A health state monitoring and management system and method based on vibration signal and multi-modal data fusion

Abstract

The application discloses a health state monitoring and management system and method based on vibration signal and multi-modal data fusion, which comprises a signal acquisition board, a data processing board and an upper computer, the signal acquisition board collects multiple vibration signals through a differential charge amplification circuit, adopts a software and hardware double-path parallel processing mechanism, and realizes synchronous high-precision collection of wideband vibration signals and characteristic envelopes thereof. The data processing board takes a high-performance ZYNQ chip as a core, realizes high-speed data acquisition and preprocessing on a PL end, runs an embedded system on a PS end to perform data fusion, compressed storage and communication with the upper computer through a gigabit Ethernet. The upper computer software provides data visualization, state diagnosis and parameter configuration functions. The application aims at problems of inaccurate deep mechanical fault feature capture, large data storage pressure and lack of multi-dimensional comprehensive analysis, realizes long-term, continuous, accurate and intelligent monitoring and management of health states of rotating machinery and other equipment.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring and fault diagnosis technology, and in particular to a health status monitoring and management system and method based on the fusion of vibration signals and multimodal data. Background Technology

[0002] Rotating machinery (such as aircraft engines, industrial gearboxes, and wind turbines) is a core piece of equipment in industrial production and transportation; its health status directly affects the operational safety and production efficiency of the entire system. Vibration signal analysis is widely recognized as one of the most effective methods for diagnosing mechanical faults.

[0003] However, existing vibration monitoring systems still have the following technical limitations in practical applications: Insufficient fault feature capture capability: Conventional data acquisition systems are limited by sampling bandwidth or only perform simple effective value calculations, making it difficult to effectively capture the high-frequency resonance and modulation envelope characteristics excited by weak faults in the early stages, such as pitting corrosion of the bearing inner ring and micro-cracks on the gear tooth surface.

[0004] Single analytical dimension: Existing systems mostly analyze vibration signals independently, failing to deeply integrate vibration signals with key operating parameters such as rotational speed and temperature, resulting in bottlenecks in the accuracy and reliability of fault diagnosis.

[0005] High data storage pressure: In order to capture high-frequency fault characteristics, vibration data needs to be continuously collected at a high sampling rate, which puts a huge pressure on local storage and remote transmission of the massive amount of data generated.

[0006] Limited system integration and environmental adaptability: In application scenarios such as aviation, ships, and rail transportation, where there are stringent requirements for equipment size, weight, and environmental adaptability, existing commercial equipment often struggles to balance high performance and high reliability. Summary of the Invention

[0007] This invention aims to overcome the shortcomings of existing technologies and provide a health status monitoring and management system and method based on the fusion of vibration signals and multimodal data. The system, through a dual-channel parallel acquisition mechanism, heterogeneous computing architecture, and intelligent data compression strategy, aims to achieve long-term, continuous, accurate, and intelligent online monitoring and management of the health status of critical equipment such as rotating machinery.

[0008] The technical solution of the present invention: A health status monitoring and management system based on the fusion of vibration signals and multimodal data includes a signal acquisition module and a data processing module.

[0009] The signal acquisition module is used to acquire vibration signals from the target device. This module includes a first signal processing channel and a second signal processing channel connected to the output of the same vibration sensor. The first signal processing channel is configured to process the vibration signal to generate raw waveform data reflecting the overall vibration profile, thus providing high-fidelity, complete signal details for subsequent spectral analysis. The second signal processing channel is configured to perform analog domain processing on the same vibration signal to extract and generate an envelope signal characterizing the vibration impact features, thereby leveraging the real-time advantages of analog circuits to achieve early extraction of sensitive features for early faults.

[0010] The data processing module communicates with the signal acquisition module and includes a programmable logic device (PLD) and an embedded processor. The PLD is configured to receive raw waveform data and perform real-time digital signal processing on it, utilizing its parallel processing capabilities to ensure deterministic real-time processing of high-speed data streams. The embedded processor interacts with the PLD, configured to receive the processed data and envelope signal, and perform fusion analysis based on the received data to generate state assessment information, thereby leveraging the processor's flexibility in executing complex algorithms and managing data.

[0011] Through the synergistic cooperation of the aforementioned dual-channel parallel acquisition mechanism and heterogeneous processing architecture, the system can efficiently extract fault feature envelopes while capturing the full details of vibration signals, balancing the depth of fault diagnosis with real-time processing efficiency, and significantly improving the detection capability and diagnostic accuracy of early and subtle faults.

[0012] Furthermore, the first signal processing channel includes a first buffer amplifier and a first analog-to-digital converter (ADC). The first ADC is connected to the data processing module via an SPI interface to ensure high-precision analog-to-digital conversion and stable transmission of the original waveform data. The second signal processing channel includes a high-pass filter, a band-pass filter, a rectifier circuit, a low-pass filter, and a second ADC connected in sequence. The second ADC is connected to the data processing module via an SPI interface, thereby directly extracting the envelope of the high-frequency impact component in the vibration signal in the analog domain, obtaining the low-frequency envelope signal reflecting the fault characteristics without occupying digital processor resources.

[0013] Furthermore, the signal acquisition module also includes a speed signal conditioning circuit and at least one temperature sensor. The speed signal conditioning circuit is used to acquire the speed signal of the target device, including a pulse level conversion circuit to convert speed sensor signals of different amplitudes into standard logic levels for reliable acquisition by the digital processing unit. The temperature sensor is connected to the data processing module via an I2C bus to monitor the board temperature, thereby enabling monitoring of the system's own health status and improving the reliability of the system during long-term operation in harsh environments.

[0014] Furthermore, the data processing module also includes a high-speed optocoupler isolation circuit, which is used to isolate and level-convert the input speed pulse signal before sending it to a programmable logic device or embedded processor, thereby effectively suppressing the impact of field interference signals on the core processing circuit and improving the anti-interference capability of speed signal acquisition and the electromagnetic compatibility of the system.

[0015] Furthermore, the embedded processor is configured to: determine the real-time rotational speed value based on the rotational speed signal, and dynamically adjust the parameters of the real-time digital signal processing executed by the programmable logic device based on the real-time rotational speed value, so that the center frequency of the real-time digital signal processing frequency band follows the real-time rotational speed value or its harmonics, thereby achieving order tracking analysis. Even under variable speed conditions, it can accurately capture the fault characteristic frequencies related to rotational speed, significantly improving the diagnostic adaptability and feature extraction accuracy of variable speed operating equipment. Simultaneously, the embedded processor is also configured to fuse the spectral analysis results of the original waveform data, the time-domain characteristics of the envelope signal, and the real-time rotational speed value to extract multi-dimensional feature vectors, and generate state assessment information based on these multi-dimensional feature vectors. By comprehensively analyzing the vibration frequency domain information, time-domain impact characteristics, and operating parameters, it overcomes the limitations of a single analysis dimension, effectively reducing the false alarm rate and improving the accuracy of fault type identification.

[0016] Furthermore, the data processing module also includes DDR memory and non-volatile memory. The DDR memory, connected to the embedded processor, is used for program execution and data caching, providing ample high-speed memory space for complex diagnostic algorithms. The non-volatile memory, also connected to the embedded processor, stores raw waveform data, envelope signals, and status assessment information. This non-volatile memory includes eMMC memory and NOR Flash memory to ensure long-term data retention after system power failure, meeting the requirements for long-term continuous recording and traceability of device health status.

[0017] Furthermore, the embedded processor is also configured to perform a data extraction and compression algorithm based on energy distribution characteristics on the raw waveform data and store the compressed data in the eMMC memory, thereby significantly reducing the amount of data storage while retaining the key information required for fault diagnosis, effectively extending the data recording time supported by limited storage space, and adapting to application scenarios of long-term online monitoring.

[0018] Furthermore, the embedded processor is also configured to run a remote upgrade module, which receives upgrade data packets via an Ethernet interface and writes the received data into a NOR Flash memory to enable remote updates of the system program. This gives the system the ability to perform on-site firmware upgrades, reducing equipment maintenance costs and downtime, and is especially suitable for rotating machinery monitoring nodes deployed in remote or hard-to-access locations.

[0019] Furthermore, programmable logic devices and embedded processors are integrated into the same heterogeneous computing chip. The heterogeneous computing chip internally uses an AXI bus to achieve memory-mapped data interaction between the programmable logic devices and the embedded processor. This replaces board-level bus connections with chip-level high-bandwidth, low-latency internal communication, significantly improving data throughput efficiency and system integration while reducing device size and power consumption. The programmable logic device is configured to implement at least one of a clock management module, an IIC driver module, an SPI driver module, and a serial port module, performing low-level interface timing control in hardware and freeing up processor resources. The embedded processor is configured to implement at least one of an AXI bus driver module, a data processing module, a storage management module, a gigabit network driver, and a communication protocol module, executing complex task scheduling and network services in software. This collaborative partitioning of the aforementioned hardware and software functional modules within the heterogeneous chip enables high-speed signal acquisition and preprocessing, as well as intelligent diagnostic analysis tasks, to be executed efficiently in parallel, improving the overall real-time performance, reliability, and intelligence level of the system.

[0020] This invention also provides a health status monitoring and management method based on the fusion of vibration signals and multimodal data, comprising the following steps: Synchronous acquisition steps: The output signal of the same vibration sensor is received, and the original waveform data and envelope signal are generated by the first and second parallel channels respectively. Thus, the complete time-domain waveform of the vibration signal and its impact envelope characteristics are obtained at the same time, providing time-aligned dual-source data for subsequent time-frequency domain correlation analysis.

[0021] Real-time processing steps: Perform digital signal processing on the original waveform data in the programmable logic device, and use the parallel hardware architecture of the programmable logic device to realize low-latency filtering and spectrum transformation of high-speed data stream.

[0022] Fusion diagnostic steps: In the embedded processor, the waveform data and envelope signal after digital signal processing are correlated in the time and frequency domains, and multi-dimensional feature vectors are generated by combining the acquired rotation speed signal to output status assessment information for characterizing the health status of the equipment. Through the complementary fusion of multi-modal data, the problem of insufficient characterization capability of a single signal source in a specific fault mode is effectively overcome, and accurate and reliable assessment of the health status of rotating machinery is achieved.

[0023] The beneficial effects of this invention are: 1. Through the unique dual-channel parallel acquisition mechanism of "software channel + hardware envelope channel", it ensures the comprehensive capture of everything from the original waveform reflecting the basic state of equipment operation to the envelope characteristics reflecting early weak faults, which significantly improves the detection rate and diagnostic accuracy of early faults. 2. By linking the speed signal with digital signal processing, adaptive filtering with dynamic adjustment of center frequency according to speed is realized, ensuring that fault characteristics can still be accurately extracted under variable speed conditions; 3. By using an intelligent compression algorithm based on signal characteristics, the data storage space is effectively reduced and the continuous recording time of effective data is extended while ensuring the integrity of the information required for diagnosis. 4. Adopting a heterogeneous computing architecture of programmable logic devices and embedded processors, it integrates high-speed data acquisition, real-time signal processing, multi-modal data fusion, intelligent diagnosis and data management into one, realizing full-process automation and intelligence of the monitoring system. Attached Figure Description

[0024] Figure 1 This is a block diagram of the system hardware architecture of the present invention.

[0025] Figure 2 This is a schematic diagram of the data processing module structure of the present invention.

[0026] Figure 3 This is a schematic diagram of the main interface of the host computer software of the present invention. Detailed Implementation

[0027] The system of this invention mainly consists of two parts: a signal acquisition board and a data processing board. These two parts are electrically interconnected via board-to-board connectors and work together to complete signal acquisition, processing, storage, and communication functions. The entire system is powered and monitored by a unified power supply and management module.

[0028] Signal acquisition board: The signal acquisition board is the analog front end of the system interface with external sensors. It is responsible for high-fidelity conditioning and acquisition of multiple vibration signals and speed signals.

[0029] Furthermore, the vibration signal processing link is The signal acquisition board provides four vibration signal input channels. Each channel includes a differential charge acquisition circuit consisting of a low-offset, JFET-input operational amplifier, used to convert the weak charge signal output from the piezoelectric accelerometer into a voltage signal. This voltage signal is then split into two parallel paths for processing: The first signal processing channel: This channel sequentially passes through buffer amplification and high-precision ADC sampling. The buffer amplifier is used for impedance matching and signal driving; the ADC is an analog-to-digital converter that sends the converted raw waveform data to the data processing board via an SPI interface. This channel aims to preserve the complete time-frequency domain details of the vibration signal for subsequent detailed analysis.

[0030] The second signal processing channel: This channel sequentially processes signals through high-pass filtering, band-pass filtering amplification, full-bridge rectification, and low-pass filtering. Finally, after sampling by another ADC, the extracted envelope signal is sent to the data processing board via the SPI interface. This channel utilizes analog circuitry to extract signal envelope features sensitive to early faults in real time, reducing the computational burden on the back-end digital processor.

[0031] Furthermore, the link between the speed signal and temperature monitoring is as follows: Speed ​​signal conditioning: A speed signal input is provided, which is output to the data processing board as a general-purpose I / O signal after passing through a pulse level conversion circuit. This circuit is used to convert speed sensor signals of different amplitudes into standard logic levels.

[0032] Temperature monitoring: At least one temperature sensor is installed on the board, which is connected to the data processing board via an I2C bus to monitor the operating temperature of the signal acquisition board in real time, providing a basis for system health management.

[0033] Furthermore, the power supply link is as follows: The signal acquisition board receives a DC 28V input from the system power supply. This input first passes through overvoltage and overcurrent protection, reverse connection protection, and EMI filtering circuits before being sent to the isolated DC-DC module. The isolated DC-DC module generates multiple isolated power supplies, such as +5V, +15V, and -15V, which are then filtered to provide clean, low-noise operating voltages for the various analog and digital circuits on the board. This power supply architecture effectively isolates digital noise from interfering with sensitive analog circuits.

[0034] Data processing board: The data processing board is the core of the system's digital processing. It is built around a heterogeneous computing chip (i.e., a System-on-a-Chip, such as the Xilinx ZYNQ series chips) that integrates programmable logic (PL) and a processing system (PS).

[0035] Furthermore, the internal functions of the SOC are divided as follows: The PL (Programmable Logic) and PS (Processing System) within the SOC undertake tasks at different levels: The PL section utilizes its hardware parallelism to implement the underlying interface drivers and timing control logic for modules such as clock management, IIC driver (for communication with the temperature sensor), SPI driver (for communication with the ADC), and serial port. The PL is responsible for high-speed, real-time acquisition of raw waveform data and envelope signals from the signal acquisition board.

[0036] The PS (Power Supply) section runs the embedded operating system, enabling complex upper-layer applications. This includes an AXI bus driver module for high-speed data interaction with the PL (Power Supply), an AXIBRAM data transfer module for efficient data movement, a data processing module for executing data fusion and diagnostic algorithms, a NORFLASH data read / write module, a remote upgrade module, a DDR3 memory management module, an eMMC data storage module, and a Gigabit Ethernet driver and communication protocol module.

[0037] Furthermore, the onboard peripherals and interfaces include: Memory: The SOC's PS side is expanded with 4Gb DDR3 memory as system running memory; two eMMC memory chips with a total capacity of 128GB are expanded for large-capacity, non-volatile storage of compressed vibration data and diagnostic records; and a 512Mb NOR Flash memory chip is expanded for storing system boot programs and application firmware.

[0038] Communication and debugging interfaces: An onboard gigabit network interface (including a PHY chip and network transformer) is used for high-speed data and control command exchange with a host computer. An RS422 debugging interface is provided for local system maintenance.

[0039] Speed ​​signal isolation input: The speed I / O signal from the signal acquisition board is first converted and isolated by the high-speed optocoupler on the data processing board before being sent to the I / O pin of the SOC to protect the core chip from external interference.

[0040] Local power supply: The data processing board draws power from the system power supply and generates various low-voltage power supplies such as 1.0V, 1.5V, 1.8V and 3.3V required by the SOC and peripherals through the on-board non-isolated DC-DC converter.

[0041] The specific steps of the health status monitoring and management method based on the fusion of vibration signals and multimodal data in this invention are as follows: Step 1: Synchronous Data Acquisition After the system powers on, the PL section of the SOC initializes the ADC and temperature sensor on the signal acquisition board via SPI and I2C interfaces. The charge signals from the four vibration sensors are amplified differentially and then enter the first and second signal processing channels in parallel. Under the control of the PL's SPI driver module, the ADCs of the first and second channels synchronously perform analog-to-digital conversion, generating the original waveform data stream and envelope signal stream, respectively.

[0042] Step Two: Real-time Processing and Fusion The PL section rapidly writes the acquired raw waveform data stream into a designated circular buffer in the PS-side DDR3 memory via the AXI bus. The PS-side data processing module reads the data from the DDR3 and performs a Fast Fourier Transform to obtain the spectrum. Simultaneously, this module also reads the envelope signal data forwarded by the PL and calculates the rotational speed data from the pulse periods captured by the PL.

[0043] Step 3: Intelligent Diagnosis and Storage The data processing module on the PS side fuses the calculated spectral characteristics, peak value, kurtosis, and other time-domain features of the envelope signal, along with the real-time rotational speed value, to form a multi-dimensional feature vector. This feature vector is then input into a pre-defined fault diagnosis model, such as a support vector machine or neural network model. The model outputs current "healthy," "alarm," or "dangerous" status assessment information.

[0044] Meanwhile, in order to save storage space, the data processing module on the PS side can perform a data extraction and compression algorithm on the high-frequency sampled raw waveform data. The compressed data, along with the feature vector and state evaluation results, is written into the onboard eMMC memory through the eMMC data storage module.

[0045] Step 4: Communication and Remote Upgrade The gigabit network driver and communication protocol module on the PS side establishes communication with the host computer through the gigabit network interface. The system periodically or upon request from the host computer uploads status assessment information, real-time waveform data, or historical compressed data.

[0046] Furthermore, when system firmware updates are required, the host computer sends the upgrade package to the device via the network. The remote upgrade module on the PS side receives and verifies the upgrade package, writes it to a designated area of ​​the NOR Flash memory, and completes the remote update of the system program, greatly facilitating on-site maintenance.

[0047] Example 1 See Figure 1 and Figure 2 This embodiment provides a complete health status monitoring and management system based on the fusion of vibration signals and multimodal data.

[0048] Hardware configuration and connection relationships: The system consists of a signal acquisition board and a data processing board stacked together via a high-density board-to-board connector.

[0049] The signal acquisition board integrates four identical vibration signal conditioning and acquisition links. Taking one of them as an example, the charge signal output by the piezoelectric accelerometer is input via an SMA connector and first enters a differential charge acquisition circuit composed of an AD8627 operational amplifier, outputting a voltage signal proportional to the acceleration. This voltage signal is split into two: one is buffered and amplified by an OPA376 and then sent to a 24-bit ADC chip AD7124-8 for ADC sampling. The AD7124-8 transmits the raw waveform data to the data processing board via the SPI bus; the other passes sequentially through a 1kHz high-pass filter, a center-frequency adjustable OPA1612 band-pass filter, a full-bridge precision rectifier circuit composed of an OPA1612 band-pass filter and diodes, and a 500Hz low-pass filter to extract the envelope signal. This envelope signal is sampled by a 16-bit synchronous sampling ADC chip AD7606, which also outputs data via the SPI bus.

[0050] The rotation speed signal on the board is converted from -10V to +10V pulses to 0 to 3.3V via a pulse level conversion circuit, and then connected to the board-to-board connector through I / O pins. Two temperature sensors, TMP101, are connected in parallel via an I2C bus.

[0051] In terms of power supply, the external DC28V input is fed into the Vicor PI31 series isolated DC-DC module after passing through the protection filter circuit. The module outputs isolated +5V, +15V and -15V, which are then filtered by LC filters to power the analog and digital devices on the board.

[0052] The core of the data processing board is the Xilinx ZYNQ-7020 SOC chip. Its PL section is implemented using Verilog code. Figure 2 The SPI driver module communicates with AD7124 and AD7606, reads the TMP101 temperature value through the IIC driver module, includes a clock management module, and a serial port module for RS422 debugging. The speed I / O signal from the signal acquisition board is first isolated by a high-speed optocoupler HCPL-0631 before being sent to the PL's I / O pin for pulse period measurement.

[0053] The PS (Power Supply) section of the SOC runs an embedded Linux system. The PL (Power Processor) uses the AXI bus driver module and AXI BRAM data transfer module to rapidly transfer the acquired vibration data to the external 4Gb DDR3 memory on the PS via DMA. The PS's data processing module executes FFT, feature extraction, data fusion, and diagnostic algorithms. The eMMC data storage module writes data to two eMMC chips with a total capacity of 128GB. The system firmware and application programs are stored in 512Mb NOR Flash and can be upgraded via a remote upgrade module. Finally, the processing results and waveform data are uploaded to the host computer via the onboard Gigabit Ethernet PHY chip through the Gigabit Ethernet driver and communication protocol module.

[0054] The onboard non-isolated DC-DC converter further converts the +5V power supply from the signal acquisition board into 1.0V, 1.5V, 1.8V and 3.3V to power the SOC and its peripherals.

[0055] Workflow example: When the system is powered on, ZYNQ's PS first boots from NOR Flash, loads FSBL, U-Boot and Linux kernel, and then loads the bitstream file of PL to complete PL configuration.

[0056] The application starts, initializes the peripherals, reads temperature information via I2C, and configures the ADC to work at the preset sampling rate via SPI.

[0057] Entering the main loop, the PL continuously acquires the raw waveforms and envelope data of the four vibrations and measures the rotational speed pulse period.

[0058] The data processing module on the PS side performs a data processing loop every second: it reads the latest data block from DDR3, calculates the spectrum and envelope features, combines them with the rotation speed value to form a feature vector, and inputs it into the SVM model to obtain the state evaluation result.

[0059] If the status is "normal", the original waveform is compressed by 1 / 10 of the binary decompression and then stored in eMMC; if the status is "alarm" or "danger", all original data is retained without compression.

[0060] The host computer can view the real-time status and waveforms of the device at any time via the network, and can issue commands to read back historical data in the eMMC or initiate a remote upgrade process.

Claims

1. A health status monitoring and management system based on the fusion of vibration signals and multimodal data, characterized in that, include: The signal acquisition module is used to acquire vibration signals of the target device. The signal acquisition module includes a first signal processing channel and a second signal processing channel connected to the output terminal of the same vibration sensor. The first signal processing channel is configured to process the vibration signal to generate raw waveform data that reflects the overall picture of the vibration. The second signal processing channel is configured to perform analog domain processing on the same vibration signal to extract and generate an envelope signal characterizing the vibration and impact features; The data processing module, which communicates with the signal acquisition module, includes a programmable logic device and an embedded processor; The programmable logic device is configured to receive raw waveform data and perform real-time digital signal processing on the raw waveform data; The embedded processor interacts with the programmable logic device and is configured to receive data and envelope signals processed by the programmable logic device, and perform fusion analysis based on the received data to generate state assessment information.

2. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 1, characterized in that: The first signal processing channel includes a first buffer amplifier and a first analog-to-digital converter (ADC), which is connected to the data processing module via an SPI interface. The second signal processing channel includes a high-pass filter, a band-pass filter, a rectifier circuit, a low-pass filter, and a second ADC connected in sequence, which is connected to the data processing module via an SPI interface.

3. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 1, characterized in that, The signal acquisition module also includes: The speed signal conditioning circuit is used to acquire the speed signal of the target device. The speed signal conditioning circuit includes a pulse level conversion circuit. At least one temperature sensor is connected to the data processing module via an I2C bus to monitor the board temperature.

4. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 3, characterized in that: The data processing module also includes a high-speed optocoupler isolation circuit, which is used to isolate and level-convert the input rotation speed pulse signal before sending it to a programmable logic device or embedded processor.

5. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 3, characterized in that: The embedded processor is also configured to: The real-time speed value is determined based on the speed signal, and the parameters of the real-time digital signal processing executed by the programmable logic device are dynamically adjusted according to the real-time speed value, so that the center frequency of the frequency band of the real-time digital signal processing follows the real-time speed value or its harmonics. The spectrum analysis results of the original waveform data, the time-domain characteristics of the envelope signal, and the real-time speed value are fused together to extract multi-dimensional feature vectors, and state evaluation information is generated based on the multi-dimensional feature vectors.

6. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 1, characterized in that: The data processing module further includes: DDR memory, connected to the embedded processor, is used for program execution and data caching; Non-volatile memory, connected to the embedded processor, is used to store raw waveform data, envelope signals, and state evaluation information. Non-volatile memory includes eMMC memory and NOR Flash memory.

7. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 6, characterized in that: The embedded processor also performs a data extraction and compression algorithm based on energy distribution characteristics on the raw waveform data and stores the compressed data in the eMMC memory.

8. The health status monitoring and management system based on vibration signal and multimodal data fusion according to claim 6, characterized in that: The embedded processor also runs a remote upgrade module, which receives upgrade data packets via an Ethernet interface and writes the received data into the NOR Flash memory to enable remote updates of the system program.

9. The health status monitoring and management system based on vibration signal and multimodal data fusion according to any one of claims 1 to 8, characterized in that, Programmable logic devices and embedded processors are integrated into the same heterogeneous computing chip. The heterogeneous computing chip uses an AXI bus to realize memory-mapped data interaction between the programmable logic devices and the embedded processor. The programmable logic device is configured to implement at least one of a clock management module, an IIC driver module, an SPI driver module, and a serial port module. The embedded processor is configured to implement at least one of an AXI bus driver module, a data processing module, a storage management module, a gigabit network driver, and a communication protocol module.

10. A health status monitoring and management method based on the fusion of vibration signals and multimodal data, applied to the health status monitoring and management system as described in any one of claims 1 to 9, characterized in that, Includes the following steps: Synchronous acquisition: Receives the output signal from the same vibration sensor, and generates raw waveform data and envelope signal respectively via the first and second parallel channels; Real-time processing: Performing digital signal processing on raw waveform data in a programmable logic device; Fusion diagnostics: In an embedded processor, waveform data and envelope signals that have undergone digital signal processing are correlated in the time and frequency domains, and multi-dimensional feature vectors are generated by combining them with the acquired rotation speed signals to output status assessment information that characterizes the health status of the equipment.

Images