Mining intrinsic safety type vibration data acquisition device and processing method

Abstract

The invention relates to the technical field of mining equipment, in particular to a mining intrinsic safety type vibration data acquisition device and a processing method. A plurality of IEPE interface circuits, wherein each IEPE interface circuit is used for being connected with one intrinsic safety type IEPE vibration sensor; a plurality of signal conditioning circuits; a multi-channel ADC synchronous acquisition circuit; the ZYNQ core module is connected with the multi-channel ADC synchronous acquisition circuit and comprises an FPGA part and an ARM part, the device size is greatly reduced while the safety requirement is met, a plurality of intrinsic safety type vibration sensors can be connected at the same time, strict synchronization of data of all channels is guaranteed in a hardware mode, and the reliability of the device is improved. And on-site real-time processing of vibration data and automatic extraction of fault features are realized, real-time early warning is realized, and the real-time performance and accuracy of fault diagnosis of mining equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of mining equipment technology, and in particular to an intrinsically safe vibration data acquisition device and processing method for mining applications. Background Technology

[0002] Mechanical equipment such as coal mining machines, tunneling machines, and scraper conveyors are core equipment in coal mine production, and their operating status directly affects the safety and efficiency of the entire mining operation. Due to the harsh underground environment and drastic load changes in coal mines, key components of these machines, such as bearings, gearboxes, and motors, are subjected to high-intensity mechanical stress and impact loads for extended periods, making them highly susceptible to progressive failures such as fatigue wear, pitting, and cracking. Vibration monitoring technology, as a mainstream method for equipment condition monitoring and fault diagnosis, can effectively identify early fault characteristics by collecting and analyzing vibration signals during equipment operation, providing a scientific basis for predictive maintenance. Currently, mine vibration monitoring mainly relies on two methods: one is manual inspection, where maintenance personnel periodically check the equipment status using hearing, touch, or handheld vibration meters; the other is using ordinary industrial vibration sensors to build a monitoring system and transmitting data to a ground center for analysis. With the advancement of intelligent coal mine construction, higher requirements are placed on the real-time performance, accuracy, and reliability of equipment condition monitoring.

[0003] However, existing technologies face numerous technical challenges in practical applications. First, manual inspection is heavily influenced by human experience, making continuous monitoring difficult. Furthermore, inspectors working in hazardous areas pose safety risks, failing to meet the demands of modern mines for safe production and intelligent management. Second, most existing sensor monitoring solutions employ single-point or asynchronous multi-point monitoring. Inconsistent data collection times at different points prevent effective correlation analysis, hindering accurate assessment of fault propagation paths and severity, and easily leading to misdiagnosis or missed diagnosis. Third, underground coal mines impose stringent intrinsically safe requirements on electrical equipment. However, commercially available intrinsically safe IEPE vibration sensors lack compatible intrinsically safe acquisition devices. Existing explosion-proof acquisition equipment is bulky, difficult to install, and challenging to deploy flexibly in space-constrained mining environments. In addition, most existing data acquisition devices only collect and forward data, lacking the ability to perform real-time signal processing and intelligent analysis on-site. A large amount of raw vibration data needs to be transmitted to the ground center for offline analysis, which not only puts a huge strain on the bandwidth of the underground communication network, but also fails to provide timely early warning of faults. Often, by the time the data analysis is completed, the equipment has already been severely damaged, thus losing the significance of predictive maintenance. Summary of the Invention

[0004] The main objective of this invention is to provide an intrinsically safe vibration data acquisition device and processing method for mining applications, aiming to solve the technical problems mentioned in the background section.

[0005] This invention proposes an intrinsically safe vibration data acquisition device for mining, comprising: The power supply circuit is used to provide a stable voltage that meets the intrinsic safety requirements for the internal modules of the intrinsically safe IEPE vibration sensor and device. Multiple IEPE interface circuits, each IEPE interface circuit is used to connect an intrinsically safe IEPE vibration sensor, including an input interface voltage limiting protection unit and a DC blocking unit to remove the DC bias of the sensor output signal. The IEPE interface circuit is connected to the power supply circuit through a constant current source excitation circuit. Multiple signal conditioning circuits are connected one-to-one with the IEPE interface circuit to buffer and convert single-ended signals to differential amplification of sensor signals. The multi-channel ADC synchronous acquisition circuit is connected to all signal conditioning circuits and is used to synchronously acquire vibration signals from multiple channels and convert them into digital signals. The ZYNQ core module, connected to the multi-channel ADC synchronous acquisition circuit, includes an FPGA section and an ARM section. The FPGA section is used to configure ADC sampling parameters and transfer the acquired data, while the ARM section is used to run data processing algorithms. Both the WIFI module and the Gigabit Ethernet module are connected to the ZYNQ core module and are used to transmit the processed data to the host computer wirelessly or via wired means.

[0006] Preferably, the power supply circuit includes an intrinsically safe DC12V input circuit, a DC12V boost to DC24.5V circuit, a DC24.5V buck to DC24V circuit, a dual voltage and current limiting circuit, and a DC12V to DC6V circuit. The DC12V boost circuit front end, the buck circuit front end, and the intrinsically safe DC12V input terminal are all equipped with unidirectional conduction diodes, and the power supply circuit is encapsulated as an independent module. The signal conditioning circuit includes an input buffer and a single-ended to differential amplifier. The input buffer provides high input impedance and low output impedance, and the single-ended to differential amplifier is used to convert a single-ended signal into a differential signal around a 2.5V common-mode voltage.

[0007] This invention also provides an intrinsically safe vibration data processing method for mining applications, comprising: The sampling frequency and sampling duration configured in the ZYNQ core module are obtained, and the multi-channel ADC synchronous acquisition circuit is controlled to acquire the raw vibration data of multiple intrinsically safe IEPE vibration sensors. The raw vibration data is then preprocessed by digital filtering and outlier removal to obtain synchronous vibration data. The synchronous vibration data is subjected to time-frequency transformation to extract time-domain and frequency-domain features. The kurtosis value in the time-domain features is used to determine whether envelope analysis is triggered. If the kurtosis value exceeds a preset threshold, the Hilbert transform is performed on the corresponding channel data to extract the envelope spectrum. The frequency components that match the preset fault feature frequency library are searched in the envelope spectrum. Based on the comparison between the amplitude of the frequency component and the preset alarm threshold, a device health status diagnosis result is generated, and the diagnosis result and feature data are sent to the host computer via a WIFI module or a gigabit Ethernet module.

[0008] Preferably, the step of performing digital filtering and outlier removal preprocessing on the original vibration data to obtain synchronous vibration data includes: The original vibration data sequence of each channel is obtained, and the power frequency interference is filtered out by an IIR notch filter to obtain the first filtered data. The first filtered data is subjected to a 1Hz high-pass filter to remove the trend term and DC component, resulting in the second filtered data. The statistical distribution characteristics of the second filtered data are obtained, and its mean and standard deviation are calculated. Data points that exceed the mean ± 3 times the standard deviation are marked as outliers and replaced by linear interpolation of adjacent data points to obtain the cleaned data. The precise clock of the FPGA is used to add a timestamp to the cleaned data of each channel. With the time base of the first channel as a reference, cubic spline interpolation is performed on the data of other channels to resample and compensate for the small time deviation between channels, so as to obtain synchronous vibration data of multiple channels.

[0009] Preferably, the step of extracting time-domain features and frequency-domain features includes: For each channel in the synchronous vibration data, the difference between its maximum and minimum values ​​is calculated as the peak value, its root mean square value is calculated as the effective value, the ratio of its fourth central moment to the square of its variance is calculated as the kurtosis, the ratio of its effective value to the rectified average value is calculated as the waveform factor, and the ratio of its peak value to the rectified average value is calculated as the impulse factor, thus constructing a time-domain feature vector. The synchronous vibration data of each channel is windowed, and the Hanning window is used to reduce spectral leakage to obtain windowed data; Perform a Fast Fourier Transform on the windowed data to convert the time-domain signal to the frequency domain, obtaining the spectrum data; Extract the top K frequency components with the largest amplitudes and their corresponding amplitudes from the spectrum data, and calculate the centroid frequency and total spectral energy based on the spectrum data to construct a frequency domain feature vector.

[0010] Preferably, the step of determining whether to trigger envelope analysis based on the kurtosis value in the time-domain features, and extracting the envelope spectrum by performing Hilbert transform on the corresponding channel data, includes: Obtain the kurtosis value for each channel and compare it with a preset kurtosis threshold; If the kurtosis value of any channel is greater than the kurtosis threshold, the channel is marked as a suspected faulty channel, and the envelope analysis process is triggered. The synchronous vibration data of the suspected faulty channel were subjected to Hilbert transform to obtain its analytical signal; The magnitude of the analytic signal is calculated to obtain the envelope signal, and the envelope signal is subjected to a fast Fourier transform to obtain the envelope spectrum. The amplitude values ​​corresponding to each fault frequency in the preset fault feature frequency library are extracted from the envelope spectrum to generate a fault feature amplitude vector.

[0011] Preferably, the step of generating a device health status diagnosis result based on the comparison result of the amplitude of the frequency component with a preset alarm threshold includes: Obtain a preset fault characteristic frequency library, which includes the inner ring fault frequency, outer ring fault frequency, rolling element fault frequency, and cage fault frequency of the device to be monitored. Obtain pre-stored multi-level alarm thresholds, which include attention thresholds and warning thresholds corresponding to each fault frequency; compare each amplitude in the fault feature amplitude vector with the attention threshold and warning threshold for the corresponding fault frequency; If the amplitude of any fault frequency point exceeds the attention threshold but is lower than the warning threshold, an attention-level diagnostic result is generated and the suspected fault type is marked. If the amplitude at any fault frequency point exceeds the warning threshold, a warning-level diagnostic result is generated, and the fault type and severity are output. If the kurtosis values ​​of all channels do not exceed the threshold and the amplitude of all fault frequency points is lower than the attention threshold, a normal-level diagnostic result is generated.

[0012] The present invention is further configured to include steps of post-processing of diagnostic results and adjustment of transmission strategy: Based on the generated equipment health status diagnosis results, the data storage strategy is dynamically adjusted: if the diagnosis result is normal, only the time domain feature vector and frequency domain feature vector are stored to the local TF card; if the diagnosis result is attention level or warning level, the corresponding original vibration data and envelope spectrum are stored together. Real-time monitoring of the network status of the WIFI module and Gigabit Ethernet module, obtaining current network bandwidth and connection stability parameters; The transmission strategy is dynamically adjusted based on the network status and diagnostic results: diagnostic results and feature data are transmitted first through the currently available network; when the network bandwidth is greater than a preset threshold, raw vibration data is uploaded as needed; when the network is unstable, the system switches to a backup network and enables the breakpoint resume mechanism.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of an intrinsically safe vibration data processing method for mining.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a mining intrinsically safe vibration data processing method.

[0015] The beneficial effects of this invention are as follows: This invention adopts an integrated multi-stage intrinsically safe power supply design and sets unidirectional conduction diodes at key nodes. While meeting the stringent explosion-proof requirements of underground coal mines, it significantly reduces the size of the device, allowing for convenient installation inside space-constrained equipment such as coal mining machines and tunneling machines. This overcomes the shortcomings of existing explosion-proof equipment, which is bulky and difficult to install. By integrating a multi-channel IEPE interface circuit and a multi-channel ADC synchronous acquisition circuit, this invention can simultaneously connect multiple intrinsically safe vibration sensors and ensure strict synchronization of data from each channel in hardware. This solves the problem of inconsistent data time at multiple monitoring points in existing technologies, providing a high-quality data foundation for accurate fault diagnosis based on multi-sensor information fusion. More importantly, this invention innovatively adopts an FPGA+ARM heterogeneous architecture for the ZYNQ core module. The FPGA is responsible for high-speed data acquisition and real-time transfer, while the ARM processor runs an intelligent diagnostic algorithm that includes kurtosis triggering mechanism, Hilbert envelope analysis, and graded threshold determination. This enables real-time on-site processing of vibration data and automatic extraction of fault features. Only the diagnostic results and key feature data are uploaded to the ground center, significantly reducing the transmission pressure on the underground communication network. At the same time, it enables immediate early warning, improves the real-time performance and accuracy of fault diagnosis for mining equipment, and provides a reliable guarantee for intelligent operation and maintenance and safe production in coal mines. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the device structure according to an embodiment of this application.

[0017] Figure 2 This is a flowchart of a method according to an embodiment of this application.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] like Figure 1 As shown, this application provides an intrinsically safe vibration data acquisition device for mining, comprising: The power supply circuit is used to provide a stable voltage that meets the intrinsic safety requirements for the internal modules of the intrinsically safe IEPE vibration sensor and device. Multiple IEPE interface circuits, each IEPE interface circuit is used to connect an intrinsically safe IEPE vibration sensor, including an input interface voltage limiting protection unit and a DC blocking unit to remove the DC bias of the sensor output signal. The IEPE interface circuit is connected to the power supply circuit through a constant current source excitation circuit. Multiple signal conditioning circuits are connected one-to-one with the IEPE interface circuit to buffer and convert single-ended signals to differential amplification of sensor signals. The multi-channel ADC synchronous acquisition circuit is connected to all signal conditioning circuits and is used to synchronously acquire vibration signals from multiple channels and convert them into digital signals. The ZYNQ core module, connected to the multi-channel ADC synchronous acquisition circuit, includes an FPGA section and an ARM section. The FPGA section is used to configure ADC sampling parameters and transfer the acquired data, while the ARM section is used to run data processing algorithms. Both the WIFI module and the Gigabit Ethernet module are connected to the ZYNQ core module and are used to transmit the processed data to the host computer wirelessly or via wired means.

[0021] As described in the hardware structure above, this invention constructs a vibration data acquisition platform that meets the intrinsic safety requirements for mining applications. Specifically, the power supply circuit is the energy core of the entire device, and its design fully considers the special safety requirements of underground coal mines. The intrinsically safe DC 12V input circuit serves as the external power interface, and its cascaded boost, buck, and dual voltage and current limiting circuits constitute a multi-stage power conversion network, capable of converting the input power to different voltage levels required by each module. Crucially, unidirectional conducting diodes are installed at the front end of the boost circuit, the front end of the buck circuit, and the intrinsically safe power input terminal. This design effectively prevents abnormal energy leakage and avoids generating electric sparks sufficient to ignite flammable and explosive gases in case of a fault. Simultaneously, the entire power supply circuit is encapsulated as an independent module, further isolating the circuit from the external environment and fundamentally meeting the explosion-proof requirements of intrinsically safe equipment in coal mines. Compared with the existing practice of simply using commercially available explosion-proof power supply modules in mining monitoring equipment, this power supply design is more targeted and has a higher degree of integration, significantly reducing the device size while ensuring intrinsic safety performance.

[0022] The IEPE interface circuit is a crucial bridge connecting the external intrinsically safe IEPE vibration sensor and the internal processing circuitry. Each interface circuit integrates an input interface voltage limiting protection unit and a DC blocking unit. The voltage limiting protection unit uses devices such as transient suppression diodes to quickly clamp the voltage when the sensor interface is subjected to electrostatic discharge or surge impacts, protecting the subsequent precision circuitry from damage. The DC blocking unit effectively blocks the DC bias voltage in the sensor output signal through capacitive coupling, allowing only the AC component reflecting vibration to pass through. This design resolves the technical contradiction that the intrinsically safe IEPE sensor output signal contains a large DC bias (typically 10-12V) while the subsequent ADC can only process low-voltage AC signals, while also enhancing the robustness and environmental adaptability of the interface.

[0023] Each signal conditioning circuit corresponds one-to-one with the IEPE interface circuit, undertaking the dual tasks of signal buffering and format conversion. Each signal conditioning circuit first achieves impedance transformation from high input impedance to low output impedance through an input buffer. The high input impedance ensures good matching with the sensor interface and avoids loading the sensor signal; the low output impedance enhances the driving capability, facilitating signal transmission in subsequent circuits. Subsequently, a single-ended to differential amplifier converts the buffered single-ended signal into a differential signal, using 2.5V as the common-mode voltage reference. This design offers several technical advantages: differential transmission effectively suppresses common-mode noise, improving the signal's anti-interference capability during transmission; setting the common-mode voltage to 2.5V allows the signal to oscillate within the 0-5V range in a single-supply system, avoiding the use of a negative voltage source, thus simplifying power supply design and reducing the complexity of intrinsically safe implementation. Compared to existing technologies that use single-ended transmission and require dual positive and negative power supplies, the signal conditioning circuit of this invention is simpler, more reliable, and suitable for intrinsically safe applications.

[0024] The multi-channel ADC synchronous acquisition circuit is the core component ensuring data synchronization. This circuit uses an 8-channel synchronous sampling ADC chip, with all channels sharing the same sampling clock and start signal. It can simultaneously perform sample-and-hold and analog-to-digital conversion on eight sensor signals, thus guaranteeing strict time alignment of the data from each channel. This is crucial for subsequent multi-sensor fusion analysis and fault diagnosis based on array signal processing, as even a tiny time deviation between channels can cause phase distortion, affecting the accuracy of fault feature extraction. The synchronous acquisition circuit connects to the ZYNQ core module via a parallel interface, enabling high-speed transmission of massive amounts of sampled data to the FPGA for processing.

[0025] The ZYNQ core module is the brain of the device, and its unique FPGA+ARM heterogeneous architecture provides an ideal platform for high-performance data acquisition and intelligent processing. The FPGA, with its advantages in parallel processing and hardware programmability, undertakes the task of managing high-speed data streams: it precisely configures the ADC's sampling rate, channel selection, and triggering mode through the SPI interface, efficiently transfers massive amounts of raw data to external DDR memory via DMA, and sends an interrupt signal to the ARM section once the data is ready. This hardware-level data transfer mechanism greatly reduces the processor's burden, allowing the ARM section to focus on running complex signal processing algorithms. The ARM section runs Linux or bare-metal programs, responsible for executing intelligent algorithms such as data preprocessing, feature extraction, envelope analysis, and fault diagnosis, fully leveraging the advantages of general-purpose processors in complex logic control and floating-point operations. The collaborative work of the FPGA and ARM achieves a perfect balance between high-speed data acquisition and complex signal processing, solving the problem in traditional solutions where a single processor cannot simultaneously meet real-time requirements and algorithm complexity.

[0026] The Wi-Fi module and the Gigabit Ethernet module together form a dual-mode communication system, providing a flexible and reliable channel for data transmission. The Gigabit Ethernet module is suitable for applications with good wired network coverage, enabling the transmission of massive amounts of raw vibration data with high bandwidth and low latency. The Wi-Fi module is suitable for scenarios where wiring is difficult or where mobile monitoring is required, providing the device with wireless access capabilities. More importantly, the dual-mode design allows the system to automatically switch transmission modes based on network status, seamlessly switching to wireless transmission when the wired network is interrupted, ensuring that critical diagnostic data is not lost, and greatly improving the system's reliability and environmental adaptability.

[0027] In summary, this invention constructs a complete, efficient, reliable vibration data acquisition device that meets the intrinsic safety requirements for mining applications through intrinsically safe power supply circuit design, IEPE interface protection and DC isolation, impedance transformation and differential conversion for signal conditioning, synchronous acquisition by multi-channel ADC, heterogeneous collaborative processing of ZYNQ, and flexible transmission via dual-mode communication. This hardware platform solves the problems of large size, difficult installation, and poor synchronization in existing technologies.

[0028] In one embodiment of the present invention, the power supply circuit includes an intrinsically safe DC12V input circuit, a DC12V boost to DC24.5V circuit, a DC24.5V buck to DC24V circuit, a dual voltage and current limiting circuit, and a DC12V to DC6V circuit. The DC12V boost circuit front end, the buck circuit front end, and the intrinsically safe DC12V input terminal are all equipped with unidirectional conduction diodes, and the power supply circuit is encapsulated as an independent module. The signal conditioning circuit includes an input buffer and a single-ended to differential amplifier. The input buffer provides high input impedance and low output impedance, and the single-ended to differential amplifier is used to convert a single-ended signal into a differential signal around a 2.5V common-mode voltage.

[0029] In a preferred embodiment, a TVS diode is added to the interface section of an intrinsically safe IEPE vibration sensor to implement voltage limiting protection, preventing surge voltage on the transmission line from damaging the interface circuit. After the voltage limiting protection circuit, a capacitor and resistor isolate the DC bias voltage of the sensor output signal, achieving high-pass filtering and allowing only signals above 1Hz to pass. Furthermore, resistor voltage division reduces the signal output amplitude, meeting the signal amplitude requirements of the subsequent input buffer circuit. After high-pass filtering and voltage division, the input and output impedances are separated by the input buffer, maintaining high input impedance and providing low output impedance for the signal. Additionally, by providing a DC 2.5V bias voltage in the high-pass filter / voltage divider circuit and the input buffer circuit, the common-mode voltage of the signal is raised to 2.5V, avoiding the use of negative voltage and eliminating the need for additional power supply circuit protection due to intrinsic safety requirements. The single-ended to differential circuit converts the single-ended signal output from the input buffer into a signal pair that differentially oscillates around the 2.5V common-mode voltage, increasing noise suppression and improving the power supply rejection ratio. The multi-channel synchronous acquisition ADC converts the input differential analog signal into a digital signal and transmits it to the ZYNQ core module in parallel for data preprocessing and data transmission.

[0030] like Figure 2 As shown, the present invention also provides an intrinsically safe vibration data processing method for mining applications, comprising: S1, obtain the sampling frequency and sampling duration configured in the ZYNQ core module, control the multi-channel ADC synchronous acquisition circuit to acquire the raw vibration data of multiple intrinsically safe IEPE vibration sensors, and perform digital filtering and outlier removal preprocessing on the raw vibration data to obtain synchronous vibration data; S2, perform time-frequency transformation on the synchronous vibration data, extract time-domain features and frequency-domain features, and determine whether to trigger envelope analysis based on the kurtosis value in the time-domain features. If the kurtosis value exceeds a preset threshold, perform Hilbert transformation on the corresponding channel data to extract the envelope spectrum, and search for frequency components in the envelope spectrum that match the preset fault feature frequency library. S3. Based on the comparison result between the amplitude of the frequency component and the preset alarm threshold, generate the device health status diagnosis result, and send the diagnosis result and feature data to the host computer through the WIFI module or gigabit Ethernet module.

[0031] As described in steps S1-S3 above, this invention constructs a processing flow from raw vibration data acquisition to intelligent fault diagnosis. Specifically, step S1 first utilizes the FPGA portion of the ZYNQ core module to configure high-precision sampling parameters, controlling the multi-channel ADC synchronous acquisition circuit to synchronously acquire raw data from eight intrinsically safe IEPE vibration sensors at a sampling rate of up to 20kHz. This process fully leverages the advantages of FPGA in real-time data acquisition and parallel processing, ensuring strict synchronization of multi-channel data at the hardware level. After acquisition, the ARM processor performs digital filtering and outlier removal preprocessing on the raw data. The digital filtering uses an IIR notch filter to effectively suppress the 50Hz power frequency interference commonly found in underground mining environments. Outlier removal is based on the 3σ criterion to identify and correct abnormal points caused by poor sensor contact or instantaneous impact. Finally, precise time alignment of multi-channel data is achieved through cubic spline interpolation resampling. This preprocessing process solves the problem in existing technologies where inconsistent acquisition times of multiple sensors and strong noise in the signal lead to decreased accuracy in subsequent analysis, providing high-fidelity synchronous vibration data for subsequent feature extraction and fault diagnosis.

[0032] Step S2, based on obtaining high-quality synchronous vibration data, firstly extracts comprehensive time-domain and frequency-domain features from the data of each channel. Time-domain features include peak value, RMS value, kurtosis, waveform factor, and impulse factor. These features reflect the amplitude level, energy magnitude, and impact characteristics of the vibration signal from different perspectives. Kurtosis, as a dimensionless parameter, is extremely sensitive to impact signals caused by early faults. Frequency-domain features are obtained by performing a Fast Fourier Transform on the windowed data to obtain the spectrum, and then extracting the main frequency components, centroid frequency, and total spectral energy to comprehensively characterize the frequency distribution characteristics of the signal. More importantly, this invention innovatively uses kurtosis as an intelligent switch to trigger envelope analysis. That is, only when the kurtosis value of a channel exceeds a preset threshold (usually 3) is a Hilbert transform performed on the data of that channel to extract the envelope spectrum, and then the frequency components matching the preset fault characteristic frequency library are searched within the envelope spectrum. Compared with the existing approach of performing envelope analysis on all channels, this strategy significantly reduces unnecessary computational overhead, making it possible to achieve real-time analysis on resource-constrained embedded platforms. At the same time, it avoids false triggering caused by noise interference, improving the stability and reliability of the system.

[0033] Step S3 further compares the fault feature frequency amplitudes extracted from the envelope spectrum with preset multi-level alarm thresholds to generate a graded diagnostic result. Specifically, when the amplitude of a fault frequency exceeds the attention threshold but is lower than the warning threshold, a "attention" level diagnostic result is generated and the suspected fault type is marked; when the amplitude exceeds the warning threshold, a "warning" level diagnostic result is generated and the specific fault type and severity are output; when the kurtosis values ​​of all channels do not exceed the threshold and all fault frequency amplitudes are lower than the attention threshold, a "normal" level diagnostic result is generated. This graded diagnostic mechanism avoids the shortcomings of traditional binary diagnosis (normal / fault), which is prone to missed detection in early weak faults and prone to false alarms in the presence of noise interference, providing field maintenance personnel with a more refined and reliable basis for decision-making. The diagnostic results and feature data are then sent to the host computer via a WIFI module or a gigabit Ethernet module. The dual-mode communication design adapts to the complex and ever-changing environment in coal mines. When the wired network is interrupted, it can automatically switch to wireless transmission, ensuring the real-time upload of critical data. In summary, by combining steps S1-S3, this invention achieves real-time acquisition, intelligent analysis, and graded early warning of vibration in mining equipment, thereby improving the accuracy and timeliness of fault diagnosis.

[0034] In one embodiment of the present invention, the step of performing digital filtering and outlier removal preprocessing on the original vibration data to obtain synchronous vibration data includes: S11, acquire the original vibration data sequence of each channel, filter out power frequency interference through an IIR notch filter, and obtain the first filtered data; S12, perform a 1Hz high-pass filter on the first filtered data to remove the trend term and DC component, and obtain the second filtered data; S13, obtain the statistical distribution characteristics of the second filtered data, calculate its mean and standard deviation, mark data points that exceed the mean ± 3 times the standard deviation as outliers, and replace them by linear interpolation of adjacent data points to obtain cleaned data; S14 uses the precise clock of the FPGA to add a timestamp to the cleaned data of each channel, and uses the time base of the first channel as a reference to perform cubic spline interpolation resampling on the data of other channels to compensate for the small time deviation between channels and obtain synchronous vibration data of multiple channels.

[0035] As described in steps S11-S14 above, the vibration data preprocessing workflow designed in this invention aims to solve the key quality problems faced in multi-channel vibration signal acquisition in mining environments. First, step S11 addresses the problem of dense power equipment and widespread power frequency interference in coal mines by using an IIR notch filter to process the original vibration data sequence of each channel. Compared to FIR filters, IIR filters have advantages such as lower order, less computational complexity, and easier phase control, making them particularly suitable for real-time operation on embedded platforms. The notch center frequency of the filter is precisely set to 50Hz (or 60Hz, determined according to the specific regional power frequency), with an appropriate bandwidth to ensure effective suppression of power frequency interference while preserving signal components near the fault characteristic frequency as much as possible. Through this processing, first filtered data with a significantly improved signal-to-noise ratio is obtained, laying a good foundation for subsequent analysis.

[0036] Step S12 further applies a 1Hz high-pass filter to the first filtered data. This operation serves a dual purpose: firstly, it removes the DC bias component introduced by the sensor or acquisition circuit; secondly, it eliminates low-frequency trend terms caused by equipment start-up and shutdown, load changes, etc. In vibration monitoring of mining equipment, speed increases and sudden load changes during equipment startup introduce slowly changing trend components into the vibration signal. If these components are not removed, they will seriously affect the accuracy of subsequent feature extraction, especially parameters sensitive to signal amplitude such as kurtosis and RMS value. This invention uses a digital high-pass filter to achieve this function. The 1Hz cutoff frequency is set based on the analysis of the typical speed range of mining equipment (usually from hundreds to thousands of revolutions per minute), ensuring the complete preservation of the rotational frequency and its harmonics while effectively removing non-fault-related low-frequency interference.

[0037] Step S13 introduces statistical methods for outlier detection and correction. In the harsh environment of mining, sensors may generate instantaneous abnormal spikes due to excessive vibration and impact, poor contact, or electromagnetic interference. If these outliers are not processed, they will severely interfere with feature extraction and fault diagnosis. This invention, based on the 3σ criterion, first calculates the mean and standard deviation of the second filtered data, and then marks data points exceeding the mean ± 3 times the standard deviation as potential outliers. For the marked outliers, linear interpolation of adjacent normal data points is used for replacement. This method achieves a good balance between computational efficiency and correction effect, avoiding the data length inconsistency problem caused by direct removal and preventing outliers from contaminating subsequent analysis. Through this processing, cleaned data with significantly improved signal quality is obtained.

[0038] Step S14 proposes an innovative solution to the time synchronization problem in multi-channel data acquisition. Even with synchronous acquisition circuits, slight time deviations may still exist between channel data due to differences in transmission delays at the analog front-ends of each channel and differences in ADC conversion times. These deviations are sufficient to affect subsequent joint analysis. This invention fully utilizes the hardware resource advantages of the FPGA section of the ZYNQ core module to add a high-precision hardware timestamp to the cleaned data of each channel, with timestamp accuracy reaching the nanosecond level. Then, using the time reference of the first channel, cubic spline interpolation is used to resample the data of other channels. Compared with linear interpolation, cubic spline interpolation has better smoothness and higher accuracy, and can accurately reconstruct the signal values ​​of each channel at the standard time point, thereby compensating for the slight time deviations between channels. The resulting strictly synchronized multi-channel vibration data provides a high-quality data foundation for subsequent multi-channel joint analysis, array signal processing, and fault diagnosis based on multi-sensor information fusion.

[0039] In one embodiment of the present invention, the step of extracting time-domain features and frequency-domain features includes: S21, for each channel in the synchronous vibration data, calculate the difference between its maximum and minimum values ​​as the peak value, calculate its root mean square value as the effective value, calculate the ratio of its fourth central moment to the square of its variance as the kurtosis, calculate the ratio of its effective value to the rectified average value as the waveform factor, and calculate the ratio of its peak value to the rectified average value as the impulse factor, and construct a time-domain feature vector. S22, Windowing is applied to the synchronous vibration data of each channel, and Hanning window is used to reduce spectral leakage to obtain windowed data; S23, Perform a fast Fourier transform on the windowed data to convert the time-domain signal to the frequency domain and obtain spectrum data; S24. Extract the top K frequency components with the largest amplitude and their corresponding amplitudes from the spectrum data, and calculate the centroid frequency and total spectral energy based on the spectrum data to construct a frequency domain feature vector.

[0040] As described in steps S21-S24 above, after obtaining high-quality synchronous vibration data, this invention characterizes the vibration signal's properties through multi-dimensional time-domain and frequency-domain feature extraction, providing rich quantitative basis for subsequent fault diagnosis. Specifically, step S21 first calculates a series of time-domain feature parameters for the data of each channel. The peak value reflects the maximum fluctuation range of the signal and has high sensitivity to impact faults; the effective value represents the overall energy level of the signal and is a basic indicator for measuring the severity of vibration; kurtosis, as a fourth-order statistic, is extremely sensitive to the impact component in the signal. When the equipment experiences an early fault, the kurtosis value will increase significantly, and therefore it is widely used as a sensitive indicator for fault early warning; the waveform factor and impulse factor describe the waveform shape and impact characteristics of the signal from different perspectives. Combined with the effective value and peak value, they can effectively distinguish different types of fault modes. Through the combination of these five parameters, a time-domain feature vector is constructed, realizing the comprehensive quantification of the time-domain characteristics of the vibration signal.

[0041] Step S22 involves windowing the time-domain data before performing frequency domain analysis. Since the actual acquired vibration signal has a finite length, directly performing a Fourier transform would result in spectral leakage, where signal energy leaks to adjacent frequency points, leading to spectral distortion. This invention uses a Hanning window for windowing, which has the advantages of a moderate main lobe width and rapid side lobe attenuation, effectively suppressing spectral leakage and improving the accuracy of spectral analysis. The windowed data obtained after windowing retains the time-domain waveform characteristics of the original signal and creates favorable conditions for subsequent spectral analysis.

[0042] Step S23 performs a Fast Fourier Transform (FFT) on the windowed data, converting the signal from the time domain to the frequency domain to obtain spectral data. The FFT is a highly efficient digital signal processing algorithm capable of performing real-time spectral calculations on large amounts of data on embedded platforms. Spectral analysis clearly shows the energy distribution of the vibration signal at various frequency points, especially the frequency components related to the equipment's rotation frequency and its harmonics, as well as fault characteristic frequencies. Spectral data forms the basis for subsequent frequency domain feature extraction and is also a crucial foundation for fault diagnosis.

[0043] Step S24 further extracts key frequency domain features from the spectral data. First, the top K frequency components with the largest amplitudes and their corresponding amplitudes are extracted. These components often represent the main energy concentration areas of the signal and may be related to the device's frequency conversion, harmonics, or fault characteristic frequencies. The value of K can be set according to actual needs, for example, 5 to 10, ensuring the integrity of the main information while avoiding excessive feature dimensionality. Simultaneously, two global features, the center-of-gravity frequency and the total spectral energy, are calculated. The center-of-gravity frequency reflects the distribution center of signal energy in the frequency domain. When a fault occurs, the energy distribution may shift to higher or lower frequencies, causing a change in the center-of-gravity frequency. The total spectral energy is another measure of the overall signal strength, complementing the effective value in the time domain. Through the above operations, a frequency domain feature vector containing local main frequency components and global statistical features is constructed, which, together with the time domain feature vector obtained in step S21, constitutes a comprehensive description of the vibration signal.

[0044] Compared to existing technologies that use only a single feature or a simple combination of features, this invention improves the ability to distinguish fault modes by jointly extracting multi-dimensional time-domain and frequency-domain features. Time-domain features are sensitive to impact and energy changes, while frequency-domain features are sensitive to frequency distribution and changes in the energy center of gravity. The two complement each other and can more comprehensively capture subtle changes in equipment status.

[0045] In one embodiment of the present invention, the step of determining whether envelope analysis is triggered based on the kurtosis value in the time-domain features, and extracting the envelope spectrum by performing Hilbert transform on the corresponding channel data, includes: S25, obtain the kurtosis calculation value of each channel and compare it with a preset kurtosis threshold. For example, the kurtosis threshold can be set to 3. S26. If the kurtosis value of any channel is greater than the kurtosis threshold, the channel is marked as a suspected faulty channel and the envelope analysis process is triggered. S27, Perform Hilbert transform on the synchronous vibration data of the suspected faulty channel to obtain its analytical signal. ,in The Hilbert transform is calculated using the following formula: In the formula, To synchronize vibration data, Pi is a constant. For integration variables, The signal after Hilbert transform is calculated using Fast Fourier Transform. For integration variables The differential; S28, calculate the magnitude of the analytic signal to obtain the envelope signal, and perform a fast Fourier transform on the envelope signal to obtain the envelope spectrum; S29, extract the amplitude corresponding to each fault frequency in the preset fault feature frequency library from the envelope spectrum, and generate a fault feature amplitude vector.

[0046] As described in steps S25-S29 above, this invention innovatively introduces a kurtosis-triggered envelope analysis mechanism, achieving efficient utilization of computing resources and accurate extraction of fault characteristics. Step S25 first obtains the kurtosis value calculated for each channel in the preceding steps and compares it with a preset kurtosis threshold. According to vibration theory, for a random signal following a normal distribution, its kurtosis value is approximately 3; when an impulse component appears in the signal, the kurtosis value increases significantly. Therefore, setting the kurtosis threshold to 3 is a theoretical critical point, and in practical applications, it can be fine-tuned based on field experience, for example, set to 3.5 to improve noise immunity. Through this comparison, channels that may have faults can be quickly screened out, avoiding indiscriminate full-channel envelope analysis.

[0047] Step S26 further marks the selected channels and triggers the subsequent envelope analysis process. This intelligent triggering mechanism contrasts sharply with the existing technology that performs envelope analysis on all channels regardless of their state. In mining multi-channel monitoring systems, the equipment is in normal operation most of the time, and the kurtosis value is usually less than 3. At this time, envelope analysis is unnecessary, which greatly reduces the computational burden on the ARM processor. This allows the system to run stably for a long time on resource-constrained embedded platforms, while ensuring a rapid response in the event of a fault.

[0048] Step S27 performs a Hilbert transform on the data marked as potentially faulty channels to obtain their analytic signals. The Hilbert transform is an all-pass filter that transforms a real signal into a complex analytic signal; its real part is the original signal, and its imaginary part is the Hilbert transform result of the original signal. The Hilbert transform allows for convenient extraction of the instantaneous amplitude and phase of the signal, forming the basis for envelope analysis. In digital implementation, the Fast Fourier Transform (FFT) is typically used to efficiently perform the Hilbert transform, i.e., performing an FFT on the original signal to construct a one-sided spectrum, followed by an inverse FFT to obtain the analytic signal.

[0049] Step S28 involves taking the modulus of the analytical signal to obtain the envelope signal, and then performing a Fast Fourier Transform (FFT) on the envelope signal to obtain the envelope spectrum. The envelope signal reflects the amplitude changes of the modulation components in the original vibration signal, and is particularly suitable for extracting the impact modulation sidebands caused by faults. In rolling bearing faults, when the rolling element passes through the damage point, it generates periodic impacts, thereby modulating the amplitude of the system's natural vibration frequency. This modulation information is often difficult to identify directly in the original spectrum, but it will appear as clear fault characteristic frequencies and their harmonics in the envelope spectrum. By performing spectral analysis on the envelope signal, these fault characteristics can be effectively demodulated, thereby improving the sensitivity and accuracy of fault diagnosis.

[0050] Step S29 extracts the amplitude corresponding to each fault frequency in the preset fault feature frequency library from the envelope spectrum, generating a fault feature amplitude vector. The fault feature frequency library pre-stores the fault feature frequencies of various components of the equipment to be monitored (such as the inner and outer rings of bearings, rolling elements, and cages). These frequencies can be calculated based on the equipment's geometric parameters and rotational speed. During actual extraction, the maximum spectral peak is searched within a certain frequency range, centered on each fault feature frequency, and the amplitude of this peak is used as the amplitude of the corresponding fault frequency. In this way, a set of amplitude vectors corresponding to each fault type is obtained, providing a quantitative basis for subsequent graded diagnosis. In summary, this invention, through a kurtosis-triggered envelope analysis mechanism, significantly reduces computational resource consumption while ensuring fault detection sensitivity, making embedded real-time diagnosis possible. Simultaneously, envelope spectrum analysis effectively demodulates fault feature information, overcoming the limitations of traditional spectrum analysis in detecting weak faults and improving the ability to identify early faults.

[0051] In one embodiment of the present invention, the step of generating a device health status diagnosis result based on the comparison result of the amplitude of the frequency component with a preset alarm threshold includes: S31, Obtain a preset fault characteristic frequency library, which includes the inner ring fault frequency, outer ring fault frequency, rolling element fault frequency and cage fault frequency of the device to be monitored. S32, obtain the pre-stored multi-level alarm thresholds, the multi-level alarm thresholds including attention thresholds and warning thresholds corresponding to each fault frequency; compare each amplitude in the fault feature amplitude vector with the attention threshold and warning threshold of the corresponding fault frequency; S33, If the amplitude of any fault frequency point exceeds the attention threshold but is lower than the warning threshold, then an attention-level diagnostic result is generated and the suspected fault type is marked; S34, If the amplitude at any fault frequency point exceeds the warning threshold, a warning-level diagnostic result is generated, and the fault type and severity are output; S35. If the kurtosis values ​​of all channels do not exceed the threshold and the amplitude of all fault frequency points is lower than the attention threshold, then a normal-level diagnostic result is generated.

[0052] As described in steps S31-S35 above, this invention expands the traditional binary diagnosis (normal / fault) into a three-level diagnosis (normal / attention / warning) by constructing a hierarchical diagnostic mechanism, thereby improving the information content and practicality of the diagnostic results. Step S31 first obtains a preset fault characteristic frequency library. This frequency library is calculated based on the equipment's structural parameters and real-time rotational speed, and includes key fault characteristic frequencies such as the inner ring fault frequency BPFI, outer ring fault frequency BPFO, rolling element fault frequency BSF, and cage fault frequency FTF. These frequencies are calculated based on the classic bearing fault frequency formula. Considering that the rotational speed may fluctuate during the actual operation of mining equipment, the frequency library can be dynamically updated to adapt to changes in rotational speed, or the frequency values ​​can be adjusted in real time using rotational speed tracking technology.

[0053] Step S32 compares the fault feature amplitude vector generated in step S29 with preset multi-level alarm thresholds. These thresholds are set comprehensively based on historical normal data, industry standards, and field experience, and include two levels: attention threshold and warning threshold. The attention threshold is lower and used to detect early, weak fault signs; the warning threshold is higher and used to confirm obvious faults. The threshold setting can adopt an adaptive method, for example, based on the statistical distribution of the amplitude of each frequency under normal conditions, the mean plus a certain multiple of the standard deviation can be used as the attention threshold, and a higher multiple can be used as the warning threshold, thereby adapting to the differences in different equipment and operating conditions.

[0054] Step S33 handles situations between the attention threshold and the warning threshold. When the amplitude of a fault frequency point exceeds the attention threshold but has not yet reached the warning threshold, it indicates that the component has shown early signs of failure, but has not yet developed into a serious stage. At this time, an "attention" level diagnostic result is generated, and the suspected fault type is marked, such as "minor wear of the inner ring" or "early pitting of the outer ring". This level of diagnosis prompts on-site personnel to pay closer attention and arrange planned maintenance, but does not require immediate shutdown. This avoids over-maintenance due to minor faults and prevents sudden shutdowns caused by the escalation of the fault.

[0055] Step S34 handles situations where the warning threshold is exceeded. When the amplitude at a certain fault frequency point exceeds the warning threshold, it indicates that the fault has developed to a relatively serious level, which may affect the normal operation of the equipment or even cause a safety accident. At this time, a "warning" level diagnostic result is generated, clearly outputting the fault type and severity, such as "Severe inner ring peeling, immediate shutdown and maintenance recommended." This level of diagnosis triggers an emergency alarm, reminding on-site personnel to take decisive measures to avoid equipment damage and safety accidents.

[0056] Step S35 handles the normal equipment condition. When the kurtosis values ​​of all channels do not exceed the threshold and the amplitude of all fault frequency points is lower than the attention threshold, it indicates that the equipment is operating well and there are no signs of fault. At this time, a "normal" level diagnostic result is generated. This result can not only be used for daily inspection records, but also as benchmark data for subsequent adaptive threshold adjustment.

[0057] Compared to the simple binary judgment in existing technologies, the three-level diagnostic mechanism of this invention has the following significant advantages: First, it provides richer status information, enabling maintenance personnel to distinguish between normal, early-stage, and severe faults, thereby formulating more targeted maintenance strategies; Second, it achieves predictive maintenance by providing early warnings through attention thresholds, avoiding sudden downtime caused by equipment operating with defects; Third, the setting of multi-level thresholds can flexibly adapt to different equipment and operating conditions, improving the versatility and robustness of the diagnostic system. Through the organic combination of steps S31-S35, this invention achieves refined assessment and intelligent early warning of the health status of mining equipment.

[0058] In one embodiment of the present invention, the method further includes steps of post-processing diagnostic results and adjusting the transmission strategy. S41, Based on the generated equipment health status diagnosis results, dynamically adjust the data storage strategy: if the diagnosis result is normal, only the time domain feature vector and frequency domain feature vector are stored to the local TF card; if the diagnosis result is attention level or warning level, the corresponding original vibration data and envelope spectrum are stored together. S42 monitors the network status of the WIFI module and Gigabit Ethernet module in real time, and obtains the current network bandwidth and connection stability parameters. S43, dynamically adjust the transmission strategy according to the network status and diagnostic results: prioritize the transmission of diagnostic results and feature data through the currently available network; when the network bandwidth is greater than the preset threshold, upload the original vibration data as needed; when the network is unstable, switch to the backup network and enable the breakpoint resume mechanism. S44 packages the data to be transmitted according to priority, sends it to the host computer through the selected network module, and clears the local cache after successful transmission.

[0059] As described in steps S41-S44 above, this invention designs an intelligent data storage and transmission strategy, effectively solving the data management problems caused by limited storage resources and unstable network communication in mining environments. Step S41 first dynamically adjusts the data storage strategy based on the diagnostic results. During equipment operation, it is in a normal state most of the time. At this time, only feature data that reflects the health trend of the equipment needs to be saved, without retaining massive amounts of raw vibration data. When the diagnostic result is "normal," this invention only stores the time-domain feature vector and frequency-domain feature vector extracted in steps S21 and S24 to the local TF card. These feature data have low dimensionality and small data volume, and can be stored for a long time without occupying too much storage space. When the diagnostic result is "attention" or "warning," it indicates that the equipment may have a fault. At this time, it is necessary to retain the raw vibration data and envelope spectrum and other detailed information for subsequent in-depth analysis, fault backtracking, or as evidence. This differentiated storage strategy ensures fault traceability while greatly saving storage resources and extending the continuous working time of the equipment in unattended situations.

[0060] Step S42 involves real-time monitoring of the network status of the WIFI module and the Gigabit Ethernet module. The underground environment in coal mines is complex; wired networks may be interrupted due to line damage, and wireless networks may be unstable due to signal attenuation. This invention continuously monitors the network connection status to obtain key parameters such as bandwidth, latency, and packet loss rate of the currently available network, comprehensively evaluating the network's connectivity and transmission capabilities. These parameters provide real-time basis for subsequent intelligent transmission decisions.

[0061] Step S43 dynamically adjusts the transmission strategy based on network status and diagnostic results. This invention designs a multi-level transmission priority system: the highest priority is the diagnostic results themselves, which should be transmitted regardless of network conditions to ensure the ground monitoring center can obtain equipment status information in a timely manner; the second highest priority is feature data, used for trend analysis and historical records; and finally, raw vibration data, which has a large volume but relatively lower real-time requirements. When network bandwidth is sufficient and the connection is stable, all data is uploaded as needed; when network bandwidth is insufficient, diagnostic results and feature data are transmitted first, while raw vibration data is temporarily stored locally and uploaded after network recovery; when the current network is interrupted, it automatically switches to a backup network (e.g., from Ethernet to Wi-Fi, or vice versa) and enables a breakpoint resume mechanism to ensure no data loss. This adaptive transmission strategy makes full use of limited communication resources, ensuring the real-time nature and integrity of critical information.

[0062] Step S44 specifically implements data packaging and transmission. The ARM processor packages the data to be transmitted according to the priority order of diagnostic results, characteristic data, and raw vibration data, adds necessary timestamps and device identification information, and sends it to the host computer through the network module selected in step S43. After successful transmission, the data in the local cache is cleared in a timely manner to release storage space and free up resources for subsequent data acquisition. If transmission fails, retransmission or temporary storage is performed according to the strategy.

[0063] Through intelligent storage and transmission management in steps S41-S44, this invention overcomes the challenges of limited storage space and unreliable network communication in mining environments, enabling long-term continuous data recording and timely uploading of critical information. Compared with the simple cyclic overwrite storage and fixed transmission mode of existing technologies, the adaptive strategy of this invention significantly improves system reliability and data utilization, providing strong support for remote intelligent operation and maintenance of mining equipment.

[0064] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of an intrinsically safe vibration data processing method for mining.

[0065] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a mining intrinsically safe vibration data processing method.

[0066] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0067] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A mining intrinsically safe vibration data acquisition device, characterized in that, include: The power supply circuit is used to provide a stable voltage that meets the intrinsic safety requirements for the internal modules of the intrinsically safe IEPE vibration sensor and device. Multiple IEPE interface circuits, each IEPE interface circuit is used to connect an intrinsically safe IEPE vibration sensor, including an input interface voltage limiting protection unit and a DC blocking unit to remove the DC bias of the sensor output signal. The IEPE interface circuit is connected to the power supply circuit through a constant current source excitation circuit. Multiple signal conditioning circuits are connected one-to-one with the IEPE interface circuit to buffer and convert single-ended signals to differential amplification of sensor signals. The multi-channel ADC synchronous acquisition circuit is connected to all signal conditioning circuits and is used to synchronously acquire vibration signals from multiple channels and convert them into digital signals. The ZYNQ core module, connected to the multi-channel ADC synchronous acquisition circuit, includes an FPGA section and an ARM section. The FPGA section is used to configure ADC sampling parameters and transfer the acquired data, while the ARM section is used to run data processing algorithms. Both the WIFI module and the Gigabit Ethernet module are connected to the ZYNQ core module and are used to transmit the processed data to the host computer wirelessly or via wired means.

2. The intrinsically safe vibration data acquisition device for mining according to claim 1, characterized in that, The power supply circuit includes an intrinsically safe DC12V input circuit, a DC12V boost to DC24.5V circuit, a DC24.5V buck to DC24V circuit, a dual voltage and current limiting circuit, and a DC12V to DC6V circuit. The DC12V boost circuit front end, the buck circuit front end, and the intrinsically safe DC12V input terminal are all equipped with unidirectional conduction diodes, and the power supply circuit is encapsulated as an independent module. The signal conditioning circuit includes an input buffer and a single-ended to differential amplifier. The input buffer provides high input impedance and low output impedance, and the single-ended to differential amplifier is used to convert a single-ended signal into a differential signal around a 2.5V common-mode voltage.

3. A method for processing intrinsically safe vibration data in mining, based on the intrinsically safe vibration data acquisition device for mining as described in claim 1, characterized in that, include: The sampling frequency and sampling duration configured in the ZYNQ core module are obtained, and the multi-channel ADC synchronous acquisition circuit is controlled to acquire the raw vibration data of multiple intrinsically safe IEPE vibration sensors. The raw vibration data is then preprocessed by digital filtering and outlier removal to obtain synchronous vibration data. The synchronous vibration data is subjected to time-frequency transformation to extract time-domain and frequency-domain features. The kurtosis value in the time-domain features is used to determine whether envelope analysis is triggered. If the kurtosis value exceeds a preset threshold, the Hilbert transform is performed on the corresponding channel data to extract the envelope spectrum. The frequency components that match the preset fault feature frequency library are searched in the envelope spectrum. Based on the comparison between the amplitude of the frequency component and the preset alarm threshold, a device health status diagnosis result is generated, and the diagnosis result and feature data are sent to the host computer via a WIFI module or a gigabit Ethernet module.

4. The intrinsically safe vibration data processing method for mining according to claim 3, characterized in that, The steps of performing digital filtering and outlier removal preprocessing on the raw vibration data to obtain synchronous vibration data include: The original vibration data sequence of each channel is obtained, and the power frequency interference is filtered out by an IIR notch filter to obtain the first filtered data. The first filtered data is subjected to a 1Hz high-pass filter to remove the trend term and DC component, resulting in the second filtered data. The statistical distribution characteristics of the second filtered data are obtained, and its mean and standard deviation are calculated. Data points that exceed the mean ± 3 times the standard deviation are marked as outliers and replaced by linear interpolation of adjacent data points to obtain the cleaned data. The precise clock of the FPGA is used to add a timestamp to the cleaned data of each channel. With the time base of the first channel as a reference, cubic spline interpolation is performed on the data of other channels to resample and compensate for the small time deviation between channels, so as to obtain synchronous vibration data of multiple channels.

5. The intrinsically safe vibration data processing method for mining according to claim 3, characterized in that, The steps for extracting time-domain and frequency-domain features include: For each channel in the synchronous vibration data, the difference between its maximum and minimum values ​​is calculated as the peak value, its root mean square value is calculated as the effective value, the ratio of its fourth central moment to the square of its variance is calculated as the kurtosis, the ratio of its effective value to the rectified average value is calculated as the waveform factor, and the ratio of its peak value to the rectified average value is calculated as the impulse factor, thus constructing a time-domain feature vector. The synchronous vibration data of each channel is windowed, and the Hanning window is used to reduce spectral leakage to obtain windowed data; Perform a Fast Fourier Transform on the windowed data to convert the time-domain signal to the frequency domain, obtaining the spectrum data; Extract the top K frequency components with the largest amplitudes and their corresponding amplitudes from the spectrum data, and calculate the centroid frequency and total spectral energy based on the spectrum data to construct a frequency domain feature vector.

6. The intrinsically safe vibration data processing method for mining according to claim 3, characterized in that, The step of determining whether to trigger envelope analysis based on the kurtosis value in the time domain features, and extracting the envelope spectrum by performing Hilbert transform on the corresponding channel data, includes: Obtain the kurtosis value for each channel and compare it with a preset kurtosis threshold; If the kurtosis value of any channel is greater than the kurtosis threshold, the channel is marked as a suspected faulty channel, and the envelope analysis process is triggered. The synchronous vibration data of the suspected faulty channel were subjected to Hilbert transform to obtain its analytical signal; The magnitude of the analytic signal is calculated to obtain the envelope signal, and the envelope signal is subjected to a fast Fourier transform to obtain the envelope spectrum. The amplitude values ​​corresponding to each fault frequency in the preset fault feature frequency library are extracted from the envelope spectrum to generate a fault feature amplitude vector.

7. The intrinsically safe vibration data processing method for mining according to claim 6, characterized in that, The step of generating a device health status diagnosis result based on the comparison result of the amplitude of the frequency component with the preset alarm threshold includes: Obtain a preset fault characteristic frequency library, which includes the inner ring fault frequency, outer ring fault frequency, rolling element fault frequency, and cage fault frequency of the device to be monitored. Obtain pre-stored multi-level alarm thresholds, which include attention thresholds and warning thresholds corresponding to each fault frequency; compare each amplitude in the fault feature amplitude vector with the attention threshold and warning threshold for the corresponding fault frequency; If the amplitude of any fault frequency point exceeds the attention threshold but is lower than the warning threshold, an attention-level diagnostic result is generated and the suspected fault type is marked. If the amplitude at any fault frequency point exceeds the warning threshold, a warning-level diagnostic result is generated, and the fault type and severity are output. If the kurtosis values ​​of all channels do not exceed the threshold and the amplitude of all fault frequency points is lower than the attention threshold, a normal-level diagnostic result is generated.

8. The intrinsically safe vibration data processing method for mining according to claim 7, characterized in that, It also includes steps for post-processing diagnostic results and adjusting transmission strategies: Based on the generated equipment health status diagnosis results, the data storage strategy is dynamically adjusted: if the diagnosis result is normal, only the time domain feature vector and frequency domain feature vector are stored to the local TF card; if the diagnosis result is attention level or warning level, the corresponding original vibration data and envelope spectrum are stored together. Real-time monitoring of the network status of the WIFI module and Gigabit Ethernet module, obtaining current network bandwidth and connection stability parameters; The transmission strategy is dynamically adjusted based on the network status and diagnostic results: diagnostic results and feature data are transmitted first through the currently available network; when the network bandwidth is greater than a preset threshold, raw vibration data is uploaded as needed; when the network is unstable, the system switches to a backup network and enables the breakpoint resume mechanism.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 3 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 3 to 8.

Images