A broadband anti-interference low-power multi-mode intelligent track vibration monitoring sensor

Abstract

The present application provides a kind of broadband anti-interference low-power consumption multi-mode intelligent track vibration monitoring sensor, it is related to track vibration monitoring technical field, shell base is equipped with the sensor fixed plate and shear flexible circuit board being connected, top is battery compartment, the upper portion of battery compartment is provided circuit board, shear flexible circuit board passes through battery compartment hollow and is connected with circuit board, microphone is integrated on circuit board;External vibration is transmitted to shear flexible circuit board through shell base, which transmits vibration signal to acceleration sensor in real time, acceleration sensor converts vibration signal into electrical signal, which is transmitted to circuit board through flexible circuit board, microphone synchronously collects environmental noise signal;Self-adaptive signal conditioning circuit carries out dynamic filtering to electrical signal, extracts broadband vibration characteristics, edge computing unit carries out weighted fusion to characteristic value, generates health index, determines track state and triggers early warning.The present application realizes high-precision, low-power consumption, multi-mode vibration and noise synchronous monitoring.

Description

Technical Field

[0001] This invention relates to the field of track vibration monitoring technology, and in particular to a wideband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor. Background Technology

[0002] With the acceleration of urbanization, subways, as an efficient and convenient mode of public transportation, have been widely used globally. However, the vibration and noise problems generated during subway operation are becoming increasingly prominent, potentially affecting not only the safety and durability of tunnel structures but also negatively impacting buildings along the line, precision instruments, and passenger comfort. Especially in densely populated urban centers, subway vibrations propagate through the tracks and ground, potentially triggering micro-vibrations in surrounding buildings, and long-term accumulation can even lead to structural fatigue damage. Furthermore, subway noise also interferes with the quality of life for residents along the line. Therefore, real-time and accurate monitoring of subway track vibration (including vibration acceleration and displacement) and environmental noise has become a crucial issue for ensuring the safe operation of rail transit and the quality of the urban environment.

[0003] Traditional vibration monitoring technologies primarily rely on large, fixed monitoring equipment or periodic manual inspections. While these methods provide some data support, they have significant limitations: First, fixed monitoring equipment is typically bulky and complex to install, making it difficult to deploy flexibly within the limited space of subway tunnels; second, manual inspections are inefficient, unable to achieve continuous, real-time data acquisition, and struggle to detect sudden vibration anomalies promptly; furthermore, traditional monitoring systems have limited data processing capabilities, usually requiring raw data to be transmitted to a backend server for analysis, resulting in response delays and failing to meet the demands of real-time early warning. With the increasing operational mileage of subways and the rising monitoring requirements, the shortcomings of traditional methods are becoming increasingly apparent, necessitating a more efficient and intelligent monitoring solution.

[0004] In recent years, the rapid development of IoT, edge computing, and micro-sensor technologies has brought new opportunities to the field of rail transit monitoring. By integrating high-precision sensing, data acquisition, and real-time analysis functions into miniaturized, low-power smart sensors, continuous monitoring and immediate analysis of track vibration and noise can be achieved. The introduction of edge computing technology enables sensors to perform data preprocessing and feature extraction locally, significantly reducing data transmission volume, improving system response speed, and reducing the load on cloud servers. Furthermore, the application of intelligent algorithms (such as machine learning) further improves the accuracy and efficiency of vibration signal analysis, enabling the system to automatically identify abnormal vibration patterns and trigger early warnings.

[0005] However, existing intelligent vibration sensors still face many challenges in the complex subway operating environment. For example, the strong electromagnetic interference and high humidity in subway tunnels place stringent requirements on the stability and anti-interference capabilities of the sensors. Simultaneously, to meet long-term monitoring needs, the sensors must possess low power consumption characteristics and support multiple operating modes (such as continuous acquisition and conditional triggering) to adapt to different monitoring scenarios. Furthermore, accurate acquisition and analysis of vibration data require sensors with high sensitivity and wide bandwidth response characteristics, which places high demands on both hardware design and signal processing algorithms.

[0006] Currently, due to the complex operating environment of the subway, traditional vibration sensors face problems such as strong electromagnetic interference, high humidity, and insufficient long-term stability, resulting in reduced data reliability. At the same time, existing equipment has limitations in power consumption control, multi-mode acquisition, and wide-band response, making it difficult to meet the accurate monitoring needs of subway floating slab tracks. Specifically, this manifests as: (1) poor environmental adaptability: strong electromagnetic interference and humid environment in tunnels can easily lead to sensor signal distortion or failure; (2) contradiction between power consumption and battery life: traditional sensors operate at high power consumption continuously, making it difficult to balance long-term monitoring and low energy consumption requirements; (3) single data acquisition mode: lack of intelligent triggering mechanism (such as automatic recording when vibration exceeds the limit), resulting in the accumulation of invalid data; (4) insufficient frequency response and sensitivity: it is difficult to simultaneously capture low-frequency floating slab vibration (4-13Hz) and high-frequency rail vibration (400-780 Hz).

[0007] Therefore, there is an urgent need to provide an intelligent track detection sensor that can achieve high-precision, low-power, multi-mode synchronous monitoring of vibration and noise, providing reliable data support for intelligent operation and maintenance of subways. Summary of the Invention

[0008] To address the aforementioned issues, this invention provides a wideband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor. Through anti-interference hardware design, adaptive signal conditioning, and edge intelligent analysis, it achieves high-precision, low-power, multi-mode synchronous monitoring of vibration and noise, effectively solving the above problems and providing reliable data support for intelligent operation and maintenance of subways.

[0009] To achieve the above objectives, the technical solution of the present invention is implemented as follows: A wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor includes: a housing base fixed to a locking plate; a sensor fixing plate and a shear flexible circuit board are provided on the upper surface of the housing base; the sensor fixing plate and the shear flexible circuit board are fixedly connected; a battery compartment is fixedly connected above the housing base; a circuit board is fixedly disposed above the battery compartment; the shear flexible circuit board passes through a reserved hole in the battery compartment and is electrically connected to the circuit board; a microphone is integrated on the circuit board; a metal shielding shell is provided around the battery compartment; a top cover is closed on the top of the circuit board; and the housing base, the metal shielding shell, and the top cover are tightly connected. The circuit board is a low-power MCU control module circuit board, and integrates an adaptive signal conditioning circuit, an edge computing unit, a wireless communication module and a Huffman coding compression module. The circuit board is connected to an accelerometer, and the accelerometer is electrically connected to the shearing flexible circuit board. External vibrations are transmitted to the shear flexible circuit board via the outer shell base. The shear flexible circuit board transmits the vibration signal to the accelerometer in real time. The accelerometer converts the vibration signal into an electrical signal, which is then transmitted to the circuit board via the flexible circuit board. The microphone simultaneously collects environmental noise signals. The adaptive signal conditioning circuit dynamically filters the electrical signal to extract 4Hz-1kHz wideband vibration features. The edge computing unit performs weighted fusion of the extracted feature values ​​to generate a health index, determine the track status, and trigger an early warning. The Huffman coding compression module compresses the monitored vibration signal and environmental noise signal and transmits them to the cloud platform via the wireless communication module.

[0010] As a further improvement of the present invention, the outer casing base is fixed to the sensor fixing plate and the shearing flexible circuit board by fixing screws respectively, the side of the battery compartment is provided with waterproof rubber, and the battery is embedded inside. The battery is fixed in the battery compartment by the battery cover plate and fixing screws.

[0011] As a further improvement of the present invention, the locking plate is provided with a left locking buckle and a right locking buckle on both sides. The left locking buckle and the right locking buckle are locked with the locking plate through locking studs and locking cups to form a locking mechanism with adjustable clamping force, which fixes the sensor on the small iron plate base on the bottom surface of the track.

[0012] As a further improvement of the present invention, a light guide plate and an antenna cover are fixed on the top cover of the shell. The top cover of the shell is also provided with an interface hole. A Type-C interface is fixedly installed in the interface hole. The Type-C interface is electrically connected to the circuit board and is used for debugging or data export.

[0013] As a further improvement of the present invention, the adaptive signal conditioning circuit is equipped with a dynamic filtering module, which can automatically identify and suppress 50Hz power frequency interference and random noise in the tunnel, and extract 4Hz-1kHz wideband vibration features by combining the FFT fast Fourier transform and order analysis module.

[0014] As a further improvement of the present invention, the control module circuit board has a built-in multi-mode acquisition trigger module, which supports three working modes: threshold trigger, timed acquisition, and continuous acquisition. The wake-up acceleration threshold for threshold trigger is 1g, and the sleep acceleration threshold is 0.75g.

[0015] As a further improvement of the present invention, the edge computing unit has a built-in lightweight machine learning model that performs weighted fusion of the frequency band energy, dominant frequency amplitude, and order feature value of the vibration signal to generate a health index HI with a score of 0-100. Based on the health index HI, the track vibration is determined to be in three states: normal, warning, and abnormal, and the corresponding warning level is triggered.

[0016] As a further improvement of the present invention, the wireless communication module is a LoRaWAN or 5G RedCap module, which supports the deployment of Mesh wireless self-organizing network; the Huffman coding compression module first performs differential coding on the vibration data, and then compresses the data packet through the Huffman coding table before it is transmitted by the wireless communication module.

[0017] As a further improvement of the present invention, the acceleration sensor adopts a high-performance MEMS accelerometer with a bandwidth covering 0-8kHz, which can simultaneously acquire low-frequency vibration data of the floating plate (4-13Hz) and high-frequency vibration data of the rail (400-780Hz).

[0018] As a further improvement of the present invention, the control module circuit board has a built-in temperature compensation model, which can automatically correct the sensitivity drift of the accelerometer in an ambient temperature range of -20℃ to 60℃, and is equipped with a built-in PZT exciter, which can automatically calibrate the sensor accuracy weekly.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention integrates a high-performance vibration sensing module, an adaptive signal conditioning circuit, and an embedded edge computing unit, enabling real-time monitoring of vibration acceleration and environmental noise in subway floating slab tracks. It optimizes data acquisition efficiency through multi-mode acquisition strategies (threshold triggering, timed recording, etc.). Its miniaturized design facilitates concealed deployment, while its robust casing and anti-interference design ensure long-term reliability in harsh environments. This wideband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor not only overcomes the shortcomings of traditional monitoring technologies but also provides strong technical support for the health management, vibration reduction and noise reduction optimization, and intelligent operation and maintenance of rail transit.

[0020] Compared to traditional solutions that are prone to failure in humid, electromagnetic interference, and drastic temperature changes within tunnels, this invention employs a metal shielding housing to effectively isolate strong electromagnetic interference generated by frequency converters, motors, and other components within the tunnel, achieving an IP67 protection rating and ensuring that the sensor's internal circuitry remains unaffected in humid and dusty environments. Furthermore, its adaptive signal conditioning algorithm dynamically identifies and filters 50 Hz power frequency interference and specific noise, rather than relying on fixed hardware filters, thus coping with complex and variable electromagnetic environments. The software incorporates a temperature compensation algorithm for temperatures ranging from -20℃ to 60℃, automatically correcting sensor sensitivity drift caused by temperature changes and ensuring data accuracy.

[0021] This invention addresses the power consumption and battery life trade-off problem in existing solutions by employing a "silent sensing, event-driven, and efficient data feedback" strategy. This includes a dual-threshold wake-up mechanism: most of the time, nodes are in deep sleep, with low-power circuitry monitoring for threshold exceedances. Only when a train approaches and vibrations reach the wake-up threshold does the system activate at full power. Once the train moves away, vibrations fall below the sleep threshold, and the system quickly enters sleep mode. This transforms the "continuous high power consumption" mode into a "momentary high power consumption + long-term ultra-low power consumption" mode, extending battery life from a few hours / days to over a week, thus resolving the biggest bottleneck in deployment and maintenance.

[0022] This invention addresses the issue of limited data acquisition modes in existing solutions by supporting three configurable acquisition modes: a threshold-triggered mode for capturing events such as train passage, which is highly efficient and energy-saving; a timed acquisition mode for actively collecting background environmental vibrations during train-free periods to assess slow structural changes; and a continuous acquisition mode that responds to cloud commands for fault diagnosis or in-depth research. This invention meets the needs of all scenarios, from routine monitoring to event capture and specialized diagnosis, with flexible and intelligent acquisition strategies that avoid data redundancy or omissions. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the appearance of a wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor disclosed in one embodiment of the present invention; Figure 2 This is a cross-sectional structural schematic diagram of a wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor disclosed in one embodiment of the present invention. Figure 3 This is a schematic diagram of the base of a wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor disclosed in one embodiment of the present invention.

[0024] Explanation of reference numerals in the attached figures: 1. Housing base; 2. Flexible circuit board; 3. Sensor fixing block; 4. First fixing screw; 5. Second fixing screw; 6. Third fixing screw; 7. Battery; 8. Battery compartment; 9. Battery cover; 10. Fourth fixing screw; 11. Fifth fixing screw; 12. Control module circuit board; 13. Sixth fixing screw; 14. Light guide plate; 15. Microphone; 16. Antenna cover; 17. Type-C interface; 18. Top cover; 19. Seventh fixing screw; 20. Waterproof rubber; 21. Locking plate; 22. Left locking buckle; 23. Right locking buckle; 24. Locking stud; 25. Locking cup; 26. Button. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0026] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and 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, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0027] In the description of this invention, it should 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 will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0028] The present invention will now be described in further detail with reference to the accompanying drawings: like Figure 1As shown, a wideband anti-interference low-power multi-mode intelligent track vibration monitoring sensor includes: a housing base 1 fixed on a locking plate 21, a sensor fixing plate and a shear flexible circuit board 2 on the upper surface of the housing base 1, the sensor fixing plate and the shear flexible circuit board 2 being fixedly connected, a battery compartment 7 being fixedly connected above the housing base 1, a circuit board being fixedly disposed above the battery compartment 7, the shear flexible circuit board 2 passing through a reserved hole in the battery compartment 7 and being electrically connected to the circuit board, a microphone 15 being integrated on the circuit board, a metal shielding shell being provided around the battery compartment 7, and a top cover 18 being covered above the circuit board, the housing base 1, the metal shielding shell and the top cover 18 being tightly connected; Specifically, such as Figure 3 As shown, the flexible circuit board 2 of the intelligent track vibration monitoring sensor is connected to the sensor fixing block 3 via a first fixing screw 4. The sensor fixing block 3 is connected to the housing base 1 via a second fixing screw 5. The flexible circuit board 2 of the intelligent track vibration monitoring sensor is connected to the housing base 1 via a third fixing screw 6. Figure 2 As shown, the battery compartment module includes a battery compartment 8, a battery cover 9, a fourth fixing screw 10, and waterproof rubber 20. The battery is fixed inside the battery compartment 8 by the battery compartment cover 9 and the fourth fixing screw 10, and the waterproof rubber 20 is fixed to the side of the battery compartment 8; Figure 2 As shown, the sensor control module includes a control module circuit board 12, a sixth fixing screw 13, a microphone 15, a Type-C interface 17, and a button 26. The circuit board 12 is fixed to the battery compartment 8 by the sixth fixing screw 13. The flexible circuit board 2 passes through the pre-reserved hole in the battery compartment 8 and is connected to the corresponding position of the circuit board 12. The battery compartment 8 is connected to the housing base 1 by the sixth fixing screw 13. Figure 1 , 2 The monitoring sensor top cover shown includes a light guide plate 14, a microphone 15, an antenna cover 16, a Type-C interface 17, a top cover 18, a seventh fixing screw 19, and waterproof rubber 20. The antenna cover 16 and the light guide plate 14 are fixed to the top cover 18 at corresponding positions using double-sided adhesive on their own surfaces. The top cover 18 is fixed to the battery compartment 8 using the seventh fixing screw 19. Figure 1 As shown, a locking plate 21 is provided under the base 1 of the outer casing. The left locking buckle 22 and the right locking buckle 23 are respectively placed on both sides of the monitoring sensor and locked with locking studs 24 and locking cups 25.

[0029] The outer casing base 1 is made of high-strength aluminum alloy with anodized surface to enhance corrosion resistance. Its bottom is rigidly connected to the sensor mounting plate via precision-machined threaded holes, ensuring the stability of vibration signal transmission. The sheared flexible circuit board 2 is made of polyimide substrate with a thickness controlled below 0.1mm. It is bonded between the sensor mounting plate and the outer casing base 1 using high thermal conductivity epoxy resin adhesive. Its flexible structure can effectively attenuate mechanical noise under vibration. The battery compartment 7 adopts a sealed cylindrical cavity design, embedding a high-energy-density lithium battery 7 inside. The battery 7 cover is tightly fitted to the inner wall of the battery compartment 7 with 20 waterproof rubber rings. The compression of the 20 waterproof rubber rings is controlled at 30% to ensure IP67 protection rating. The metal shielding shell is stamped from beryllium copper alloy with a thickness of only 0.3mm. It is closed to the outer casing base 1 by laser welding. Its grounding terminal is connected to the circuit board ground layer through spring contacts, forming a complete electromagnetic shielding circuit. The light guide plate 14 and the antenna cover 16 are fixed to the inside of the top cover 18 with double-sided adhesive. The external interface hole is embedded with a Type-C interface, and a waterproof aviation plug is used to achieve a sealed connection between the interface and the circuit board.

[0030] The circuit board is a low-power MCU control module circuit board 12, which integrates an adaptive signal conditioning circuit, an edge computing unit, a wireless communication module and a Huffman coding compression module. The circuit board is connected to an accelerometer, and the accelerometer is electrically connected to the shear flexible circuit board 2. External vibrations are transmitted to the flexible shear circuit board 2 via the outer shell base 1. The flexible shear circuit board 2 transmits the vibration signal to the accelerometer in real time. The accelerometer converts the vibration signal into an electrical signal, which is then transmitted to the circuit board via the flexible circuit board. The microphone 15 simultaneously collects the ambient noise signal. The adaptive signal conditioning circuit dynamically filters the electrical signal and extracts the 4Hz-1kHz wideband vibration features. The edge computing unit performs weighted fusion of the extracted feature values ​​to generate a health index, determine the track status, and trigger an early warning. The Huffman coding compression module compresses the monitored vibration signal and ambient noise signal and transmits them to the cloud platform via the wireless communication module. Specifically, the accelerometer uses the Analog Devices ADXL355 triaxial MEMS chip, with a measurement range covering ±20g and a noise density as low as 20μg / √Hz. It communicates with a low-power MCU (STM32U5 series) via a high-speed SPI interface. The microphone 15 can be a Knowles SPH0641LM4H-1 digital microphone 15, with a sensitivity of -26dBFS and a sampling rate strictly synchronized with the accelerometer (up to 32kHz). The adaptive signal conditioning circuit integrates a dynamic filtering module, employing the LMS adaptive filtering algorithm to suppress 50Hz power frequency interference and random noise generated by fans and frequency converters in the tunnel in real time, with the filter convergence time controlled within 50ms. The FFT operation is implemented based on a fixed-point DSP library, using a Hanning window function to reduce spectral leakage. Order analysis converts the spectral abscissa to the order unit of the rotating component through synchronous resampling of the rotational speed pulse. The edge computing unit incorporates a lightweight machine learning model (TensorFlow Lite Micro framework) with only 20KB of parameters and an inference time of ≤10ms. It generates the Health Index (HI) by weighted fusion of frequency band energy (60%), dominant frequency amplitude (30%), and order deviation (10%). A Huffman coding compression module differentially encodes the vibration signal before constructing a dynamic coding table, achieving a stable compression ratio of ≥3:1. The wireless communication module supports dual-mode transmission of LoRaWAN (spreading factors SF7-SF12) and 5G RedCap (uplink rate 5Mbps). The Mesh network topology uses the AODV routing protocol, with a maximum hop count between nodes of ≤3. The low-power MCU (STM32U5) achieves multi-mode switching via RTC clock and interrupt mechanisms, with current ≤5μA in deep sleep mode and peak current ≤500mA during full-speed sampling. The threshold trigger logic uses a voltage comparator (LM393) to monitor the acceleration signal in real time, with a dual-threshold switching response time of ≤20ms. The timed acquisition mode is triggered by an RTC alarm interrupt, with a time error of ±5ppm. The continuous acquisition mode can be remotely activated via LoRaWAN MAC commands, with a data packet size ≤1KB and a transmission latency ≤100ms.

[0031] In some embodiments, the outer casing base 1 is fixed to the sensor fixing plate and the shear flexible circuit board 2 respectively by the second fixing screw 5 and the third fixing screw 6. The side of the battery compartment 7 is provided with waterproof rubber 20, and the battery 7 is embedded inside. The battery 7 is fixed in the battery compartment 7 by the battery 7 cover plate and the fourth fixing screw 10.

[0032] In some embodiments, locking buckles 22 on the left and 23 on the right are provided on both sides of the locking plate 21. The locking buckles 22 and 23 are locked with the locking plate 21 by locking studs 24 and locking cups 25 to form a locking mechanism with adjustable clamping force, which fixes the sensor on the small iron plate base on the bottom of the track.

[0033] In some embodiments, a light guide plate 14 and an antenna cover 16 are fixed on the top cover 18. The top cover 18 is also provided with an interface hole, in which a Type-C interface is fixedly provided. The Type-C interface is electrically connected to the circuit board for debugging or data export.

[0034] In some embodiments, the adaptive signal conditioning circuit is equipped with a dynamic filtering module, which can automatically identify and suppress 50Hz power frequency interference and random noise in the tunnel. Combined with the FFT fast Fourier transform and order analysis module, it extracts 4Hz-1kHz wideband vibration features.

[0035] In some embodiments, the control module circuit board 12 has a built-in multi-mode acquisition trigger module that supports three working modes: threshold trigger, timed acquisition, and continuous acquisition. The wake-up acceleration threshold for threshold trigger is 1g, and the sleep acceleration threshold is 0.75g.

[0036] The multi-mode triggering module is directly connected to the accelerometer via the GPIO port of the low-power MCU. In threshold triggering mode, when the MCU enters sleep mode, it only maintains power to the acceleration interrupt monitoring circuit. The timed acquisition mode is driven by the RTC clock, triggering an interrupt at a fixed time each day, and performing 10 seconds of data acquisition after waking up. The continuous acquisition mode is remotely activated via LoRaWAN MAC commands, supporting full-speed sampling (32kHz) and segmented uploading (uploading data packets every 5 seconds). Threshold triggering mode: wakes the system when acceleration exceeds 1g, and remains in sleep mode when below 0.75g. Timed acquisition mode: automatically wakes up at 2 AM daily to record background vibration. Continuous acquisition mode: responds to cloud commands to sample at full speed (32kHz) and upload raw data. Specifically, the indicator lights in each mode display the following: (1) Continuous data acquisition mode: Operation indicator light: Solid on (indicates the device is in continuous operation). Data logging light: Flashes regularly (indicating that data is being continuously written). (2) Condition-triggered acquisition mode: Standby status (trigger threshold not reached): Operation indicator light: Flashing slowly (indicating the device is in standby monitoring mode). Data logging light: Solid on (indicates the device is ready but no data is being recorded). Triggered state (reaching the collection threshold): Operation indicator light: flashing rapidly (indicating that the trigger condition has been met) Data logging light: Flashing synchronously (indicating that trigger data is being recorded). In some embodiments, the edge computing unit has a built-in lightweight machine learning model that performs weighted fusion of the frequency band energy, dominant frequency amplitude, and order feature value of the vibration signal to generate a health index HI with a score of 0-100. Based on the health index HI, the track vibration is determined to be in three states: normal, warning, and abnormal, and the corresponding warning level is triggered.

[0037] Specifically, the LoRaWAN module uses the Semtech SX1262 chip, with an adjustable spreading factor of SF7-SF12, a transmission distance of ≥5km (line-of-sight), and a Mesh network node capacity of ≥50. The 5G RedCap module supports the NR-Light protocol, with an uplink rate of 5Mbps and an air interface latency of ≤10ms. The Huffman coding implementation process is as follows: differential operation on 16-bit ADC sampled values ​​→ statistical distribution of the difference frequency → generation of a dynamic coding table → compression of data packets, with a typical compression ratio of 3.2:1.

[0038] In some embodiments, the wireless communication module is a LoRaWAN or 5G RedCap module, supporting Mesh wireless self-organizing network deployment; the Huffman coding compression module first performs differential coding on the vibration data, and then compresses the data packets through the Huffman coding table before transmitting them by the wireless communication module.

[0039] In some embodiments, the accelerometer employs a high-performance MEMS accelerometer with a bandwidth covering 0-8kHz, capable of simultaneously acquiring low-frequency vibration data of the floating plate (4-13Hz) and high-frequency vibration data of the rail (400-780Hz).

[0040] Specifically, the MEMS accelerometer adopts a piezoresistive structure with a measurement range of ±50g, nonlinearity ≤0.5%, and lateral sensitivity ≤5%. The temperature compensation model is based on a BP neural network, with input parameters being temperature sensor (TMP117) data and output being a sensitivity correction coefficient. After compensation, the sensitivity drift is ≤0.1% / ℃. The PZT exciter automatically triggers at 3 AM every week, applying a 20Vpp sinusoidal excitation, and after calibration, the zero-point error is ≤±1mg.

[0041] In some embodiments, the control module circuit board 12 has a built-in temperature compensation model that can automatically correct the sensitivity drift of the accelerometer in ambient temperatures ranging from -20°C to 60°C, and is equipped with a built-in PZT exciter that can automatically calibrate the sensor accuracy weekly.

[0042] Specifically, the temperature compensation algorithm employs polynomial fitting with a compensation accuracy of ±0.5%FS; the PZT exciter (PICeramic CERAMIC-080308) has a voltage drive range of 0-30Vpp, and a calibration cycle of 7 days ±2 hours. The automatic calibration process is as follows: disconnect external vibration input → apply 20Vpp sinusoidal excitation → record the frequency response curve → calculate the zero-point offset → update calibration parameters. Extreme environment testing shows that after the sensor is kept in a -40℃ cryogenic chamber for 24 hours, it can still work normally after restarting, with a data loss rate of ≤0.1%. Example 1:

[0043] The broadband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor of the present invention includes the following core structural features: The outer casing base 1 is made of high-strength aluminum alloy casting with anodized surface treatment. The bottom is rigidly connected to the sensor fixing plate and the shear flexible circuit board 2 by the second fixing screw 5 to ensure stable transmission of vibration signals.

[0044] The flexible circuit board 2 is made of polyimide substrate (0.1mm thick) and is bonded to the sensor fixing plate and the housing base 1 with high thermal conductivity epoxy resin. The flexible structure can attenuate mechanical vibration noise.

[0045] Battery 7 compartment: Cylindrical aluminum alloy cavity, inner wall sprayed with insulating ceramic coating, waterproof rubber 20 rings (compression amount 30%) seals battery 7, and is fixed by battery 7 cover plate and fourth fixing screw 10.

[0046] Metal shielding housing: beryllium copper alloy stamping (thickness 0.3mm), closed with the housing base 1 by laser welding, and the grounding terminal is connected to the circuit board ground layer through spring contacts.

[0047] Locking mechanism: The locking plate 21 is a U-shaped channel steel structure (10mm thick) with a pre-set glue groove at the bottom; the locking buckle assembly (locking buckle left 22, right, locking stud 24, locking cup 25) slides with the locking plate 21 through the threaded hole; during installation, the pre-embedded iron plate base is pre-coated with strong adhesive, the locking buckle clamps the outer shell base 1, and the tapered thread of the locking cup 25 cooperates with the internal thread of the locking buckle to form a self-locking anti-loosening structure.

[0048] In this structure, the sheared flexible circuit board 2 reduces vibration transmission noise, and the metal shielding shell suppresses electromagnetic interference; the locking mechanism ensures the long-term stability of the sensor under track vibration and temperature changes, adapting to concrete / steel sleeper installation scenarios. Example 2:

[0049] The broadband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor of this invention is used to monitor the vibration of subway tracks. The specific process includes: Step 1: Before starting work, install a small iron plate base at the corresponding position on the bottom of the track, and fix the vibration monitoring sensor to the small iron plate base with strong adhesive; clamp the sensor housing base 1 from both sides by locking the left 22 and right sides with the locking buckle, screw in the locking stud 24 and push the locking buckle to the center until it fits tightly, and the tapered thread of the locking cup head 25 and the internal thread of the locking buckle cooperate to form a self-locking anti-loosening structure.

[0050] Step 2: Initialization and Multi-mode Triggering.

[0051] a. System power-on / reset: Software initializes all peripherals: configures ADC (analog-to-digital converter), timers, wireless modules (LoRaWAN / 5G), file system, etc.; loads preset parameters from Flash: trigger threshold (1 g for wake-up, 0.75 g for sleep), filter coefficients, acquisition plan, etc.

[0052] b. Waiting for a trigger event (low power state): The main loop enters low power mode, and the dynamic power management unit shuts down unnecessary clocks and peripherals, maintaining only basic operation.

[0053] The software continuously monitors three trigger conditions, corresponding to three working modes: Threshold Trigger (Main Mode): A background task reads acceleration values ​​at a low frequency (100 Hz) and compares them to a wake-up acceleration threshold of 1g. Once exceeded, the system is immediately woken up and enters threshold trigger mode.

[0054] Timed data acquisition: When the system's internal real-time clock (RTC) reaches the preset time point (2:00 AM every day), an interrupt is generated, waking up the system and entering the timed data acquisition mode.

[0055] Continuous data acquisition (command mode): When a command is received from the cloud via a wireless network, the system is awakened and enters continuous data acquisition mode (used for debugging or special monitoring tasks).

[0056] Step 3: External vibration is transmitted through the outer casing base 1 to the shear flexible circuit board 2, and then to the accelerometer. When the subway approaches the monitoring point where the detection sensor is installed, the acceleration exceeds 1g to wake up the system. The low-power management unit immediately sends a wake-up signal to the control module circuit board 12: the ADC and timer are started in full channel, the MCU switches to working mode, the vibration sensing module (MEMS accelerometer, sampling rate switched to 32kHz) and the noise sensing module (microphone 15, sampling rate 32kHz) are woken up, and the wireless communication module switches from sleep mode to standby mode to prepare for data transmission.

[0057] Step 4, Multimodal Data Acquisition: The vibration sensing module acquires vibration acceleration signals with a bandwidth of 0-8kHz, covering low-frequency vibration of the floating slab (4-13Hz) and high-frequency vibration of the rail (400-780Hz); the noise sensing module simultaneously acquires environmental noise signals from 20Hz to 20kHz; both types of signals are transmitted to the circuit board via the sheared flexible circuit board 2 (flexible vibration damping conduction to avoid mechanical vibration interference), completing the synchronous acquisition.

[0058] Step 5, Adaptive Signal Conditioning: The circuit board performs a 1024-point FFT analysis on the first 100ms of raw data to identify 50Hz power frequency interference and tunnel fan frequency (approximately 50Hz) noise; dynamically generates notch filter parameters to suppress interference in subsequent data streams; converts the spectrum abscissa to order through order analysis (speed pulse synchronous resampling) to stably track the characteristic frequencies of the rotating components of the train wheelset and extracts 4Hz-1kHz wideband vibration features (including spectrum centroid, sideband features, and dominant order amplitude).

[0059] Step 6, Edge Computing Diagnosis: Feature Value Extraction: The circuit board extracts core feature values ​​such as frequency band energy and dominant frequency amplitude from the conditioned data; Health Index Generation: The built-in lightweight decision tree model weights and fuses the feature values, outputting a health index (HI) of 0-100; Graded Diagnosis: If HI≥80 (normal): Only the feature value log is recorded locally, and no warning is triggered; If 60≤HI<80 (Caution): A local low-level warning is triggered (light guide plate 14 lights up yellow), and the feature value + 1 second of raw data is recorded; If HI<60 (abnormal): A high-level alarm is triggered (light guide plate 14 lights up red), and the feature value + 5 seconds of raw waveform data is recorded.

[0060] Step 7, Data Compression and Transmission: Data filtering: Only upload diagnostic results (health index + warning level) and abnormal data fragments (HI<80); do not upload raw data under normal conditions; Huffman coding compression: perform differential coding (convert absolute values ​​to relative values) on the raw waveform data to be uploaded; count the frequency of numerical occurrences and construct a Huffman coding table (high-frequency values ​​are represented by short codewords); complete data compression, with a compression ratio of 3:1~5:1; wireless transmission: the compressed data is uploaded to the cloud platform through a LoRaWAN Mesh self-organizing network, with a packet loss rate of <1% during transmission.

[0061] Step 8, Sleep Recovery: After the subway passes, if the track vibration acceleration value remains below 0.75g (sleep threshold) for 10 seconds, the low power management unit triggers a sleep command, including: shutting down the vibration sensing module, noise sensing module, ADC, and unnecessary timer channels; the MCU switches back to Stop mode, keeping only the threshold monitoring circuit running; the wireless communication module is turned off, and the sensor returns to deep sleep state, waiting for the next trigger.

[0062] Advantages of this invention: This invention enables wideband, high-precision sensing, providing comprehensive coverage for track vibration monitoring needs. Equipped with a high-performance MEMS accelerometer with a bandwidth of 0-8kHz, it can simultaneously and accurately acquire low-frequency vibrations of the floating slab (4-13Hz) and high-frequency vibrations of the rail (400-780Hz). Combined with 1024-point FFT and order analysis, it achieves wideband feature extraction from 4Hz to 1kHz. A single sensor can simultaneously monitor the health status of the track structure and rotating components, overcoming the limitations of insufficient frequency response and sensitivity of traditional sensors, and meeting the full-frequency vibration monitoring needs of rail transit.

[0063] This invention features a multi-dimensional anti-interference design, adaptable to the harsh working environment of tunnels. At the hardware level, it employs a metal shielded shell with IP67 protection, combined with waterproof rubber and grounding, effectively isolating strong electromagnetic interference within the tunnel and resisting the effects of humidity, dust, and temperature fluctuations. At the software level, it dynamically identifies and suppresses 50Hz power frequency interference and random tunnel noise through an adaptive signal conditioning algorithm. It also incorporates a temperature compensation model ranging from -20℃ to 60℃ to automatically correct sensor sensitivity drift. This three-pronged anti-interference solution, encompassing hardware protection, software filtering, and environmental calibration, ensures long-term stable operation of the sensor in complex tunnel environments and distortion-free data acquisition.

[0064] This invention employs intelligent multi-mode data acquisition, balancing data integrity and power consumption control. It supports three configurable operating modes: threshold triggering, timed acquisition, and continuous acquisition. Threshold triggering utilizes a dual-threshold mechanism of 1g wake-up / 0.75g sleep to accurately capture key events such as train passage. Timed acquisition proactively acquires background vibration data during periods without train traffic, while continuous acquisition responds to cloud commands to meet debugging / specialized monitoring needs. Simultaneously, combined with intelligent data decision-making logic, normal data retains only feature values, while abnormal data retains the original waveform. This avoids the accumulation of invalid data and prevents the omission of critical monitoring information, achieving an optimal balance between acquisition efficiency and data value.

[0065] This invention employs an ultra-low power design, significantly improving battery life and ease of maintenance. It features a low-power MCU (STM32U5) and a dedicated low-power management unit, utilizing a "silent sensing, event-driven" strategy. When no events occur, it patrols at a low sampling rate of 1-100Hz, with a deep sleep current of <10μA, maintaining only the threshold monitoring circuit. Only after an event is triggered does it switch to full-speed sampling at 32kHz. Simultaneously, it processes data locally through edge computing, uploading only diagnostic results or abnormal data fragments. Combined with Huffman coding compression technology to reduce wireless transmission power consumption, this extends sensor battery life to over one month, addressing the pain points of traditional sensors' continuous high power consumption and short battery life, and reducing on-site deployment and maintenance costs.

[0066] This invention achieves real-time diagnosis and rapid early warning of monitoring data through edge intelligent analysis. The control module circuit board integrates an embedded edge computing unit with a built-in lightweight decision tree model. It can analyze vibration characteristic values ​​locally, generate a health index of 0-100, and perform a three-level classification diagnosis: a health index ≥80 indicates normal, 60 ≤ health index <80 indicates attention, and a health index <60 indicates abnormal. Different levels correspond to different early warning strategies. In case of an abnormality, it can automatically upload raw waveform data for in-depth cloud analysis. This eliminates the need to transmit massive amounts of raw data to a backend server, significantly reducing cloud processing pressure and enabling real-time identification and rapid early warning of vibration anomalies with a significantly reduced response latency.

[0067] This invention utilizes cross-modal collaborative monitoring to improve the accuracy of fault diagnosis. It integrates a vibration sensing module and a noise sensing module to simultaneously acquire track vibration acceleration signals and environmental noise signals. By treating vibration and noise as different representations of the same monitoring event, spatiotemporal correlation and correlation analysis can be performed in the cloud. This effectively distinguishes the location and type of fault sources (e.g., on / offboard, structural / acoustic problems), achieving more accurate fault location and identification. The diagnostic accuracy is far higher than that of a single-modal sensor, providing a more comprehensive basis for track fault investigation.

[0068] This invention establishes an edge-cloud collaborative architecture to construct an efficient and intelligent monitoring closed loop. It clearly defines the task division among the end (sensing layer: synchronous vibration / noise acquisition), edge (edge ​​layer: local diagnosis, data compression, intelligent triggering), and cloud (cloud layer: cross-modal correlation, in-depth data mining, decision-making). The event-driven mechanism of the edge layer provides precise time triggers for cloud analysis, and the clean signal after adaptive conditioning at the front end lays the data foundation for back-end cross-modal analysis, forming a complete closed loop of "sensing-transmission-diagnosis-decision." This upgrades sensors from simple data loggers to on-site intelligent monitoring terminals, driving the transformation of rail transit operation and maintenance from "passive maintenance" to "proactive prevention."

[0069] This invention features a miniaturized and easily deployable design, adaptable to the limited installation space of rail tracks. It employs a modular structure design, with components compactly connected using screws, double-sided adhesive, and other methods, resulting in a miniaturized overall size. Equipped with a dedicated locking mechanism, a small iron plate base can be installed at a pre-set position on the bottom of the track, and then the sensor can be fixed to the base with strong adhesive, eliminating the need for complex wiring. Simultaneously, it supports LoRaWAN / 5G RedCap for building a mesh wireless self-organizing network, providing coverage without blind spots, and ensuring that a single node failure does not affect the overall network operation. Installation efficiency is improved by 70%, allowing for flexible and concealed deployment in the limited space of subway tunnels, meeting the on-site installation needs of rail transit.

[0070] This invention achieves efficient data transmission while reducing bandwidth and transmission costs. It employs a lightweight data compression algorithm based on Huffman coding. First, it differentially encodes the vibration data to increase redundancy. Then, it constructs an encoding table based on the frequency of numerical values, representing high-frequency values ​​with short codewords, achieving a high compression ratio and significantly reducing wireless transmission bandwidth requirements. Simultaneously, it combines intelligent data filtering through edge computing, uploading only critical data and reducing invalid data transmission. This significantly reduces wireless transmission bandwidth and power consumption costs while ensuring the effective transmission of monitoring information.

[0071] This invention ensures long-term monitoring accuracy through self-calibration and high-reliability design. The sensor features a built-in PZT exciter, enabling automatic weekly calibration. Combined with a temperature compensation model and dynamic calibration technology, it continuously corrects sensor sensitivity, effectively solving the "sensing inaccuracy" problem caused by sensitivity degradation and drift during long-term use. Simultaneously, the hardware employs modular integration, ensuring robust and reliable connections. The wireless mesh self-organizing network possesses disaster recovery capabilities; even if a single node fails, data can be transmitted via dynamic routing. These three aspects—hardware reliability, data accuracy calibration, and network disaster recovery—ensure long-term stable sensor operation and continuously accurate monitoring data.

[0072] The above are merely preferred embodiments of the present invention and do not limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A wideband, anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor, characterized in that, include: A housing base is fixed to a locking plate. The upper surface of the housing base is provided with a sensor fixing plate and a shearing flexible circuit board. The sensor fixing plate is fixedly connected to the shearing flexible circuit board. A battery compartment is fixedly connected to the top of the housing base. A circuit board is fixedly installed on the top of the battery compartment. The shearing flexible circuit board passes through a reserved hole in the battery compartment and is electrically connected to the circuit board. A microphone is integrated on the circuit board. A metal shielding shell is provided around the battery compartment. A top cover is closed on the top of the circuit board. The housing base, the metal shielding shell, and the top cover are tightly connected. The circuit board is a low-power MCU control module circuit board, and integrates an adaptive signal conditioning circuit, an edge computing unit, a wireless communication module and a Huffman coding compression module. The circuit board is connected to an accelerometer, and the accelerometer is electrically connected to the shearing flexible circuit board. External vibrations are transmitted to the shear flexible circuit board via the outer shell base. The shear flexible circuit board transmits the vibration signal to the accelerometer in real time. The accelerometer converts the vibration signal into an electrical signal, which is then transmitted to the circuit board via the flexible circuit board. The microphone simultaneously collects environmental noise signals. The adaptive signal conditioning circuit dynamically filters the electrical signal to extract 4Hz-1kHz wideband vibration features. The edge computing unit performs weighted fusion of the extracted feature values ​​to generate a health index, determine the track status, and trigger an early warning. The Huffman coding compression module compresses the monitored vibration signal and environmental noise signal and transmits them to the cloud platform via the wireless communication module.

2. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The outer casing base is fixed to the sensor mounting plate and the sheared flexible circuit board by fixing screws. The side of the battery compartment is provided with waterproof rubber, and the battery is embedded inside. The battery is fixed in the battery compartment by the battery cover plate and fixing screws.

3. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The locking plate is equipped with a left locking buckle and a right locking buckle on both sides. The left locking buckle and the right locking buckle are locked with the locking plate through locking studs and locking cups to form an adjustable clamping force locking mechanism, which fixes the sensor on the small iron plate base on the bottom of the track.

4. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, A light guide plate and an antenna cover are fixed on the top cover of the housing. The top cover of the housing also has an interface hole, in which a Type-C interface is fixedly installed. The Type-C interface is electrically connected to the circuit board and is used for debugging or data export.

5. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The adaptive signal conditioning circuit is equipped with a dynamic filtering module, which can automatically identify and suppress 50Hz power frequency interference and random noise in the tunnel. Combined with the FFT fast Fourier transform and order analysis module, it extracts 4Hz-1kHz wideband vibration features.

6. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The control module circuit board has a built-in multi-mode acquisition trigger module, which supports three working modes: threshold trigger, timed acquisition, and continuous acquisition. The wake-up acceleration threshold for threshold trigger is 1g, and the sleep acceleration threshold is 0.75g.

7. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The edge computing unit has a built-in lightweight machine learning model that performs weighted fusion of the frequency band energy, dominant frequency amplitude, and order feature value of the vibration signal to generate a health index HI with a score of 0-100. Based on the health index HI, the track vibration is determined to be in three states: normal, warning, and abnormal, and the corresponding warning level is triggered.

8. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The wireless communication module is a LoRaWAN or 5G RedCap module, supporting Mesh wireless self-organizing network deployment; the Huffman coding compression module first performs differential coding on the vibration data, and then compresses the data packets through a Huffman coding table before transmitting them through the wireless communication module.

9. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The accelerometer uses a high-performance MEMS accelerometer with a bandwidth covering 0-8kHz, and can simultaneously acquire low-frequency vibration data of the floating plate (4-13Hz) and high-frequency vibration data of the rail (400-780Hz).

10. The wideband anti-interference, low-power, multi-mode intelligent track vibration monitoring sensor according to claim 1, characterized in that, The control module circuit board has a built-in temperature compensation model that can automatically correct the sensitivity drift of the accelerometer in ambient temperatures ranging from -20℃ to 60℃. It also has a built-in PZT exciter that can automatically calibrate the sensor accuracy weekly.

Images