Tunnel fan health diagnosis and early warning system and method based on multi-source fusion

Abstract

The invention relates to the technical field of tunnel monitoring, and discloses a tunnel fan health diagnosis and early warning system and method based on multi-source fusion, comprising a sensing layer, a data acquisition and transmission layer, a data processing and fusion layer, an intelligent diagnosis and early warning layer and an application layer which are arranged in sequence from bottom to top by adopting a layered architecture, all the layers cooperate to achieve tunnel fan collision safety and equipment health two-dimensional management and control. Vibration, displacement, strain, temperature and power multi-source monitoring data are fused, single-parameter monitoring limitation is broken through, wind turbine typical faults such as bearing damage, coupling misalignment and foundation looseness and specific risks such as over-limit vehicle collision are covered, and a full-dimension monitoring system is constructed. By means of a deep belief network (DBN), a convolutional neural network (CNN) and an MATLAB core algorithm, and in combination with a fault-feature mapping library and a multi-evidence combined judgment mechanism, the fault diagnosis accuracy is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel monitoring technology, specifically to a tunnel ventilation fan health diagnosis and early warning system and method based on multi-source fusion. Background Technology

[0002] As the core equipment of the tunnel ventilation system, the safe and stable operation of tunnel ventilation fans is directly related to air circulation, fire emergency response, and the safety of life and property of passing vehicles and personnel. With the expansion of urban tunnel construction and the increase in service life, especially in densely populated cities with high traffic volume (such as the area where the Jiangyin Jingjiang Yangtze River Tunnel is located), tunnel ventilation fans face complex operating environments such as high temperature, high humidity, and vibration interference for a long time. At the same time, they are subjected to multiple effects such as vehicle piston wind, their own mechanical wear, and structural loads, which can easily lead to typical failures such as bearing damage, coupling misalignment, rotor imbalance, and foundation loosening. There is also a unique risk of fan displacement and falling caused by collisions with oversized vehicles, which seriously threatens the operational safety of the tunnel.

[0003] The current operation and maintenance management model for tunnel ventilation fans still has many technical limitations, making it difficult to meet the needs of intelligent and refined safety assurance. Limitations of traditional monitoring methods: Currently, the maintenance of tunnel ventilation fans mostly relies on manual periodic inspections, post-incident repairs, or simple single-parameter threshold alarms (such as excessive vibration). This approach suffers from problems such as high blindness, long inspection cycles, and strong subjectivity. It is difficult to detect sudden faults and hidden dangers, cannot provide early warnings of potential faults, and cannot assess the trend of equipment performance degradation. This results in delayed fault detection and insufficient early warning capabilities, which has become a bottleneck in improving the level of tunnel safety management.

[0004] The limitations of single-sensor monitoring: Existing monitoring methods mostly rely on a single sensor to collect single parameters such as vibration, temperature, or current, lacking collaborative analysis of multi-source data. However, wind turbine failures are often caused by the coupling of multiple factors. For example, wind turbine misalignment can cause changes in vibration and may also lead to an increase in current. Relying solely on single-parameter analysis is prone to misjudgment or missed detection due to incomplete information, making it difficult to accurately determine the type and root cause of complex faults.

[0005] Data silos: While some monitoring systems can collect various types of data, these data are stored independently and managed in a scattered manner, without effective correlation and comprehensive analysis, forming "information silos." This makes it impossible to form a unified understanding of the overall health status of the wind turbine through multi-dimensional data fusion, and to comprehensively reflect the equipment's operating status and safety risks.

[0006] Lack of predictive maintenance capabilities: Most existing technologies can only perform "condition monitoring," meaning they can only issue alarms after a failure occurs or detect problems during inspections, and cannot achieve "health prediction." They cannot detect early signs of failure in advance, nor can they predict the remaining service life of critical components and future failure risks, leading to reactive maintenance work and making predictive maintenance difficult.

[0007] With the deepening of the "Transportation Powerhouse" and "Smart City" strategies, higher requirements have been placed on the intelligent operation and maintenance and full life-cycle safety management of infrastructure. Therefore, developing a comprehensive, intelligent, and predictive tunnel ventilation fan health monitoring solution to overcome existing technological limitations and achieve accurate diagnosis, early warning, and remaining life prediction of fan failures has become an urgent need in the field of tunnel operation safety assurance. Summary of the Invention

[0008] To address the aforementioned shortcomings of existing technologies, this invention provides a tunnel ventilation fan health diagnosis and early warning system and method based on multi-source fusion.

[0009] This invention provides the following technical solution: a tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion, which includes a layered architecture, from bottom to top: a perception layer, a data acquisition and transmission layer, a data processing and fusion layer, an intelligent diagnosis and early warning layer, and an application layer. Each layer works together to achieve dual-dimensional control of tunnel ventilation fan collision safety and equipment health. Sensing layer: Deploy vibration acceleration sensors, displacement sensors, strain gauges, temperature sensors, current transformers and power metering and control instruments to collect parameters such as fan vibration, tilt displacement, structural strain, temperature and three-phase voltage, current and power, covering the monitoring needs of oversized vehicle collisions, tunnel fans with loose foundations, and mechanical and electrical faults. Data acquisition and transmission layer: includes field instrument boxes, switches and Category 6a shielded cables. The instrument boxes integrate dual power supply modules, communication control modules and function modules, and are deployed in the middle pipe gallery to realize synchronous acquisition, preprocessing and wired transmission of multi-source data, and support seamless integration with the existing tunnel integrated monitoring system. Data processing and fusion layer: Employs spatiotemporal alignment and feature fusion algorithms to unify data timestamps, establish a mapping relationship between sensor locations and fan bearings, impellers, and hoisting structures, extract key features such as vibration peak value, kurtosis, temperature rise rate, and current effective value, and construct a comprehensive health feature vector; Intelligent Diagnosis and Early Warning Layer: Built-in fault diagnosis model based on Deep Belief Network (DBN) and Convolutional Neural Network (CNN) is used to analyze the comprehensive health feature vector, automatically identify various fault types such as bearing damage, coupling misalignment, rotor imbalance, foundation looseness and uneven motor air gap, evaluate the comprehensive health index of the equipment, and generate graded early warning information in combination with an adaptive threshold mechanism; Application layer: Includes a computer web client and a mobile APP client based on B / S architecture, providing machine overview, monitoring center, early warning center, advanced analysis and intelligent diagnosis function modules, supporting equipment status visualization, early warning processing, in-depth map analysis, diagnostic report generation and export, as well as macro-display on a smart screen based on geographic information system.

[0010] Preferably, each wind turbine is equipped with two vibration acceleration sensors, which are fixedly installed on the horizontal and vertical directions of the wind turbine casing by welding brackets, respectively; each wind turbine is equipped with two displacement sensors, which are deployed near the wind turbine tower to monitor the overall tilt displacement of the wind turbine; each wind turbine is equipped with eight strain gauges, which are installed on the surface of the concrete support structure near the wind turbine embedded parts; the temperature sensor adopts a digital temperature chip, which is attached to the wind turbine motor casing and fixed with high-temperature adhesive, and covered with thermal insulation cotton; the power monitoring equipment includes current transformers and power metering and control instruments installed in the wind turbine distribution box to collect three-phase voltage, current, power, power factor and power parameters.

[0011] Preferably, the method for performing data fusion in the data processing and fusion layer specifically includes: Time synchronization: Establishing a unified time reference for multi-source data such as vibration, temperature, and electricity; Spatial correlation: Establish a mapping relationship between the installation location of each sensor and the key components of the wind turbine bearings, impellers, and hoisting structure; Feature extraction: Extract kurtosis and waveform indices from vibration signals, calculate temperature rise rate from temperature data, and extract harmonic features from power data; Feature fusion: The extracted multi-dimensional features are integrated into a comprehensive health feature vector through a weighted fusion algorithm.

[0012] Preferably, the intelligent diagnosis and early warning layer is configured to perform the following core functions; Fault diagnosis: Based on multi-parameter correlation analysis and a pre-set fault-feature mapping library, it automatically identifies and locates various fault types and locations, including bearing damage, rotor imbalance, coupling misalignment, and foundation loosening. Health assessment and prediction: Based on time series analysis of equipment performance degradation trends, calculate and output a comprehensive health index on a scale of 0-100, and combine it with degradation models to predict the remaining service life of key components; Tiered early warning: Establish multi-level early warning thresholds including low reporting, high reporting, and very high reporting. When the monitored value exceeds the limit, the corresponding level of early warning will be automatically triggered, and early warning information including fault details, trend charts and handling suggestions will be pushed. Operation and maintenance decision support: Automatically generates operation and maintenance recommendation reports containing maintenance measures and spare parts lists based on diagnostic conclusions, and provides maintenance priority ranking for multi-device operation and maintenance tasks.

[0013] Preferably, the application layer is specifically configured to provide multi-terminal collaborative interactive functions, including: supporting synchronous access via computer web terminal and mobile APP terminal, and displaying corresponding device status, early warning information and diagnostic reports based on user roles and permission control; in the monitoring center module, providing two device view display modes, list and grid, and supporting multi-view viewing based on three-dimensional model and full display of measurement point distribution; providing early warning information and diagnostic report export function, supporting the generation and export of PDF or Word format documents containing fault details, analysis graphs and maintenance suggestions.

[0014] Preferably, the fault diagnosis model in the intelligent diagnosis and early warning layer adopts a multi-evidence joint judgment and confidence assessment mechanism, specifically configured as follows: Establish a database of typical fault characteristics, including bearing failure, coupling misalignment, rotor imbalance, loose foundation, and motor electrical faults; During the diagnosis process, vibration spectrum characteristics, temperature change trends, current harmonic components and structural strain data are analyzed simultaneously, and multi-source parameters are used as correlation evidence for fusion judgment. Based on the deep belief network or convolutional neural network algorithm, the occurrence confidence of each potential fault type is calculated, and a diagnostic conclusion with a confidence level higher than a preset threshold is output. The diagnostic conclusions are linked and stored with the corresponding historical cases, feature maps, and indicator contribution analyses to support manual review and continuous optimization of the model algorithm.

[0015] Preferably, it also includes a data management optimization module, used to continuously optimize system performance based on historical operation and maintenance data, receive and store early warning review results, manual processing records and fault repair feedback data from the application layer; use the early warning review results and processing records to iteratively adjust the adaptive threshold and judgment parameters of the fault diagnosis model in the intelligent diagnosis and early warning layer through machine learning; and dynamically update the monitoring weights of each sensor and the performance degradation model parameters of key components according to the long-term operation data and operating condition changes of the wind turbine.

[0016] The method for health diagnosis and early warning of tunnel ventilation fans based on multi-source fusion is as follows: S1: Multi-source data acquisition By deploying vibration acceleration sensors, displacement sensors, strain gauges, temperature sensors and current transformers on the tunnel ventilation fan, multi-source heterogeneous data on the fan's vibration, tilt displacement, structural strain, operating temperature and electrical parameters are collected simultaneously. S2: Data Preprocessing and Feature Extraction The collected multi-source heterogeneous data is time-aligned and preprocessed locally to extract key feature parameters, including vibration peak value, kurtosis, temperature change rate, current RMS value and harmonic characteristics. S3: Feature Fusion and Comprehensive Health Status Assessment The extracted key feature parameters are spatially correlated and fused to generate a comprehensive health feature vector characterizing the overall operating status and structural safety of the wind turbine. Based on a pre-set fault-feature mapping library and machine learning model, the comprehensive health feature vector is analyzed to identify whether the wind turbine has bearing failure, coupling misalignment, rotor imbalance, loose foundation or motor electrical abnormality, and the comprehensive health index of the equipment is calculated. S4: Intelligent Early Warning and Operation and Maintenance Decision Generation Based on the analysis results and preset adaptive warning thresholds, a graded warning information is generated, which includes fault type, location, severity level and confidence level; based on the diagnostic conclusions, an operation and maintenance decision report is automatically generated, which includes maintenance measures, spare parts recommendations and maintenance priorities. S5: Information Visualization and Interactive Processing The comprehensive health index, early warning information, diagnostic conclusions, and operation and maintenance decision reports are visualized and pushed through computer terminals or mobile APPs; the system receives and processes user verification and confirmation of early warning information, status updates, and maintenance record entry operations, and feeds the processing results back to the system to optimize subsequent diagnostic and early warning models.

[0017] Compared with the prior art, the present invention has the following beneficial effects: (1) By integrating multi-source monitoring data of vibration, displacement, strain, temperature and power, the limitations of single-parameter monitoring are overcome, covering typical wind turbine faults such as bearing damage, coupling misalignment, and foundation loosening, as well as unique risks such as collisions with oversized vehicles, and constructing a full-dimensional monitoring system. With the help of deep belief networks (DBN), convolutional neural networks (CNN) and MATLAB core algorithms, combined with fault-feature mapping library and multi-evidence joint judgment mechanism, the accuracy of fault diagnosis is greatly improved, and the fault type, location and severity level can be accurately identified, effectively avoiding false alarms and missed alarms.

[0018] (2) Abandoning the traditional passive alarm mode, the system captures early signs of failure from equipment operation data through time series analysis and performance degradation models, calculates a comprehensive health index on a scale of 0-100, and accurately predicts the remaining service life of key components. This enables a shift from post-maintenance and periodic inspections to proactive prediction and preventative maintenance, avoiding the risk of worsening failures in advance and providing a scientific basis for the formulation of operation and maintenance plans.

[0019] (3) Build a multi-terminal collaborative platform for computer Web terminal and mobile APP terminal, integrating integrated functional modules such as machine overview, intelligent diagnosis, and advanced analysis, supporting real-time visualization of equipment status, automatic push of early warning information, and one-click export of diagnostic reports; the system automatically completes data collection, feature extraction, fault analysis and maintenance suggestion generation, greatly reducing the reliance on human experience, improving maintenance response efficiency by more than 50%, and alleviating the work pressure of maintenance personnel.

[0020] (4) Through precise early warning and predictive maintenance, losses caused by unplanned downtime and excessive maintenance can be avoided, the service life of key components of the fan can be extended, and the operation and maintenance costs of the equipment throughout its entire life cycle can be significantly reduced. At the same time, it can effectively prevent traffic accidents and secondary disasters caused by fan falls and downtime due to malfunctions, ensure the continuous and stable operation of the tunnel ventilation system, build a solid defense for the safety of drivers and passengers and the smooth flow of urban traffic, and achieve significant social benefits. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the wind turbine fault diagnosis system of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. In order to keep the following description of the embodiments of this disclosure clear and concise, detailed descriptions of known functions and known components are omitted to avoid unnecessarily obscuring the concept of the present invention.

[0023] This implementation uses the jet fan monitoring of the right tunnel of the Jiangyin Jingjiang Yangtze River Tunnel as an application scenario. The tunnel is approximately 2.66km long, with a single-tube clear width of 13.6m and a passage clearance of 4.5m. The fans are suspended jet fans that operate in a complex environment of high temperature, high humidity (relative humidity ≤95%), and vibration interference, and must strictly comply with the requirements of the standard "Vibration Detection and Limits of Ventilation Fans" (JB / T8689-2014). The system covers 65 jet fans, each equipped with a complete sensing, acquisition, transmission, and analysis unit, and is ultimately connected to the existing integrated monitoring platform of the tunnel for unified management and control.

[0024] See Figure 1 It can be seen that the tunnel ventilation health diagnosis and early warning system based on multi-source fusion consists of [the following components].

[0025] (I) Perception Layer: Sensor Deployment and Selection The sensing layer deploys sensors based on a two-dimensional approach: "structural safety" and "equipment health." Specific selection and installation requirements are as follows: Vibration acceleration sensor Selection: The HD-YD-226 piezoelectric accelerometer is adopted, with a sensitivity of 100mV / g, a frequency response of 1Hz~9KHz, a range of ±70g, a protection rating of IP67, an operating temperature of -40℃~+120℃, and an output method of 5 / 8-24UNF two-core socket with two-core shielded cable. Deployment: Two sensors are configured for each fan, which are fixed to the fan casing horizontally and vertically by welding brackets. Before installation, the surface of the casing is polished to the metal color, and the flatness of the bracket is checked with a level to ensure that the sensor is in close contact with the surface of the equipment. The sensor collects the fan vibration signal to identify faults such as bearing wear and impeller imbalance.

[0026] Displacement sensor Selection: Linear range 25mm, linear range 3~28mm, power supply -24VDC (±20%), power consumption current ≤15mA, output mode -2VDC~-18VDC, sensitivity 0.64V / mm±2%, nonlinearity ≤1.0%, operating temperature -20℃~+60℃; Deployment: Two units are configured for each wind turbine and installed near the turbine tower. They are fixed with custom brackets. During initial installation, a high-precision electronic level is used to zero the turbine and monitor the overall tilt displacement of the wind turbine in real time to provide early warning of risks such as foundation loosening and structural deformation.

[0027] strain gauge Selection: Vibrating wire strain gauge, standard range ±1500με, nonlinearity linear ≤0.5%FS, polynomial ≤0.1%FS, sensitivity 1με, operating temperature range -20℃ to +80℃, temperature measurement accuracy ±0.5℃.

[0028] Deployment: Eight units are configured for each wind turbine and installed on the concrete structure surface near the wind turbine's embedded parts. Before installation, the fireproof board on the tunnel roof is cut (single section 25cm×25cm). The concrete surface is then ground, cleaned, and positioned. Special adhesive is used for bonding. After curing, waterproof sealant is applied for protection, and structural stress and strain changes are monitored.

[0029] Temperature sensor Selection: Uses DS18B20 digital temperature chip, range -55℃~+125℃, accuracy ±0.5℃, sensitivity 0.1℃, outputs 485 Modbus RTU digital signal; Deployment: Attached to the fan motor housing and fixed with high-temperature adhesive, covered with external insulation cotton to reduce environmental interference, accurately monitors the motor operating temperature, and provides early warning of overheating faults.

[0030] Power monitoring equipment Selection: Current transformer (range adapted to the rated current of the fan), SEP310 power metering and control instrument, which supports the acquisition of three-phase voltage, current, power, power factor, energy, frequency and other parameters, has 4-channel passive switch monitoring and 2-channel relay output, supports 4~20mA analog output and RS485 communication interface; Deployment: The current transformer is installed in the main circuit of the three-phase cable inside the fan distribution box. The power metering and control instrument is installed on the guide rail inside the cabinet and connected to the transformer and power supply through the wiring terminals to collect power parameters in real time to analyze electrical faults.

[0031] (II) Data Acquisition and Transmission Layer: Data acquisition and transmission implementation Field Instrument Box Deployment Model Selection: Made of cold-rolled steel plate, 1.5mm thick, IP65 protection rating, double-door structure (tempered glass inlaid on the outer layer), 12 M30 openings and M6 grounding bolts at the bottom, built-in dual power supply module, communication control module, vibration monitoring module (8500F-ZG type), settlement monitoring module (8500F-CJ type), data acquisition module, conversion module, supporting centralized processing of multi-source data.

[0032] Installation: Deployed in the middle of the tunnel utility tunnel, avoiding other pipelines and equipment boxes, and powered by the nearest jet fan power control box (AC220V). The internal modules are arranged in the order of "power - data acquisition - communication". Cables are connected and fixed through the bottom opening to ensure stable operation of the equipment.

[0033] Data transmission scheme Cable selection: Category 6A shielded network cable (CAT6A-SFTP) is used, and flame-retardant copper core cable is used for power cables. Sensor cables are protected by PE conduit, which has anti-interference, moisture-proof and wear-resistant properties.

[0034] Transmission Link: The signals from each sensor are connected to the field instrument box via shielded cables. The communication module inside the instrument box is connected to the tunnel switch via a wired network, and then transmitted to the urban construction tunnel and bridge server via optical fiber to achieve real-time data upload with a transmission delay of ≤100ms. It supports seamless integration with the existing integrated monitoring system of the tunnel via Modbus RTU (RS485) and Ethernet interface.

[0035] Data acquisition equipment monitoring The data acquisition equipment uses the 8500F-TDI model, which supports wired networking and monitors the communication status (connected, disconnected, inactive) and the last communication time in real time. The administrator account can view the communication statistics of different frame types, which facilitates quick troubleshooting of data transmission anomalies.

[0036] (III) Data Processing and Fusion Layer: Data Processing and Fusion Implementation Data preprocessing Time synchronization: The NTP protocol is used to add a unified timestamp to all sensor data, with a time deviation of ≤10ms, to ensure the spatiotemporal consistency of multi-source data such as vibration, temperature, and power.

[0037] Noise reduction processing: The vibration signal is denoised using a wavelet thresholding algorithm to remove environmental interference, while the temperature and power data are filtered using a moving average to eliminate fluctuation errors and ensure data quality.

[0038] Feature extraction and fusion Spatial association: Establish a mapping relationship between sensor locations and key components of the wind turbine (e.g., horizontal vibration sensors correspond to wind turbine bearings, strain gauges correspond to embedded parts, and power monitoring equipment corresponds to motors) to clarify data ownership.

[0039] Feature extraction: Wavelet packet transform was used to extract peak value, kurtosis, waveform indices, first harmonic energy, and other features from vibration signals; temperature data was analyzed using the sliding window method to calculate the temperature rise rate (unit: ℃ / min) and average temperature; harmonic features were extracted from power data using Fourier transform, and the effective value of current and power factor deviation were calculated.

[0040] Feature fusion: A weighted fusion algorithm is adopted, and weights are assigned according to the degree of fault impact (vibration feature weight 0.4, strain feature weight 0.25, temperature feature weight 0.15, and electrical feature weight 0.2) to construct a 12-dimensional comprehensive health feature vector, providing high-quality data support for fault diagnosis.

[0041] (iv) Intelligent Diagnosis and Early Warning Layer: Implementation of Diagnosis and Early Warning Fault Diagnosis Model Training and Deployment Model Construction: A fusion model is built based on Deep Belief Network (DBN) and Convolutional Neural Network (CNN). The input is a 12-dimensional comprehensive health feature vector, and the output is the fault type, location, severity level, and confidence level. The model integrates the core algorithm of MATLAB and can accurately identify bearing faults (outer ring damage, rolling element damage, cage damage, inner ring damage, poor lubrication), coupling misalignment, rotor imbalance, loose foundation, uneven air gap between motor rotor and stator, etc.

[0042] Training data includes normal operation data of the wind turbine, simulated fault data and historical fault data from the field (such as the case of low horizontal vibration alarm of jet fan-07-A and high horizontal vibration alarm of jet fan-08-C), totaling more than 100,000 samples. After training, the model's diagnostic accuracy is ≥90%.

[0043] Fault-feature mapping library: Preset feature thresholds and associated parameters for various typical faults, such as "abrupt vibration kurtosis index + slow rise in bearing temperature" corresponding to early bearing damage, and "surge in vibration amplitude + strain value exceeding limit + abnormal displacement" corresponding to foundation loosening.

[0044] Health assessment and prediction Health index calculation: The weighted summation formula is used (health score = vibration health score × 0.4 + structural health score × 0.3 + electrical health score × 0.2 + temperature health score × 0.1). The score range is 0-100 points, where 85-100 points is healthy, 70-84 points is sub-healthy, 50-69 points is faulty, and <50 points is dangerous. The health status is updated in real time and displayed visually.

[0045] Remaining service life prediction: Based on the analysis of the health index degradation trend using a time series ARIMA model, and combined with the component life database, the remaining service life of key components (bearings, motors) is predicted with an error of ≤10%.

[0046] Tiered early warning mechanism Warning thresholds: The three-level thresholds for low warning, high warning, and high-high warning are dynamically adjusted (e.g., low vibration acceleration warning 0.3g, high warning 0.5g, and high-high warning 0.7g). The thresholds are automatically corrected based on the operating years and conditions of the wind turbine, and manual fine-tuning is supported.

[0047] Warning triggering logic: When the monitored and collected values ​​(acceleration - peak value, velocity - effective value, displacement - peak-to-peak value, and the rest are effective values) exceed the set threshold, an early warning is triggered. If the same measuring point index meets the triggering conditions repeatedly within 2 hours, the warning will not be repeated. It can be triggered again after 2 hours.

[0048] Warning information includes warning ID, measurement point name, real-time value, baseline value, percentage exceeding the limit, trend chart, fault details, and handling suggestions, and is pushed through multiple terminals.

[0049] (v) Application Layer: Implementation of Multi-Terminal Interaction Functional modules: Machine Overview: Displays statistics on the health status (healthy, sub-healthy, faulty, dangerous), warning status (normal, low warning, high warning, high-high warning), and operating status (operating, stopped) of all wind turbines. It supports project switching and time interval filtering (e.g., 2025-04-23 to 2025-05-23). ​​It displays the latest 5 diagnostic and warning data, including warning statistics, warning equipment ranking, warning trends, and machine operating status distribution charts.

[0050] Monitoring Center: Supports multi-dimensional searches by project name, machine type, machine name, and machine number, and provides both list and grid display modes; the machine details page allows for 360° viewing of 3D models (front view, side view, top view), full display of measurement points, and displays of monitoring information (collected values ​​and percentage bar charts), measurement point trend charts (last month), health status time distribution, pending warnings, warning analysis / fault analysis, warning records, and diagnostic records.

[0051] Early Warning Center: Supports searching by early warning indicator, early warning ID, early warning time, project name, machine name, machine number, and measurement point name; filtering by early warning level, processing status, and early warning review; the early warning details page allows viewing trend charts of multiple indicators (valid value, peak value, peak-to-peak value, etc.); provides early warning review (valid / invalid / to be observed), processing description entry, and image upload functions; and supports advanced analysis jumps.

[0052] Advanced Analysis: Provides spectral analysis functions such as time domain analysis, frequency domain analysis, envelope demodulation, filtering analysis, EMD analysis, time-frequency analysis, cepstral analysis, waterfall plot, order analysis, correlation analysis, axis trajectory, and polar coordinate plot. Supports single-point / multi-point analysis. The spectrum can be zoomed in, dragged, and downloaded. It integrates automatic identification of fault characteristic frequencies, custom frequency identification, and automatic identification of harmonics.

[0053] Intelligent Diagnosis: Displays the distribution of device health levels, the top 5 devices with abnormal diagnoses, and the daily trend of abnormal diagnoses. It supports searching by diagnosis time period. The diagnosis details page displays the fault conclusion, confidence level, maintenance suggestions, related fault cause explanations and graph data, and supports exporting reports in PDF / Word format.

[0054] Workbench: Includes an early warning center, intelligent diagnosis, and advanced analysis modules; the early warning center supports searching by early warning name, number, early warning ID, and measurement point name, and allows viewing early warning details and processing early warnings (review, input instructions, and upload images); intelligent diagnosis allows viewing diagnostic records and details, and supports processing diagnoses; advanced analysis supports single-point / multi-point analysis, and allows adding analysis charts, filtering statistical intervals, and measurement point indicators.

[0055] (vi) Implementation of the data management optimization module Data storage: The monitoring data is stored using a distributed database. The original vibration data is saved for 3 months, while the characteristic data and diagnostic results are permanently saved. Data traceability and historical query are supported to ensure that the operation and maintenance data is searchable throughout the entire life cycle.

[0056] Model optimization: The fault-feature mapping library is updated quarterly using the results of early warning review and manual processing records. The diagnostic model parameters and early warning thresholds are adjusted iteratively through machine learning to reduce the false alarm rate to below 5% and improve diagnostic accuracy.

[0057] Weight update: The feature weights of each sensor are dynamically adjusted based on the operating years of the wind turbine, ambient humidity, and frequency of failures to ensure that the health assessment is accurately adapted to the actual operating conditions.

[0058] The above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the present invention. The scope of protection of the present invention is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to the present invention within its spirit and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of the present invention.

Claims

1. A tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion, characterized in that: This includes a layered architecture, from bottom to top consisting of a perception layer, a data acquisition and transmission layer, a data processing and fusion layer, an intelligent diagnosis and early warning layer, and an application layer. Each layer works together to achieve dual-dimensional control of tunnel fan collision safety and equipment health. Sensing layer: Deploy vibration acceleration sensors, displacement sensors, strain gauges, temperature sensors, current transformers and power metering and control instruments to collect parameters such as fan vibration, tilt displacement, structural strain, temperature and three-phase voltage, current and power, covering the monitoring needs of oversized vehicle collisions, tunnel fans with loose foundations, and mechanical and electrical faults. Data acquisition and transmission layer: includes field instrument boxes, switches and Category 6a shielded cables. The instrument boxes integrate dual power supply modules, communication control modules and function modules, and are deployed in the middle pipe gallery to realize synchronous acquisition, preprocessing and wired transmission of multi-source data, and support seamless integration with the existing tunnel integrated monitoring system. Data processing and fusion layer: Employs spatiotemporal alignment and feature fusion algorithms to unify data timestamps, establish a mapping relationship between sensor locations and fan bearings, impellers, and hoisting structures, extract key features such as vibration peak value, kurtosis, temperature rise rate, and current effective value, and construct a comprehensive health feature vector; Intelligent Diagnosis and Early Warning Layer: Built-in fault diagnosis model based on Deep Belief Network (DBN) and Convolutional Neural Network (CNN) is used to analyze the comprehensive health feature vector, automatically identify various fault types such as bearing damage, coupling misalignment, rotor imbalance, foundation looseness and uneven motor air gap, evaluate the comprehensive health index of the equipment, and generate graded early warning information in combination with an adaptive threshold mechanism; Application layer: Includes a computer web client and a mobile APP client based on B / S architecture, providing machine overview, monitoring center, early warning center, advanced analysis and intelligent diagnosis function modules, supporting equipment status visualization, early warning processing, in-depth map analysis, diagnostic report generation and export, as well as macro-display on a smart screen based on geographic information system.

2. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that: Two vibration acceleration sensors are configured for each wind turbine, which are fixedly installed on the horizontal and vertical directions of the wind turbine casing via welded brackets, respectively; two displacement sensors are configured for each wind turbine, deployed near the wind turbine tower, to monitor the overall tilt displacement of the wind turbine; eight strain gauges are configured for each wind turbine, installed on the surface of the concrete support structure near the wind turbine embedded parts; the temperature sensor uses a digital temperature chip, attached to the wind turbine motor casing and fixed with high-temperature adhesive, and covered with thermal insulation cotton; the power monitoring equipment includes current transformers and power metering and control instruments installed in the wind turbine distribution box, used to collect three-phase voltage, current, power, power factor and power parameters.

3. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that, The specific methods for data fusion performed by the data processing and fusion layer include: Time synchronization: Establishing a unified time reference for multi-source data such as vibration, temperature, and electricity; Spatial correlation: Establish a mapping relationship between the installation location of each sensor and the key components of the wind turbine bearings, impellers, and hoisting structure; Feature extraction: Extract kurtosis and waveform indices from vibration signals, calculate temperature rise rate from temperature data, and extract harmonic features from power data; Feature fusion: The extracted multi-dimensional features are integrated into a comprehensive health feature vector through a weighted fusion algorithm.

4. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that: The intelligent diagnosis and early warning layer is configured to perform the following core functions; Fault diagnosis: Based on multi-parameter correlation analysis and a pre-set fault-feature mapping library, it automatically identifies and locates various fault types and locations, including bearing damage, rotor imbalance, coupling misalignment, and foundation loosening. Health assessment and prediction: Based on time series analysis of equipment performance degradation trends, calculate and output a comprehensive health index on a scale of 0-100, and combine it with degradation models to predict the remaining service life of key components; Tiered early warning: Establish multi-level early warning thresholds including low reporting, high reporting, and very high reporting. When the monitored value exceeds the limit, the corresponding level of early warning will be automatically triggered, and early warning information including fault details, trend charts and handling suggestions will be pushed. Operation and maintenance decision support: Automatically generates operation and maintenance recommendation reports containing maintenance measures and spare parts lists based on diagnostic conclusions, and provides maintenance priority ranking for multi-device operation and maintenance tasks.

5. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that: The application layer is specifically configured to provide multi-terminal collaborative interactive functions, including: supporting synchronous access via computer web terminal and mobile APP terminal, and displaying corresponding device status, early warning information and diagnostic reports based on user roles and permission control; in the monitoring center module, providing two device view display modes, list and grid, and supporting multi-view viewing based on 3D model and full display of measurement point distribution; providing early warning information and diagnostic report export function, supporting the generation and export of PDF or Word format documents containing fault details, analysis graphs and maintenance suggestions.

6. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that: The fault diagnosis model in the intelligent diagnosis and early warning layer adopts a multi-evidence joint judgment and confidence assessment mechanism, specifically configured as follows: Establish a database of typical fault characteristics, including bearing failure, coupling misalignment, rotor imbalance, loose foundation, and motor electrical faults; During the diagnosis process, vibration spectrum characteristics, temperature change trends, current harmonic components and structural strain data are analyzed simultaneously, and multi-source parameters are used as correlation evidence for fusion judgment. Based on the deep belief network or convolutional neural network algorithm, the occurrence confidence of each potential fault type is calculated, and a diagnostic conclusion with a confidence level higher than a preset threshold is output. The diagnostic conclusions are linked and stored with the corresponding historical cases, feature maps, and indicator contribution analyses to support manual review and continuous optimization of the model algorithm.

7. The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion according to claim 1, characterized in that: It also includes a data management and optimization module, which is used to continuously optimize system performance based on historical operation and maintenance data, receive and store early warning review results, manual processing records and fault repair feedback data from the application layer; use the early warning review results and processing records to iteratively adjust the adaptive threshold and judgment parameters of the fault diagnosis model in the intelligent diagnosis and early warning layer through machine learning; and dynamically update the monitoring weights of each sensor and the performance degradation model parameters of key components according to the long-term operation data and operating condition changes of the wind turbine.

8. A method for health diagnosis and early warning of tunnel ventilation fans based on multi-source fusion, characterized in that, The tunnel ventilation fan health diagnosis and early warning system based on multi-source fusion as described in any one of claims 1-7 is operated as follows: S1: Multi-source data acquisition By deploying vibration acceleration sensors, displacement sensors, strain gauges, temperature sensors and current transformers on the tunnel ventilation fan, multi-source heterogeneous data on the fan's vibration, tilt displacement, structural strain, operating temperature and electrical parameters are collected simultaneously. S2: Data Preprocessing and Feature Extraction The collected multi-source heterogeneous data is time-aligned and preprocessed locally to extract key feature parameters, including vibration peak value, kurtosis, temperature change rate, current RMS value and harmonic characteristics. S3: Feature Fusion and Comprehensive Health Status Assessment The extracted key feature parameters are spatially correlated and fused to generate a comprehensive health feature vector characterizing the overall operating status and structural safety of the wind turbine. Based on a pre-set fault-feature mapping library and machine learning model, the comprehensive health feature vector is analyzed to identify whether the wind turbine has bearing failure, coupling misalignment, rotor imbalance, loose foundation or motor electrical abnormality, and the comprehensive health index of the equipment is calculated. S4: Intelligent Early Warning and Operation and Maintenance Decision Generation Based on the analysis results and the preset adaptive warning threshold, hierarchical warning information including fault type, location, severity level and confidence level is generated; Based on the diagnostic results, an operation and maintenance decision report is automatically generated, which includes maintenance measures, spare parts recommendations, and repair priorities. S5: Information Visualization and Interactive Processing The comprehensive health index, early warning information, diagnostic conclusions, and operation and maintenance decision reports are visualized and pushed through computer terminals or mobile APPs; the system receives and processes user verification and confirmation of early warning information, status updates, and maintenance record entry operations, and feeds the processing results back to the system to optimize subsequent diagnostic and early warning models.

Images