Intelligent monitoring system for stress state of metro vibration isolator and monitoring method thereof

Abstract

The application discloses a subway vibration isolator stress state intelligent monitoring system and a monitoring method thereof, and relates to the field of subway vibration isolator monitoring.The system comprises an intelligent vibration isolator, a patrol instrument and a remote monitoring platform, and a three-level linkage system of equipment-collection-platform is constructed.The intelligent vibration isolator measures the compression amount of a disc spring piece group through electromagnetic induction non-contact measurement, completes data preprocessing and feature extraction through edge computing, and realizes long endurance through multiple low-power consumption designs.The patrol instrument completes instruction forwarding, data collection and firmware upgrading.The remote monitoring platform accurately identifies four types of diseases based on a multi-channel LSTM model, predicts stress trends, realizes graded alarm and intelligent operation and maintenance.The application improves monitoring precision and efficiency, reduces operation and maintenance costs, greatly improves the intelligent level of subway vibration isolator operation and maintenance, and guarantees rail transit operation safety.

Description

Technical Field

[0001] This invention relates to the field of rail transit equipment monitoring technology, specifically to an intelligent monitoring system and method for monitoring the stress state of subway vibration isolators. Background Technology

[0002] Subway floating slab vibration isolation systems widely use steel springs or rubber isolators. Traditional monitoring methods mainly rely on manual periodic inspections and single-point measurements. Current technology typically involves installing force sensors within pre-embedded outer sleeves, recording the stress state through probes contacting the isolator sleeve, and uploading the data to a remote terminal. This method has the following significant drawbacks: 1. Low monitoring efficiency: Manual inspections are time-consuming and labor-intensive, with long intervals, making it difficult to capture continuous changes in the isolator's stress in real time; single-point measurements cannot comprehensively reflect the overall stress state of the isolator, resulting in biased assessments. 2. Insufficient intelligence: Existing systems have low integration, with each module operating independently, lacking intelligent data analysis and early warning capabilities, and unable to predict potential isolator failures. 3. High maintenance costs: Some wireless monitoring devices have high power consumption and short battery life, requiring frequent charging or battery replacement, or even complete equipment replacement, increasing maintenance costs and workload. 4. Sensor vulnerability: Probe wear due to long-term contact leads to decreased sensitivity, affecting monitoring accuracy and reliability.

[0003] Therefore, there is an urgent need for a vibration isolator stress state monitoring system and method that can achieve real-time, intelligent, low-power, and long-endurance operation to improve the intelligence level and safety of subway track system operation and maintenance. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent monitoring system and method for monitoring the stress state of subway vibration isolators.

[0005] This invention discloses an intelligent monitoring system for the stress state of subway vibration isolators, including intelligent vibration isolators, inspection instruments and remote monitoring platforms, which work together to form a three-level linkage intelligent monitoring system; Multiple intelligent vibration isolators are respectively installed under the floating slab of the subway and are communicatively connected to the inspection instrument. Each intelligent vibration isolator has a built-in disc spring assembly. The compression of the disc spring assembly is collected through a non-contact measurement method and converted into vertical support reaction force data of the subway floating slab acting on the vibration isolator. The support reaction force data is preprocessed, features are extracted, and then stored. The intelligent vibration isolator is configured with a low-power operating mode, which only performs data acquisition and processing operations during a set time period and remains in a dormant state at other times. The inspection device is communicatively connected to the remote monitoring platform, and is used to receive operation instructions issued by the remote monitoring platform and send the instructions to the corresponding intelligent vibration isolator; and to collect the vertical support reaction data stored in the intelligent vibration isolator, package the data and upload it to the remote monitoring platform; The remote monitoring platform is used to receive, parse and store the vertical support reaction force data uploaded by the inspection instrument, perform intelligent analysis of the stress state of the vibration isolator based on a deep learning model, identify the fault type and predict the stress change trend, and issue remote operation commands to the inspection instrument for the intelligent vibration isolator.

[0006] As a further improvement of the present invention, the intelligent vibration isolator includes a vibration isolator top seat, a vibration isolator base, a disc spring assembly, a sealing sleeve, and a main unit; The top seat of the vibration isolator and the base of the vibration isolator are coaxially arranged vertically. The sealing sleeve is fitted on the outside of the adjacent parts of the two, so that the top seat of the vibration isolator and the base of the vibration isolator are sealed and connected to form a closed cavity. The closed cavity is filled with damping fluid. The disc spring assembly is located in the closed cavity and is elastically supported between the top seat of the vibration isolator and the base of the vibration isolator. The top seat of the vibration isolator can float vertically relative to the base of the vibration isolator with the elastic deformation of the disc spring assembly. The vibration isolator top seat has a downward-facing first protrusion, and the vibration isolator base has an upward-facing second protrusion coaxial with the first protrusion; the bottom end of the first protrusion forms a main unit mounting cavity, which is filled with epoxy resin potting compound, and the main unit is mounted in the main unit mounting cavity through a main unit cover plate; the main unit includes a battery, a main board, a sensor, and a sensor probe arranged sequentially from top to bottom and electrically connected, and the main unit cover plate has a through hole corresponding to the sensor probe; the sensor probe is arranged towards the second protrusion through the through hole, and obtains the compression amount of the disc spring assembly by monitoring the relative distance with the second protrusion, and the main board converts the compression amount into vertical support reaction force data of the subway floating plate acting on the vibration isolator.

[0007] This invention discloses an intelligent monitoring method for the stress state of subway vibration isolators, which is applied to the aforementioned intelligent monitoring system for the stress state of subway vibration isolators, comprising: Step 1: The intelligent vibration isolator collects the force data of the vibration isolator during the preset working time period, and stores it after preprocessing and feature extraction; during the non-working time period, the intelligent vibration isolator is in a low-power sleep state. Step 2: After receiving the data collection command from the remote monitoring platform, the inspection instrument establishes a wireless connection with the intelligent vibration isolator, reads the stored force data, and uploads the force data to the remote monitoring platform. Step 3: The remote monitoring platform processes the force data received from multiple intelligent vibration isolators and uses a pre-trained deep learning model to identify the health status of the vibration isolators. The deep learning model is a multi-channel LSTM model based on multi-sensor time-series data. Step 4: The remote monitoring platform generates visual information based on the diagnostic results and issues an alert when an abnormal state is identified.

[0008] As a further improvement of the present invention, in step 1, the force data collected by the intelligent vibration isolator includes dynamic force data and static force data: Dynamic force data: During the period when a vehicle passes by, when the sensor detects vibration, it wakes up the host, measures the maximum value of the compression of the disc spring assembly, and converts it into the maximum value of the dynamic force of the vibration isolator; Static stress data: At the time when no vehicle passes, the compression of the disc spring assembly is directly collected and converted into the static stress value of the vibration isolator.

[0009] As a further improvement of the present invention, in step 1, the preprocessing of the vertical support reaction force data by the intelligent vibration isolator includes: using bandpass filtering and wavelet denoising technology to remove stray current interference, and combining the STE short-time average energy algorithm to identify and extract the effective data segment when the train passes. The feature extraction includes: calculating the RMS root mean square value, kurtosis, and power spectral density (PSD) obtained by FFT fast Fourier transform of the support reaction force data, and transforming the original waveform data into a lightweight feature vector representing the health status of the vibration isolator.

[0010] As a further improvement of the present invention, in step 1, the low-power sleep state of the intelligent vibration isolator is achieved in the following way: The host is woken up by a sensor trigger. And / or use I / O ports to control the power supply of the sensor, and turn off the sensor's standby power consumption; And / or enable the corresponding function only during the set time period, and keep it in sleep mode the rest of the time.

[0011] As a further improvement of the present invention, in step 2, after the inspection instrument establishes a wireless connection with the intelligent vibration isolator, the method further includes: Perform time synchronization and send the current accurate time to the intelligent vibration isolator; Send data transmission instructions to the intelligent vibration isolator and receive the data stored therein for the day; If a firmware upgrade command is available, the firmware package data to be upgraded will be sent to the smart vibration isolator. After the data interaction is completed, the intelligent vibration isolator returns to sleep mode.

[0012] As a further improvement of the present invention, in step 3, the remote monitoring platform processes the force data, specifically including: The stress time series data of multiple intelligent vibration isolators are downsampled to extract daily average features; The extracted daily average features are normalized. A sliding window method was used to construct a multi-channel sample, each sample containing the daily mean features of M vibration isolators over N consecutive days, where N≥7 and M≥2. The multi-channel samples are input into the multi-channel LSTM model, which is a stacked LSTM structure, and is used to extract the temporal features of the multi-channel samples.

[0013] As a further improvement of the present invention, after extracting the time-series features, it further includes: The features of each time step output by the LSTM layer are input into the temporal attention layer. By calculating the attention weight of each time step, the LSTM output is weighted and summed to obtain the feature vector focused on the key time step. The weighted feature vector is input into the classifier, which outputs the probability that the health status of the vibration isolator belongs to a preset variety of defects. The preset variety of defects includes normal state, uneven stress state, suspended / settled state, and broken spring state.

[0014] As a further improvement of the present invention, the criterion for judging the uneven force state is: the force variance among the M vibration isolators is greater than a preset threshold. The criteria for determining the suspended / settled state are: the force on a single vibration isolator is continuously lower than 50% of its own average or higher than 150% of its own average, and the duration is not less than 7 days; The criteria for determining the spring breakage state are: a single vibration isolator experiences a step change in force, with the change amplitude exceeding 80%; The training process of the multi-channel LSTM model adopts a loss function with class weights. The class weight of the broken spring state is higher than that of the uneven force, suspended / sinking state, and the class weight of the uneven force, suspended / sinking state is higher than that of the normal state. In addition, an early stopping mechanism is introduced into the training process. If the loss of the validation set does not decrease for a preset number of consecutive rounds, the training is stopped.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention revolutionizes the force acquisition method of vibration isolators from the monitoring hardware perspective. It adopts non-contact measurement technology based on electromagnetic induction to obtain the compression of the disc spring assembly, completely abandoning the traditional easily worn probe contact structure and fundamentally solving the problem of sensor sensitivity decay with use. At the same time, the intelligent vibration isolator has built-in damping fluid to suppress resonance, and the main unit mounting cavity is filled with epoxy resin potting compound to ensure sealing. Combined with the mechanical property formula of the disc spring, it can achieve accurate conversion of compression to force value. It can also simultaneously collect static and dynamic dual-dimensional force data, which greatly improves the accuracy, stability and data comprehensiveness of vibration isolator force monitoring.

[0016] This invention constructs a three-level intelligent monitoring system linking intelligent vibration isolators, inspection instruments, and remote monitoring platforms. It realizes digital management of the entire vibration isolator monitoring process. The intelligent vibration isolator can autonomously complete data preprocessing, feature extraction, and local storage. The inspection instrument can read data in batches, forward remote commands, and perform remote firmware upgrades. This replaces the traditional manual periodic inspection and single-point measurement mode, significantly reducing manpower and time costs and solving the problems of low efficiency, inability to capture continuous changes in force, and biased evaluation results of manual inspection. At the same time, the intelligent vibration isolator is designed with multiple low-power modes such as sensor-triggered wake-up, I / O port-controlled power supply, and timed operation, achieving long-lasting operation without frequent charging or battery replacement. Combined with the remote firmware upgrade function, it avoids the need for complete equipment replacement, significantly reducing equipment maintenance costs and on-site maintenance workload.

[0017] This invention integrates edge computing and deep learning technologies to achieve intelligent diagnosis, trend prediction, and precise operation and maintenance of vibration isolators. The intelligent vibration isolator completes data noise reduction, effective segment extraction, and feature extraction through edge computing. The remote monitoring platform, based on a multi-channel LSTM model combined with a temporal attention layer, accurately identifies four types of defects: normal, uneven stress, suspension / settlement, and broken spring. The remote monitoring platform has comprehensive data storage, 3D digital twin visualization, multi-role permission management, and custom report generation functions, and constructs a vibration isolator health status database, providing scientific data support for long-term performance monitoring and maintenance strategy formulation. This significantly improves the intelligence and scientific level of subway vibration isolator operation and maintenance, effectively reduces safety accidents caused by vibration isolator defects, and ensures the safety and efficiency of railway operation. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the structure of an intelligent monitoring system for the stress state of subway vibration isolators disclosed in one embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent vibration isolator in a subway vibration isolator stress state intelligent monitoring system disclosed in an embodiment of the present invention; Figure 3 This is a side sectional view of the intelligent vibration isolator of the intelligent monitoring system for the stress state of subway vibration isolators disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of the measurement principle of an intelligent vibration isolator in a subway vibration isolator stress state intelligent monitoring system disclosed in one embodiment of the present invention; Figure 5 This is a schematic diagram of the inspection instrument of the intelligent monitoring system for the stress state of subway vibration isolators disclosed in one embodiment of the present invention; Figure 6 This is a front view of the inspection instrument of the intelligent monitoring system for the stress state of subway vibration isolators disclosed in one embodiment of the present invention; Figure 7This is a flowchart of the inspection instrument of the intelligent monitoring system for the stress state of subway vibration isolators disclosed in one embodiment of the present invention; Figure 8 This is a flowchart of an intelligent monitoring method for the stress state of a subway vibration isolator, as disclosed in one embodiment of the present invention.

[0019] In the picture: 1. Intelligent vibration isolator; 1-1. Vibration isolator top mount; 1-1-1. First protrusion; 1-1-2. Wiring groove; 1-1-3. Main unit mounting cavity; 1-2. Vibration isolator base; 1-2-1. Second protrusion; 1-3. Sealing sleeve; 1-4. First sealing ring; 1-5. Second sealing ring; 1-6. Main unit; 1-6-1. Battery; 1-6-2. Main board; 1-6-3. Sensor; 1-6-4. Sensor probe; 1-7. Main unit sealing plate; 1-8. Disc spring assembly; 1-9. Enclosed accommodating cavity; 1-10. Vibration isolator indicator light; 2. Inspection instrument; 2-1. Display screen; 2-2. Power switch; 2-3. Charging and data interface; 2-4. Inspection instrument indicator light; 2-5. Bluetooth external high-gain antenna; 3. Remote monitoring platform. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0022] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1:

[0024] like Figure 1 As shown, the intelligent monitoring system for the stress state of a subway vibration isolator provided by the present invention includes an intelligent vibration isolator 1, an inspection instrument 2, and a remote monitoring platform 3, which together form a three-level linkage intelligent monitoring system. Multiple intelligent vibration isolators 1 are respectively installed under the subway floating slab and are communicatively connected to the inspection instrument 2. Each intelligent vibration isolator 1 has a built-in disc spring assembly 1-8, which collects the compression of the disc spring assembly 1-8 through a non-contact measurement method, converts it into vertical support reaction force data of the subway floating slab acting on the vibration isolator, and stores the support reaction force data after preprocessing and feature extraction. The intelligent vibration isolator 1 is configured with a low-power operating mode, operating only for a set time. The first stage performs data acquisition and processing operations, and remains in sleep mode at other times. The inspection instrument 2 is connected to the remote monitoring platform 3 to receive operation instructions issued by the remote monitoring platform 3 and send the instructions to the corresponding intelligent vibration isolator 1. It also collects the vertical support reaction force data stored in the intelligent vibration isolator 1, packages the data and uploads it to the remote monitoring platform 3. The remote monitoring platform 3 receives, parses and stores the vertical support reaction force data uploaded by the inspection instrument, performs intelligent analysis of the force state of the vibration isolator based on a deep learning model, identifies the fault type and predicts the force change trend, and can issue remote operation instructions to the inspection instrument 2 for the intelligent vibration isolator 1.

[0025] Specifically: like Figure 2-4As shown, in the above embodiment, preferably, the intelligent vibration isolator 1 includes a vibration isolator top seat 1-1, a vibration isolator base 1-2, a disc spring assembly 1-8, a sealing sleeve 1-3, a first sealing ring 1-4, a second sealing ring 1-5, a main unit 1-6, a main unit sealing plate 1-7, and a vibration isolator indicator light 1-10. The vibration isolator top seat 1-1 and the vibration isolator base 1-2 are coaxially arranged vertically, and the sealing sleeve 1-3 is fitted onto the outer side of the adjacent portions of the two, so that the vibration isolator top seat... 1-1. The isolator base 1-2 is sealed together to form a closed cavity 1-9, which is filled with damping fluid to suppress resonance and accelerate vibration decay. The disc spring assembly 1-8 is located within the closed cavity 1-9 and elastically supported between the isolator top seat 1-1 and the isolator base 1-2. The isolator top seat 1-1 can float vertically relative to the isolator base 1-2 with the elastic deformation of the disc spring assembly 1-8. The base 1-2 of the vibration isolator has a first protrusion 1-1-1 at the bottom and a second protrusion 1-2-1 coaxial with the first protrusion 1-1-1 at the top. The bottom end of the first protrusion 1-1-1 forms a main unit mounting cavity 1-1-3, which is filled with epoxy resin potting compound to ensure sealing. The main unit 1-6 is installed in the main unit mounting cavity 1-1-3 through the main unit sealing plate 1-7. The main unit 1-6 includes a battery 1-6-1, a main board 1-6-2, a sensor 1-6-3, and a sensor probe 1-6-4 arranged sequentially from top to bottom and electrically connected. The main unit sealing plate 1-7 has a through hole corresponding to the sensor probe 1-6-4. The sensor probe 1-6-4 is arranged towards the second protrusion 1-2-1 through the through hole. The compression amount of the disc spring assembly 1-8 is obtained by monitoring the relative distance with the second protrusion 1-2-1. The main board 1-6-2 converts the compression amount into vertical support reaction force data of the subway floating plate acting on the vibration isolator. The vibration isolator indicator light 1-10 is installed on the upper end of the vibration isolator top seat 1-1, and a wire channel 1-1-2 is vertically provided in the first protrusion 1-1-1 to connect the vibration isolator indicator light 1-10 with the main unit mounting cavity 1-1-3. The wire of the vibration isolator indicator light 1-10 is connected to the main board 1-6-2 of the main unit mounting cavity 1-1-3 through the wire channel 1-1-2.

[0026] In the above embodiments, preferably, the sealing sleeve 1-3 is fitted with a first sealing ring 1-4 and a second sealing ring 1-5 on the side sealing plate of the vibration isolator top seat 1-1 and the vibration isolator base 1-2. The first sealing ring 1-4 and the second sealing ring 1-5 are used to tighten the sealing sleeve 1-3 with the vibration isolator top seat 1-1 and the vibration isolator base 1-2.

[0027] In the above embodiments, preferably, the motherboard 1-6-2, sensor 1-6-3, and sensor probe 1-6-4 are designed as an integrated unit.

[0028] In the above embodiments, preferably, the disc spring group 1-8 consists of multiple disc springs stacked vertically within the enclosed accommodating cavity 1-9. Among the multiple disc springs, the bottom disc spring is fixed to the vibration isolator base 1-2 through contact with its outer periphery; the upper disc spring is fixed to the first protrusion 1-1-1 of the vibration isolator top seat 1-1 through its inner diameter being the same as that of the first protrusion 1-1-1. In this embodiment, the disc springs are mainly arranged in parallel with the same direction, and if necessary, a combination of series and parallel arrangements with opposite directions is used to meet the requirements of high load-bearing capacity, low natural frequency, and stable stiffness characteristics in the subway vibration isolation environment. The conversion between the compression and force values ​​of the multiple disc springs follows the mechanical characteristics of the disc springs and is quickly calculated according to the linearized simplified formula. At this time, the force and compression are approximately proportional: F=Kd*λ, where F is the axial pressure borne by a single disc spring, in N; Kd is the equivalent stiffness coefficient of a single disc spring, in N / m; and λ is the compression of a single disc spring, in m. When connected in parallel, F = n * Ktotal * h, Ktotal = n * Kd, where n is the number of disc reeds connected in parallel; when connected in series, F = n * Ktotal * h, Ktotal = Kd / n, where n is the number of disc reeds connected in series; when a mixed combination of series and parallel is used, F = n * Ktotal * h, Ktotal = K groups / n2, K groups = n1 * Kd, where n1 is the number of disc reeds connected in parallel in each group, and n2 is the number of groups connected in series.

[0029] like Figure 2-3 As shown, in the above embodiment, preferably, the sensor probe 1-6-4 measures the distance to the second protrusion 1-2-1 of the vibration isolator base 1-2 via electromagnetic induction. The sensor probe 1-6-4 integrates an electromagnetic excitation unit, a receiving unit, and processing circuitry. During operation, the excitation unit generates an alternating electromagnetic field around the sensor probe 1-6-4. This electromagnetic field induces a response with the surrounding metal structure, and the intensity of the induced field is related to the field strength at its location. When the distance between the sensor probe 1-6-4 and the second protrusion 1-2-1 of the vibration isolator base 1-2 changes, the induced field strength in the base also changes. The change in induced field strength can be monitored by the induction coil in the sensor probe 1-6-4, thereby achieving relative distance measurement.

[0030] In the above embodiment, preferably, the intelligent vibration isolators 1 are arranged as follows: six intelligent vibration isolators 1 are placed below each subway floating slab, arranged in two rows below the floating slab. Adaptive positioning holes and mounting blocks are provided below the floating slab corresponding to the intelligent vibration isolators 1.

[0031] like Figure 5-7As shown, in the above embodiment, preferably, the inspection device 2 has a display screen 2-1 on its front, an indicator light 2-4 on the upper right corner of the display screen 2-1, a power switch 2-2 and a charging and data interface 2-3 on its side, and an external high-gain Bluetooth antenna 2-5 on its top. In this embodiment, the inspection device 2 is used for command processing and data transmission. Command processing mainly includes control command processing (turning on Bluetooth, turning off Bluetooth, changing working time periods), data transmission command processing (data transmission), and firmware package command processing (system upgrade). When performing control command processing, the inspection device 2 receives Bluetooth control commands sent by the remote monitoring platform 3 to turn Bluetooth on / off and start / stop the inspection work. After receiving a Bluetooth control command from the host, the inspection device 2 determines whether to connect / disconnect Bluetooth based on the command type. When processing firmware package commands, the inspection device 2, upon receiving the firmware package reception command, sends firmware upgrade packages in batches to the host devices 1-6 of the intelligent vibration isolator 1. When processing data transmission commands, the inspection device 2, upon receiving a request for sensor data command, waits for the host devices 1-6 to upload the stored monitoring data. After receiving and packaging the data, it sends it to the remote monitoring platform 3, waits for the Bluetooth to turn off command, and ends the inspection work.

[0032] In the above embodiments, preferably, the remote monitoring platform 3 adopts a front-end and back-end separation architecture, supporting high-concurrency data processing, visualization, and system management functions. The back-end uses the Spring Boot 3.2 + MyBatis Plus + TDengine 3.0 technology stack, deployed on a Kubernetes cluster, to achieve elastic scaling and high availability. Specifically, time-series data storage uses TDengine, specifically for processing high-frequency time-series data such as vibration isolator force data; business data storage uses MySQL (InnoDB engine, master-slave replication architecture) to store business data such as device information, user information, and maintenance records. The platform supports two data access methods: MQTT protocol and HTTP API. The inspection device 2 uploads the collected data to the remote monitoring platform 3 via its built-in 4G / 5G network (built-in eSIM card), and the message queue uses RabbitMQ for peak shaving and valley filling to ensure stable processing during peak data periods. The remote monitoring platform 3 has a built-in custom protocol parser (implemented in Python script) that converts the binary data uploaded by the inspection device 2 into JSON format. Fields include device_id (device number), timestamp, force_value (force value), and battery_level (battery level). Format validation and deduplication are performed before data entry to ensure data quality. The front-end uses a Vue 3 + Element Plus + ECharts5 technology stack, supporting responsive layouts for both PC and mobile devices. Three.js is used to implement a 3D digital twin scene, dynamically displaying the geographical distribution of devices. Clicking on a device icon allows viewing real-time data, including force curves, battery status, and alarm information. Key metrics include the number of online devices, the number of alarms, and data throughput, displayed using dynamic charts (bar charts, line charts, and dashboards) with a refresh rate of 1 second. The device details page displays basic device information (model, installation location, maintenance records) on the left and real-time data curves (force trend over the last 24 hours), historical alarm lists (supporting time and status filtering), and sensor waveforms (raw voltage signals for fault analysis) on the right, displayed in tabs. In terms of system management, the remote monitoring platform 3 adopts the RBAC (Role-Based Access Control) model, defining three roles: super administrator, operations supervisor, and inspector, each with different permissions. Operation logs are recorded and stored in Elasticsearch, supporting keyword search and audit analysis. The platform supports multi-level group management (e.g., line → site → section → device), allowing batch allocation of devices via drag-and-drop. After grouping, data statistics within each group (average, maximum, alarm rate) can be performed, and multi-group comparative analysis (e.g., force distribution of vibration isolators in different sections) is supported.In addition, the remote monitoring platform 3 supports custom report templates, allowing users to select devices, attributes (stress, power consumption, temperature), and time spans (hours / days / months / years) to generate PDF / Excel reports. These reports include trend charts, statistical tables, alarm summaries, and more, and support scheduled email sending. The remote monitoring platform 3 also features a command issuance function, enabling it to send remote operation commands to the inspection device 2 for the intelligent vibration isolator 1, including enabling / disabling Bluetooth, data transmission, firmware upgrades, and changing operating time periods. Example 2:

[0033] like Figure 8 As shown, this embodiment provides an intelligent monitoring method for the stress state of a subway vibration isolator. This method is applied to the intelligent monitoring system described in Embodiment 1 and specifically includes the following steps.

[0034] Step 1: Data acquisition and preprocessing of intelligent vibration isolator 1 The intelligent vibration isolator 1 automatically collects the force data of the isolator during a preset working period, and stores the data after preprocessing and feature extraction. To reduce power consumption, the intelligent vibration isolator 1 only performs data acquisition and processing operations during the set time period, and remains in a low-power sleep state at other times. Specifically, this includes: 1.1 Acquisition of Dynamic and Static Force Data The force data collected by the intelligent vibration isolator 1 includes dynamic force data and static force data: Dynamic force data acquisition: During periods when trains pass by (e.g., 10:00-10:30 AM daily), the main unit 1-6 is in a standby state. When sensor 1-6-3 detects vibration, it immediately wakes up the main unit 1-6. The main unit 1-6 then activates sensor probe 1-6-4 to continuously measure the compression of the disc spring assembly 1-8 and records the maximum compression value during that time period. This value is then converted into the maximum dynamic force of the vibration isolator using the mechanical characteristic formula of the disc spring assembly 1-8 (such as the stiffness coefficient of parallel, series, or series-parallel combinations). After the data acquisition is complete, the main unit returns to sleep mode.

[0035] Static force data acquisition: At times when no trains pass (e.g., 4:00 AM daily), the main unit 1-6 is woken up at regular intervals. The sensor probe 1-6-4 measures the compression of the disc spring assembly 1-8 once and converts it into the static force value of the vibration isolator. Then the main unit 1-6 goes into sleep mode again.

[0036] 1.2 Data Preprocessing and Feature Extraction The mainboard 1-6-2 of the intelligent vibration isolator 1 performs edge computing preprocessing and feature extraction on the collected raw force data: Preprocessing: First, bandpass filtering and wavelet denoising techniques are used to remove interference signals caused by stray currents during subway operation. Then, the short-time average energy (STE) algorithm is used to accurately identify and extract the effective data segments when the train passes by, and remove invalid noise segments.

[0037] Feature extraction: The root mean square (RMS) value reflecting the total energy is calculated for the effective data segment, the kurtosis of the early damage impact characteristics is captured, and the power spectral density (PSD) is obtained through fast Fourier transform (FFT). These features transform the raw waveform data into a lightweight feature vector representing the health status of the vibration isolator, significantly reducing the amount of data stored and transmitted.

[0038] 1.3 Low-power sleep mode The low-power sleep state of the intelligent vibration isolator 1 is achieved through one or more combinations of the following methods: Sensor 1-6-3 is used to trigger the wake-up of host 1-6, and it remains in deep sleep when there is no vibration. Use the I / O port to control the power supply of sensor 1-6-3, and completely cut off the standby power consumption of sensor 1-6-3 during sleep period; The corresponding function is only activated during the preset working time period; all modules are in a dormant state at other times.

[0039] Step 2: Data Upload and Interaction of Inspection Instrument After receiving the data acquisition command from the remote monitoring platform 3, the inspection device 2 establishes a wireless Bluetooth connection with the intelligent vibration isolator 1 in the corresponding area. After the connection is established, the following interaction process is executed: Time synchronization: The inspection instrument 2 sends the current accurate time to the intelligent vibration isolator 1, and the intelligent vibration isolator 1 updates its internal clock to ensure the accuracy of the timestamps of subsequent data collection.

[0040] Data transmission: The inspection instrument 2 sends a data transmission command to the intelligent vibration isolator 1, and the intelligent vibration isolator 1 sends its stored data for the day (including the maximum dynamic force value, static force value and preprocessed feature vector) to the inspection instrument 2.

[0041] Firmware upgrade: If the remote monitoring platform 3 issues a firmware upgrade command, the inspection instrument 2 will send the firmware package data to be upgraded to the intelligent vibration isolator 1 in batches. The intelligent vibration isolator 1 will receive and store the firmware package and automatically complete the upgrade when it is woken up again.

[0042] Entering sleep mode: After data interaction and firmware upgrade are completed, the inspection instrument 2 sends a sleep command to the intelligent vibration isolator 1, and the intelligent vibration isolator 1 re-enters the low-power sleep state.

[0043] The inspection device 2 packages all the collected data and uploads it to the remote monitoring platform 3 via a 4G / 5G network.

[0044] Step 3: Remote monitoring platform data processing and intelligent diagnosis The remote monitoring platform 3 receives and parses the data uploaded by the inspection device 2, performing centralized processing and intelligent analysis. Specifically, this includes: 3.1 Data Preprocessing and Sample Construction Downsampling: The stress time series data of each smart vibration isolator 1 is downsampled to extract the daily average feature. That is, the arithmetic mean of all sampled values ​​of each smart vibration isolator 1 within a single day is calculated to form daily granular stress data. In this embodiment, taking 6 smart vibration isolators deployed in parallel as an example, 6 sets of daily average time series data for 365 days are obtained after one year of monitoring.

[0045] Normalization: Z-score normalization is performed independently on the time-series data of each vibration isolator. The original daily mean is subtracted from the mean of the daily mean for the entire year, and then divided by the standard deviation to eliminate differences in the baseline force values ​​of different vibration isolators. The normalization parameters are calculated only based on the training set and are used for subsequent validation and test sets.

[0046] Sliding window sample construction: A multi-channel sample was constructed using the sliding window method. The window length was set to 30 days, and the sliding step size was 1 day. Each sample contained the daily average features of 6 vibration isolators over 30 consecutive days, forming a tensor of shape (number of samples, 30, 6). In this embodiment, a total of 335 samples were obtained.

[0047] 3.2 Deep Learning Model Structure and Training The remote monitoring platform 3 has a built-in pre-trained multi-channel long short-term memory (LSTM) network model. The structure and training process of this model are as follows: Model structure: Stacked LSTM backbone: A two-layer LSTM is used. The first layer has an input dimension of 6 (corresponding to 6 vibration isolator channels), a hidden layer dimension of 64, and a dropout rate of 0.2. The second layer has an input dimension of 64, a hidden layer dimension of 64, and a dropout rate of 0.2. Long-term dependency features of multi-channel time series data are extracted through the two-layer LSTM.

[0048] Temporal attention layer: The features of each time step output by the LSTM are input into the linear transformation layer, the attention weight of each time step is calculated, and after softmax normalization, the LSTM output is weighted and summed to obtain the feature vector focusing on the key time steps of the disease.

[0049] Classification Header: The weighted feature vectors are passed sequentially through a batch normalization layer, a first fully connected layer (64→128), a ReLU activation function, a second batch normalization layer, and an output layer (128→4), outputting the logits values ​​corresponding to the four health states.

[0050] Training process: Dataset partitioning: The 335 samples were divided into a training set (234 samples), a validation set (67 samples), and a test set (34 samples) in a ratio of 7:2:1.

[0051] Tag definition: A multi-tag classification system is used, with tags created according to engineering rules. Normal state: All vibration isolators are under stable stress, with no sudden changes or continuous shifts.

[0052] Uneven stress state: The stress variance of the 6 vibration isolators is greater than 1500 N².

[0053] Suspended / Settled State: The force on a single vibration isolator is continuously lower than 50% of its average value (suspended) or higher than 150% (settled), and this condition lasts for no less than 7 days.

[0054] Spring breakage condition: The force on a single vibration isolator undergoes a step change, with an amplitude greater than 80%.

[0055] Training parameters: The AdamW optimizer is used, with an initial learning rate of 1e-4 and weight decay of 1e-5; the learning rate is scheduled using StepLR, multiplied by 0.9 every 10 rounds; the loss function is BCEWithLogitsLoss with class weights, and the weights for each class are set as follows: normal: 1.0, uneven force: 5.0, suspended / sinking: 5.0, broken spring: 10.0; an early stopping mechanism is introduced, and training stops if the validation set loss does not decrease for 5 consecutive rounds, with a maximum of 50 training rounds.

[0056] 3.3 Model Reasoning and Disease Diagnosis For the real-time uploaded stress data, the remote monitoring platform 3 constructs input samples using the same preprocessing procedures as the training phase (downsampling, normalization, sliding window). These samples are then input into the trained model to obtain the logits values ​​for each type of defect, which are converted into probabilities using the Sigmoid function. A probability threshold of 0.5 is set; categories with probabilities greater than 0.5 are considered to be in the current health state of the vibration isolator.

[0057] Uneven stress state: If the force variance of M vibration isolators is greater than the preset threshold (e.g., 1500 N²) and the model output probability meets the standard, it is judged as uneven stress.

[0058] Suspension / Settlement Status: If the force on a single vibration isolator is continuously lower than 50% or higher than 150% of its own average value, and the duration is ≥7 days, it is determined to be either suspension or settlement based on the model output.

[0059] Spring breakage status: If a single vibration isolator experiences a step change in force (amplitude > 80%), based on the model output, it is determined to be a spring breakage.

[0060] Normal state: If it does not fall under the above abnormalities, it is considered normal.

[0061] The test set verification showed that the model had a false negative rate of 0 for spring breakage and a false alarm rate of 1. The overall macro-F1 value reached 0.91, and the F1 value for spring breakage was 0.92. The model can output an alarm within 2 days after the fault occurs, which meets the requirements of engineering applications.

[0062] Step 4: Early Warning and Visualization The remote monitoring platform 3 generates visual information based on the diagnostic results and issues warnings when abnormal states are identified. Visualization: The platform front end uses a 3D digital twin scene to dynamically display the geographical distribution of devices. Clicking on the device icon allows you to view real-time stress curves, power consumption, alarm information, etc.; key indicators (number of online devices, number of alarms, data throughput) are refreshed with dynamic charts.

[0063] Early Warning and Maintenance Recommendations: When an abnormal condition of the vibration isolator is diagnosed, the platform immediately issues an early warning (color-coded alert, SMS, or email notification) and automatically generates maintenance recommendation work orders based on the status type. For example, for no-load / overload conditions, on-site verification is recommended within 24 hours; for light load conditions, inspection is recommended within 7 days; and for normal conditions, a maintenance report is generated quarterly. Specifically: Based on the vertical support reaction force data derived from the mechanical characteristic formula of the intelligent vibration isolator by combining the disc spring compression amount with series, parallel, or series-parallel hybrid connections, and integrating real-time torque data, a four-level status judgment rule is established: No-load: Support reaction force F < 5kN, triggering a red alarm; Light load: 5kN ≤ F < 20kN, triggering a yellow alarm; Normal: 20kN ≤ F ≤ 40kN, displaying green; Overload: F > 40kN, triggering a red alarm. An ARIMA time series model is used to train on historical stress data (7-day time window) to predict the stress trend for the next 24 hours. When the predicted value exceeds the threshold, an early warning work order is generated in advance. Based on the status inversion results and trend analysis, the system automatically generates maintenance recommendations: For no-load / overload conditions, it recommends on-site verification within 24 hours, adjusting preload or replacing the vibration isolator; for light-load conditions, it recommends scheduling an inspection within 7 days to check installation tightness; for normal conditions, it generates quarterly maintenance reports to record status trends. The platform registers all in-service vibration isolator health status intelligent monitoring devices, and system management functions include user access control and vibration isolator command issuance. Data storage and reporting: All monitoring data, diagnostic results, and operation and maintenance records are stored in time-series databases and business databases, supporting multi-level group statistics, custom report generation, and scheduled sending, providing data support for long-term performance analysis and maintenance strategy optimization.

[0064] Advantages of this invention: This invention revolutionizes the force acquisition method of vibration isolators from the monitoring hardware perspective. It adopts non-contact measurement technology of electromagnetic induction to obtain the compression of disc spring group 1-8, completely abandoning the traditional easily worn probe contact structure and fundamentally solving the problem of sensor sensitivity decay with use. At the same time, the intelligent vibration isolator 1 has built-in damping fluid to suppress resonance, and the main unit mounting cavity 1-1-3 is filled with epoxy resin potting compound to ensure sealing. Combined with the mechanical characteristic formula of disc spring, it realizes the accurate conversion of compression to force value. It can also simultaneously collect static and dynamic dual-dimensional force data, which greatly improves the accuracy, stability and data comprehensiveness of vibration isolator force monitoring.

[0065] This invention constructs a three-level intelligent monitoring system linking an intelligent vibration isolator 1, an inspection instrument 2, and a remote monitoring platform 3. This system enables digital management of the entire vibration isolator monitoring process. The intelligent vibration isolator 1 can autonomously complete data preprocessing, feature extraction, and local storage. The inspection instrument 2 can read data in batches, forward remote commands, and perform remote firmware upgrades. This replaces the traditional manual periodic inspection and single-point measurement mode, significantly reducing manpower and time costs. It solves the problems of low efficiency, inability to capture continuous changes in force, and biased evaluation results of manual inspections. At the same time, the intelligent vibration isolator 1 is designed with multiple low-power modes, including sensor-triggered wake-up, I / O port-controlled power supply, and timed operation, enabling long-lasting operation without frequent charging or battery replacement. Combined with the remote firmware upgrade function, it avoids the need for complete equipment replacement, significantly reducing equipment maintenance costs and on-site maintenance workload.

[0066] This invention integrates edge computing and deep learning technologies to achieve intelligent diagnosis, trend prediction, and precise operation and maintenance of vibration isolators. The intelligent vibration isolator 1 completes data noise reduction, effective segment extraction, and feature extraction through edge computing. The remote monitoring platform 3, based on a multi-channel LSTM model combined with a temporal attention layer, accurately identifies four types of defects: normal, uneven stress, suspension / settlement, and broken spring. The remote monitoring platform 3 has comprehensive data storage, 3D digital twin visualization, multi-role permission management, and custom report generation functions, and constructs a vibration isolator health status database, providing scientific data support for long-term performance monitoring and maintenance strategy formulation. This significantly improves the intelligence and scientific level of subway vibration isolator operation and maintenance, effectively reduces safety accidents caused by vibration isolator defects, and ensures the safety and efficiency of railway operation.

[0067] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent monitoring system for the stress state of subway vibration isolators, characterized in that, It includes intelligent vibration isolators, inspection instruments, and remote monitoring platforms, which work together to form a three-level linkage intelligent monitoring system; Multiple intelligent vibration isolators are respectively installed under the floating slab of the subway and are communicatively connected to the inspection instrument. Each intelligent vibration isolator has a built-in disc spring assembly. The compression of the disc spring assembly is collected through a non-contact measurement method and converted into vertical support reaction force data of the subway floating slab acting on the vibration isolator. The support reaction force data is preprocessed, features are extracted, and then stored. The intelligent vibration isolator is configured with a low-power operating mode, which only performs data acquisition and processing operations during a set time period and remains in a dormant state at other times. The inspection device is communicatively connected to the remote monitoring platform, and is used to receive operation instructions issued by the remote monitoring platform and send the instructions to the corresponding intelligent vibration isolator; and to collect the vertical support reaction data stored in the intelligent vibration isolator, package the data and upload it to the remote monitoring platform; The remote monitoring platform is used to receive, parse and store the vertical support reaction force data uploaded by the inspection instrument, perform intelligent analysis of the stress state of the vibration isolator based on a deep learning model, identify the fault type and predict the stress change trend, and issue remote operation commands to the inspection instrument for the intelligent vibration isolator.

2. The intelligent monitoring system for the stress state of subway vibration isolators according to claim 1, characterized in that, The intelligent vibration isolator includes a vibration isolator top mount, a vibration isolator base, a disc spring assembly, a sealing sleeve, and a main unit; The top seat of the vibration isolator and the base of the vibration isolator are coaxially arranged vertically. The sealing sleeve is fitted on the outside of the adjacent parts of the two, so that the top seat of the vibration isolator and the base of the vibration isolator are sealed and connected to form a closed cavity. The closed cavity is filled with damping fluid. The disc spring assembly is located in the closed cavity and is elastically supported between the top seat of the vibration isolator and the base of the vibration isolator. The top seat of the vibration isolator can float vertically relative to the base of the vibration isolator with the elastic deformation of the disc spring assembly. The vibration isolator top seat has a downward-facing first protrusion, and the vibration isolator base has an upward-facing second protrusion coaxial with the first protrusion; the bottom end of the first protrusion forms a main unit mounting cavity, which is filled with epoxy resin potting compound, and the main unit is mounted in the main unit mounting cavity through a main unit cover plate; the main unit includes a battery, a main board, a sensor, and a sensor probe arranged sequentially from top to bottom and electrically connected, and the main unit cover plate has a through hole corresponding to the sensor probe; the sensor probe is arranged towards the second protrusion through the through hole, and obtains the compression amount of the disc spring assembly by monitoring the relative distance with the second protrusion, and the main board converts the compression amount into vertical support reaction force data of the subway floating plate acting on the vibration isolator.

3. A method for intelligent monitoring of the stress state of a subway vibration isolator, applied to the intelligent monitoring system for the stress state of a subway vibration isolator as described in any one of claims 1-2, characterized in that, include: Step 1: The intelligent vibration isolator collects the force data of the vibration isolator within a preset working time period, and stores it after preprocessing and feature extraction; During non-working periods, the intelligent vibration isolator is in a low-power sleep state; Step 2: After receiving the data collection command from the remote monitoring platform, the inspection instrument establishes a wireless connection with the intelligent vibration isolator, reads the stored force data, and uploads the force data to the remote monitoring platform. Step 3: The remote monitoring platform processes the force data received from multiple intelligent vibration isolators and uses a pre-trained deep learning model to identify the health status of the vibration isolators. The deep learning model is a multi-channel LSTM model based on multi-sensor time-series data. Step 4: The remote monitoring platform generates visual information based on the diagnostic results and issues an alert when an abnormal state is identified.

4. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 3, characterized in that, In step 1, the force data collected by the intelligent vibration isolator includes dynamic force data and static force data: Dynamic force data: During the period when a vehicle passes by, when the sensor detects vibration, it wakes up the host, measures the maximum value of the compression of the disc spring assembly, and converts it into the maximum value of the dynamic force of the vibration isolator; Static stress data: At the time when no vehicle passes, the compression of the disc spring assembly is directly collected and converted into the static stress value of the vibration isolator.

5. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 3, characterized in that, In step 1, the preprocessing of the vertical support reaction force data by the intelligent vibration isolator includes: using bandpass filtering and wavelet denoising technology to remove stray current interference, and combining the STE short-time average energy algorithm to identify and extract the effective data segment when the train passes. The feature extraction includes: calculating the RMS root mean square value, kurtosis, and power spectral density (PSD) obtained by FFT fast Fourier transform of the support reaction force data, and transforming the original waveform data into a lightweight feature vector representing the health status of the vibration isolator.

6. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 3, characterized in that, In step 1, the low-power sleep state of the intelligent vibration isolator is achieved in the following way: The host is woken up by a sensor trigger. And / or use I / O ports to control the power supply of the sensor, and turn off the sensor's standby power consumption; And / or enable the corresponding function only during the set time period, and keep it in sleep mode the rest of the time.

7. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 3, characterized in that, In step 2, after the inspection device establishes a wireless connection with the intelligent vibration isolator, the following steps are also included: Perform time synchronization and send the current accurate time to the intelligent vibration isolator; Send data transmission instructions to the intelligent vibration isolator and receive the data stored therein for the day; If a firmware upgrade command is available, the firmware package data to be upgraded will be sent to the smart vibration isolator. After the data interaction is completed, the intelligent vibration isolator returns to sleep mode.

8. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 3, characterized in that, In step 3, the remote monitoring platform processes the force data, specifically including: The stress time series data of multiple intelligent vibration isolators are downsampled to extract daily average features; The extracted daily average features are normalized. A sliding window method was used to construct a multi-channel sample, each sample containing the daily mean features of M vibration isolators over N consecutive days, where N≥7 and M≥2. The multi-channel samples are input into the multi-channel LSTM model, which is a stacked LSTM structure, and is used to extract the temporal features of the multi-channel samples.

9. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 8, characterized in that, After extracting time-series features, the following are also included: The features of each time step output by the LSTM layer are input into the temporal attention layer. By calculating the attention weight of each time step, the LSTM output is weighted and summed to obtain the feature vector focused on the key time step. The weighted feature vector is input into the classifier, which outputs the probability that the health status of the vibration isolator belongs to a preset variety of defects. The preset variety of defects includes normal state, uneven stress state, suspended / settled state, and broken spring state.

10. The intelligent monitoring method for the stress state of subway vibration isolators according to claim 9, characterized in that, The criterion for judging the uneven force state is: the force variance among the M vibration isolators is greater than a preset threshold. The criteria for determining the suspended / settled state are: the force on a single vibration isolator is continuously lower than 50% of its own average or higher than 150% of its own average, and the duration is not less than 7 days; The criteria for determining the spring breakage state are: a single vibration isolator experiences a step change in force, with the change amplitude exceeding 80%; The training process of the multi-channel LSTM model adopts a loss function with class weights. The class weight of the broken spring state is higher than that of the uneven force, suspended / sinking state, and the class weight of the uneven force, suspended / sinking state is higher than that of the normal state. In addition, an early stopping mechanism is introduced into the training process. If the loss of the validation set does not decrease for a preset number of consecutive rounds, the training is stopped.

Images