threshold and air pressure balance margin for maintaining underground space environmental standards are calculated. The calculation process for the airflow gain threshold is as follows: first, determine the minimum standard airflow required to maintain daily air exchange and emergency fire smoke extraction in the underground space; then, measure the actual airflow at the end of the current duct network; subtract the current actual airflow from the standard airflow, and the difference is the airflow gain threshold. If the current actual airflow already meets the standard, the airflow gain threshold is set to 0. The calculation process for the air pressure balance margin is as follows: first, determine the standard air pressure range for each node of the underground ventilation duct network; then, measure the actual air pressure at the end of the current duct network; subtract the current actual air pressure from the median of the standard air pressure, and the difference is the air pressure balance margin. If the current actual air pressure is within the standard range, the air pressure balance margin is set to 0; if the actual air pressure is higher than the standard range, the air pressure balance margin is set to a negative value to guide a reduction in air pressure; if the actual air pressure is lower than the standard range, the air pressure balance margin is set to a positive value to guide an increase in air pressure.

[0119] Based on the identified hazard type, the calculated airflow gain threshold and air pressure balance margin are corrected. For example, if the airflow decreases due to impeller dust accumulation imbalance, the airflow gain threshold needs to be appropriately increased, while the air pressure balance margin is adjusted to ensure stable pipeline air pressure. If the air pressure increases due to valve lubrication blockage, the air pressure balance margin needs to be reduced, while the airflow gain threshold is fine-tuned to avoid excessive airflow fluctuations. The corrected airflow gain threshold and air pressure balance margin are combined to clearly indicate the adjustment range of the fan frequency converter speed regulation and the adjustment angle of the valve opening, ultimately forming a dynamic compensation command for system airflow and air pressure.

[0120] Step 5.3 involves encoding the maintenance work order and the system's dynamic airflow and pressure compensation instructions using communication protocols and encapsulating their execution timing. This information is then sent to the centralized control cabinet to synchronously trigger fan frequency conversion speed regulation, valve opening adjustment, and work order push notifications to the maintenance terminal. Specifically, since the centralized control cabinet for the underground ventilation system uses an industrial-grade communication protocol, the text and parameter information of the maintenance work order and compensation instructions must be converted into an encoding format that the centralized control cabinet can recognize and parse. During the encoding process, it is crucial to ensure that key information such as fault location coordinates, hazard type identification, airflow gain threshold, and air pressure balance margin are not lost or distorted. Simultaneously, the encoding must be compatible with the communication rate and data transmission format of the centralized control cabinet to avoid instruction transmission failures or parsing errors, ensuring accurate instruction delivery. Based on the operating logic of the underground ventilation system, the order of maintenance work order push notifications and compensation instruction execution is rationally arranged to avoid mutual interference and ensure stable system operation. The specific execution sequence is as follows: First, the maintenance work order code information is pushed to the maintenance personnel's terminal to ensure that the maintenance personnel receive the maintenance tasks in a timely manner and understand the fault situation and operation requirements; then, the system air volume and air pressure dynamic compensation instruction code information is sent to the centralized control cabinet, triggering the centralized control cabinet to perform the corresponding adjustment operation; finally, the centralized control cabinet synchronously triggers the fan frequency conversion speed regulation and valve opening adjustment, adjusts the fan speed according to the air volume gain threshold in the compensation instruction, and adjusts the valve opening of equipment such as fireproof smoke exhaust valves and static pressure boxes according to the air pressure balance margin, to ensure that the air volume and air pressure of the pipeline network quickly reach the compensation target and maintain the normal operation of the system.

[0121] After encapsulation, the encoded maintenance work order and compensation instruction are sent to the centralized control cabinet via industrial communication lines. After receiving the instruction, the centralized control cabinet immediately parses the instruction content and simultaneously executes operations such as pushing maintenance work orders to the operation and maintenance terminal, adjusting the frequency conversion speed of the fan, and adjusting the valve opening, so as to achieve the coordinated advancement of maintenance tasks and compensation adjustments.

[0122] Step 5.4: Real-time acquisition of the pipeline end pressure feedback value and equipment response current after the execution of the system's dynamic airflow and pressure compensation command, and comparison and verification of the deviation with the expected compensation target. If the deviation exceeds the allowable threshold, the compensation parameters are iteratively corrected and reissued. If the deviation is within the allowable threshold, the current intervention parameters are locked and the equipment health status baseline is updated, completing the predictive maintenance and closed-loop health management of the ventilation system's end equipment. Specifically, after the system's dynamic airflow and pressure compensation command is executed, two types of core feedback data are collected in real time through the sensors of the underground ventilation system: one is the pipeline end pressure feedback value, used to determine whether the pipeline pressure has reached the expected target after the compensation command is executed; the other is the equipment response current, used to determine whether the operating status of equipment such as fans and valves is normal after compensation adjustment, avoiding excessive equipment load and further deterioration due to over-adjustment.

[0123] After collecting feedback data, it is compared and verified against preset expected compensation targets. These expected compensation targets are the feedback values ​​corresponding to the determined standard air volume and standard air pressure, as well as the response current range during normal equipment operation. The specific process for deviation comparison and verification is as follows: the difference between the air pressure feedback value at the end of the pipeline and the expected air pressure target, and the difference between the equipment response current and the expected current range are calculated respectively. The absolute values ​​of the two differences are taken. If both absolute values ​​are less than or equal to preset allowable thresholds, which are calibrated according to equipment performance and environmental standards, the allowable air pressure threshold is ±5% of the standard air pressure, and the allowable current threshold is ±3% of the equipment's rated current. This indicates that the compensation command has been executed effectively, the equipment is operating well, and no adjustment of intervention parameters is required.

[0124] If the absolute value of any difference exceeds the allowable threshold, it indicates that the compensation command execution effect has not met expectations. This may be due to a deviation in the judgment of the impact of the hazard type or inaccurate calculation of the compensation parameters. In this case, the compensation parameter iterative correction process is immediately triggered: based on the magnitude of the deviation, the air volume gain threshold and the air pressure balance margin are adjusted. For example, if the air pressure feedback value is lower than the expected target and exceeds the allowable threshold, the air pressure balance margin is appropriately increased; if the equipment response current is higher than the expected range and exceeds the allowable threshold, the air volume gain threshold is appropriately decreased, and the fan speed is reduced. After the correction is completed, the adjusted compensation command is re-encoded with the communication protocol and encapsulated with the execution timing, and sent to the centralized control cabinet to execute the compensation adjustment operation again until the deviation between the feedback data and the expected compensation target is within the allowable threshold.

[0125] When the deviation is within the allowable threshold, the current intervention parameters are locked, namely the final airflow gain threshold, air pressure balance margin, fan speed, valve opening, etc., while the equipment health status baseline is updated. The equipment health status baseline update process involves using the current equipment operating parameters, including vibration signal, air pressure, current, speed, and real-time health index, as the new health status benchmark to replace the original health baseline. This new baseline is used for subsequent equipment degradation trend analysis, fault warning, and feature comparison to ensure the accuracy of subsequent monitoring and analysis. It adapts to the health status of the equipment after compensation and adjustment. From equipment operating parameter acquisition, feature extraction, fault warning, and maintenance decision-making to compensation adjustment, effect verification, and baseline update, a complete predictive maintenance and closed-loop health management process is formed.

[0126] In this embodiment of the invention, the fault warning level and the predicted remaining service life are input into a pre-set maintenance decision matrix to generate a structured maintenance work order. Based on the hidden danger type identifier and the current attenuation margin of the equipment performance degradation trajectory, the air volume gain threshold and the air pressure balance margin are calculated to combine the system's air volume and air pressure dynamic compensation command. After communication protocol encoding and timing encapsulation, the command is sent to the centralized control cabinet to simultaneously trigger the fan frequency conversion speed regulation and valve opening adjustment. The method also includes real-time collection of the air pressure feedback value at the end of the pipeline network and the equipment response current for deviation comparison and verification to trigger iterative correction of compensation parameters or lock intervention parameters and update the health status baseline. Therefore, this method overcomes the problem of mutual interference between fault warning and actual control execution in the operation and maintenance of traditional ventilation systems. The technical problems of disconnection, lack of dynamic air volume and pressure adaptive compensation mechanism based on equipment degradation status, lack of closed-loop feedback verification in the intervention process which easily leads to compensation failure or over-adjustment, and difficulty in continuously ensuring ventilation efficiency and fire emergency smoke exhaust standards in underground spaces during equipment operation with defects or maintenance windows have been addressed. This has resulted in the achievement of deep collaboration between predictive maintenance decision-making and system operation control, and the construction of a closed-loop health management system covering the entire chain of early warning, dynamic compensation, real-time feedback, parameter iteration, and baseline update. This has maintained the dynamic operational balance and emergency response capability of the ventilation network during the equipment performance degradation period, improved the accuracy of operation and maintenance intervention and the reliability of system operation, and reduced unplanned downtime losses and environmental safety management risks.

[0127] like Figure 2 As shown, embodiments of the present invention also provide a fault prediction and health management system for terminal equipment of a ventilation system, comprising:

[0128] The module is used to extract key time-frequency features that characterize the mechanical operation of the fan and the airflow state in the pipeline by using the standardized operational dataset as the processing object; and to fit and construct a three-dimensional spatial mapping ellipsoid based on the spatial distribution coordinates of three key monitoring nodes and the aeromechanical impedance characteristics of each node.

[0129] The calculation module is used to perform adaptive mesh generation and numerical simulation of flow-vibration energy transfer on a three-dimensional spatially mapped ellipsoid along the principal axis, and obtain spatial phase coupling correction values.

[0130] The analysis module is used to perform phase alignment and energy weight calibration of key time-frequency features through spatial phase coupling correction values, complete multi-dimensional feature fusion, and construct a dynamic operating feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operating feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained;

[0131] The prediction module is used to extrapolate the nonlinear trend along the continuous time axis using the real-time health index as the fitting benchmark to obtain the equipment performance degradation trajectory; the equipment performance degradation trajectory is matched and mapped with the preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and the fault warning level and remaining service life prediction results are deduced.

[0132] The compensation module is used to obtain maintenance work orders and dynamic compensation instructions for system air volume and air pressure based on the fault warning level and the prediction results of the remaining service life, so as to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

[0133] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0134] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0135] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0136] I. Overview of Experimental Conditions

[0137] This experimental example uses the centralized ventilation system of an underground parking garage in a large commercial complex as the simulation object, selecting the coupling end of the centrifugal fan bearing housing, the sealing area of ​​the fireproof smoke exhaust valve plate shaft, and the outlet surface of the static pressure box guide array as three key monitoring nodes. The operational data acquisition frequency is 1kHz, continuously collecting operational parameters for 180 days, covering dimensions such as bearing vibration acceleration, fan speed, outlet air pressure, motor current, pipe temperature, and airflow velocity. The three principal axis parameters of the three-dimensional spatially mapped ellipsoid are obtained by solving the eigenvalues ​​and eigenvectors of the multi-source feature covariance matrix. Adaptive mesh generation is performed along the principal axis directions according to the impedance gradient distribution, with a total of 5000 meshes. The fault mechanism feature library includes five standard degradation templates: bearing wear, impeller imbalance, valve body lubrication jamming, belt loosening, and motor insulation aging. The real-time health index is calculated using a nonlinear decay function, with a safety failure threshold set at 40 points. The predictive maintenance decision matrix matches maintenance priorities and intervention strategies based on the fault warning level (Level 1 Attention to Level 4 Severe).

[0138] Step 3, as follows Figure 3 and Figure 4 As shown (Fault Early Warning and Predictive Maintenance): Using a real-time health index of 62 points on day 170 as the fitting benchmark, nonlinear trend extrapolation is performed along the continuous time axis, predicting that the equipment will drop to the severe fault line (40 points) around day 195. The performance degradation trajectory is matched with the fault mechanism feature library, and the specific early latent fault corresponding to the current deterioration stage is identified as "early bearing micro-wear," with a fault early warning level determined as Level 3 (Warning). Maintenance work orders are generated based on the fault early warning level (including bearing housing coupling end positioning coordinates, bearing micro-wear hazard identification, and a 25-day expected operation window). Simultaneously, the system airflow gain threshold and air pressure balance margin are calculated, and dynamic airflow and air pressure compensation commands are generated and sent to the centralized control cabinet. During 180 days of operation, a total of 22 alarms were triggered, including 12 Level 1 alerts, 6 Level 2 warnings, 3 Level 3 warnings, and 1 Level 4 severe alarm.

[0139] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for fault prediction and health management of terminal equipment in a ventilation system, characterized in that, The method includes: Step 1: Real-time acquisition of multi-dimensional operating parameters of the terminal equipment of the underground space ventilation system during continuous operation. Three key monitoring nodes are selected: the coupling end of the centrifugal fan bearing seat, the sealing area of ​​the fireproof smoke exhaust valve plate shaft, and the outlet surface of the static pressure box guide array. The spatial topological coordinates and physical boundary attributes of each node are recorded synchronously. The multi-dimensional operating parameters are then subjected to spatiotemporal alignment, noise filtering, and dimensional normalization to obtain a standardized operating dataset. An adaptive time-frequency decomposition operation is performed on the standardized operating dataset to separate the mechanical vibration frequency band and the pneumatic vibration frequency band. The instantaneous amplitude envelope, dominant frequency offset, and phase difference information of each frequency band are extracted and combined to obtain key time-frequency characteristics representing the mechanical operation of the fan and the airflow state in the duct. The key time-frequency feature vector set is transformed with the spatial distribution coordinates of three key monitoring nodes to obtain a spatial reference vector. The aerodynamic-mechanical impedance characteristics of each node are quantized into frequency domain impedance weighting coefficients. The key time-frequency feature vector set is weighted and projected using the impedance weighting coefficients to obtain a multi-source feature covariance matrix. Using the multi-source feature covariance matrix as the processing object, eigenvalues ​​and eigenvectors are extracted as spatial topological constraints. The principal axis direction parameters and semi-axis length parameters of the ellipsoid quadratic surface equation are solved. The principal axis direction parameters and semi-axis length parameters are mapped to a three-dimensional Cartesian coordinate system, and a three-dimensional spatial mapping ellipsoid characterizing the boundary of the coupled operating state of the equipment is fitted and constructed. Step 2: Adaptive mesh generation and flow-vibration energy transfer numerical simulation are performed on the three-dimensional spatially mapped ellipsoid along the principal axis to obtain the spatial phase coupling correction value; Step 3: Phase alignment and energy weight calibration of key time-frequency features are performed using spatial phase coupling correction values ​​to complete multi-dimensional feature fusion and construct a dynamic operation feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operation feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained; Step 4: Using the real-time health index as the fitting benchmark, perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory; match and map the equipment performance degradation trajectory with the preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and deduce the fault warning level and remaining service life prediction results. Step 5: Based on the fault warning level and the predicted remaining service life, maintenance work orders and dynamic compensation instructions for system air volume and air pressure are obtained to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

2. The method for fault prediction and health management of terminal equipment in a ventilation system according to claim 1, characterized in that, Step 2: Adaptive mesh generation and flow-vibration energy transfer numerical simulation are performed on the three-dimensional spatially mapped ellipsoid along the principal axis to obtain spatial phase coupling correction values, including: Along the principal axis of the three-dimensional spatially mapped ellipsoid, a variable density adaptive meshing operation is performed based on the aerodynamic-mechanical impedance gradient distribution of the nodes to obtain a three-dimensional discrete mesh topology. The airflow pulsation load and mechanical excitation load are applied to the mesh nodes as virtual excitation sources to perform a numerical simulation of flow-vibration energy transfer and obtain the set of multi-directional normal stress and shear stress components of each mesh element section under transient conditions. Using the set of multi-directional normal stress and shear stress components as input data, a time-series stress state evolution sequence is constructed, the stress extreme point sequence is extracted, and the coordinates are mapped to a two-dimensional stress polar coordinate system to obtain the stress state geometric transformation trajectory of the periodic evolution of stress state. Geometric feature tracing is performed on the stress state geometric transformation trajectory to extract the spatiotemporal translation vector of the trajectory center coordinates and the extreme envelope of the trajectory radius; based on the geometric mapping relationship between the spatiotemporal translation vector and the extreme envelope, the principal stress amplitude distribution field and the spatial tilt tensor of the maximum shear stress surface of each grid element are solved in reverse. Substituting the principal stress amplitude distribution field and the spatial tilt tensor into the material fatigue constitutive relation, and combining the fluid-structure interaction boundary conditions, the mechanical vibration phase lag angle and the aerodynamic response energy attenuation coefficient are calculated; tensor synthesis and normalization operations are performed on the mechanical vibration phase lag angle and the aerodynamic response energy attenuation coefficient to obtain the spatial phase coupling correction value.

3. The method for fault prediction and health management of terminal equipment in a ventilation system according to claim 2, characterized in that, Step 3: Phase alignment and energy weight calibration of key time-frequency features are performed by spatial phase coupling correction values ​​to complete multi-dimensional feature fusion and construct a dynamic operation feature sequence of the equipment; By performing sliding window time-series evolution analysis on the dynamic operating characteristic sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained, including: The phase lag compensation parameter and energy attenuation weight parameter are extracted from the spatial phase coupling correction value. The phase lag compensation parameter is applied to the mechanical vibration phase angle and pneumatic pulsation phase angle of key time-frequency characteristics to achieve phase alignment between the fan operation signal and the pipeline airflow signal. A multi-band energy allocation matrix is ​​constructed using energy attenuation weight parameters. The key time-frequency features after phase alignment are weighted and superimposed with the energy allocation matrix according to the feature frequency band to complete the multi-dimensional feature fusion and obtain the dynamic operation feature sequence of the equipment. Set a sliding time window with a fixed time span and step interval, input the dynamic operation feature sequence of the equipment into the sliding time window in sequence, extract the statistical moments and frequency domain energy ratio of the feature vectors in each time window, calculate the distance divergence and evolution difference rate of the feature vectors of adjacent time windows, and construct the temporal evolution difference matrix. Principal component extraction is performed on the temporal evolution difference matrix to obtain the characteristic energy decay ratio in the principal evolution direction. The characteristic energy decay ratio is then mapped to the standard health baseline to obtain the real-time health index of the current degradation level of the device through a nonlinear decay function.

4. The method for fault prediction and health management of terminal equipment in a ventilation system according to claim 3, characterized in that, Step 4: Using the real-time health index as the fitting benchmark, perform nonlinear trend extrapolation along the continuous time axis to obtain the equipment performance degradation trajectory; match and map the equipment performance degradation trajectory with the preset fault mechanism feature library to analyze the specific early latent faults corresponding to the current deterioration stage, and deduce the fault warning level and remaining service life prediction results, including: The real-time health index in the continuous time series is smoothed and denoised, and the inflection point of health status decay and the node of rate change are extracted. Using the decay inflection point as the segment boundary, nonlinear trend extrapolation is performed along the continuous time axis to obtain the equipment performance degradation trajectory. The equipment performance degradation trajectory is divided into multiple degradation feature segments. The attenuation slope, curvature change rate and frequency domain jitter features of each degradation feature segment are extracted and input into a preset fault mechanism feature library. The multidimensional spatial similarity with the standard degradation template in the library is calculated to obtain the trajectory matching similarity vector. Extreme value retrieval and clustering are performed on the trajectory matching similarity vector to filter out the fault mechanism category corresponding to the similarity peak. Combined with the aerodynamic-mechanical impedance characteristics of the equipment, the specific early latent faults corresponding to the current deterioration stage are analyzed. Based on the damage evolution rate of specific early latent faults and the remaining attenuation margin of equipment performance degradation trajectory, combined with the preset safety failure threshold, life attenuation integral extrapolation is performed to divide risk intervals to obtain fault warning levels, and the predicted results of the remaining service life of the equipment are output.

5. The method for fault prediction and health management of terminal equipment in a ventilation system according to claim 4, characterized in that, Step 5, based on the fault warning level and remaining service life prediction results, generates maintenance work orders and system airflow and pressure dynamic compensation instructions to complete predictive maintenance and closed-loop health management of the ventilation system's terminal equipment, including: The fault warning level and the remaining service life prediction results are input into the preset maintenance decision matrix, and the corresponding equipment maintenance priority and intervention strategy are matched to obtain a maintenance work order that includes fault location coordinates, hidden danger type identification, expected operation window period and standardized handling process. Based on the hazard type identifier in the maintenance work order and the current attenuation margin of the equipment performance degradation trajectory, the air volume gain threshold and air pressure balance margin required to maintain the underground space environmental standard are calculated, and the system air volume and air pressure dynamic compensation command is obtained by combining them. The maintenance work order and the system air volume and air pressure dynamic compensation instruction are encoded with communication protocol and encapsulated with execution timing, and sent to the centralized control cabinet to synchronously trigger the fan frequency conversion speed regulation, valve opening adjustment and maintenance terminal work order push. The system collects the feedback value of the air pressure at the end of the pipeline and the response current of the equipment after the execution of the dynamic compensation command for air volume and air pressure in real time. It then compares and verifies the deviation with the expected compensation target. If the deviation exceeds the allowable threshold, it triggers the iterative correction of the compensation parameters and reissues them. If the deviation is within the allowable threshold, it locks the current intervention parameters and updates the baseline of the equipment health status, thus completing the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

6. A fault prediction and health management system for terminal equipment of a ventilation system, wherein the system implements the method as described in any one of claims 1 to 5, characterized in that, include: The module is used to extract key time-frequency features that characterize the mechanical operation of the fan and the airflow state in the pipeline by using the standardized operational dataset as the processing object. Based on the spatial distribution coordinates of three key monitoring nodes and the aero-mechanical impedance characteristics of each node, a three-dimensional spatial mapping ellipsoid is fitted and constructed. The calculation module is used to perform adaptive mesh generation and numerical simulation of flow-vibration energy transfer on a three-dimensional spatially mapped ellipsoid along the principal axis, and obtain spatial phase coupling correction values. The analysis module is used to perform phase alignment and energy weight calibration of key time-frequency features through spatial phase coupling correction values, complete multi-dimensional feature fusion, and construct a dynamic operating feature sequence of the equipment; by performing sliding window time-series evolution analysis on the dynamic operating feature sequence of the equipment, a real-time health index of the current degree of equipment degradation is obtained; The prediction module is used to extrapolate the nonlinear trend along a continuous time axis using a real-time health index as a fitting benchmark to obtain the trajectory of equipment performance degradation. The equipment performance degradation trajectory is matched and mapped with a preset fault mechanism feature library to analyze the specific early hidden faults corresponding to the current deterioration stage, and the fault warning level and remaining service life prediction results are deduced. The compensation module is used to obtain maintenance work orders and dynamic compensation instructions for system air volume and air pressure based on the fault warning level and the prediction results of the remaining service life, so as to complete the predictive maintenance and closed-loop health management of the terminal equipment of the ventilation system.

7. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images