Belt conveyor idler failure detection method and system

Abstract

The application belongs to the technical field of idler fault detection, and discloses a belt conveyor idler fault detection method and system, which solves the problems of low detection efficiency, poor reliability and low accuracy in the prior art. The application adopts an optical fiber sensing + voiceprint recognition scheme to perform fault detection, so that idler fault detection can be completed by arranging optical fibers along the conveyor. Through an AI voiceprint recognition model, different fault types such as bearing damage and drum skin wear can be accurately distinguished, and environmental interference noise such as wind and rain can be effectively filtered out, so that the diagnosis accuracy is high. Therefore, the application realizes full automation and 7x24 hours of uninterrupted online monitoring, completely replacing the inefficient, high-risk and high-missed-inspection-rate manual inspection mode. Therefore, the application is very suitable for large-scale application and promotion.

Description

Technical Field

[0001] This invention belongs to the field of belt conveyor idler fault detection technology based on image recognition, and specifically relates to a method and system for detecting belt conveyor idler faults. Background Technology

[0002] Belt conveyors are key equipment used for transporting bulk materials in industries such as mining, ports, power, and chemicals. During operation, numerous idlers carrying the belt and materials operate under harsh conditions of heavy load and dust, making them one of the components with the highest failure rate. Common idler failures include bearing damage, cylinder wear, bending deformation, and weld cracking.

[0003] Currently, monitoring the operating status of idler rollers mainly relies on periodic manual inspections. This involves inspectors using tools they carry (such as listening rods) to approach the idler rollers and judge whether their operating sounds are abnormal based on hearing and experience. However, this traditional inspection method has many drawbacks: First, the inspection workload is large and inefficient, and the judgment results are highly dependent on the inspector's personal experience, making it subjective and difficult to guarantee accuracy. Second, manual inspections are periodic operations, making it impossible to achieve 24 / 7 uninterrupted real-time monitoring of the equipment, and difficult to detect sudden or intermittent faults in a timely manner. Furthermore, conducting close-range inspections in hazardous environments with high noise and dust, such as mines and chemical plants, poses a significant threat to the personal safety of inspectors and is prone to missed inspections due to environmental factors.

[0004] Therefore, there is an urgent need in this field for a roller fault detection technology that can achieve automation, real-time online operation, high-precision identification, and accurate positioning, in order to replace the traditional manual inspection method and improve the intelligence level and safety of belt conveyor operation and maintenance. Summary of the Invention

[0005] The present invention aims to solve the technical problems of low detection efficiency, poor real-time performance, insufficient security and easy missed detection caused by the reliance on manual inspection in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention first provides a method for detecting faults in idler rollers of belt conveyors, comprising the following steps: S1. Signal acquisition steps: Install sensing optical fibers along the frame of the belt conveyor to be tested, emit probe light pulses into the sensing optical fibers, and collect the back Rayleigh scattering light signals returned from each position point in the sensing optical fibers. S2. Signal demodulation and positioning steps: Based on phase-sensitive optical time-domain reflectometry, the backscattered Rayleigh light signal is demodulated to obtain vibration signals corresponding to each position point on the sensing fiber, and the occurrence position of each vibration signal is determined according to the principle of optical time-domain reflectometry. S3. Acoustic feature extraction step: Extract acoustic features from the vibration signal to characterize the running state of the idler roller; S4. Intelligent fault identification step: Input the voiceprint features into the pre-trained idler roller fault identification model, and the model outputs the corresponding idler roller fault type identification result.

[0007] Furthermore, the voiceprint features include at least one of Mel spectrum, contrast spectrum, and higher-order spectrum.

[0008] Furthermore, the idler roller fault identification model is a deep learning model built based on a convolutional neural network or a recurrent neural network.

[0009] Furthermore, the method also includes a sound restoration step: restoring the vibration signal into an audio signal and playing it for remote verification.

[0010] Furthermore, in the signal acquisition step, multiple enhanced sensors are spaced apart on the sensing optical fiber to enhance the acquisition sensitivity of the vibration of the idler roller running sound; the enhanced sensors include a suspended disc fiber shaft that is separate from the mounting base.

[0011] This invention also provides a belt conveyor idler roller fault detection system, comprising: a sensing optical fiber, used to be laid along the frame of the belt conveyor under test to sense vibration signals generated by the operation of the idler roller; a signal processing host, connected to the sensing optical fiber, used to transmit probe light pulses to the sensing optical fiber and receive the backscattered Rayleigh light signals returned by it, demodulate and locate them based on phase-sensitive optical time-domain reflectometry, and output a vibration signal corresponding to the position; and a data processing platform, communicatively connected to the signal processing host, used to receive the vibration signal, extract the acoustic signature features therein, and use a pre-trained idler roller fault identification model to identify the acoustic signature features in order to output the fault type.

[0012] Furthermore, the system also includes multiple enhanced sensors connected in series on the sensing optical fiber. Each enhanced sensor includes: a base for mounting to the frame of the belt conveyor; and a fiber coil, separate from the base and suspended in mid-air, for winding and fixing a portion of the sensing optical fiber to enhance the sensitivity for acquiring acoustic vibration signals propagating in the air. The enhanced sensors employ a flexible mounting method, further reducing vibration. The "partial" refers to a section of the sensing optical fiber used for acoustic vibration signal acquisition, such as a 0.5m to 5m length of sensing optical fiber wound around it, with the suspended winding method improving the acoustic vibration acquisition sensitivity.

[0013] Furthermore, the signal processing host is a phase-sensitive optical time-domain reflectometry system based on pulse digital code modulation technology.

[0014] Furthermore, the data processing platform also includes a sound restoration module for restoring the vibration signal into an audio signal.

[0015] Furthermore, the data processing platform also includes an interface for linkage with the video surveillance system, which is used to retrieve video footage from the corresponding location for verification when a fault type is output.

[0016] Compared with the prior art, the belt conveyor idler roller fault detection method and system of the present invention has the following advantages: 1. It has achieved fully automated, 24 / 7 uninterrupted online monitoring, completely replacing the inefficient, high-risk, and high-miss-rate manual inspection method.

[0017] 2. Combining Φ-OTDR technology, it achieves meter-level or even higher precision positioning of faulty idlers. At the same time, through the AI ​​voiceprint recognition model, it can accurately distinguish different fault types such as bearing damage and drum wear, and effectively filter out environmental interference noise such as wind and rain, resulting in high diagnostic accuracy.

[0018] 3. The innovative "sound restoration" function allows maintenance personnel to "listen" to the on-site sound from the remote control room and can link with video monitoring, realizing an "immersive" remote review and diagnosis, which greatly improves the accuracy of fault diagnosis.

[0019] 4. The system can utilize spare fiber cores in existing communication optical cables as sensing optical fibers. The front-end sensing part (optical fiber and enhanced sensor) does not require power supply and has intrinsic safety characteristics. It is particularly suitable for flammable and explosive hazardous environments such as coal mines and chemical plants, while significantly reducing deployment costs and construction complexity.

[0020] 5. Through the application of a unique suspended disc fiber shaft design and PDC modulation technology, the structural vibration interference of the frame is effectively isolated, the ability to collect weak sound waves in the air is enhanced, and the system's high sensitivity and strong anti-interference ability are ensured in complex industrial environments. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of the belt conveyor idler roller fault detection method provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a belt conveyor idler roller fault detection system provided in an embodiment of the present invention; Figure 3 A schematic diagram of the structure of the enhanced sensor is provided for an embodiment of the present invention; Figure 4 This is a schematic diagram of a data processing platform provided in an embodiment of the present invention.

[0022] 100 sensing fiber, 200 signal processing host, 300 data processing platform, 110 enhanced sensor, 111 base, 112 fiber optic disc, 310 data receiving and preprocessing module, 320 voiceprint feature extraction module, 330 intelligent fault identification module, 340 sound restoration module, and 350 linkage verification module. Detailed Implementation

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0024] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0025] Example 1: like Figures 1 to 4 As shown, this embodiment provides a belt conveyor idler roller fault detection system. The system mainly comprises three core components: a sensing fiber optic cable 100, a signal processing host 200, and a data processing platform 300 (e.g., ...). Figure 2 (As shown).

[0026] 1. System composition and connection relationships: The sensing fiber optic cable 100, acting as the "sensory nerve" of the system, is laid along the frame of the conveyor belt to be tested. In actual deployment, spare fiber cores from existing communication optical cables in the mining area or factory area can be used, or a single-mode or multimode optical fiber can be specially laid. Preferably, to enhance the sensing capability of weak acoustic signals, multiple enhanced sensors 110 can be connected in series at intervals on the sensing fiber optic cable 100.

[0027] Signal processing host 200: Serving as the system's "demodulation hub," it is deployed in a safe area such as a monitoring room or underground chamber, and connected to one end of the sensing fiber optic cable 100. Internally, it integrates core optoelectronic devices such as a narrow-linewidth laser, an acousto-optic modulator, an erbium-doped fiber amplifier, a photodetector, and a high-speed data acquisition card, forming a complete phase-sensitive optical time-domain reflectometer (Φ-OTDR) system. In a preferred embodiment, this host is a high-performance Φ-OTDR system based on pulse digital code (PDC) modulation technology to improve the system's signal-to-noise ratio and detection range.

[0028] Data processing platform 300: As the "intelligent brain" of the system, it can be a local high-performance server or a cloud server deployed in the cloud. It establishes a data connection with the signal processing host 200 through communication links such as industrial Ethernet, 4G / 5G wireless network or fiber optic, and performs real-time data interaction.

[0029] 2. Structure and function of core components: Sensitizer 110: To address the issues of fiber optic cables being susceptible to structural vibration interference and insensitive to airborne sound transmission when directly attached to the rack, this embodiment provides a preferred sensitizer structure. For example... Figure 3 As shown, the sensor includes a base 111 and a fiber optic spindle 112. The base 111 is used for fixed installation on the frame of the belt conveyor. The fiber optic spindle 112 is physically separated from the base 111 and is suspended in the air. It is connected to the base 111 via a slender connecting rod or directly fixed to an independent support next to the frame. A section of the sensing optical fiber 100 is wound and fixed on the suspended fiber optic spindle 112. This suspended design allows the optical fiber on the fiber optic spindle 112 to primarily sense sound waves propagating through the air (i.e., the sound waves generated during the operation of the idler rollers), thereby effectively isolating structural noise from the frame and acting as a mechanical filter.

[0030] The signal processing host 200 works by injecting highly coherent probe light pulses into the sensing fiber 100 and continuously collecting interference signals of backscattered Rayleigh light caused by external vibrations (such as sound) at various points in the fiber. When the idler roller produces abnormal sounds, the refractive index at that location changes, causing a phase change in the backscattered light. Through phase demodulation technology, the signal processing host 200 can extract this phase change information, thereby reconstructing the original vibration signal at that location. Simultaneously, based on the principle of optical time-domain reflectometry, by calculating the time difference between the transmitted pulse and the received signal, the specific location of the vibration can be accurately pinpointed (e.g., the third idler roller 1.5 km from the host's starting point). Using PDC modulation technology, fading noise can be suppressed by encoding the pulse sequence, further improving positioning accuracy and signal-to-noise ratio, achieving a positioning accuracy within ±1 meter.

[0031] 3. Functional modules of the data processing platform 300: like Figure 4 As shown, the data processing platform 300 runs a specially designed software system, which includes the following key functional modules: Data receiving and preprocessing module 310: responsible for receiving massive vibration signal data from signal processing host 200 in real time and performing preprocessing such as noise reduction and normalization.

[0032] The acoustic signature feature extraction module 320 employs signal processing algorithms to extract acoustic signature features characterizing the idler roller's operating state from the pre-processed vibration signal. These features include not only traditional time-domain and frequency-domain features, but more importantly, the Mel spectrum, which effectively simulates human hearing characteristics, converting a one-dimensional vibration signal into a two-dimensional time-frequency image. Furthermore, as an alternative or supplementary solution, features such as contrast spectrum and higher-order spectrum (e.g., bispectral) can be extracted to more comprehensively reflect the nonlinearity and non-Gaussianity of faults, enhancing the ability to distinguish between different fault types (e.g., bearing pitting and cage fracture). This component effectively extracts acoustic spectrum information, using a multi-modal acoustic spectrum recognition algorithm based on Mel spectrum, contrast spectrum, and higher-order spectrum to extract higher-order spectral features, achieving near-zero false negatives and extremely low false positives.

[0033] Intelligent Fault Identification Module 330: This module incorporates a pre-trained idler roller fault identification model. In this embodiment, the model is constructed using a Convolutional Neural Network (CNN), such as a lightweight network structure like ResNet or MobileNet. The model takes the acoustic signature features extracted from the front end (such as Mel spectrograms) as input. After processing through multiple layers of convolution, pooling, and fully connected layers, the output layer uses the Softmax function to provide probability values ​​for various states of the current idler roller, such as "normal," "bearing damage," "drum wear," and "seal failure," and outputs the category with the highest probability as the final identification result. As a variant, this model can also use a Recurrent Neural Network (RNN) and its variants (such as LSTM and GRU) to process the temporal characteristics of the vibration signal.

[0034] Sound restoration module 340: To achieve "human-machine collaboration" verification, this module utilizes the physical homology between vibration signals and audio signals, and uses a digital-to-analog conversion algorithm to restore the vibration signal into an audible audio signal. Maintenance personnel can then click on the corresponding idler roller location in the monitoring center to "listen" to the real-time sound on-site via headphones or speakers, achieving remote "diagnosis."

[0035] Linkage Verification Module 350: This module provides an interface for linkage with the video surveillance system already deployed on-site. When the intelligent fault identification module 330 outputs a fault alarm for a certain idler roller, the linkage verification module 350 automatically retrieves real-time video footage or historical recordings from nearby cameras based on the fault location information, and displays them in split-screen format on the monitoring screen for maintenance personnel to perform visual verification and confirm whether there are any externally visible abnormalities such as smoke or jamming, forming an immersive diagnostic experience that combines sound and image.

[0036] Example 2: This embodiment provides a method for detecting faults in idler rollers of a belt conveyor based on the system described in Embodiment 1. For example... Figure 1 As shown, the method includes the following steps: S1. Signal acquisition steps: A sensing fiber optic cable 100 is laid along the frame of the belt conveyor under test. Optionally, in areas with dense idlers or critical locations, the fiber optic cable is suspended in mid-air by a sensor 110. The signal processing host 200 continuously emits high-coherence probe light pulses into the sensing fiber optic cable 100 and collects the backscattered Rayleigh light signals returned from various locations in the fiber in real time.

[0037] S2. Signal demodulation and positioning steps: The signal processing host 200 uses phase-sensitive optical time-domain reflectometry (Φ-OTDR) technology to demodulate the received backscattered light signal to extract a vibration signal proportional to the intensity of external vibration. Simultaneously, based on the time difference between the emission time of the light pulse and the reception time of the corresponding scattered light, the location of the vibration signal is precisely calculated using the speed of light propagation in the optical fiber. This location information, along with the vibration signal data, is then packaged and sent to the data processing platform 300.

[0038] S3. Voiceprint feature extraction steps: After receiving the vibration signal, the voiceprint feature extraction module 320 of the data processing platform 300 first performs preprocessing such as framing and windowing. Then, it performs a Fast Fourier Transform (FFT) on each frame of the signal to obtain its spectrum, and then filters it through a Mel filter bank to finally generate a Mel spectrogram. This step transforms the complex original vibration waveform into two-dimensional image features suitable for processing by deep learning models. In scenarios requiring higher recognition accuracy, higher-order spectral features of the signal can also be calculated in parallel and fused with the Mel spectrogram.

[0039] S4. Intelligent Fault Identification Steps: The Mel-ray spectrogram generated in step S3 is input into the pre-trained CNN model in the intelligent fault identification module 330. This model has been trained using a large number of idler roller fault acoustic samples collected in the field and simulated in the laboratory. After forward propagation calculation, the model outputs a multi-dimensional vector indicating the probability of each fault type corresponding to the current input signal. For example, if the output result is [normal 5%, bearing damage: 92%, drum wear: 3%], the system determines that the idler roller has a bearing damage fault and issues an alarm message.

[0040] S5. Remote Immersive Review Steps (Optional): After the system issues an alarm, the linkage process is automatically triggered: The sound restoration module 340 restores the original vibration signal to audio, allowing experts to remotely "listen" and confirm.

[0041] The linkage verification module 350 calls on the PTZ camera near the fault location, automatically adjusts the focus to align with the faulty roller, and pushes the real-time video image to the monitoring center.

[0042] On the same interface, maintenance personnel can observe the real-time video of the faulty idler while listening to the "click" noise played by the sound restoration module, enabling remote immersive verification, ultimately confirming the fault and arranging a maintenance plan.

[0043] Working principle: The working principle of this invention is based on the concept of "vibration as fingerprint." Each idler roller has its own specific acoustic signature during normal operation. When faults such as bearing wear or cylinder shell cracks occur, the vibration signal (i.e., acoustic signature) generated will change, producing new frequency components or modulation phenomena. This invention utilizes distributed optical fiber acoustic sensing (DAS) technology, using Φ-OTDR technology to turn the entire optical fiber into tens of thousands of "virtual microphones," enabling parallel and real-time acquisition of the operating status of hundreds or thousands of idler rollers. Subsequently, leveraging the powerful feature extraction and pattern recognition capabilities of AI deep learning models, it automatically identifies abnormal acoustic signatures representing different fault types from massive amounts of data, thereby achieving accurate early warning and location of idler roller faults.

[0044] Verification of beneficial effects: The system of this invention was deployed on a 5-kilometer-long main conveyor belt in a coal mine, covering approximately 8,000 sets of idlers. After three months of operational verification: Real-time performance: The system enables 24 / 7 uninterrupted monitoring, changing the previous manual inspection mode of 2 hours per shift.

[0045] Accuracy: A total of 12 idler roller fault warnings were issued during the period. After manual verification, 11 of them were genuine faults (9 cases of early bearing damage and 2 cases of severe wear on the roller skin), and only 1 was a false alarm caused by a large stone impact. The fault identification accuracy rate reached 91.7%. This successfully avoided a major safety accident caused by belt tearing due to idler roller jamming.

[0046] Safety: The sensor is partially passive and inherently safe, making it suitable for gaseous environments in underground coal mines. Maintenance personnel can complete all diagnostic work from the monitoring center, significantly reducing safety risks for personnel working underground.

[0047] Economic efficiency: Utilizing backup optical fiber in the underground ring network as the sensing medium significantly reduces cable laying costs. Precise preventative maintenance avoids excessive repairs and unplanned downtime, saving maintenance costs.

[0048] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention, such as replacing the CNN model with a Transformer model, or replacing the Mel spectrum with wavelet packet decomposition features, should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting faults in idler rollers of a belt conveyor, characterized in that, Includes the following steps: Signal acquisition steps: Install sensing optical fibers along the frame of the belt conveyor to be tested, emit probe light pulses into the sensing optical fibers, and collect the back Rayleigh scattering light signals returned from each position point in the sensing optical fibers. Signal demodulation and positioning steps: Based on phase-sensitive optical time-domain reflectometry, the backscattered Rayleigh light signal is demodulated to obtain vibration signals corresponding to each position point on the sensing fiber, and the occurrence position of each vibration signal is determined according to the principle of optical time-domain reflectometry. Acoustic signature extraction step: Extract acoustic signature features from the vibration signal to characterize the running state of the idler roller; Intelligent fault identification steps: Input the acoustic features into the pre-trained idler roller fault identification model, and the model outputs the corresponding idler roller fault type identification result.

2. The method for detecting faults in belt conveyor idlers according to claim 1, characterized in that, The voiceprint features include at least one of Mel spectrum, contrast spectrum, and higher-order spectrum.

3. The method for detecting faults in belt conveyor idlers according to claim 1 or 2, characterized in that, The idler roller fault identification model is a deep learning model built on convolutional neural networks or recurrent neural networks.

4. The method for detecting faults in belt conveyor idlers according to claim 1 or 2, characterized in that, It also includes a sound restoration step: restoring the vibration signal into an audio signal and playing it for remote verification.

5. The method for detecting faults in belt conveyor idlers according to claim 1 or 2, characterized in that, In the signal acquisition step, multiple enhanced sensors are spaced apart on the sensing optical fiber to enhance the acquisition sensitivity of the vibration of the idler roller running sound; the enhanced sensors include a suspended disc fiber shaft that is separate from the mounting base.

6. A fault detection system for belt conveyor idlers, characterized in that, include: The sensing fiber is laid along the frame of the belt conveyor to be tested to sense the vibration signals generated by the movement of the idlers. The signal processing host is connected to the sensing optical fiber and is used to send probe light pulses to the sensing optical fiber and receive the back Rayleigh scattering light signals returned by it. Based on phase-sensitive optical time-domain reflectometry, it demodulates and locates the signals and outputs vibration signals corresponding to the position. The data processing platform is communicatively connected to the signal processing host and is used to receive the vibration signal, extract the acoustic features therein, and use a pre-trained idler roller fault identification model to identify the acoustic features in order to output the fault type.

7. The belt conveyor idler roller fault detection system according to claim 6, characterized in that, It also includes multiple enhanced sensors connected in series on the sensing optical fiber, the enhanced sensors comprising: A base for mounting onto the frame of a belt conveyor; The fiber optic spindle, separate from the base and suspended in mid-air, is used to wind and fix part of the sensing fiber to enhance the sensitivity of acquiring acoustic vibration signals propagating in the air.

8. The belt conveyor idler roller fault detection system according to claim 6 or 7, characterized in that, The signal processing host is a phase-sensitive optical time-domain reflectometry system based on pulse digital code modulation technology.

9. The belt conveyor idler roller fault detection system according to claim 6 or 7, characterized in that, The data processing platform also includes a sound restoration module, used to restore the vibration signal to an audio signal.

10. The belt conveyor idler roller fault detection system according to claim 6 or 7, characterized in that, The data processing platform also includes an interface for linking with the video surveillance system, which is used to retrieve video footage from the corresponding location for verification when a fault type is output.

Images