Carrier roller fault monitoring method and device based on AI data annotation and electronic equipment
By using AI-based data annotation methods and time-frequency analysis and deep learning algorithms, real-time monitoring and accurate identification of idler roller faults were achieved, solving the problem of low efficiency in manual inspection and improving the monitoring effect of idler roller operation status.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YHD
- Filing Date
- 2026-02-10
- Publication Date
- 2026-04-24
AI Technical Summary
In the existing technology, the monitoring of faults in the operation of belt conveyor idlers relies on manual inspection, which has problems such as low efficiency, discontinuous monitoring, inaccurate fault location, and untimely information feedback, leading to safety hazards.
By employing an AI-based data annotation method, the acoustic vibration signals of idler rollers are acquired, and time-frequency fusion data is generated using time-frequency analysis and deep learning algorithms. Combined with normal and abnormal data sample sets, real-time monitoring and identification of idler roller faults are achieved.
It enables rapid location and accurate identification of idler roller faults, reduces reliance on manual inspections, improves the real-time performance and accuracy of monitoring, and ensures the safety and performance of the equipment.
Smart Images
Figure CN121913293A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of equipment fault diagnosis technology, specifically to a method, device, and electronic equipment for monitoring idler roller faults based on AI data annotation. Background Technology
[0002] Belt conveyors are key equipment in production and transportation in coal mines, power plants, and industrial logistics. Due to their large load capacity, they are prone to wear and tear, leading to idler roller failures. Idler roller malfunctions can cause belt misalignment, resulting in serious production safety accidents such as wear, tearing, burning, and even fires. Currently, monitoring the operational status of belt conveyor idler rollers still relies primarily on manual inspection. Although some companies have adopted intelligent robots for automated inspection, problems remain, including low inspection efficiency, discontinuous monitoring, inaccurate location of idler roller malfunctions, and untimely feedback of fault information. Therefore, a real-time, efficient, and reliable monitoring method is urgently needed to improve the overall performance and safety of conveyor idler rollers. Summary of the Invention
[0003] To address the aforementioned technical problems, this invention provides a method, apparatus, and electronic device for monitoring idler roller faults based on AI data annotation.
[0004] To achieve the above objectives, the present invention adopts the following technical solution:
[0005] This invention provides a method for monitoring idler roller faults based on AI data annotation. For a target belt conveyor, steps S1 to S3 are used to construct normal and abnormal data sample sets for data comparison. Step A then enables real-time monitoring and identification of idler roller faults.
[0006] Step S1: Obtain the acoustic vibration signal generated by each idler roller in the running state, and generate the original vibration data corresponding to each idler roller through preprocessing operation. The running state includes normal state and various fault states.
[0007] Step S2: Based on the original vibration data, the vibration waveform data and noise frequency data corresponding to each idler are generated through time-frequency analysis and frequency domain analysis. The vibration waveform data and noise frequency data of each idler are then fused using a time-frequency fusion depth algorithm to generate time-frequency fusion data corresponding to each idler.
[0008] Step S3: Based on the preset state fluctuation range corresponding to the normal state and each different fault state, the state label of each time-frequency fusion data is determined by AI data annotation. Then, the time-frequency fusion data is grouped according to the state label to form a normal data sample set and an abnormal data sample set. The abnormal data sample set is composed of a subset of samples from different fault states.
[0009] Step A: Acquire the acoustic vibration signals of each idler roller in real time. Generate time-frequency fusion data for each idler roller according to steps S1 to S2. Use AI data annotation to compare each time-frequency fusion data with the normal data sample set constructed in step S3. Combine with the preset fluctuation range, determine whether there is time-frequency fusion data that does not meet the preset fault threshold, i.e., fault data. If so, use AI data annotation to compare the fault data with the abnormal data sample set constructed in step S3, and combine with the preset thresholds for different faults to determine the fault type. Otherwise, return to step A to perform real-time monitoring of the idler roller.
[0010] Further, step S1 acquires the acoustic vibration signal generated by each idler roller during operation as follows:
[0011] Step S11: The pulsed laser signal is transmitted to the target belt conveyor using a sensing optical cable, and the reverse detection optical signal in the sensing optical cable is received. The reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller.
[0012] Step S12: Based on the photovoltaic effect and photoelectric effect, the received reverse probe light signal is converted into a photoelectric signal and then quantized and encoded to obtain the corresponding analog current and voltage signal.
[0013] Step S13: The analog current and voltage signals are converted from analog to digital and then denoised and demodulated by an analog-to-digital converter to convert them into vibration signals. Furthermore, based on the Rayleigh scattering characteristics of the optical signal, the spatial position of each idler roller in the length direction of the sensing optical cable is calibrated according to the speed of light and the time difference of light signal return, thereby obtaining the acoustic vibration signal generated by each idler roller in the running state.
[0014] Further, the preprocessing operation in step S1 is as follows: for the acoustic vibration signal generated by each idler roller in operation, the acoustic vibration signal is normalized by Hilbert transform, and the normalized acoustic vibration signal is windowed and framed by Hamming window function. Then, the complementary set empirical mode decomposition deep learning algorithm is used for noise reduction and demodulation, thereby generating the original vibration data corresponding to each idler roller.
[0015] Another aspect of the present invention provides an idler roller fault monitoring device based on AI data annotation, comprising:
[0016] The acquisition module is used to acquire the acoustic vibration signals generated in real time by each idler of the target belt conveyor during operation, and the acoustic vibration signals are obtained based on the reverse detection optical signals in the sensing optical cable.
[0017] The data processing module is used to preprocess and perform time-frequency fusion analysis on the acoustic vibration signal to generate time-frequency fusion data corresponding to each idler roller.
[0018] The AI data labeling module is used to determine the state labels of each time-frequency fusion data and form a normal data sample set and an abnormal data sample set. Then, for the time-frequency fusion data to be analyzed, based on the preset state fluctuation range under normal conditions, the time-frequency fusion data that does not meet the preset state fluctuation range is selected, i.e., fault data. The fault data is compared with the abnormal data sample set, and the fault type is determined by combining the preset thresholds of different faults.
[0019] Furthermore, the acquisition module includes:
[0020] The transmission module is used to transmit pulsed laser signals to the idler rollers via a sensing optical cable;
[0021] The receiving module is used to receive the reverse detection optical signal generated in the sensing optical cable, and the reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller;
[0022] The first conversion module is used to convert the received reverse probe light signal into a photoelectric signal, and after quantization and encoding, obtain the corresponding analog current and voltage signal.
[0023] The conversion module is used to convert analog current and voltage signals into vibration signals, and combined with the Rayleigh scattering characteristics of optical signals, to obtain the acoustic vibration signals generated by each idler roller in operation.
[0024] Furthermore, the present invention also provides an electronic device for monitoring idler roller faults based on AI data annotation, including a memory and a processor. The memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the aforementioned method for monitoring idler roller faults based on AI data annotation.
[0025] The beneficial effects of adopting the above technical solution are as follows:
[0026] (1) Based on the Rayleigh scattering characteristics of light signals, this invention can calibrate the spatial position of each idler in the direction of the length of the sensing optical cable according to the speed of light and the time difference of light signal return, and obtain the acoustic vibration signal generated by each idler in the running state, thereby realizing long-distance all-round monitoring, full coverage of idler operation status monitoring, and all-weather online real-time monitoring, and realizing rapid location of the time and location of idler failure.
[0027] (2) This invention reduces the reliance on manual inspection by comparing data based on AI data annotation, improves the accuracy of fault detection, and realizes intelligent fault judgment. Attached Figure Description
[0028] Figure 1 This is a flowchart of the idler roller fault monitoring method based on AI data annotation of the present invention;
[0029] Figure 2 This is a schematic diagram of the idler roller fault monitoring device based on AI data annotation according to the present invention;
[0030] Figure 3 This is a schematic diagram of the electronic device for monitoring idler roller faults based on AI data annotation, as described in this invention. Detailed Implementation
[0031] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.
[0032] Example 1:
[0033] refer to Figure 1 This embodiment describes an AI-based data annotation-based idler roller fault monitoring method. For a target belt conveyor, steps S1 to S3 are used to construct normal and abnormal data sample sets for data comparison. Step A then enables real-time monitoring and identification of idler roller faults.
[0034] Step S1: Acquire the acoustic vibration signals generated by each idler roller under normal and different fault conditions, and generate the original vibration data corresponding to each idler roller through preprocessing operation. The fault conditions are achieved by replacing the idler rollers of the belt conveyor with different types of faults.
[0035] Step S2: Based on the original vibration data, the vibration waveform data and noise frequency data corresponding to each idler are generated through time-frequency analysis and frequency domain analysis. The vibration waveform data and noise frequency data of each idler are then fused using a time-frequency fusion depth algorithm to generate time-frequency fusion data corresponding to each idler.
[0036] Step S3: Based on the preset state fluctuation range corresponding to the normal state and each different fault state, the state label of each time-frequency fusion data is determined by AI data annotation. Then, the time-frequency fusion data is grouped according to the state label to form a normal data sample set and an abnormal data sample set. The abnormal data sample set is composed of a subset of samples from different fault states.
[0037] Step A: Acquire the acoustic vibration signals of each idler roller in real time. Generate time-frequency fusion data for each idler roller according to steps S1 to S2. Use AI data annotation to compare each time-frequency fusion data with the normal data sample set constructed in step S3. Combine with the preset fluctuation range, determine whether there is time-frequency fusion data that does not meet the preset fault threshold, i.e., fault data. If so, use AI data annotation to compare the fault data with the abnormal data sample set constructed in step S3, and combine with the preset thresholds for different faults to determine the fault type. Otherwise, return to step A to perform real-time monitoring of the idler roller.
[0038] Further, step S1 acquires the acoustic vibration signal generated by each idler roller during operation as follows:
[0039] Step S11: The pulsed laser signal is transmitted to the target belt conveyor using a sensing optical cable, and the reverse detection optical signal in the sensing optical cable is received. The reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller.
[0040] Step S12: Based on the photovoltaic effect and photoelectric effect, the received reverse probe light signal is converted into a photoelectric signal and then quantized and encoded to obtain the corresponding analog current and voltage signal.
[0041] Step S13: The analog current and voltage signals are converted from analog to digital and then denoised and demodulated by an analog-to-digital converter to convert them into vibration signals. Furthermore, based on the Rayleigh scattering characteristics of the optical signal, the spatial position of each idler roller in the length direction of the sensing optical cable is calibrated according to the speed of light and the time difference of light signal return, thereby obtaining the acoustic vibration signal generated by each idler roller in the running state.
[0042] Further, the preprocessing operation in step S1 is as follows: for the acoustic vibration signal generated by each idler roller in operation, the acoustic vibration signal is normalized by Hilbert transform, and the normalized acoustic vibration signal is windowed and framed by Hamming window function. Then, the complementary set empirical mode decomposition deep learning algorithm is used for noise reduction and demodulation, thereby generating the original vibration data corresponding to each idler roller.
[0043] Furthermore, in this embodiment, for each original vibration data, time-domain analysis and frequency-domain analysis are performed using Hilbert spectrum and Hilbert marginal spectrum to generate vibration waveform data and noise frequency data generated by each idler roller due to vibration, respectively. The extreme points of vibration waveform data and noise frequency data are obtained by time-frequency data fusion depth algorithm, each extreme point is fitted with time curve, and the local features of time-frequency data are superimposed to generate time-frequency fusion data corresponding to each idler roller.
[0044] Example 2:
[0045] refer to Figure 2 The idler roller fault monitoring device based on AI data annotation described in this embodiment includes:
[0046] The acquisition module is used to acquire the acoustic vibration signals generated in real time by each idler of the target belt conveyor during operation, and the acoustic vibration signals are obtained based on the reverse detection optical signals in the sensing optical cable.
[0047] The data processing module is used to preprocess and perform time-frequency fusion analysis on the acoustic vibration signal to generate time-frequency fusion data corresponding to each idler roller.
[0048] The AI data labeling module is used to determine the state labels of each time-frequency fusion data and form a normal data sample set and an abnormal data sample set. Then, for the real-time generated time-frequency fusion data to be analyzed, based on the preset state fluctuation range under normal conditions, it filters out time-frequency fusion data that does not meet the preset state fluctuation range, i.e., fault data. Then, it compares the fault data with the abnormal data sample set and, combined with the preset thresholds for different faults, determines the fault type.
[0049] In this embodiment, an optical fiber acoustic wave detection device, comprising a light source emitter, an optical coupler, an optical modulator, an optical signal amplifier, an isolator, a photoelectric converter, an analog-to-digital converter, and an optical fiber connector, is used to acquire the acoustic vibration signals generated in real time by each idler of the target belt conveyor during operation.
[0050] Specifically, the target belt conveyor 1 includes n idlers. A sensing optical cable 2 with an armored, built-in multi-core fiber optic sensor is installed along the entire length of the target belt conveyor and deployed at the fixed supports of 1 to n idlers. The other end of the sensing optical cable is connected to a fiber optic acoustic wave monitoring device 3. The fiber optic acoustic wave monitoring device 3 has at least one detection interface that can connect to at least one sensing optical cable 2. By deploying each sensing optical cable 2 on different belt conveyors, monitoring of idlers on multiple belt conveyors can be achieved simultaneously.
[0051] The fiber optic acoustic wave monitoring device 3 uses a light source transmitter to emit continuous optical signals. Based on optical coherence and heterodyne detection technology, an optical coupler splits the optical signal into a probe optical signal and a local optical signal. The probe optical signal is a signal light modulated by frequency, amplitude, and phase, while the local optical signal is the original frequency, original amplitude, and original phase of the light, serving as a reference light. An acousto-optic modulator converts the probe optical signal into an optical pulse signal, which is then amplified by an optical signal amplifier, isolated, and injected into the sensing optical cable 2. The detection light signal inside the sensing optical cable 2 will detect the regular rotational vibration of each point of 1 to n idlers, and transmit the generated vibration signal and audio signal backward to the optical coupler, where it beats with the original local light. The photoelectric converter decomposes the light signal and converts it into different levels based on the photovoltaic effect and photoelectric effect. After quantization and encoding, it forms a processable analog current and voltage signal. The analog current and voltage signal is converted into digital signal, and noise reduction and demodulation are performed by the analog-to-digital converter to convert the analog current and voltage signal into a vibration signal. Furthermore, based on the Rayleigh scattering characteristics of the light signal, the spatial position of each idler along the length of the sensing optical cable is determined according to the speed of light and the time difference of light signal return, thereby obtaining the acoustic vibration signal generated by each idler in the running state.
[0052] Furthermore, the acquisition module includes:
[0053] The transmission module is used to transmit pulsed laser signals to the idler rollers via a sensing optical cable;
[0054] The receiving module is used to receive the reverse detection optical signal generated in the sensing optical cable, and the reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller;
[0055] The first conversion module is used to convert the received reverse probe light signal into a photoelectric signal, and after quantization and encoding, obtain the corresponding analog current and voltage signal.
[0056] The conversion module is used to convert analog current and voltage signals into vibration signals, and combined with the Rayleigh scattering characteristics of optical signals, to obtain the acoustic vibration signals generated by each idler roller in operation.
[0057] Example 3:
[0058] Based on Example 2, refer to Figure 3 This embodiment provides an electronic device for monitoring idler roller faults based on AI data annotation, including a memory and a processor. The memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the idler roller fault monitoring method based on AI data annotation. As a specific implementation of this embodiment, the electronic device is a personal computer or a mobile terminal.
[0059] Specifically, the sensing optical cable is connected to a personal computer 5 via a data transmission device 4, which can be a switch, router, gateway, protocol converter, or network cable. This device transmits the collected acoustic vibration signals to the personal computer 5 for data analysis and fault diagnosis. All data involved in the analysis is stored in a pre-built database on the personal computer 5. The personal computer 5 has a built-in data processing module and an AI data annotation module responsible for further processing and analysis of the data. Based on preset algorithms, the personal computer 5 can identify different fault modes and determine the location and time of the faulty idler roller, further feeding back the analysis results to the operator for effective maintenance decisions. This electronic device can also be integrated with a digital twin platform to display and analyze data through 3D visualization, charts, and audio, thereby improving the efficiency and accuracy of fault diagnosis sample data annotation.
[0060] Specifically, the digital twin platform constructs a 1:1 three-dimensional virtual model of the idler roller fault monitoring device based on AI data annotation, and integrates the relevant data of 1 to n idler rollers labeled with AI data to display them in real time with three-dimensional dynamic effects, including the spatial position of 1 to n idler rollers, vibration waveform charts, audio, fault information and other data.
[0061] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A method for monitoring idler roller faults based on AI data annotation, characterized in that, For the target belt conveyor, construct normal data sample sets and abnormal data sample sets for data comparison according to steps S1 to S3, and realize real-time monitoring and identification of idler roller faults according to step A: Step S1: Obtain the acoustic vibration signal generated by each idler roller in the running state, and generate the original vibration data corresponding to each idler roller through preprocessing operation. The running state includes normal state and various fault states. Step S2: Based on the original vibration data, the vibration waveform data and noise frequency data corresponding to each idler are generated through time-frequency analysis and frequency domain analysis. The vibration waveform data and noise frequency data of each idler are then fused using a time-frequency fusion depth algorithm to generate time-frequency fusion data corresponding to each idler. Step S3: Based on the preset state fluctuation range corresponding to the normal state and each different fault state, the state label of each time-frequency fusion data is determined by AI data annotation. Then, the time-frequency fusion data is grouped according to the state label to form a normal data sample set and an abnormal data sample set. The abnormal data sample set is composed of a subset of samples from different fault states. Step A: Acquire the acoustic vibration signals of each idler roller in real time. Generate time-frequency fusion data for each idler roller according to steps S1 to S2. Use AI data annotation to compare each time-frequency fusion data with the normal data sample set constructed in step S3. Combine with the preset fluctuation range, determine whether there is time-frequency fusion data that does not meet the preset fault threshold, i.e., fault data. If so, use AI data annotation to compare the fault data with the abnormal data sample set constructed in step S3, and combine with the preset thresholds for different faults to determine the fault type. Otherwise, return to step A to perform real-time monitoring of the idler roller.
2. The method for monitoring idler roller faults based on AI data annotation according to claim 1, characterized in that, Step S1 acquires the acoustic vibration signal generated by each idler roller during operation as follows: Step S11: The pulsed laser signal is transmitted to the target belt conveyor using a sensing optical cable, and the reverse detection optical signal in the sensing optical cable is received. The reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller. Step S12: Based on the photovoltaic effect and photoelectric effect, the received reverse probe light signal is converted into a photoelectric signal and then quantized and encoded to obtain the corresponding analog current and voltage signal. Step S13: The analog current and voltage signals are converted from analog to digital and then denoised and demodulated by an analog-to-digital converter to convert them into vibration signals. Furthermore, based on the Rayleigh scattering characteristics of the optical signal, the spatial position of each idler roller in the length direction of the sensing optical cable is calibrated according to the speed of light and the time difference of light signal return, thereby obtaining the acoustic vibration signal generated by each idler roller in the running state.
3. The method for monitoring idler roller faults based on AI data annotation according to claim 1, characterized in that, The preprocessing operation in step S1 is as follows: for the acoustic vibration signal generated by each idler roller in operation, the acoustic vibration signal is normalized by Hilbert transform, and the normalized acoustic vibration signal is windowed and framed by Hamming window function. Then, the complementary set empirical mode decomposition deep learning algorithm is used for noise reduction and demodulation, thereby generating the original vibration data corresponding to each idler roller.
4. A roller fault monitoring device based on AI data annotation, characterized in that, include: The acquisition module is used to acquire the acoustic vibration signals generated in real time by each idler of the target belt conveyor during operation, and the acoustic vibration signals are obtained based on the reverse detection optical signals in the sensing optical cable. The data processing module is used to preprocess and perform time-frequency fusion analysis on the acoustic vibration signal to generate time-frequency fusion data corresponding to each idler roller. The AI data labeling module is used to determine the status labels of each time-frequency fusion data and form a normal data sample set and an abnormal data sample set. Then, for the time-frequency fusion data to be analyzed, based on the preset status fluctuation range under normal conditions, the time-frequency fusion data that does not meet the preset status fluctuation range is selected, i.e. fault data. The fault data is compared with the abnormal data sample set, and the fault type is determined by combining the preset thresholds for different faults.
5. The idler roller fault monitoring device based on AI data annotation according to claim 4, characterized in that, The acquisition module includes: The transmission module is used to transmit pulsed laser signals to the idler rollers via a sensing optical cable; The receiving module is used to receive the reverse detection optical signal generated in the sensing optical cable, and the reverse detection optical signal is generated by the rotational vibration generated in real time during the operation of the idler roller; The first conversion module is used to convert the received reverse probe light signal into a photoelectric signal, and after quantization and encoding, obtain the corresponding analog current and voltage signal. The conversion module is used to convert analog current and voltage signals into vibration signals, and combined with the Rayleigh scattering characteristics of optical signals, to obtain the acoustic vibration signals generated by each idler roller in operation.
6. An electronic device for monitoring idler roller faults based on AI data annotation, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the idler roller fault monitoring method based on AI data annotation according to any one of claims 1 to 3.