Digital twin-driven multi-mode passive sensing train bogie monitoring method and system

Abstract

The invention discloses a digital twin-driven multi-mode passive sensing train bogie monitoring method and system, and belongs to the technical field of rail transit vehicles. The system is composed of a vehicle-mounted excitation receiving unit, a distributed multi-mode passive sensor network and a comprehensive diagnosis unit. The vehicle-mounted excitation and receiving unit transmits single-frequency continuous wave radio-frequency signals to construct an electromagnetic energy field and collect frequency spectrum signals, the passive sensing unit induces energy to work, converts measured physical quantities into circuit parameter changes and modulates reflected signals, the comprehensive signal processing and diagnosis unit analyzes and extracts features through high-speed FFT, decouples dynamic and static components, and outputs the dynamic and static components. Health state evaluation is completed based on the three-level evaluation system. The system realizes passive wireless maintenance-free monitoring, breaks through the limitation that traditional passive sensing only measures a single static quantity, avoids signal collision, improves the fault diagnosis accuracy, and meets the monitoring requirement of the whole life cycle of a train.

Description

Technical Field

[0001] This application relates to the field of rail transit vehicle technology, and in particular to a digital twin-driven multimodal passive sensing method and system for monitoring train bogies. Background Technology

[0002] The operational safety of rail transit vehicles directly depends on the service condition of key bogie components. Currently, monitoring of train bogie components mainly relies on wired sensors and active wireless sensors. Wired monitoring methods face challenges in the complex mechanical environment of the bogie, including difficulties in wiring, cable fatigue and breakage, and electromagnetic interference, making them difficult to deploy on a large scale. Active wireless sensors are limited by battery life and size, and their maintenance costs are extremely high in the high-temperature, high-vibration environment under the train, with frequent battery replacements, making it difficult to meet the maintenance-free monitoring requirements throughout the train's entire lifecycle.

[0003] While existing passive sensing technologies based on surface acoustic wave (SAW) or radio frequency identification (RFID) have addressed the power supply issue to some extent, they are mostly limited to low-frequency sampling and monitoring of single static physical quantities. They lack integration with digital twin technology, failing to achieve real-time mapping between physical sensor data and virtual models. This results in isolated monitoring data and difficulties in tracing the entire lifecycle status. Under high-speed train operating conditions, static data cannot reflect the dynamic vibration, impact, and high-frequency fault characteristics of structural components. Furthermore, when multiple passive sensors are densely deployed within the confined space of a bogie, traditional technologies often face signal collision and recognition delay issues, lacking a comprehensive monitoring method that can simultaneously consider multimodal static parameters, high-frequency dynamic vibration characteristics, and multi-node parallel acquisition.

[0004] Existing technology discloses a passive wireless vibration sensing and monitoring system for freight train bogies. This system comprises monitoring sensors, a bogie monitoring system, an axle-end tachometer generator system, a trackside data base station, and a remote monitoring and diagnostic center. It powers the onboard passive wireless bogie monitoring system and provides speed signals via a miniature tachometer generator installed at the axle end of the freight train. The bogie monitoring system synchronously collects data and transmits it to the trackside data base station, which then transmits it to the remote monitoring and diagnostic center. However, this technical solution utilizes a miniature tachometer generator for power generation, meaning the equipment still requires batteries and a power source to operate. It fails to truly achieve the requirement of passive, maintenance-free monitoring and does not construct a digital twin model of the bogie, thus hindering fault tracing, lifespan prediction, and optimization decision-making through virtual-physical collaboration. Summary of the Invention

[0005] To address the technical problems of complex wiring in traditional wired monitoring, high maintenance costs in active wireless monitoring, and the limitations of existing passive sensing technologies in monitoring only single static physical quantities, the susceptibility of multi-sensor signals to collisions, and low fault diagnosis accuracy, this invention provides a digital twin-driven multimodal passive sensing method and system for monitoring train bogies. The technical solution is as follows: On the one hand, a digital twin-driven multimodal passive sensing train bogie monitoring system is provided, including an on-board excitation and receiving unit, a distributed passive multimodal sensing network, and a comprehensive signal processing and diagnostic unit; The vehicle-mounted excitation and receiving unit is used to transmit at a frequency of The single-frequency continuous wave radio frequency signal is used to construct an electromagnetic energy field covering the bogie area, and the electromagnetic spectrum signal in the area is collected in real time. The distributed passive multimodal sensing network consists of multiple passive sensing units. These passive sensing units operate by sensing electromagnetic energy, converting the measured physical quantity into the circuit parameter capacitance. or inductor The change in this modulates the reflected signal, generating a sideband frequency offset containing physical quantity information. ,and The sideband signal satisfies ; The integrated signal processing and diagnostic unit is used to perform high-speed FFT analysis on the acquired spectral signals and extract... Based on its time-varying characteristics, frequency domain filtering technology is used to decouple the quasi-static and dynamic components of the signal, demodulate the static values ​​and dynamic waveforms of each physical parameter, and complete the bogie health status assessment based on a three-level evaluation system.

[0006] Optionally, the vehicle-mounted excitation and reception unit includes an excitation subunit and a reception subunit; The excitation subunit includes at least a radio frequency signal source, a power amplifier, and a transmitting antenna to provide a stable radio electromagnetic energy field. The excitation subunit supports frequency and power adjustment. The receiving subunit includes a high-gain receiving antenna, a low-noise amplifier, a mixer, and a high-speed analog-to-digital converter module. It features an embedded real-time spectrum analyzer architecture, enabling it to perform fast Fourier transforms at high frequencies, with a system sampling rate of... Satisfying the Nyquist sampling theorem to achieve maximum capture The dynamic fluctuation of the measured parameter at a frequency of / 2.

[0007] Optionally, the distributed passive multimodal sensing network includes a passive pressure sensing unit, a passive temperature sensing unit, a passive humidity sensing unit, and a passive electromagnetic sensing unit. The passive pressure sensing unit uses a pressure-sensitive capacitor or a piezoelectric stack structure as the sensing element, and the bolt preload is... Changes cause capacitance value Change, and thus Changes have occurred; The passive temperature sensing unit uses a temperature-sensitive capacitor material as the sensing element to detect ambient temperature. Changes cause capacitance value Change, and thus Changes have occurred; The passive humidity sensing unit uses a humidity-sensitive capacitor material as the sensing element; changes in ambient humidity cause changes in capacitance. Change, and thus Changes have occurred; The passive electromagnetic sensing unit uses a magnetic inductor as the sensing element, and the external magnetic field strength... Changes alter the permeability of the magnetic core This leads to an increase in inductance. Change, make An offset has occurred.

[0008] Optionally, the passive pressure sensing unit is deployed at the bolts of the key connection parts of the bogie, wherein the bolts of the key connection parts include at least the axle box end cover mounting bolts, gearbox hanger node bolts, and primary / secondary suspension system connection bolts of the bogie, for monitoring the axial preload of the bolts at the key connection parts; The passive temperature sensing unit is deployed on the outer surface of the bogie friction components and high-power electrical components. The outer surface includes at least the outer surface of the bogie axle box bearing seat, the traction motor housing, and the outer wall of the gearbox oil sump, and is used to monitor the thermal state of the friction components and high-power electrical components. The passive humidity sensing unit is deployed in the sealed area and electrical protection area of ​​the bogie's oil lubrication components. The area includes at least the vicinity of the bogie's gearbox breather, the labyrinth seal of the axle box rear cover, and the interior of the electrical junction box, for monitoring the sealing performance of the oil lubrication components and the ambient humidity. The passive electromagnetic sensing unit is deployed in the traction motor and surrounding electromagnetic sensitive areas of the bogie. The electromagnetic sensitive areas include at least the vicinity of the stator winding of the traction motor and the vicinity of the speed sensor probe, and are used to monitor the electromagnetic environment of the traction motor and its surroundings.

[0009] Optionally, the health status assessment system of the integrated signal processing and diagnostic unit includes three levels: single-parameter threshold diagnosis, passive dynamic feature analysis, and multi-parameter fusion early warning. The integrated signal processing and diagnostic unit decouples signals using frequency domain filtering technology. The process of decoupling signals using frequency domain filtering technology includes: For raw time series data Applying a low-pass filter to extract quasi-static components ,right Apply a high-pass filter and de-stress to extract dynamic components. .

[0010] Optionally, the criteria for the single-parameter threshold diagnosis include: Criteria for loosening bolts Hot axis criteria Criteria for seal failure Electromagnetic anomaly criteria Deviation from the benchmark value exceeds ;in, Depend on The equivalent monitoring quantities obtained by mapping include , , , , respectively, represent the quasi-static monitoring quantities of the pressure mode, temperature mode, humidity mode, and magnetic field mode; , , For the preset threshold, , , and It is set by at least one of the following: vehicle factory calibration, historical health baseline statistics, and operation and maintenance experience, and supports adaptive updates based on operating conditions.

[0011] Optionally, the passive dynamic feature analysis refers to the dynamic components obtained from the original time series data through high-pass filtering and detrending. Feature extraction and anomaly characterization are performed using a sliding window, wherein the feature extraction includes at least one of the following three types of features: The first category is general dynamic features, which are all applied within a sliding window. The calculations include at least the root mean square value, variance, standard deviation, peak value, peak-to-peak value, absolute mean value, skewness, kurtosis, peak factor, impulse factor, margin factor, waveform factor, window energy, sliding window coefficient of variation, short window to long window energy ratio, short window to long window root mean square ratio, first-order difference sequence characteristics, and second-order difference sequence characteristics. The second category is frequency-time strongly correlated features. These features are used to characterize the periodic modulation and cyclic statistical properties of passive frequency point timing. They include at least spectral centroid, spectral bandwidth, spectral flatness, spectral entropy, spectral kurtosis, and cyclic stationary features. The cyclic stationary features include cyclic autocorrelation features and / or spectral correlation features. The third category comprises passive signal-specific features. These features are designed for the upper and lower sideband structures formed on both sides of the carrier frequency by passive modulation. The passive signal-specific features of the third category include at least one of the following features: Feature 1, sideband symmetry feature, in the first Within a sliding window, extract the sideband energy of the upper and lower sidebands. , And define the sideband imbalance index. And further define sideband symmetry ,in, To avoid extremely small positive numbers with a denominator of zero, , The spectrum within the window in the sideband neighborhood The power spectral density is obtained by summing. For passive excitation carrier frequency, This represents the offset of the sideband relative to the carrier frequency. The sideband energy integral bandwidth; Feature 2, passive modulation stability index, in the first... Estimate sideband offset frequency within a sliding window and in length of Constructing frequency stability quantities on window sequences Meanwhile, define the sideband detection indicator. And further calculate the peak drop rate. Finally, the passive modulation stability index is defined. ,in, , These are the weighting coefficients. A larger value indicates more stable passive modulation and more reliable frequency timing; among which, the sideband detection indicator is defined. The content includes: In response to the When the sideband peak of a window is detectable and exceeds the threshold ,otherwise .

[0012] Optionally, the multi-parameter fusion early warning establishes a multi-dimensional state space of "force-thermal-humidity-magnetic-vibration" and outputs hierarchical early warning results using a passive reliability adaptive weighted fusion method, specifically including: First, state space construction, including in the... Constructing a state vector within a sliding window ; in, These are the quasi-static residual eigenvectors and / or dynamic eigenvectors for pressure, temperature, humidity, and magnetic field modes, respectively. For the reason The extracted dynamic feature vectors corresponding to the "vibration" modes, and The sideband symmetry characteristic and passive modulation stability index; Second, evidence of single-modal anomalies, including constructing anomaly scores for each modality. ( ),satisfy , and when If it is a scalar, then ,in, It is a norm 2; Third, the fusion of passive credibility and adaptive weighted credential assessment, including the construction of global credibility. , The adjustment coefficient is used to determine the fusion weight. To implement differentiated weights for different modalities, it is supported to... Expanded to modal credibility ,at this time Weighted calculation of fusion anomaly scores ; Fourth, a tiered early warning algorithm will... The warning results are compared with a preset threshold and include: in Determined by health benchmark statistics, calibration experiments, or operational experience, and can be based on... Adaptive adjustments are made to suppress false alarms caused by passive reading fluctuations.

[0013] Optionally, the system adopts a frequency division multiplexing strategy to achieve anti-collision parallel acquisition of multiple passive nodes. The passive reflection spectrum is pre-planned into exclusive working large frequency bands that do not overlap. The sensing unit of each physical mode corresponds to a preset fixed exclusive working large frequency band: the pressure mode sensing unit corresponds to the f1~f2 frequency band, the temperature mode sensing unit corresponds to the f3~f4 frequency band, the humidity mode sensing unit corresponds to the f5~f6 frequency band, and the magnetic field mode sensing unit corresponds to the f7~f8 frequency band. Within the dedicated operating frequency band of the same physical mode, sub-channels are further subdivided. The sensing units at different monitoring points under each mode correspond to a unique dedicated sub-channel. The receiving unit locks the corresponding physical mode through the frequency band where the spectrum peak is located, and then locks the corresponding sensor position through the sub-channel within the frequency band, thus realizing IDless identification and parallel acquisition.

[0014] On the other hand, a digital twin-driven multimodal passive sensing method for monitoring train bogies is provided, applicable to the aforementioned system, characterized by comprising: Step S1: Construct an electromagnetic energy field covering the key areas of the bogie; the onboard excitation unit continuously emits at a frequency of [frequency value missing]. A single-frequency carrier signal; Step S2, the vehicle-mounted receiving unit uses a sampling rate Continuously execute Fast Fourier Transform to obtain real-time spectrum frames; Step S3, search for the location in each frame of spectrum. Sideband peaks on both sides, locked sideband frequency Record peak frequency points and construct original time series data of sideband frequency changes over time. ; Step S4, for Apply low-pass and high-pass filters respectively and perform de-trendization to extract quasi-static components. and dynamic components ; Step S5, will Substituting the frequency-parameter calibration curve, the pressure mode is obtained by calculation. Temperature modes Humidity mode Magnetic field modes The corresponding quasi-static monitoring quantity; Step S6, based on Perform single-parameter threshold diagnosis, based on Perform passive dynamic feature analysis and combine it with multimodal correlation matrix to complete multi-parameter fusion early warning, and realize comprehensive assessment of bogie health status.

[0015] Therefore, compared with the prior art, this application achieves at least the following technical effects.

[0016] This application discloses a digital twin-driven multimodal passive sensing method and system for monitoring train bogies, belonging to the field of rail transit vehicle technology. The system consists of an onboard excitation and receiving unit, a distributed multimodal passive sensing network, and a comprehensive diagnostic unit. The onboard excitation and receiving unit transmits a single-frequency continuous wave radio frequency signal to construct an electromagnetic energy field and collects spectral signals. The passive sensing unit senses the energy and converts the measured physical quantity into changes in circuit parameters and modulates the reflected signal. The comprehensive signal processing and diagnostic unit extracts features through high-speed FFT analysis, decouples dynamic and static components, and completes a health status assessment based on a three-level evaluation system. This system achieves passive, wireless, and maintenance-free monitoring, breaking the limitation of traditional passive sensing that only measures a single static quantity, avoiding signal collisions, improving fault diagnosis accuracy, and meeting the monitoring needs of the entire train lifecycle. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of the digital twin-driven multimodal passive sensing train bogie monitoring system of the present invention; Figure 2 This is a schematic diagram of the signal processing flow of the multi-sensor passive monitoring method for train bogies of the present invention; Figure 3 This is a flowchart illustrating the digital twin-driven multimodal passive sensing method for monitoring train bogies according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0019] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. Example 1

[0020] This embodiment provides a digital twin-driven multimodal passive sensing train bogie monitoring system, such as... Figure 1 The diagram shown is a structural schematic of the digital twin-driven multimodal passive sensing train bogie monitoring system of the present invention. The system includes an on-board excitation and receiving unit, a distributed passive multimodal sensing network, and a comprehensive signal processing and diagnostic unit.

[0021] 1. Onboard excitation and receiving unit.

[0022] Used for transmission frequency A single-frequency continuous wave radio frequency signal is used to construct an electromagnetic energy field covering the bogie area, and the electromagnetic spectrum signal in the area is collected in real time.

[0023] The vehicle-mounted excitation and receiving unit is installed on the vehicle body underframe or bogie, and one or more units may be installed.

[0024] Second, distributed passive multimodal sensor networks.

[0025] Composed of multiple passive sensing units, the passive sensing units work by sensing electromagnetic energy, converting the measured physical quantity into the circuit parameter capacitance. or inductor The change in this modulates the reflected signal, generating a sideband frequency offset containing physical quantity information. ,and The sideband signal satisfies .

[0026] The distributed passive multimodal sensor network includes passive sensing units for four modes: pressure, temperature, humidity, and magnetic field. Each unit operates through a sideband modulation mechanism and does not require battery power.

[0027] Third, integrated signal processing and diagnostic unit.

[0028] Used for high-speed FFT analysis of acquired spectral signals to extract... Based on its time-varying characteristics, frequency domain filtering technology is used to decouple the quasi-static and dynamic components of the signal, demodulate the static values ​​and dynamic waveforms of each physical parameter, and complete the bogie health status assessment based on a three-level evaluation system.

[0029] The frequency domain filtering technology of the integrated signal processing and diagnostic unit specifically involves applying low-pass filtering to the original time series data to extract quasi-static components, applying high-pass filtering and de-trending to extract dynamic components, and the three-level evaluation system includes single-parameter threshold diagnosis, passive dynamic feature analysis, and multi-parameter fusion early warning.

[0030] This embodiment is based on Figure 1 The system structure diagram shown is as follows. Figure 1 The connection relationship between the on-board excitation and receiving unit, the distributed passive multimodal sensor network and the integrated signal processing and diagnostic unit can be seen. Each passive sensing unit is distributed in key parts of the bogie.

[0031] Therefore, the technical effect of this application embodiment is that it realizes passive wireless maintenance-free monitoring of bogie status, breaks the limitation of traditional passive sensing that can only monitor a single static physical quantity, and simultaneously acquires static parameters and dynamic vibration characteristics; further, through passive dynamic feature extraction (general / frequency timing / proprietary features) and multi-parameter reliability weighted fusion, it provides "static-dynamic-fusion" three-dimensional data support for bogie fault diagnosis, and improves the comprehensiveness and accuracy of fault diagnosis. Example 2

[0032] This application provides a further description of the vehicle-mounted excitation and receiving unit.

[0033] In one possible implementation, the vehicle-mounted excitation and reception unit includes an excitation subunit and a reception subunit.

[0034] The excitation subunit includes at least a radio frequency signal source, a power amplifier, and a transmitting antenna to provide a stable radio electromagnetic energy field. The excitation subunit supports frequency and power adjustment, and those skilled in the art can add or remove components according to design requirements.

[0035] The receiver subunit includes a high-gain receiving antenna, a low-noise amplifier, a mixer, and a high-speed analog-to-digital converter module. It features an embedded real-time spectrum analyzer architecture, capable of performing fast Fourier transforms at high frequencies, and a system sampling rate of [missing information]. Satisfying the Nyquist sampling theorem to achieve maximum capture The dynamic fluctuation of the measured parameter at a frequency of / 2.

[0036] The excitation subunit can transmit a single-frequency continuous wave in the range of 200MHz to 1GHz, and the transmission power is adjustable to cover the entire bogie monitoring area. The high-speed analog-to-digital conversion module of the receiving subunit ensures high-frequency FFT operations, ensuring efficient acquisition and real-time response of the sideband signals reflected by the passive sensing unit.

[0037] This demonstrates that by adapting the power-adjustable excitation mechanism to meet the energy coverage requirements of the complex monitoring area of ​​the bogie, and leveraging the high-frequency FFT computing capability, the physical quantity change information in the sideband signal can be accurately captured. At the same time, the combination of the high-gain receiving antenna and low-noise amplifier in the receiving subunit effectively improves the acquisition capability of weak sideband signals, providing high-quality raw data for subsequent passive dynamic feature (such as spectral kurtosis and sideband symmetry) extraction and health status assessment system, reducing the impact of environmental electromagnetic interference on feature analysis and fault diagnosis, and ensuring the effectiveness of multimodal sensing data and the accuracy of diagnostic results. Example 3

[0038] This application provides further explanation of distributed passive multimodal sensor networks.

[0039] In one possible implementation, the distributed passive multimodal sensing network includes passive pressure sensing units, passive temperature sensing units, passive humidity sensing units, and passive electromagnetic sensing units. This enables passive monitoring of four physical parameters: pressure, temperature, humidity, and magnetic field. Each modal sensing unit uses a unified radio frequency interface, ensuring strong compatibility. Changes in the output physical quantities can be directly mapped to passive dynamic characteristics (such as kurtosis corresponding to the pressure mode and spectral correlation characteristics corresponding to the magnetic field mode), providing multidimensional raw data for the subsequent three-level evaluation system. Compared to single-modal passive sensing, it more accurately supports multidimensional state analysis of "force-heat-humidity-magnetism-vibration," which will be described in detail below.

[0040] 1. Passive pressure sensing unit.

[0041] Using pressure-sensitive capacitors or piezoelectric stacks as the sensing element, bolt preload... Changes cause capacitance value Change, and thus Things have changed.

[0042] Among them, the pressure-sensitive capacitor senses the preload force through changes in the electrode spacing, and the piezoelectric stack structure generates changes in electrical signals through deformation.

[0043] Optionally, the passive pressure sensing unit is deployed at the bolts of the bogie's critical connection parts. The bolts of the critical connection parts include at least the axle box end cover mounting bolts, gearbox hanger node bolts, and primary / secondary suspension system connection bolts of the bogie. It is used to monitor the axial preload of the bolts of the critical connection parts. The bolts at which the passive pressure sensing unit is deployed are all critical nodes of the bogie connection. Loose bolts can easily cause major safety hazards.

[0044] Second, passive temperature sensing unit.

[0045] Thermosensitive capacitor material is used as the sensing element, and the ambient temperature... Changes cause capacitance value Change, and thus Things have changed.

[0046] Among them, the temperature-sensitive capacitor uses temperature-sensitive ceramic material, which has a stable temperature coefficient.

[0047] The passive temperature sensing unit is deployed on the outer surface of the bogie friction components and high-power electrical components. The outer surface includes at least the outer surface of the bogie axle box bearing housing, the traction motor housing, and the outer wall of the gearbox oil sump. It is used to monitor the thermal state of the friction components and high-power electrical components. The axle box bearing housing where the passive temperature sensing unit is deployed is a high-incidence area for hot axle failures, and the traction motor and gearbox are the main heat-generating components of the bogie.

[0048] Third, passive humidity sensing unit.

[0049] Using a humidity-sensitive capacitor material as the sensing element, the capacitance value changes with the ambient humidity. Change, and thus Things have changed.

[0050] Among them, the humidity-sensitive capacitor adopts an interdigitated electrode structure and utilizes the high dielectric constant of water to achieve humidity monitoring.

[0051] The passive humidity sensing unit is deployed in the sealed area and electrical protection area of ​​the bogie's oil lubrication components. The areas include at least the vicinity of the bogie's gearbox breather, the labyrinth seal of the axle box rear cover, and the inside of the electrical junction box. It is used to monitor the sealing performance of the oil lubrication components and the ambient humidity. The passive humidity sensing unit is deployed in areas where sealing failure and moisture intrusion are likely to occur.

[0052] Fourth, passive electromagnetic sensing unit.

[0053] Using a magnetic inductor as the sensing element, the external magnetic field strength Changes alter the permeability of the magnetic core This leads to an increase in inductance. Change, make An offset has occurred.

[0054] The magnetosensitive inductor uses a magnetostrictive material to wrap the coil, and the permeability of the magnetic core changes with the strength of the external magnetic field. Each modal sensing unit operates based on a sideband modulation mechanism. The offset reflects the change in the measured physical quantity.

[0055] The passive electromagnetic sensing unit is deployed in the traction motor and surrounding electromagnetic sensitive areas of the bogie. The electromagnetic sensitive areas include at least the vicinity of the stator winding of the traction motor and the vicinity of the speed sensor probe. It is used to monitor the electromagnetic environment of the traction motor and its surroundings. The deployment position of the passive electromagnetic sensing unit can directly sense the magnetic field changes when the traction motor is working and the electromagnetic environment of the speed sensor.

[0056] Therefore, the above-mentioned sensor unit deployment locations are precisely matched with the high-incidence areas of bogie failure, enabling targeted monitoring of the status of key components; combined with parameters such as kurtosis (impact characteristics) and spectral centroid (frequency distribution) in passive dynamic feature analysis, early fault characteristics such as bolt loosening, bearing wear, and seal failure can be captured earlier, avoiding monitoring blind spots and increasing the advance warning of faults, thereby further improving the timeliness and effectiveness of the warning. Example 4

[0057] This application provides a further description of the integrated signal processing and diagnostic unit.

[0058] In one possible implementation, the health status assessment system of the integrated signal processing and diagnostic unit includes three levels: single-parameter threshold diagnosis, passive dynamic feature analysis, and multi-parameter fusion early warning.

[0059] The integrated signal processing and diagnostic unit decouples signals using frequency domain filtering techniques. The process of decoupling signals using frequency domain filtering techniques includes: For raw time series data Applying a low-pass filter to extract quasi-static components ,right Apply a high-pass filter and de-stress to extract dynamic components. Among them, the quasi-static component corresponds to the steady-state values ​​of physical quantities such as pressure and temperature, while the dynamic component corresponds to transient changes such as vibration and impact.

[0060] Criteria for single-parameter threshold diagnosis include: Criteria for loosening bolts Hot axis criteria Criteria for seal failure Electromagnetic anomaly criteria Deviation from the benchmark value exceeds , among which, by The equivalent monitoring quantities obtained by mapping include , , , , respectively, represent the quasi-static monitoring quantities for the pressure mode, temperature mode, humidity mode, and magnetic field mode. , , For the preset threshold, , , and It is set by at least one of the following: vehicle factory calibration, historical health benchmark statistics, and operation and maintenance experience, and supports adaptive updates according to operating conditions; For passive dynamic feature analysis, passive dynamic feature analysis refers to the dynamic components obtained from the original time series data after high-pass filtering and detrending. Feature extraction and anomaly representation are performed using a sliding window, and the feature extraction includes at least one of the following three types of features: The first category is general dynamic features, which are all applied within a sliding window. The calculations include at least the root mean square value, variance, standard deviation, peak value, peak-to-peak value, absolute value of mean, skewness, kurtosis, peak factor, impulse factor, margin factor, waveform factor, window energy, sliding window coefficient of variation, short window-to-long window energy ratio, short window-to-long window root mean square ratio, first-order difference sequence characteristics, and second-order difference sequence characteristics, where the peak value is the maximum absolute value, and the peak-to-peak value is the difference between the minimum and maximum values. The second category is frequency-time strongly correlated features. These features are used to characterize the periodic modulation and cyclic statistical properties of passive frequency point timing. They include at least spectral centroid, spectral bandwidth, spectral flatness, spectral entropy, spectral kurtosis, and cyclic stationary features. Cyclic stationary features include cyclic autocorrelation features and / or spectral correlation features. The third category comprises passive signal-specific features. These features are designed for the upper and lower sideband structures formed on both sides of the carrier frequency by passive modulation. The third category of passive signal-specific features includes at least one of the following features: Feature 1, sideband symmetry feature, in the first Within a sliding window, extract the sideband energy of the upper and lower sidebands. , And define the sideband imbalance index. And further define sideband symmetry ,in, To avoid extremely small positive numbers with a denominator of zero, , The spectrum within the window in the sideband neighborhood The power spectral density is obtained by summing. For passive excitation carrier frequency, This represents the offset of the sideband relative to the carrier frequency. The sideband energy integral bandwidth; Feature 2, passive modulation stability index, in the first... Estimate sideband offset frequency within a sliding window and in length of Constructing frequency stability quantities on window sequences Meanwhile, define the sideband detection indicator. And further calculate the peak drop rate. Finally, the passive modulation stability index is defined. ,in, , These are the weighting coefficients. A larger value indicates more stable passive modulation and more reliable frequency timing; among which, the sideband detection indicator is defined. The content includes: In response to the When the sideband peak of a window is detectable and exceeds the threshold ,otherwise .

[0061] Therefore, the multi-parameter fusion early warning system establishes a multi-dimensional state space of "force-thermal-humidity-magnetic-vibration" and uses a passive reliability adaptive weighted fusion method to output hierarchical early warning results, specifically including: First, state space construction, including in the... Constructing a state vector within a sliding window ; in, These are the quasi-static residual eigenvectors and / or dynamic eigenvectors for pressure, temperature, humidity, and magnetic field modes, respectively. For the reason The extracted dynamic feature vectors corresponding to the "vibration" modes, and The sideband symmetry characteristic and passive modulation stability index; Second, evidence of single-modal anomalies, including constructing anomaly scores for each modality. ( ),satisfy , and when If it is a scalar, then ,in, It is a norm 2; Third, the fusion of passive credibility and adaptive weighted credential assessment, including the construction of global credibility. , The adjustment coefficient is used to determine the fusion weight. To implement differentiated weights for different modalities, it is supported to... Expanded to modal credibility ,at this time Weighted calculation of fusion anomaly scores ; Fourth, a tiered early warning algorithm will... The warning results are compared with a preset threshold and include: in Determined by health benchmark statistics, calibration experiments, or operational experience, and can be based on... Adaptive adjustments are made to suppress false alarms caused by passive reading fluctuations.

[0062] Therefore, by introducing three types of quantitative feature extraction formulas into passive dynamic feature analysis (the sliding window calculation formula for general dynamic features, the calculation logic for spectral centroid / spectral kurtosis of frequency-time strongly correlated features, and the sideband imbalance index and modulation stability index formulas for passive-specific features), not only are the problems of traditional dynamic feature analysis relying on qualitative judgment and insufficient adaptation to passive signals solved, but the sideband symmetry feature formula can accurately identify imbalance anomalies on both sides of the carrier frequency (such as modulation deviation caused by component loosening). The modulation stability index formula, through the quantitative calculation of frequency stability and peak drop rate, effectively filters invalid data caused by passive signal fluctuations, improving the reliability of dynamic features. Simultaneously, in multi-parameter fusion early warning... By constructing a multi-dimensional state vector formula for "force-heat-humidity-magnetism-vibration", a L2 norm calculation formula for single-mode anomaly scoring, and a passive credibility weighted fusion formula, the original qualitative multi-parameter correlation logic is transformed into a quantifiable mathematical model. This model not only fully covers the monitoring data of the humidity mode, but also dynamically adjusts the weights of each mode through global credibility, avoiding early warning deviations caused by misjudgment of a single mode. Ultimately, this upgrades bogie fault diagnosis from "experience-driven" to "data-driven", improving fault identification accuracy. The early warning lead time for early faults (such as bearing micro-wear and bolt micro-loosening) is extended compared to traditional solutions, and it can adapt to signal fluctuations under different train operating conditions, significantly improving the system's adaptability to complex train conditions and diagnostic reliability. Example 5

[0063] To achieve parallel passive monitoring of multiple physical parameters and multiple monitoring points within the bogie area, this method adopts a frequency division multiplexing strategy. The passive reflection spectrum is pre-planned into several non-overlapping dedicated large-frequency bands, and the data source is distinguished by identifying the frequency position of the echo signal.

[0064] Specifically, the system employs a frequency division multiplexing strategy to achieve parallel acquisition of multiple passive nodes without collisions. The passive reflection spectrum is pre-planned into dedicated, non-overlapping large operating frequency bands. Each physical mode sensing unit corresponds to a fixed dedicated large operating frequency band, as follows: The dedicated high-frequency band for the pressure mode sensing unit is f1~f2, for the temperature mode sensing unit it is f3~f4, for the humidity mode sensing unit it is f5~f6, and for the magnetic field mode sensing unit it is f7~f8.

[0065] In one example, f1=420MHz, f2=425MHz, f3=426MHz, f4=430MHz, f5=431MHz, f6=435MHz, f7=436MHz, and f8=440MHz.

[0066] Within the dedicated operating frequency band corresponding to the same physical mode, sub-channels are further subdivided, and a unique dedicated sub-channel is assigned to the sensing unit at different monitoring points under each mode: Within the large frequency band f1~f2 of the pressure mode, the pressure sensing unit corresponding to the axle box end cover mounting bolt occupies f 11 ~f 12 Sub-channel (e.g., 420.0~420.1MHz), the pressure sensing unit corresponding to the gearbox boom node bolt occupies f 13 ~f 14 Sub-channels (e.g., 420.1~420.2MHz), the pressure sensing units corresponding to the primary / secondary suspension system connecting bolts occupy f 15 ~f 16 Sub-channels (e.g., 420.2~420.3MHz); Within the large frequency band f3~f4 of the temperature mode, the temperature sensing unit corresponding to the bearing housing of the axle box occupies f 31 ~f 32 Sub-channels (e.g., 426.0~426.1MHz), the temperature sensing unit corresponding to the traction motor housing occupies f 33 ~f 34 Sub-channels (e.g., 426.1~426.2MHz), the temperature sensing unit corresponding to the outer wall of the gearbox oil sump occupies f 35 ~f 36 Sub-channels (e.g., 426.2~426.3MHz); Within the f5~f6 frequency band of the humidity mode, the humidity sensing unit near the gearbox breather occupies f 51 ~f 52 Sub-channels (e.g., 431.0~431.1MHz), the humidity sensing unit corresponding to the labyrinth seal on the rear cover of the axle housing occupies f 53 ~f 54 Sub-channels (e.g., 431.1~431.2MHz), the corresponding humidity sensing unit inside the electrical junction box occupies f 55 ~f 56Sub-channels (e.g., 431.2~431.3MHz); Within the f7~f8 frequency band of the magnetic field mode, the magnetic field sensing unit near the stator winding of the traction motor occupies f 71 ~f 72 Sub-channels (e.g., 436.0~436.1MHz), the corresponding magnetic field sensing unit near the velocity sensor probe occupies f 73 ~f 74 Sub-channels (e.g., 436.1~436.2MHz).

[0067] Furthermore, the receiving unit performs a full-band FFT scan, first identifying the corresponding physical mode based on the large frequency band range where the spectral peak is located (e.g., if the peak falls in the f1~f2 range, it is determined to be pressure mode data; if it falls in the f3~f4 range, it is determined to be temperature mode data), and then identifying the corresponding sensor location based on the specific sub-channel location within that large frequency band (e.g., if the peak falls in the f3~f4 range, it is determined to be temperature mode data). 11 ~f 12 If the sub-channel is selected, it is determined to be the pressure sensing unit at the mounting bolt of the axle box end cover; if it falls on f 31 ~f 32 The sub-channel is identified as the temperature sensing unit at the bearing housing of the axle box, thereby achieving IDless identification and parallel data acquisition.

[0068] Therefore, this application embodiment solves the signal collision and identification delay problems in dense sensor networks by using a frequency division multiplexing strategy of large frequency band interval division and sub-channel subdivision. It does not require a complex digital ID decoding protocol. The receiving unit achieves parallel acquisition of multi-modal and multi-node data through dual positioning of the spectrum peak "large frequency band-sub-channel", which improves the system response speed and data throughput and meets the timeliness requirements of real-time train monitoring and dynamic feature analysis. Example 6

[0069] On the other hand, such as Figure 3 The flowchart shown is a digital twin-driven multimodal passive sensing train bogie monitoring method of the present invention. It provides a digital twin-driven multimodal passive sensing train bogie monitoring method applicable to the aforementioned system, including: Step S1: Construct an electromagnetic energy field covering the key areas of the bogie; the onboard excitation unit continuously emits at a frequency of [frequency value missing]. A single-frequency carrier signal; Step S2, the vehicle-mounted receiving unit uses a sampling rate Continuously execute Fast Fourier Transform to obtain real-time spectrum frames; Step S3, search for the location in each frame of spectrum. Sideband peaks on both sides, locked sideband frequency Record peak frequency points and construct original time series data of sideband frequency changes over time. ; Step S4, for Apply low-pass and high-pass filters respectively and perform de-trendization to extract quasi-static components. and dynamic components ; In step S4, the low-pass filter cutoff frequency is dynamically adjusted according to the train speed. When the train is running at low speed... Smaller Choose a smaller value to avoid misjudging low-speed vibrations as static components; when the train is running at high speed... Increase The corresponding increase ensures the accuracy of static component extraction.

[0070] Step S5, will Substituting the frequency-parameter calibration curve, the pressure mode is obtained by calculation. Temperature modes Humidity mode Magnetic field modes The corresponding quasi-static monitoring quantity; In step S5, the frequency-parameter calibration curve is established through previous experimental calibration. The one-to-one correspondence with each physical quantity.

[0071] Step S6, based on Perform single-parameter threshold diagnosis, based on Perform passive dynamic feature analysis and combine it with multimodal correlation matrix to complete multi-parameter fusion early warning, and realize comprehensive assessment of bogie health status.

[0072] Therefore, the embodiments of this application form a passive multimodal monitoring method with a full-link "excitation-acquisition-decoupling-analysis-fusion". The dynamic adjustment of sampling rate and filter cutoff frequency improves data accuracy under different working conditions. Passive dynamic feature extraction (including proprietary features such as sideband symmetry and modulation stability) enhances fault identification capability. Multi-parameter credibility weighted fusion avoids false alarms and missed alarms. Combined with a three-level evaluation system, the bogie status is upgraded from "qualitative evaluation" to "quantitative classification", and the evaluation accuracy is good.

[0073] The following is through Figure 2 The above embodiments are illustrated with accompanying drawings. Figure 2 This is a schematic diagram of the signal processing flow of the multi-sensor passive monitoring method for train bogies of the present invention. It is used to present the entire link logic from electromagnetic energy excitation to bogie health status assessment. The information flow and correlation of each link are as follows.

[0074] Figure 2 The excitation and sensing response stages are illustrated. The transmission frequency of the onboard excitation unit is [frequency missing]. A single-frequency continuous wave radio frequency signal is used to construct an electromagnetic energy field covering the bogie area (corresponding to step S1); each sensing unit in the distributed passive multimodal sensor network senses the electromagnetic energy and starts working, converting the measured physical quantity (pressure, temperature, humidity, magnetic field) into changes in circuit parameters (capacitance C or inductance L), and then modulating the reflected signal with sidebands to generate a sideband frequency offset containing information about the physical quantity. This ultimately forms a sideband signal. .

[0075] Figure 2 The signal acquisition and sequence construction stages are also illustrated. The vehicle-mounted receiving unit uses a sampling rate of Continuously perform Fast Fourier Transform (FFT) to acquire electromagnetic spectrum frames within the region in real time (corresponding to step S2); search within each frame of spectrum. The sideband peaks on both sides lock the target sideband frequency. The corresponding peak frequency points are recorded, and the original time series data of sideband frequency variation over time is constructed through N consecutive samplings. (Corresponding to step S3).

[0076] Figure 2 The signal decoupling and component extraction stages are also illustrated. The integrated signal processing and diagnostic unit... To perform frequency domain filtering and decoupling, a low-pass filter is applied to extract quasi-static components. This corresponds to the steady-state value of physical quantities (such as bolt preload and component normal temperature); on the other hand, a high-pass filter is applied and de-trendring is performed to extract dynamic components. This corresponds to transient changes in physical quantities (such as parameter fluctuations caused by vibration or impact) (corresponding to step S4).

[0077] Figure 2 The diagram also illustrates the parameter calculation and state assessment stages. Substituting the pre-calibrated "frequency-parameter" calibration curve, the pressure mode is obtained by calculation. Temperature modes Humidity mode , Static physical quantities (corresponding to step S5); based on static physical quantities, perform single-parameter threshold diagnosis (quickly determine faults such as loose bolts and hot shafts), based on Calculate the root mean square value and kurtosis value, and monitor the resonance peak at the 20Hz snake frequency (analyze the running stability, impact characteristics and resonance risk). Finally, combine the multi-modal correlation matrix to complete the multi-parameter fusion early warning and realize the comprehensive assessment of the bogie's health status (corresponding to step S6).

[0078] thus Figure 2The entire information flow forms a closed loop of "excitation-sensing-acquisition-computation-evaluation", with each link closely connected to ensure the comprehensive capture of static physical quantities and dynamic characteristics, providing complete data support for fault diagnosis.

[0079] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0080] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A digital twin-driven multimodal passive sensing train bogie monitoring system, characterized in that, It includes an onboard excitation and reception unit, a distributed passive multimodal sensor network, and an integrated signal processing and diagnostic unit; The vehicle-mounted excitation and receiving unit is used to transmit at a frequency of The single-frequency continuous wave radio frequency signal is used to construct an electromagnetic energy field covering the bogie area, and the electromagnetic spectrum signal in the area is collected in real time. The distributed passive multimodal sensing network consists of multiple passive sensing units. These passive sensing units operate by sensing electromagnetic energy, converting the measured physical quantity into the circuit parameter capacitance. or inductor The change in this modulates the reflected signal, generating a sideband frequency offset containing physical quantity information. ,and The sideband signal satisfies ; The integrated signal processing and diagnostic unit is used to perform high-speed FFT analysis on the acquired spectral signals and extract... Based on its time-varying characteristics, the quasi-static and dynamic components of the signal are decoupled through frequency domain filtering technology, and the static values ​​and dynamic waveforms of each physical parameter are demodulated. The data is then synchronized to the bogie digital twin model in real time to achieve collaborative updating of virtual and real states. Finally, the bogie health status assessment is completed based on a three-level evaluation system.

2. The system according to claim 1, characterized in that, The vehicle-mounted excitation and receiving unit includes an excitation subunit and a receiving subunit; The excitation subunit includes at least a radio frequency signal source, a power amplifier, and a transmitting antenna to provide a stable radio electromagnetic energy field. The excitation subunit supports frequency and power adjustment. The receiving subunit includes a high-gain receiving antenna, a low-noise amplifier, a mixer, and a high-speed analog-to-digital converter module. It features an embedded real-time spectrum analyzer architecture, enabling it to perform fast Fourier transforms at high frequencies, with a system sampling rate of... Satisfying the Nyquist sampling theorem to achieve maximum capture The dynamic fluctuation of the measured parameter at a frequency of / 2.

3. The system according to claim 1, characterized in that, The distributed passive multimodal sensing network includes a passive pressure sensing unit, a passive temperature sensing unit, a passive humidity sensing unit, and a passive electromagnetic sensing unit. The passive pressure sensing unit uses a pressure-sensitive capacitor or a piezoelectric stack structure as the sensing element, and the bolt preload is... Changes cause capacitance value Change, and thus Changes have occurred; The passive temperature sensing unit uses a temperature-sensitive capacitor material as the sensing element to detect ambient temperature. Changes cause capacitance value Change, and thus Changes have occurred; The passive humidity sensing unit uses a humidity-sensitive capacitor material as the sensing element; changes in ambient humidity cause changes in capacitance. Change, and thus Changes have occurred; The passive electromagnetic sensing unit uses a magnetic inductor as the sensing element, and the external magnetic field strength... Changes alter the permeability of the magnetic core This leads to an increase in inductance. Change, make An offset has occurred.

4. The system according to claim 3, characterized in that, The passive pressure sensing unit is deployed at the bolts of the key connection parts of the bogie, including at least the axle box end cover mounting bolts, gearbox hanger node bolts, and primary / secondary suspension system connection bolts of the bogie, and is used to monitor the axial preload of the bolts at the key connection parts. The passive temperature sensing unit is deployed on the outer surface of the bogie friction components and high-power electrical components. The outer surface includes at least the outer surface of the bogie axle box bearing seat, the traction motor housing, and the outer wall of the gearbox oil sump, and is used to monitor the thermal state of the friction components and high-power electrical components. The passive humidity sensing unit is deployed in the sealed area and electrical protection area of ​​the bogie's oil lubrication components. The area includes at least the vicinity of the bogie's gearbox breather, the labyrinth seal of the axle box rear cover, and the inside of the electrical junction box, for monitoring the sealing performance of the oil lubrication components and the ambient humidity. The passive electromagnetic sensing unit is deployed in the traction motor and surrounding electromagnetic sensitive areas of the bogie. The electromagnetic sensitive areas include at least the vicinity of the stator winding of the traction motor and the vicinity of the speed sensor probe, and are used to monitor the electromagnetic environment of the traction motor and its surroundings.

5. The system according to claim 1, characterized in that, The health status assessment system of the integrated signal processing and diagnostic unit includes three levels: single-parameter threshold diagnosis, passive dynamic feature analysis, and multi-parameter fusion early warning. The results of each level are synchronized to the bogie digital twin model to support the model's state iteration and simulation verification. The integrated signal processing and diagnostic unit decouples signals using frequency domain filtering technology. The process of decoupling signals using frequency domain filtering technology includes: For raw time series data Applying a low-pass filter to extract quasi-static components ,right Apply a high-pass filter and de-stress to extract dynamic components. The quasi-static and dynamic components are both used as real-time inputs to the bogie digital twin model, realizing the mapping between the physical state and the virtual model.

6. The system according to claim 5, characterized in that, The criteria for single-parameter threshold diagnosis include: Criteria for loosening bolts Hot axis criteria Criteria for seal failure Electromagnetic anomaly criteria Deviation from the benchmark value exceeds ;in, Depend on The equivalent monitoring quantities obtained by mapping include , , , , respectively, represent the quasi-static monitoring quantities of the pressure mode, temperature mode, humidity mode, and magnetic field mode; , , For the preset threshold, , , and It is set by at least one of the following: vehicle factory calibration, historical health baseline statistics, and operation and maintenance experience, and supports adaptive updates based on operating conditions.

7. The system according to claim 5, characterized in that, The passive dynamic feature analysis refers to the dynamic components obtained from the original time series data after high-pass filtering and detrending. Feature extraction and anomaly characterization are performed using a sliding window, wherein the feature extraction includes at least one of the following three types of features: The first category is general dynamic features, which are all applied within a sliding window. The calculations include at least the root mean square value, variance, standard deviation, peak value, peak-to-peak value, absolute mean value, skewness, kurtosis, peak factor, impulse factor, margin factor, waveform factor, window energy, sliding window coefficient of variation, short window to long window energy ratio, short window to long window root mean square ratio, first-order difference sequence characteristics, and second-order difference sequence characteristics. The second category is frequency-time strongly correlated features. These features are used to characterize the periodic modulation and cyclic statistical properties of passive frequency point timing. They include at least spectral centroid, spectral bandwidth, spectral flatness, spectral entropy, spectral kurtosis, and cyclic stationary features. The cyclic stationary features include cyclic autocorrelation features and / or spectral correlation features. The third category comprises passive signal-specific features. These features are designed for the upper and lower sideband structures formed on both sides of the carrier frequency by passive modulation. The passive signal-specific features of the third category include at least one of the following features: Feature 1, sideband symmetry feature, in the first Within a sliding window, extract the sideband energy of the upper and lower sidebands. , And define the sideband imbalance index. And further define sideband symmetry ,in, To avoid extremely small positive numbers with a denominator of zero, , The spectrum within the window in the sideband neighborhood The power spectral density is obtained by summing. For passive excitation carrier frequency, This represents the offset of the sideband relative to the carrier frequency. The sideband energy integral bandwidth; Feature 2, passive modulation stability index, in the first... Estimate sideband offset frequency within a sliding window and in length of Constructing frequency stability quantities on window sequences Meanwhile, define the sideband detection indicator. And further calculate the peak drop rate. Finally, the passive modulation stability index is defined. ,in, , These are the weighting coefficients. A larger value indicates more stable passive modulation and more reliable frequency timing; among which, the sideband detection indicator is defined. The content includes: In response to the When the sideband peak of a window is detectable and exceeds the threshold ,otherwise .

8. The system according to claim 7, characterized in that, The multi-parameter fusion early warning system establishes a multi-dimensional state space of "force-thermal-humidity-magnetic-vibration". This multi-dimensional state space corresponds one-to-one with the simulation dimensions of the bogie's digital twin model, and a passive reliability adaptive weighted fusion method is used to output hierarchical early warning results, specifically including: First, state space construction, including in the... Constructing a state vector within a sliding window ; in, These are the quasi-static residual eigenvectors and / or dynamic eigenvectors for pressure, temperature, humidity, and magnetic field modes, respectively. For the reason The extracted dynamic feature vectors corresponding to the "vibration" mode, and The sideband symmetry characteristic and passive modulation stability index; Second, evidence of single-modal anomalies, including constructing anomaly scores for each modality. ( ),satisfy , and when If it is a scalar, then ,in, It is a 2-norm. It is represented as a set of multidimensional state spaces, where P is the pressure in the dimension, T is the temperature in the dimension, H is the humidity in the dimension, B is the magnetic field in the dimension, and V is the vibration in the dimension; Third, the fusion of passive credibility and adaptive weighted credential assessment, including the construction of global credibility. , The adjustment coefficient is used to determine the fusion weight. To implement differentiated weights for different modalities, it is supported to... Expanded to modal credibility ,at this time Weighted calculation of fusion anomaly scores ; Fourth, a tiered early warning algorithm will... The warning results are compared with a preset threshold and include: in Determined by health benchmark statistics, calibration experiments, or operational experience, and can be based on... Adaptive adjustments are made to suppress false alarms caused by passive reading fluctuations.

9. The system according to claim 1, characterized in that, The system employs a frequency division multiplexing strategy to achieve parallel acquisition of multiple passive nodes without collision. The passive reflection spectrum is pre-planned into exclusive working large frequency bands that do not overlap. The sensing units of each physical mode correspond to preset fixed exclusive working large frequency bands: the pressure mode sensing unit corresponds to the f1~f2 frequency band, the temperature mode sensing unit corresponds to the f3~f4 frequency band, the humidity mode sensing unit corresponds to the f5~f6 frequency band, and the magnetic field mode sensing unit corresponds to the f7~f8 frequency band. Within the dedicated operating frequency band of the same physical mode, sub-channels are further subdivided. The sensing units at different monitoring points under each mode correspond to a unique dedicated sub-channel. The receiving unit locks the corresponding physical mode through the frequency band where the spectrum peak is located, and then locks the corresponding sensor position through the sub-channel within the frequency band, thus realizing IDless identification and parallel acquisition.

10. A digital twin-driven multimodal passive sensing method for monitoring train bogies, applicable to the system described in any one of claims 1 to 9, characterized in that, include: Step S1: Construct an initial digital twin model of the bogie and an electromagnetic energy field covering key areas of the bogie. The onboard excitation unit continuously transmits at a frequency of [frequency missing]. A single-frequency carrier signal; Step S2, the vehicle-mounted receiving unit uses a sampling rate Continuously execute Fast Fourier Transform to obtain real-time spectrum frames; Step S3, search for the location in each frame of spectrum. Sideband peaks on both sides, locked sideband frequency Record peak frequency points and construct original time series data of sideband frequency changes over time. ; Step S4, for Apply low-pass and high-pass filters respectively and perform de-trendization to extract quasi-static components. and dynamic components ; Step S5, will Substituting the frequency-parameter calibration curve, the pressure mode is obtained by calculation. Temperature modes Humidity mode Magnetic field modes The corresponding quasi-static monitoring quantity; Step S6, based on Perform single-parameter threshold diagnosis, based on Perform passive dynamic feature analysis and combine it with multimodal correlation matrix to complete multi-parameter fusion early warning, and realize comprehensive assessment of bogie health status.

Images