Carrier roller fault diagnosis method and system based on coupled vibration analysis
By using a coupled vibration analysis-based method and employing Φ-OTDR signal separation and a dual-flow network structure, the problems of accuracy and early identification in idler roller fault diagnosis were solved, enabling intelligent and online management of idler roller status.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-24
- Publication Date
- 2026-03-27
AI Technical Summary
Existing methods for diagnosing idler roller faults are difficult to achieve long-distance, real-time, and accurate early fault diagnosis. In particular, DAS-based methods have low diagnostic accuracy and high false alarm rate in complex environments, making it difficult to identify idler roller fault characteristics.
A method based on coupled vibration analysis is adopted, which separates the direct vibration signal and the conveyor belt conducted vibration signal by phase-sensitive optical time-domain reflectometry (Φ-OTDR), calculates their respective feature vectors, and uses a dual-stream network structure to perform feature fusion to achieve the probability distribution determination of the fault category.
It achieves accurate and robust diagnosis of idler roller faults, especially with strong early fault identification capabilities, reducing sensor deployment and maintenance costs, and adapting to high diagnostic accuracy under complex working conditions.
Smart Images
Figure CN121743947A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of industrial equipment condition monitoring and fault diagnosis technology, and particularly relates to a method and system for fault diagnosis of idler rollers based on coupled vibration analysis. Background Technology
[0002] Idler rollers are critical load-bearing components of belt conveyors. Due to their large number and long-term operation in harsh environments, they are prone to malfunctions such as bearing wear, cylinder scratches, and jamming. Idler roller failures not only lead to conveyor belt misalignment and accelerated wear, but can also cause major safety accidents such as fires and belt breakage. Therefore, real-time and accurate fault diagnosis of idlers is of great significance for ensuring the safe, stable, and efficient operation of the conveyor system.
[0003] Existing methods for diagnosing idler roller faults mainly include:
[0004] Manual inspection: relies on experience, is inefficient, cannot achieve continuous online monitoring, and is difficult to detect early faults.
[0005] Point-based monitoring based on vibration sensors: Installing acceleration sensors on critical idlers. This method has high sensor deployment and maintenance costs, is difficult to cover thousands of idlers on long-distance conveyor lines, and the sensors themselves are easily damaged in harsh environments.
[0006] Visual inspection based on infrared or visible light images is susceptible to interference from ambient light, dust, and water mist, and is difficult to diagnose internal bearing failures.
[0007] Acoustic or ultrasonic-based monitoring suffers from significant environmental noise interference and limited positioning accuracy.
[0008] Distributed optical fiber acoustic sensing (DAS) technology has been introduced into pipeline monitoring, perimeter security, and other fields in recent years. Utilizing a single ordinary communication optical fiber as a sensor, it can continuously sense vibration / acoustic signals over a range of several kilometers to tens of kilometers along a pipeline, offering advantages such as full distribution, long distance, resistance to electromagnetic interference, and intrinsic security. Preliminary research has attempted to apply DAS to conveyor inspection, but existing methods typically treat DAS signals as simple "vibration events" for detection and location, or only perform simple spectral analysis, resulting in a limited diagnostic dimension. The fundamental problem lies in the fact that DAS acquires a comprehensive vibration field signal excited by a faulty idler and propagated and modulated through a complex path (especially via the conveyor belt). Directly extracting features corresponding one-to-one with the idler fault from this signal is extremely difficult, leading to low diagnostic accuracy, high false alarm rate, and particularly difficulty in diagnosing early, subtle faults.
[0009] Therefore, there is an urgent need for a new diagnostic system and method that can fully utilize the advantages of DAS fully distributed monitoring and accurately extract and identify roller fault characteristics from complex coupled vibration signals. Summary of the Invention
[0010] Purpose of the invention: In order to solve the problems existing in the prior art, the present invention provides a method and system for diagnosing idler roller faults based on coupled vibration analysis.
[0011] Technical solution: This invention discloses a method for diagnosing idler roller faults based on coupled vibration analysis, specifically as follows:
[0012] Based on the physical parameters of the conveyor system, the phase-sensitive optical time-domain reflectometry (Φ-OTDR) signal is separated into a direct vibration signal and a conveyor belt-conducted vibration signal. The direct vibration signal is denoted as the first frequency band signal, and the conveyor belt-conducted vibration signal is denoted as the second frequency band signal.
[0013] Calculate the feature vectors of the first frequency band signal to obtain the first feature vector set;
[0014] Calculate the feature vectors of the second frequency band signal to obtain the second feature vector set;
[0015] Feature fusion is performed on the first feature vector set and the second feature vector set. Based on the fused features, the probability distribution of the fault category is obtained.
[0016] Furthermore, the separation of the phase-sensitive optical time-domain reflectometry (Φ-OTDR) signal includes preprocessing the Φ-OTDR signal; the preprocessing includes noise reduction and signal enhancement of the Φ-OTDR signal.
[0017] Furthermore, the fault characteristic frequency is calculated based on the geometric parameters and rotational speed of the idler roller bearing. ,Will As the fundamental frequency, the frequency band related to the fundamental frequency and its harmonics is used as the first frequency band signal. The frequency band with a frequency lower than the fundamental frequency and related to the longitudinal vibration mode of the conveyor belt, as well as the frequency band containing the sideband centered on the fundamental frequency and modulated by the vibration of the conveyor belt, are jointly determined as the second frequency band signal.
[0018] Furthermore, the separation of the preprocessed Φ-OTDR signal based on the physical parameters of the conveying system specifically involves: using the characteristic frequency of idler roller bearing failure... Based on, The frequency band and the main harmonic frequency band are used as the first frequency band signal, wherein the main harmonic frequency band includes , For window size, , Indicates the total number of coefficients;
[0019] Will frequency band and The sidebands on both sides serve as the second frequency band signal; among which The characteristic modulation frequency of the conveyor belt.
[0020] Furthermore, the first feature vector set includes root mean square value, peak factor, kurtosis, spectral centroid, frequency variance, and MFCC coefficients;
[0021] The second feature vector set includes the Hilbert envelope spectrum and its features. The neighborhood energy, modulation sideband energy ratio, and approximate entropy.
[0022] Furthermore, the method for calculating the modulation sideband energy ratio is as follows: in the envelope spectrum of the vibration signal transmitted by the conveyor belt, taking the fault characteristic frequency of the idler roller as the center, calculate the ratio of the total energy in the sidebands with a preset width on both sides to the energy at the center frequency point.
[0023] Furthermore, a two-stream network structure is used to fuse the first feature vector set and the second feature vector set. The two-stream network includes a first feature extraction subnetwork, a second feature extraction subnetwork, and a feature fusion unit. The first feature extraction subnetwork is used to process the first feature vector set, and the second feature extraction subnetwork is used to process the second feature vector set. The outputs of the first feature extraction subnetwork and the second feature extraction subnetwork are both input to the feature fusion unit. The feature fusion unit outputs the probability distribution of the fault category.
[0024] A roller fault diagnosis system based on coupled vibration analysis, including
[0025] The coupled vibration signal separation and feature extraction module is used to separate the Φ-OTDR signal based on the physical parameters of the conveyor system, separating it into direct vibration signal and conveyor belt conducted vibration signal; and to calculate the feature vector of the first frequency band signal to obtain the first feature vector set; and to calculate the feature vector of the second frequency band signal to obtain the second feature vector set.
[0026] The multi-source feature fusion and fault diagnosis module is used to fuse the first feature vector set and the second feature vector set, and obtain the probability distribution of the fault category based on the fused features.
[0027] The monitoring center module is used to monitor the status of the idler rollers in real time and issue alarms when a fault occurs.
[0028] A computer device includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the steps of a roller fault diagnosis method based on coupled vibration analysis as described in any one of claims 1 to 7.
[0029] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a roller fault diagnosis method based on coupled vibration analysis as described in any one of claims 1 to 7.
[0030] Beneficial effects:
[0031] (1) Deeper and more comprehensive diagnostic dimensions: For the first time, the “idler-conveyor belt coupled vibration” model was clearly proposed and utilized to separate the characteristics of direct vibration and transmitted vibration from the DAS signal, thus more completely depicting the physical nature of the fault.
[0032] (2) Strong early fault identification capability: The modulation sideband and nonlinear characteristics in the conducted vibration signal are more sensitive to early weak faults, which makes up for the weakness of the direct vibration signal under strong background noise.
[0033] (3) High anti-interference and robustness: The dual-channel feature fusion mechanism, especially the deep learning model that introduces the attention mechanism, can adaptively balance the contribution of the two types of features and maintain high diagnostic accuracy under complex working conditions (such as material impact and background mechanical vibration).
[0034] (4) Strong engineering practicality: Based on a single optical fiber, the monitoring of the entire line of idlers is realized, which greatly reduces the cost of sensor deployment and maintenance. The system structure is simple and easy to be modified and implemented on existing conveyors, truly realizing intelligent, online and precise management of the health status of idlers in long-distance belt conveyors. Attached Figure Description
[0035] Figure 1 This is a schematic diagram of the overall architecture of the idler roller fault diagnosis system provided in an embodiment of the present invention.
[0036] Figure 2 This is a block diagram of the internal structure of the coupled vibration signal separation and feature extraction module provided in an embodiment of the present invention.
[0037] Figure 3 This is a schematic diagram of the structure of the deep learning model in the multi-source feature fusion and fault diagnosis module provided in an embodiment of the present invention.
[0038] Figure 4 This is an overall flowchart of the idler roller fault diagnosis method provided in an embodiment of the present invention. Detailed Implementation
[0039] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0040] See Figure 1The overall system architecture of this invention includes: a DAS data acquisition module 100, a coupled vibration signal separation and feature extraction module 200, and a multi-source feature fusion and fault diagnosis module 300. A sensing optical fiber 101 is laid along the conveyor frame below the conveyor belt 102, penetrating the area where all idlers 103 are located. The DAS data acquisition module 100 injects probe light into the optical fiber and receives backscattered Rayleigh light, demodulating a Φ-OTDR signal containing vibration information. This signal is then sent to the coupled vibration signal separation and feature extraction module 200 and the multi-source feature fusion and fault diagnosis module 300 for processing.
[0041] See Figure 2 The coupled vibration signal separation and feature extraction module 200 specifically includes:
[0042] Signal preprocessing unit 201: Performs wavelet threshold denoising and moving average filtering on the original Φ-OTDR signal to suppress high-frequency environmental noise and baseline drift.
[0043] Frequency band segmentation unit 202: based on the fault characteristic frequency of the idler roller bearing (This can be calculated based on known bearing parameters and belt speed, serving as a benchmark.) and f bpf The main harmonic frequency band of the fundamental frequency (with Centered on (k=1, 2, ..., K), with each center on both sides The set of windows; where K is usually 3 to 5. The first frequency band (based on frequency resolution and rotational speed fluctuation) is used to extract the direct vibration characteristics of the idler roller. (Setting...) Frequency bands (covering low-order longitudinal modes of the conveyor belt) and Side strips on both sides ( The second frequency band (typically a few Hz to tens of Hz) is used to extract the vibration characteristics transmitted by the conveyor belt.
[0044] This embodiment sets the bearing failure characteristic frequency f as the standard. bpf The first frequency band centered on is This range adopts f bpf The ±20% relative bandwidth centered on the theoretical frequency is due to the following reasons: Under actual operating conditions, the idler roller speed is affected by belt speed fluctuations, slippage, load changes, etc., and the bearing geometric parameters have manufacturing tolerances and assembly errors, resulting in the actual fault characteristic frequency being relatively lower than the theoretical frequency. bpf This results in a certain degree of drift; the lower limit is set to 0.8f. bpf This can cover the characteristic frequency drift towards the lower frequency side and the low-side adjacent components generated by modulation, avoiding the truncation of effective fault energy; if changed to 0.9f bpfA narrower lower limit makes it easier to miss detections at low speeds or during slippage. The upper limit is set to 1.2f. bpf It can cover the drift towards the high-frequency side while controlling the bandwidth to be too wide, thereby reducing the probability of introducing adjacent non-faulty components (such as structural resonance, material impact, and excitation from other components), improving the characteristic signal-to-noise ratio and discrimination stability, and reducing computational redundancy caused by irrelevant frequencies.
[0045] At the same time, the low frequency range is set as The basis for extracting conveyor belt transmitted vibration / modulation information is that fault excitation often manifests as "f" in the transmission path. bpf As carrier frequency, with low frequency f m In the form of modulation, the modulation frequency and the low-order modes in the longitudinal direction of the conveyor belt are usually concentrated at a frequency much lower than f. bpf The low-frequency region; the upper limit of the low frequency is limited to 0.5f. bpf This ensures that the low-frequency characteristic band is consistent with f. bpf The nearby carrier frequency / harmonic intervals do not overlap, which facilitates the effective separation of direct vibration components and conducted vibration components. At the same time, it suppresses the contamination of low-frequency modulation characteristics by high-frequency random noise and irrelevant impacts, thereby improving robustness and generalization ability under different belt speeds and load conditions.
[0046] Dual-channel feature calculation unit 203:
[0047] On the first frequency band signal, calculate: time-domain statistical characteristics, spectral characteristics, and Mel frequency cepstral coefficients. The time-domain statistical characteristics and spectral characteristics include: root mean square value, peak factor, kurtosis and other time-domain statistics; spectral centroid, frequency variance, etc.
[0048] Calculate the Hilbert envelope spectrum and its characteristics on the second frequency band signal. Neighborhood energy at a given location; modulation sideband energy ratio (formula is...) ,in and for Both sides ± Spectral energy within the range for The energy at the location); nonlinear complexity features such as approximate entropy constitute the second feature vector set.
[0049] See Figure 3The deep learning diagnostic model in the multi-source feature fusion and fault diagnosis module 300 adopts a two-stream network structure. The first feature extraction sub-network 301 processes the first feature vector set. The second feature extraction sub-network 302 processes the second feature vector set. The feature fusion unit 303 receives the high-order features output by the two sub-networks, calculates their respective weights through an attention layer, and performs weighted concatenation. The final classifier 304 (fully connected layer + Softmax) outputs the probability distribution of fault categories, namely "healthy", "bearing outer ring fault", "bearing inner ring fault", "ball bearing fault", "cylinder imbalance", etc.
[0050] See Figure 4 The process of the method of the present invention is as follows:
[0051] Step S401: The DAS system continuously acquires Φ-OTDR data and extracts time-series signals according to the spatial channel (corresponding to each idler roller position).
[0052] Step S402: Preprocess the signal (denoise reduction, enhancement).
[0053] Step S403: Based on the current idler roller speed information, adaptively calculate... And perform frequency band segmentation to obtain direct vibration components and transmitted vibration components.
[0054] Step S404: Extract the first and second feature vector sets from the two components in parallel.
[0055] Step S405: Input the two sets of feature vectors into the trained two-stream deep learning model.
[0056] Step S406: The model performs feature fusion and inference, and outputs the fault diagnosis result of the idler roller.
[0057] Step S407: Visualize the diagnostic results (including location, fault type, and severity) and report them to the monitoring center.
[0058] This invention, through the aforementioned system and method, achieves accurate and robust diagnosis of idler roller faults, especially early-stage faults. Those skilled in the art can adjust the specific parameters of the frequency band division, feature selection, network structure, etc., without departing from the principles of this invention; such adjustments should also be considered within the scope of protection of this invention.
[0059] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.
Claims
1. A method for diagnosing idler roller faults based on coupled vibration analysis, characterized in that, Specifically: Based on the physical parameters of the conveyor system, the phase-sensitive optical time-domain reflectometry (Φ-OTDR) signal is separated into a direct vibration signal and a conveyor belt-conducted vibration signal. The direct vibration signal is denoted as the first frequency band signal, and the conveyor belt-conducted vibration signal is denoted as the second frequency band signal. Calculate the feature vectors of the first frequency band signal to obtain the first feature vector set; Calculate the feature vectors of the second frequency band signal to obtain the second feature vector set; Feature fusion is performed on the first feature vector set and the second feature vector set. Based on the fused features, the probability distribution of the fault category is obtained.
2. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 1, characterized in that, The separation of the phase-sensitive optical time-domain reflectometry (Φ-OTDR) signal includes preprocessing the Φ-OTDR signal; the preprocessing includes noise reduction and signal enhancement of the Φ-OTDR signal.
3. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 1, characterized in that, Calculate the fault characteristic frequency based on the geometric parameters and rotational speed of the idler roller bearing. ,Will As the fundamental frequency, the frequency band related to the fundamental frequency and its harmonics is used as the first frequency band signal. The frequency band with a frequency lower than the fundamental frequency and related to the longitudinal vibration mode of the conveyor belt, as well as the frequency band containing the sideband centered on the fundamental frequency and modulated by the vibration of the conveyor belt, are jointly determined as the second frequency band signal.
4. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 1, characterized in that, The separation of the preprocessed Φ-OTDR signal based on the physical parameters of the conveying system specifically involves: using the characteristic frequency of idler bearing failure... Based on, The frequency band and the main harmonic frequency band are used as the first frequency band signal, wherein the main harmonic frequency band includes , For window size, , Indicates the total number of coefficients; Will frequency band and The sidebands on both sides serve as the second frequency band signal; among which The characteristic modulation frequency of the conveyor belt.
5. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 1, characterized in that, The first set of eigenvectors includes root mean square value, peak factor, kurtosis, spectral centroid, frequency variance, and MFCC coefficients; The second feature vector set includes the Hilbert envelope spectrum and its features. The neighborhood energy, modulation sideband energy ratio, and approximate entropy.
6. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 5, characterized in that, The method for calculating the modulation sideband energy ratio is as follows: in the envelope spectrum of the vibration signal transmitted by the conveyor belt, with the fault characteristic frequency of the idler roller as the center, calculate the ratio of the total energy in the sidebands with a preset width on both sides to the energy at the center frequency point.
7. The method for diagnosing idler roller faults based on coupled vibration analysis according to claim 1, characterized in that, A two-stream network structure is used to fuse the first feature vector set and the second feature vector set. The two-stream network includes a first feature extraction subnetwork, a second feature extraction subnetwork, and a feature fusion unit. The first feature extraction subnetwork is used to process the first feature vector set, and the second feature extraction subnetwork is used to process the second feature vector set. The output of the first feature extraction subnetwork and the attribute of the second feature extraction subnetwork are both input to the feature fusion unit; The feature fusion unit outputs the probability distribution of fault categories.
8. The roller fault diagnosis system based on coupled vibration analysis according to claim 1, characterized in that, include The coupled vibration signal separation and feature extraction module is used to separate the Φ-OTDR signal based on the physical parameters of the conveyor system, separating it into direct vibration signal and conveyor belt conducted vibration signal; and to calculate the feature vector of the first frequency band signal to obtain the first feature vector set; and to calculate the feature vector of the second frequency band signal to obtain the second feature vector set. The multi-source feature fusion and fault diagnosis module is used to fuse the first feature vector set and the second feature vector set, and obtain the probability distribution of the fault category based on the fused features. The monitoring center module is used to monitor the status of the idler rollers in real time and issue alarms when a fault occurs.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the roller fault diagnosis method based on coupled vibration analysis as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the roller fault diagnosis method based on coupled vibration analysis as described in any one of claims 1 to 7.