Belt conveyor predictive maintenance system, method, sensor network and storage unit

Abstract

This application discloses a predictive maintenance system, method, sensor network, and storage unit for belt conveyors. It includes a multi-core optical cable and several silicon photonic triaxial accelerometers connected to different fiber branches therein. The application feeds detection light into the silicon photonic triaxial accelerometers via a silicon photonic demodulation and control unit. Based on the reflection of the detection light, it calculates the acceleration at each sensor's location to obtain the vibration state changes in the time domain along the belt conveyor in three orthogonal directions. Correspondingly, it extracts features, identifies faults, and generates predictive maintenance strategies. This application can proactively and in real-time assess the risk of abnormal belt conveyor faults and provide early warnings. Its sensing nodes are completely passive and intrinsically safe; it has a wide frequency response and high sensitivity; it uses a multi-point triaxial synchronous acquisition method to obtain equipment attitude information; it is suitable for long-distance, large-scale deployment with controllable costs; and it can also support predictive maintenance based on big data and artificial intelligence.

Description

Technical Field

[0001] This application belongs to the field of industrial conveying equipment technology, and in particular relates to a predictive maintenance system, method, sensor network and storage unit for belt conveyors. Background Technology

[0002] With the development of industries such as smart mines, smart ports, and smart chemicals, belt conveyors (hereinafter referred to as "belt conveyors"), as core equipment for bulk material transportation, exhibit characteristics such as long conveying distances, large conveying capacities, harsh operating environments, and high downtime losses. In large-scale belt conveyor systems, the number of idlers can reach tens of thousands, and abnormal conditions such as jamming, bearing damage, detachment, and uneven wear are among the most significant and common failures. If these abnormal conditions are not detected and addressed in a timely manner, they can lead to minor issues such as localized belt wear and increased energy consumption, or even serious problems such as belt tearing, idler fires, and even fires and explosions.

[0003] The following technical solutions are currently mainly used for condition monitoring and maintenance of belt conveyors: 1. Manual Inspection Plan In most cases, maintenance teams still rely on three shifts to conduct inspections, judging whether rollers, belts, and drive components are abnormal by listening, looking, and touching. This approach has several drawbacks: high manpower investment and labor intensity; heavy reliance on personnel experience, resulting in poor consistency; long inspection cycles, making it difficult to detect early faults in a timely manner; and low level of intelligence, failing to support predictive maintenance and data-driven management.

[0004] 2. Electrical Active Sensor Solution Some systems use electrical accelerometers, acoustic sensors, or temperature sensors as data sources to monitor the operating status of belt conveyors. In this scheme, the operating data of the belt conveyor is transmitted to the control center in real time via wired or wireless means for interpretation and early warning. This scheme has the following prominent problems: Wired methods require laying a large number of power supply and signal cables, resulting in limited load nodes, high costs, and complex construction; on long-distance belt conveyors, cables are prone to forming lightning conductors, and there is a risk of electric sparks in the live environment; wireless methods require battery power, but the detection period supported by batteries is limited, while vibration and acoustic signals often need to be continuously collected, so the sensor detection units deployed in the wireless method have extremely high battery capacity requirements, resulting in high maintenance costs; in addition, regardless of the data transmission method used, the electrical sensors and cables on which they are based suffer from poor durability and insufficient reliability in environments with high dust, high humidity, high corrosion, and strong electromagnetic interference. Summary of the Invention

[0005] The problem this application aims to solve is the difficulty in monitoring the condition of industrial conveying equipment and the inability to effectively achieve predictive maintenance. It provides an intelligent operation and maintenance and safety management system for long-distance belt conveyors in mines, ports, power plants, metallurgy, chemical industry and other fields. Based on the vibration data of the belt conveyor in all directions collected by fiber optic silicon photonics triaxial accelerometer, it predicts the fault point and performs timely proactive maintenance on the key parts of the system.

[0006] To solve the above-mentioned technical problems, the technical solution adopted in this application is as follows: a predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers, comprising: a multi-core optical cable arranged along the belt conveyor; silicon photonics triaxial accelerometers fixedly installed in the mechanical structure along the belt conveyor, each silicon photonics triaxial accelerometer being communicatively connected to the end of at least one fiber core in the multi-core optical cable; and a silicon photonics demodulation and control unit arranged at the other end of the multi-core optical cable, used to send detection light to the multi-core optical cable, receive reflected light from each silicon photonics triaxial accelerometer connected to the other end of the fiber, calculate the acceleration information in each direction at the location of each silicon photonics triaxial accelerometer based on the reflected light, obtain the vibration state changes in the three orthogonal directions in the time domain at each location along the belt conveyor, and thereby provide early warning of abnormal fault risks of the belt conveyor.

[0007] Optionally, in the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described above, the silicon photonics triaxial accelerometers are arranged at intervals on the idler frame or the side of the side frame of the belt conveyor.

[0008] Optionally, in the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described above, each of the multi-core optical cables has a pre-reserved interface at a preset interval for one or more fiber cores to be led out. The fiber cores led out from the interface are respectively connected to the corresponding silicon photonics triaxial accelerometer and / or distributed optical fiber vibration (DAS) device to detect the triaxial acceleration vector and / or vibration field information at different positions along the belt conveyor.

[0009] Optionally, in the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described above, the vibration state changes in the three orthogonal directions at each position along the belt conveyor include any one or any combination of the following: the pitch angle θ = arctan(ax_g / sqrt(ay_g^2 + az_g^2)) at the position of the silicon photonics triaxial accelerometer, and its variation with time and space; the roll angle φ = arctan(ay_g / sqrt(ax_g^2 + az_g^2)) at the position of the silicon photonics triaxial accelerometer, and its variation with time and space; the components of the vibration acceleration in each direction at the position of the silicon photonics triaxial accelerometer, ax_v(t), ay_v(t), az_v(t), and their variation with time and space; and the total root mean square acceleration a_rms = sqrt((1 / T) · ∫[0→T] ( ax_v(t)^2 + ay_v(t)^2 + az_v(t)^2 ) dt ), and its changes with time and space; based on the total root mean square acceleration a_rms corresponding to each position, the acceleration spectrum A(f_k) is obtained through discrete sampling and fast Fourier transform (FFT); the spectral peak value, frequency band energy and bandwidth of the acceleration spectrum A(f_k) in the idler rotation frequency, the harmonic frequency corresponding to the idler rotation, the belt fundamental frequency and its harmonic frequency band, and their changes with time and space; where ax_g, ay_g and az_g respectively represent the gravity components in the x, y and z directions corresponding to the position of the silicon photonics triaxial accelerometer obtained by low-pass filtering.

[0010] Optionally, in the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described above, the step of providing early warning of abnormal fault risks of the belt conveyor based on the vibration state changes in the three orthogonal directions at various positions along the belt includes: constructing feature vectors F = [a_rms, kurtosis, peak factor, band energy 1, band energy 2, ...] for each node of the silicon photonics triaxial accelerometer based on the spectral peak value, band energy, and bandwidth of the acceleration spectrum A(f_k) in the idler rotation frequency, the harmonic frequency corresponding to the idler rotation, the belt fundamental frequency, and its harmonic bands; and constructing a health index HI = Σ[i] (w_i·(F_i - F_i_ref) / Based on the abnormal changes in the feature vector F of each location node and its corresponding health index HI, and combined with the discrete cross-correlation function R_ij(τ) between different location nodes, the location s of the abnormal fault risk along the belt conveyor is estimated by the least squares method with the goal of minimizing the residual, and an early warning is triggered for the location s; the goal of minimizing the residual is min_sΣ[k=1→m] ( t_k - t_0 - | s - s_k | / v )^2; where F_i_ref is the baseline value of any i-th feature F_i in the feature vector F under normal conditions, w_i is the corresponding weight, s_k is the geometric coordinate of the k-th location node, t_0 is the time when the abnormal change occurs at the corresponding reference position, t_k is the linear regression of the k-th location node, m is the total number of location nodes, and v is the equivalent propagation speed or running speed estimated based on the cross-correlation time difference between adjacent location nodes.

[0011] Meanwhile, this application also provides a predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers. The steps include: sending detection light to a multi-core optical cable, receiving reflected light from each silicon photonics triaxial accelerometer connected to the other end of the optical fiber, calculating the axial acceleration information of each silicon photonics triaxial accelerometer based on the reflected light to obtain the vibration state changes in the time domain of the three orthogonal directions at each position along the belt conveyor, so as to provide early warning of abnormal fault risks of the belt conveyor; wherein, each silicon photonics triaxial accelerometer is fixedly installed in the mechanical structure along the belt conveyor, and is connected to the silicon photonics demodulation and control unit through different fiber cores of the unified multi-core optical cable to perform the above steps.

[0012] Optionally, the predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers, as described above, when not connected to DAS data, relies solely on the feature vectors, health indices, and cross-correlation results between nodes for anomaly identification and location; when connected to DAS data, it also follows these steps to provide early warning of abnormal fault risks of the belt conveyor in conjunction with DAS data: while collecting the axial acceleration information of each silicon photonics triaxial accelerometer (2) at its location, it simultaneously collects the triaxial vibration data and attitude data of each silicon photonics triaxial accelerometer node along the belt conveyor. The system collects continuous spatial vibration field data from the distributed optical fiber vibration device (DAS); performs continuous spatial anomaly screening on the acquired DAS data to identify candidate anomaly sections; establishes a positional mapping relationship between the installation position of each silicon photonic triaxial accelerometer node and the DAS coordinates; extracts triaxial vibration features, attitude features, and characteristic frequency band information from nodes within or near the candidate anomaly sections; and performs time synchronization, spatial registration, and confidence fusion of the triaxial vibration features, attitude features, and characteristic frequency band information corresponding to the nodes in the section with the DAS features to output the fault type, fault location, and fault level.

[0013] Optionally, the predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers, as described above, wherein when providing early warning of abnormal fault risks of the belt conveyor based on the changes in vibration state in the three orthogonal directions at various positions along the belt conveyor in the time domain, the multi-source fusion artificial intelligence model specifically used for the belt conveyor scenario includes: a node feature encoding module, used to encode the feature vector F_i(t), attitude angle θ_i(t), φ_i(t), and health index HI of each node; a DAS feature encoding module, used to encode the time-frequency energy distribution, abnormal segment length, spatial continuity, and propagation direction characteristics in the continuous spatial vibration field; and a fusion decision module, used to perform time synchronization, spatial registration, noise suppression, and weighted fusion on the above encoding results, and output the fault type, fault location, fault level, trend warning, and remaining life estimation results for idler bearing faults, jamming, uneven wear, belt misalignment, or idler frame deformation.

[0014] To address the aforementioned technical problems, this application also provides a fiber optic silicon photonics triaxial accelerometer network, comprising: a multi-core optical cable with interfaces reserved at preset intervals for one or more fiber cores to emerge; silicon photonics triaxial accelerometers, each communicatively connected to the end of the fiber core at each interface in the multi-core optical cable; the fiber optic silicon photonics triaxial accelerometer network system receives detection light rays sent by a silicon photonics demodulation and control unit, and feeds back the reflected light rays from each silicon photonics triaxial accelerometer to the silicon photonics demodulation and control unit, so that the silicon photonics demodulation and control unit can calculate the axial acceleration information of each silicon photonics triaxial accelerometer based on the reflected light rays, obtain the vibration state changes in the three orthogonal directions in the time domain at each position along the conveyor belt, and thereby provide early warning of abnormal fault risks of the conveyor belt.

[0015] Furthermore, based on the above-mentioned scheme, this application also provides a computer storage unit, which stores a program that can be executed by the silicon photonics demodulation and control unit. When the program is executed, it triggers the silicon photonics demodulation and control unit to send detection light to the multi-core optical cable connected to it, and causes the silicon photonics demodulation and control unit to calculate the acceleration information of each silicon photonics triaxial accelerometer at the location of each silicon photonics triaxial accelerometer based on the reflected light of each silicon photonics triaxial accelerometer connected to the multi-core optical cable, thereby obtaining the vibration state changes in the time domain of the three orthogonal directions at each position along the belt conveyor, so as to provide early warning of abnormal failure risks of the belt conveyor. Beneficial effects

[0016] This application discloses a predictive maintenance system, method, sensor network, and storage unit for belt conveyors. It includes a multi-core optical cable arranged along the belt conveyor and several silicon photonics triaxial accelerometers fixedly installed along the side frame or idler frame of the belt conveyor and connected to different optical fiber branches in the optical cable. This application feeds detection light to the optical fiber and the silicon photonics triaxial accelerometers connected to its ends through a silicon photonics demodulation and control unit located in a centralized control center or nearby electrical control room. Based on the reflection of the detection light, the acceleration at each sensor location is calculated to obtain the vibration state changes in the time domain in three orthogonal directions at each location along the belt conveyor, enabling corresponding feature extraction, fault identification, and predictive maintenance strategy generation. This application can proactively and in real-time assess the risk of abnormal belt conveyor faults and provide early warnings; its sensing nodes are completely passive and intrinsically safe; it has a wide frequency response and high sensitivity; it adopts a multi-point triaxial synchronous acquisition method to obtain equipment attitude information; it is suitable for long-distance, large-scale deployment with controllable costs; and it can also support predictive maintenance based on big data and artificial intelligence.

[0017] The system provided in this application deploys multi-core optical fibers along the conveyor belt, connecting multiple passive silicon photonics triaxial accelerometer nodes via branch fibers. A silicon photonics demodulation and control unit is set up in the central control room to perform high-frequency response demodulation on the triaxial acceleration signals of each node. It can utilize the low-frequency components of acceleration for gravity decomposition, calculate the pitch and roll angles of the idler frame where the node is located, and achieve online monitoring of belt misalignment and structural deformation. It constructs a health index using the spectral characteristics, time-domain statistics, and cross-correlation characteristics of adjacent nodes of the high-frequency components of acceleration, enabling the identification and location of abnormalities such as idler failures and foreign object jamming. Furthermore, this application can further utilize the idle cores in the multi-core optical fibers to connect additional distributed fiber optic vibration (DAS) devices, achieving synergy between point-based high sensitivity and continuous distributed monitoring. This system adopts an all-fiber, passive sensing structure, possessing advantages such as intrinsic safety, resistance to electromagnetic interference, suitability for long-distance deployment, and high level of intelligence. It can be widely applied to predictive maintenance and safe operation management of conveyor belts in mining, ports, power plants, and chemical industries. Attached Figure Description

[0018] The embodiments of this application will be described in further detail below with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer, as described in this application. Figure 2 This is a schematic diagram of the internal working unit of the silicon photonics demodulation and control unit in the system of this application. Figure 3 This is a schematic diagram illustrating the internal structure and working principle of the silicon photonics triaxial accelerometer used in the system of this application. Figure 4 This application provides a dustproof, moisture-proof, and corrosion-resistant encapsulation structure for multi-core optical fiber and silicon photonics triaxial accelerometer nodes in some implementations of this application.

[0019] In the diagram, 1 represents a multi-core optical cable; 2 represents a silicon photonics triaxial accelerometer; 3 represents a silicon photonics demodulation and control unit; 4 represents a belt conveyor; and 5 represents a drive roller. Detailed Implementation

[0020] To make the objectives and technical solutions of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0021] Those skilled in the art will understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0022] The term "connection" as used in this application can mean a direct connection between components or an indirect connection between components through other components.

[0023] The meaning of "and / or" as used in this application includes situations where each exists alone or both exist simultaneously.

[0024] This application addresses the problems of poor inherent safety, difficult deployment, insufficient sensitivity, weak spatial positioning capability, and low level of intelligence in existing belt conveyor condition monitoring technologies. It provides a predictive maintenance system, method, sensor network, and storage unit for belt conveyors. This system enables the deployment of a large number of passive silicon photonics triaxial acceleration sensor nodes along the entire belt conveyor line. It achieves integrated transmission and sensing through low-cost multi-core optical fiber, constructs rich features using high frequency response, triaxial information, and attitude information, and can work in conjunction with distributed fiber optic vibration (DAS) to support artificial intelligence models for predictive maintenance of various faults such as roller abnormalities, belt misalignment, and foreign object jamming.

[0025] Specifically, the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers provided in this application includes: a fiber optic silicon photonics triaxial accelerometer network, and a silicon photonics demodulation and control unit 3 connected and interacting with it.

[0026] The fiber-optic silicon photonic triaxial accelerometer network includes: One or more multi-core optical cables 1, with interfaces reserved at each preset interval for one or more optical fiber cores to be led out; Several silicon photonic triaxial accelerometers 2 are respectively connected to the inner core of the optical fiber at each interface of the multi-core optical cable 1.

[0027] Therefore, the silicon photonics demodulation and control unit 3, which is located at the other end of the multi-core optical cable 1, can send detection light to the multi-core optical cable 1 connected to it according to the program instructions stored in its storage unit, and receive the reflected light fed back by each silicon photonics triaxial accelerometer 2 connected to the other end of the multi-core optical cable 1. Based on the reflected light, it calculates the acceleration information of each silicon photonics triaxial accelerometer 2 at its location, and obtains the vibration state changes in the time domain of the three orthogonal directions at each position along the belt conveyor, so as to provide early warning of abnormal fault risks of the belt conveyor.

[0028] In the above system, each silicon photonics triaxial accelerometer 2 is arranged at intervals on the idler frame or the side of the side frame of the belt conveyor. The installation position on the side frame allows each silicon photonics triaxial accelerometer 2 to be fixed to the side of the belt conveyor, effectively avoiding interference. The idler frame supports the rollers below the belt conveyor used to drive the belt. Although arranging the silicon photonics triaxial accelerometer 2 on it is more susceptible to interference, it is closer to the vibration area of ​​the belt conveyor. Therefore, the vibration data obtained at this position can more directly reflect the real-time operating status of the belt conveyor.

[0029] Specifically, in order to provide feedback on the optical signals of the triaxial acceleration generated by the operation of each silicon photonics triaxial accelerometer 2 at its location, it is necessary to follow the instructions... Figure 3 The arrangement shown is implemented in the silicon substrate of each silicon photonic triaxial accelerometer 2: The coupling component connects to optical fibers externally and internally distributes the detection light provided by the single-mode or multi-mode optical fibers to three mutually orthogonal detection optical paths. In specific applications, the multi-core optical cable 1 provided in this application can flexibly select single-mode or multi-mode optical fibers for its inner core according to the setup requirements. Generally, for cost-sensitive scenarios or scenarios requiring long-distance detection deployment, single-mode optical fibers are usually preferred to transmit the detection optical signal provided by the control end to silicon photonic triaxial accelerometers 2 preset at different positions on the conveyor belt. In each silicon photonic triaxial accelerometer 2, three detection optical paths corresponding to the three axes are respectively set on its silicon substrate: A silicon-based micromirror receives one of the detection beams along a triaxial axis. The detection beam corresponding to that axis is incident onto a side surface of a reflective mass block suspended and fixed in the silicon substrate via an elastic connection device, corresponding to the detection beam path of that axis. The micromirror receives the reflected beam from the reflective mass block and feeds the reflected beam corresponding to the detection beam along that axis back into an optical fiber via a coupling component. In this structure, the reflective mass block is elastically supported inside the sensor by a spring or elastic arm, and can be displaced in response to the vibration of the sensor installation position. The mutually orthogonal surfaces of the mass block, which are used to provide the reflection of the above-mentioned three-axis optical path, are respectively provided with a monotonically changing reflectivity gradient distribution along the incident direction of the detection light on their respective sides.

[0030] Therefore, in the aforementioned sensing structure, the silicon-based micromirrors corresponding to the detection optical paths in each direction can open at different time period windows by oscillating at a preset frequency. This allows the detection light rays corresponding to each axis to be sequentially incident onto the corresponding detection surface of the reflective mass block. The reflectivity distribution of the reflective mass block on this detection surface reflects the change in the position of the reflected light spot through the gradient distribution of the reflected light. Thus, the different reflectivities distributed on the surface of the reflective mass block superimpose corresponding displacement-power components into the reflected light rays along each axis. The change in the absorption rate of the incident light at the position of the reflected light spot correspondingly reduces the intensity of the reflected light. This, combined with the window period of the detection fiber, forms a detection signal of acceleration in each reflection direction. Therefore, the system can directly determine which direction of the three axes the current signal corresponds to based on the time period of the received reflected signal, and determine the reflectivity of the reflective mass block surface where the reflected light spot is located based on the power intensity of the reflected signal. Based on the gradient distribution of the reflectivity of the side surface, the system can calculate the reflection position corresponding to the reflected light spot, and calculate the magnitude of the acceleration in the detection direction corresponding to the axis based on the change of the reflection position over time.

[0031] The above calculation process for optical signals and acceleration in different directions can be achieved through... Figure 1 right side or Figure 2 The silicon photonics demodulation and control unit 3 shown on the left is implemented. This demodulation unit can connect to several triaxial fiber optic accelerometers based on silicon photonics microstructures via a multi-core communication optical cable, as shown on the right side of the figure. By sending corresponding detection light to each sensor individually or simultaneously, and by installing each sensor at different positions on the belt conveyor, real-time detection of acceleration changes at different sensor installation positions can be achieved.

[0032] Therefore, the fiber-optic silicon photonics triaxial accelerometer network of this application can simultaneously achieve real-time sensing and detection over long distances and at multiple points using a single optical cable, offering low cost and ease of engineering. Taking a silicon photonics accelerometer sensor where each core of the fiber is independently connected to a different location on a conveyor belt using a multi-core communication optical cable as an example, its silicon photonics demodulation and control unit 3 can flexibly choose to communicate with a server or third-party platform via wired or wireless means, enabling real-time calculation and analysis of acceleration at different detection locations within the computer room. The number of cores in engineering optical cables typically ranges from 4 to 288, with prices generally ranging from 1 to 10 yuan per meter. Taking 288 cores as an example, the average cost per meter of each fiber is approximately 0.03 yuan, which is a significant cost advantage compared to traditional electrical signal-based sensing and detection methods. Furthermore, the advantages of this application increase with longer detection distances and the increased number of sensors required.

[0033] To achieve feeding into the aforementioned different optical axes and demodulating the reflected light to process the acceleration information carried therein, this application typically refers to... Figure 3 The structure shown is configured with a connection structure for the silicon photonics module. Simultaneously with emitting detection light, the module processes the detection signal from the sensor, received via single-mode or multimode fiber, to extract the acceleration information. The connection structure of this silicon photonics module includes: Fiber optic connectors, of which SC fiber optic connectors may be selected, are used to connect the fiber optic cores stripped from the interfaces set at different positions along the belt conveyor of the multi-core optical cable 1. The circulator has a unidirectional optical path switch with a ring connection. It is generally connected to the laser in the silicon photonic demodulation and control unit 3 at the first port, to the optical fiber connector at the second port to detect the optical signal of the silicon photonic triaxial accelerometer 2, and to the photodetector at the third port to receive the reflected light signal carrying the position of the reflected light spot fed back by the silicon photonic triaxial accelerometer 2. Thus, the laser feeds detection light into the optical fiber through a circulator and an optical fiber connector; The reflected light detected by the silicon photonics triaxial accelerometer 2 is received through an optical fiber connector, transferred by a circulator to a photodetector to extract the corresponding detection signal, and supplied to the conditioning control unit. Thus, by identifying and calculating the reflection power of the reflected signal, acceleration data matching the change in the position of the reflected light spot is obtained.

[0034] In the aforementioned silicon photonics module: the laser and modulation control interact to determine the emission wavelength of the detection light received from the sensor and match it with the reflected wavelength received by the device. The laser feeds the detection light of the corresponding wavelength unidirectionally through a circulator into the optical connector, feeding the light into the corresponding sensor, and then through the sensor's internal structure to the corresponding position on the surface of the reflective mass block. Because the cubic mass block is elastically connected to the silicon substrate of the sensor only through spring structures such as elastic cantilever or support beams, it is held in a suspended posture within the sensor. Therefore, it can be driven by gravity in real time, displacing relative to the surrounding silicon substrate according to the rotation of the sensor. The incident light in the sensor will be incident at various positions with different reflectivity due to the rotation of the mass block. Due to the gradient setting of the reflectivity of the mass block surface, it will absorb a portion of the incident light and reflect a corresponding proportion of optical power. Thus, the intensity of the reflected detection light can reflect the position of the reflection point on the mass block surface according to the magnitude of the optical power, thereby detecting acceleration and rotation state by the change in reflection position per unit period. The reflected light from the superimposed reflected light power enters the silicon photonics module through the fiber optic connector, and reaches the photodetector through unidirectional transmission via the circulator. By extracting the reflected light that matches the wavelength of the current incident laser, and by demodulating and calculating the changes in its optical power intensity, real-time acceleration data can be obtained.

[0035] Therefore, the predictive maintenance system for belt conveyors based on fiber optic silicon photonics triaxial accelerometers of this application can be connected to a sensor network via a multi-core fiber optic cable 1. Multiple silicon photonics triaxial accelerometers in the network, in conjunction with a silicon photonics demodulation and control unit, can monitor the status of nodes at different locations. Finally, through data processing and diagnostic devices, based on the spectral peak value, band energy, and bandwidth of the acceleration spectrum A(f_k) acquired by the sensors within the idler rotation frequency, the corresponding harmonic frequency of the idler rotation, the belt fundamental frequency, and its harmonic bands, feature vectors F = [a_rms, kurtosis, peak factor, band energy 1, band energy 2, … ] are constructed for each node at the location of the silicon photonics triaxial accelerometer 2. By comparing this feature vector with the feature vector F of each node at the location of the silicon photonics triaxial accelerometer 2 under normal operating conditions, a health index HI = Σ[i] (w_i · (F_i - F_i_ref ) / F_i_ref); Finally, based on the abnormal changes of the feature vector F of each location node and its corresponding health index HI, combined with the discrete cross-correlation function R_ij(τ) between different location nodes, the least squares method is used to estimate the location s of the abnormal fault risk along the belt conveyor device with the objective of minimizing the residual min_s Σ[k=1→m] ( t_k - t_0- | s - s_k | / v )^2, and trigger an early warning for that location s. Wherein, F_i_ref is the baseline value of any i-th feature F_i in the feature vector F under normal conditions, w_i is the corresponding weight, s_k is the geometric coordinate of the k-th location node, t_0 is the time when the abnormal change occurs at the corresponding reference position, t_k is the linear regression of the k-th location node, m is the total number of location nodes, and v is the equivalent propagation speed or running speed estimated based on the cross-correlation time difference between adjacent location nodes.

[0036] The objective function used in this embodiment is not limited to the above-described form. It only provides a function form that minimizes the overall error between the actual arrival time of the anomaly detected by each node and the theoretical arrival time derived from the candidate fault location, so as to facilitate the estimation of the location of the anomaly in the belt direction.

[0037] Since the silicon photonic triaxial accelerometer node used in this application is a passive device, it directly guides incident light into the silicon photonic accelerometer. Acceleration information is acquired through the coordinated control of a synchronous micromirror and a mass block reflecting the gradient, and then transmitted back to the demodulator. The multi-core optical fiber only transmits optical signals and does not require on-site power supply, thus enabling intrinsically safe monitoring of belt conveyors. Therefore, this application effectively avoids the shortcomings of existing detection technologies, such as unstable power supply and high maintenance costs in wireless transmission methods; and the risks of electromagnetic interference, lightning strikes, and electrical sparks in wired transmission methods under live conditions.

[0038] This application integrates the following into a multi-channel silicon fiber silicon photonics demodulator / demodulator / control unit 3: Figure 2 The multi-channel optical sensing system shown has the characteristics of mass production and low cost, and can be widely used in industrial scenarios of long-distance belt conveyors.

[0039] Specifically, in this application, the multi-core optical cable 1 can use 4 to 288 single-mode optical fibers. The multi-core optical fibers can be laid on the side frame or roller frame of the belt conveyor. During the factory pre-processing stage, at least one core is extracted every 5 to 20 m as a branch optical fiber and connected to the corresponding silicon photonics triaxial acceleration sensing node.

[0040] In this system, the triaxial acceleration sensing node connected to each lead-out end of the optical fiber uses a triaxial acceleration measurement sensing unit that can cover a frequency response range from 0 Hz to over 4 kHz by modulating the reflected light of the detection light. This allows the node to measure the absolute attitude of its position using the gravity component and to capture high-frequency vibration and impact signals during the operation of the conveyor belt.

[0041] The data processing and diagnostic device set in the silicon photonics demodulation and control unit 3 terminal can gradually realize the attitude monitoring and deviation diagnosis of the belt conveyor idler frame or side frame according to the low frequency components ax_g, ay_g, and az_g of the triaxial acceleration output by any silicon photonics triaxial acceleration sensor node in the following manner: 1. Based on the optical signals fed back by each sensing unit, the triaxial acceleration and attitude of the sensing unit at its location are measured and calculated.

[0042] Based on the changes in light intensity over time corresponding to each axis in the feedback signals from each sensor, the triaxial acceleration vector output by each silicon photonics triaxial acceleration sensing node in the local coordinate system is calculated: a(t) = [ ax(t), ay(t), az(t) ]ᵀ When the belt conveyor is stationary or running smoothly at low speed, the gravitational acceleration g is the main low-frequency component of the acceleration signal, and the following approximate relationship exists: ax^2 + ay^2 + az^2 ≈ g^2 Where g is the nominal value of gravitational acceleration, approximately 9.81 m / s². By extracting the gravity components ax_g, ay_g, and az_g through low-pass filtering, the pitch angle θ and roll angle φ of the support or roller frame at the node installation location can be calculated as follows: Pitch angle θ = arctan( ax_g / sqrt( ay_g^2 + az_g^2 ) ) Roll angle φ = arctan( ay_g / sqrt( ax_g^2 + az_g^2 ) ) Then, by monitoring the spatial distribution and temporal variation of attitude data {θ_i, φ_i} consisting of pitch angle and roll angle at multiple sensor nodes along the belt conveyor, online monitoring of idler frame sinking, deformation, uneven foundation settlement, and belt conveyor deviation trend can be achieved.

[0043] 2. Extract the vibration characteristics of each node based on the attitude changes at each node's location, and calculate the corresponding health index.

[0044] The data processing and diagnostic device extracts the vibration signals corresponding to the three axes at different times based on the attitude changes of each silicon photonics triaxial acceleration sensing node over time. It performs bandpass filtering and spectral analysis on the vibration signals to calculate the root mean square acceleration, band energy, peak factor, kurtosis, and other characteristics corresponding to the vibration. Based on this, it constructs the health index HI corresponding to the node according to the following formula, and evaluates the health status of the corresponding conveyor roller and structure in the following manner: Bandpass filtering (e.g., 1 Hz to 4 kHz) is applied to the triaxial acceleration signal of each node to obtain the vibration acceleration components ax_v(t), ay_v(t), and az_v(t). Within a time window T, the total root mean square acceleration a_rms of that node is calculated as follows: a_rms = sqrt( (1 / T) · ∫[0→T] ( ax_v(t)^2 + ay_v(t)^2 + az_v(t)^2) dt ) Discrete sampling and Fast Fourier Transform (FFT) of the total root mean square acceleration yields the acceleration spectrum A(f_k) at the node where the sensor is located. Its general form can be expressed as: A(f_k) = Σ[n=0→N-1] ( a[n] · exp( -j · 2π · k · n / N ) ) Within key frequency bands such as idler roller rotation frequency and its harmonics, belt fundamental frequency and its harmonics, features such as spectral peak value, frequency band energy, and bandwidth are extracted. Combined with time-domain statistics such as peak value, root mean square, peak factor, and kurtosis, a feature vector F is constructed for each node. F = [ a_rms, kurtosis, peak factor, band energy 1, band energy 2, … ] Therefore, this system can construct a health index HI corresponding to each sensor node to characterize the health level of the conveyor belt structure near that node: HI = Σ[i] ( w_i · ( F_i - F_i_ref ) / F_i_ref ) Where F_i_ref is the baseline value of the i-th feature element in the feature vector under normal conditions, and w_i is the weight corresponding to the feature element.

[0045] 3. Based on the correlation of data between different sensor nodes, calculate and locate the fault location.

[0046] The principle of the cross-correlation algorithm for positioning and velocity estimation is as follows.

[0047] Along the belt conveyor direction, two adjacent silicon photonics triaxial accelerometer nodes are denoted as node i and node j, respectively, with a distance L_ij between them. The discrete signals on a certain vibration channel are x_i[n] and x_j[n]. The discrete cross-correlation function R_ij(τ) of the two sequences is defined as: R_ij(τ) = Σ[n] ( x_i[n] · x_j[n + τ] ) When there is a periodic impact source on the belt (such as a defect in a roller or a belt joint), R_ij(τ) will reach its maximum value at a certain delay τ_hat, which represents the propagation time difference of the same excitation at two measurement points.

[0048] Therefore, the data processing and diagnostic device can effectively estimate the belt running speed v ≈ L_ij / τ_hat based on the cross-correlation function of vibration signals between two or more adjacent silicon photonics triaxial accelerometer nodes, and combine the geometric position of each node to achieve spatial localization of abnormal events such as idler roller failure and foreign object jamming. When m nodes along the belt conveyor detect a transient abnormal event with arrival times of t_1, t_2, …, t_m at those nodes, the position s of the event in the belt direction can be estimated by minimizing the residual of the following formula using the least squares method. Based on this, an early warning of abnormal failure risk of the belt conveyor can be issued: min_s Σ[k=1→m] ( t_k - t_0 - | s - s_k | / v )^2 Where s_k is the geometric position of the k-th node, and t_0 is the time when the event occurs at the reference position.

[0049] Without access to DAS data, the method relies solely on the feature vectors, health indices, and cross-correlation results of each location node for anomaly identification and localization. However, considering the discontinuity of the detection information from the silicon photonics triaxial accelerometer, in a more preferred implementation, the belt conveyor predictive maintenance system based on a fiber optic silicon photonics triaxial accelerometer provided in this application can be further integrated with multi-core optical fibers and distributed fiber vibration (DAS) to improve the detection accuracy and precision of belt conveyor fault conditions through coordinated monitoring.

[0050] In this implementation, multi-core single-mode optical fiber can be used as the integrated transmission and sensing carrier. During the factory pre-processing stage, one or more fiber branches are reserved at fixed intervals to connect silicon photonics triaxial accelerometer nodes. The remaining idle fiber cores can serve as redundant channels to connect distributed optical vibration (DAS) devices. The DAS provides continuously spatially distributed vibration field information, which, combined with the high-sensitivity triaxial, attitude, and characteristic frequency band information provided by the silicon photonics sensing nodes, complements each other from their respective dimensions. This allows for continuous spatial anomaly screening of the acquired DAS data, identifying candidate anomaly sections. A positional mapping relationship is established between the installation location of each silicon photonics triaxial accelerometer node and the DAS coordinates. Triaxial vibration characteristics, attitude characteristics, and characteristic frequency band information are extracted from nodes within or near the candidate anomaly sections. Through comprehensive processing and diagnosis of both types of data, and by integrating the silicon photonics triaxial accelerometer node data with the DAS data... Joint data analysis, by synchronizing, spatially registering and fusing the triaxial vibration characteristics, attitude characteristics and characteristic frequency band information corresponding to the segment nodes with DAS characteristics in time, can more effectively improve the fault location accuracy and diagnostic reliability.

[0051] The following section details the specific principles and implementation process of join analysis.

[0052] The silicon photonic triaxial accelerometer based on the foregoing embodiments of this application reflects changes in the position and attitude of the silicon photonic structure in the reflected signal of the detection light, thereby achieving real-time detection of the attitude and vibration of each node along the conveyor belt. The distributed fiber optic vibration measurement (DAS) scheme superimposed in this embodiment is a fiber optic sensing system that uses optical fiber as the sensor; that is, the idle optical core itself serves as the sensor.

[0053] If the DAS is used independently for detection, the following problems often arise: Due to the rigidity of the silica material of the optical fiber itself, commonly known as glass fiber, when it receives external vibrations, the signal inside the optical cable will propagate to the left and right sides along the axis, affecting a wider area of ​​the optical fiber, generating crosstalk, and easily causing false alarms. The point-type sensor based on the aforementioned embodiment only uses optical fiber as the transmission unit, not as the sensing element, thus offering greater accuracy and consistency. However, considering the limited number of detection points that can be connected in the optical fiber, the point-distribution method in the aforementioned embodiment may not be able to achieve large-scale coverage in large conveyor belt systems, and may also be too costly. Therefore, this embodiment combines the two, using a fiber optic silicon photonic triaxial accelerometer for focused monitoring and a DAS deployed within the same optical fiber system for comprehensive measurement, thus organically combining the two to achieve better detection results.

[0054] In this embodiment, the silicon photonics demodulation and control unit 3 can collect fiber optic triaxial acceleration data in the manner of the previous embodiment, and simultaneously collect triaxial vibration data and attitude data of each silicon photonics triaxial acceleration sensing node along the belt conveyor, as well as DAS data such as the continuous spatial vibration field output by the distributed fiber optic vibration device. Then, the DAS data is matched with the positional relationship of the synchronous fiber optic triaxial acceleration, and the data of the two types of sensors are labeled to distinguish between the point data provided by the fiber optic silicon photonics triaxial acceleration sensor and the cable-type data provided by the DAS system. After normalizing the two types of data, continuous spatial anomaly screening is first performed along the conveyor belt based on DAS data to identify candidate anomaly sections. Then, silicon photonics triaxial accelerometer nodes within or near these candidate anomaly sections are used to extract triaxial vibration characteristics, attitude characteristics, and characteristic frequency band information. These are then uniformly fed into a model library for training. Artificial intelligence models, such as deep neural networks, random forests, or graph neural networks, are used to synchronize the two types of data in time, register them spatially, and jointly determine the data. The model extracts the correspondence between different data and various fault conditions and corresponding fault locations. This simultaneous interaction of data from both sensors improves fault location accuracy and diagnostic reliability, and outputs corresponding detection results and early warning information. Specifically, this artificial intelligence model can be pre-trained based on historical operating data, simulation data, and on-site calibration data to predict the development trend and remaining life of conveyor belt idler roller faults, thereby providing early warnings and predictive maintenance before specific faults occur. However, even without a DAS device, or when DAS data cannot be effectively obtained due to special reasons, this system can still complete fault identification and location based on the feature vectors of each node, health index, and cross-correlation results between nodes.

[0055] In this embodiment, when predicting the risk of abnormal failures of the belt conveyor based on the changes in vibration state in the three orthogonal directions at various locations along the conveyor line, the multi-source fusion artificial intelligence model specifically used for the belt conveyor scenario can drive predictive maintenance in the following ways: The feature preprocessing layer is used to denoise, bandpass filter, normalize, and divide the triaxial vibration signals and DAS signals acquired on site into time windows; The node feature encoding module is used to extract node features and continuous spatial features by combining prior information such as roller rotation frequency, frequency multiplication, belt fundamental frequency and its harmonics, and to encode the feature vector F_i(t), attitude angle θ_i(t), φ_i(t) and health index HI of each node; The DAS feature encoding module is used to encode the time-frequency energy distribution, anomalous segment length, spatial continuity, and propagation direction features in a continuous spatial vibration field, and to extract the anomalous segment length, time-frequency energy distribution, spatial continuity, and propagation direction features in a continuous spatial vibration field. The fusion decision module is used to perform time synchronization, spatial registration, noise suppression, and weight fusion on the above-mentioned encoding results based on the geometric position, adjacency relationship, and mapping relationship of each node with DAS coordinates. Through processing by models such as deep neural networks, random forests, or graph neural networks, it outputs the fault type, fault location, and fault level of idler bearing failure, jamming, uneven wear, belt misalignment, or idler frame deformation. It provides trend warnings for belt misalignment and idler frame deformation, and predicts the fault development trend and remaining life by combining time series. The warning or alarm information is sent to the belt conveyor control system or upper monitoring system, and the remaining life estimation results are output. This enables predictive maintenance and maintenance window recommendations for the belt conveyor, thereby reducing the false alarm rate under high dust, high humidity, and strong background noise conditions.

[0056] The model used in this embodiment is preferably trained using a combination of pre-labeled operational samples collected on-site, historical operational data, simulation samples, and common fault samples. It can also establish sub-models or adjust model weights for different operating conditions such as mines, docks, power plants, or chemical plants. Its feature processing layer can reduce the false alarm rate under high-noise conditions by suppressing low-frequency background disturbances and highlighting the relevant characteristics of roller rotation frequency, harmonics, and belt fundamental frequency through a pre-processing frequency band of 1Hz to 4kHz.

[0057] In other implementations, to meet the sealing and durability requirements of long-term operation in mining, dock, and chemical environments, the system of this application can also add, for example, multi-core fiber and silicon photonics triaxial accelerometer nodes. Figure 4The dustproof, moisture-proof, and corrosion-resistant encapsulation structure shown is designed to ensure stable operation of the equipment. Specifically, it includes a locking nut, a middle sleeve, an adapter, a silicone ring, and a rubber ring, located between the multi-core optical cable branch outlet and the silicon photonics triaxial accelerometer, or at the connection point between the sensor's pigtail and the branch optical fiber, for connection, sealing, and stress relief. This structure connects the front end of the locking nut to the outer shell of the sensor node and the rear end to the outer sheath, sealing sleeve, or pigtail protective sleeve of the multi-core optical cable, thus clamping and sealing the optical cable. The silicone ring, middle sleeve, adapter, and rubber ring are correspondingly placed between the front and rear locking nuts to achieve enclosure and protection of the optical structure. The silicone ring is placed between the middle sleeve and the locking nut, and the rubber ring is connected to the adapter to seal the gap between the adapter and the middle sleeve. This achieves sealing, moisture-proofing, corrosion protection, and stress relief while the pigtail is led out, rather than completely encapsulating the sensor device inside the rotating fiber optic connector. This encapsulation structure is positioned between the branch interface of the multi-core optical cable 1 and the silicon photonics triaxial accelerometer 2. It connects and seals the branch optical fiber from the multi-core optical cable 1 to the pigtail or input end of the silicon photonics triaxial accelerometer 2, utilizing silicone rings, rubber rings, and other structures to provide a seal while simultaneously releasing stress. The rear side of the locking nut connects to the entry end of the branch optical fiber and clamps the optical cable sheath, while the front side connects to the housing or sensor mounting base. Seals are used to seal and fix the pigtail, connector, or sensor lead-out end, achieving dustproof, moisture-proof, corrosion-proof, and mechanical shock resistance. The sealing structure is corrosion-resistant, and the seal meets IP68 standards. The fiber optic sensor string is prefabricated in the factory and integrally formed with the optical cable, requiring only fixing and termination on-site, significantly simplifying installation. The silicon photonics sensing unit has an insulated surface, preventing charge accumulation and effectively avoiding electrostatic damage. The entire system uses only optical fiber, is non-conductive (no lightning strikes), and is suitable for complex long-distance conveyor belt detection scenarios.

[0058] In summary, this application achieves the following through a multi-core fiber optic sensor network, a silicon photonics triaxial accelerometer node, a silicon photonics demodulation and control unit, a host computer and an artificial intelligence diagnostic server, and an optional distributed fiber optic vibration (DAS) device: (1) Safe and passive sensing detection; (2) Multi-point triaxial acceleration and attitude integrated monitoring; (3) By using rich bandwidth and feature information, multi-dimensional features can be obtained to improve the accuracy of early warning; (4) Flexible deployment can be achieved at a lower cost through multi-core optical fibers; (5) Supports DAS collaborative monitoring; (6) Facilitates predictive maintenance and centralized management of artificial intelligence.

[0059] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer, characterized in that, include: A multi-core optical cable (1) is arranged along the belt conveyor (4); The silicon photonics triaxial accelerometer (2) is fixedly installed in the mechanical structure along the belt conveyor. Each silicon photonics triaxial accelerometer (2) is communicatively connected to the end of at least one fiber core in the multi-core optical cable (1). The silicon photonic demodulation and control unit (3) is located at the other end of the multi-core optical cable (1) and is used to send detection light to the multi-core optical cable (1), receive the reflected light fed back by each silicon photonic triaxial accelerometer (2) connected to the other end of the optical fiber, calculate the acceleration information of each silicon photonic triaxial accelerometer (2) at its location based on the reflected light, obtain the vibration state changes in the three orthogonal directions in the time domain at each location along the belt conveyor, and use this to provide early warning of abnormal fault risks of the belt conveyor.

2. The predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer as described in claim 1, characterized in that, The silicon photonics triaxial accelerometer (2) is arranged at intervals on the idler frame or the side of the side frame of the belt conveyor.

3. The predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer as described in claim 2, characterized in that, The multi-core optical cable (1) has an interface reserved at each preset interval for one or more optical fiber cores to be led out. The optical fiber cores led out from the interface are respectively connected to the corresponding silicon photonic triaxial acceleration sensor (2) and / or distributed optical fiber vibration device to detect the triaxial acceleration vector and / or vibration field information at different positions along the belt conveyor.

4. The predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer as described in claim 3, characterized in that, The vibration state changes in the time domain in the three orthogonal directions at various locations along the belt conveyor include any one or any combination of the following: The pitch angle θ of the position of the silicon photonics triaxial accelerometer (2) is θ = arctan(ax_g / sqrt(ay_g^2 + az_g^2)), and its variation with time and space. The roll angle φ = arctan( ay_g / sqrt( ax_g^2+ az_g^2 )) at the location of the silicon photonics triaxial accelerometer (2), and its variation with time and space; The components of the vibration acceleration in all directions at the location of the silicon photonics triaxial accelerometer (2), namely ax_v(t), ay_v(t), and az_v(t), and their changes with time and space; The total root mean square acceleration a_rms = sqrt( (1 / T)· ∫[0→T] ( ax_v(t)^2 + ay_v(t)^2 + az_v(t)^2 ) dt ) at the location of the silicon photonics triaxial accelerometer (2), and its variation with time and space; Based on the total root mean square acceleration a_rms at each location, the acceleration spectrum A(f_k) is obtained through discrete sampling and fast Fourier transform (FFT). The spectral peak value, frequency band energy and bandwidth of the acceleration spectrum A(f_k) in the idler rotation frequency, the harmonic frequency corresponding to the idler rotation, the belt fundamental frequency and its harmonic frequency band, and the changes of their distribution over time and space; Where ax_g, ay_g, and az_g represent the gravitational components in the x, y, and z axes corresponding to the position of the silicon photonics triaxial accelerometer (2) obtained by low-pass filtering.

5. The predictive maintenance system for belt conveyors based on a fiber optic silicon photonics triaxial accelerometer as described in claim 4, characterized in that, The steps for early warning of abnormal fault risks of belt conveyors based on the vibration state changes in the three orthogonal directions at various locations along the conveyor line include: Based on the spectral peak value, frequency band energy, and bandwidth of the acceleration spectrum A(f_k) in the idler rotation frequency, the harmonic frequency corresponding to the idler rotation, the belt fundamental frequency, and its harmonic frequency band, the feature vector F = [ a_rms, kurtosis, peak factor, frequency band energy 1, frequency band energy 2, … ] of each silicon photonics triaxial acceleration sensing device (2) is constructed respectively. Based on the feature vector F of each silicon photonics triaxial accelerometer (2) at its location node under normal operating conditions, the health index corresponding to each node is constructed.

6. A predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers, characterized in that the steps include... include: Send detection light to the multi-core optical cable (1), receive the reflected light fed back by each silicon photonic triaxial accelerometer (2) connected to the other end of the optical fiber, calculate the acceleration information of each silicon photonic triaxial accelerometer (2) at its location based on the reflected light, obtain the vibration state changes in the time domain of the three orthogonal directions at each position along the belt conveyor, and use this to provide early warning of abnormal fault risks of the belt conveyor. Each silicon photonics triaxial accelerometer (2) is fixedly installed in the mechanical structure along the belt conveyor, and is connected to the silicon photonics demodulation and control unit (3) through different fiber cores of a unified multi-core optical cable (1) to perform the above steps.

7. The predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described in claim 6, characterized in that, When DAS data is not accessed, the method only identifies and locates anomalies based on the feature vectors, health indices, and cross-correlation results between each location node. When accessing DAS data, the method also follows these steps to provide early warning of abnormal fault risks of the belt conveyor in conjunction with the DAS data: While collecting the acceleration information of each silicon photonics triaxial accelerometer (2) at its location, the triaxial vibration data, attitude data and continuous spatial vibration field data of each silicon photonics triaxial accelerometer node along the belt conveyor are collected simultaneously. Continuous spatial anomaly screening is performed on the acquired DAS data to identify candidate anomaly segments; Based on the installation position of each silicon photonics triaxial accelerometer node and the DAS coordinates, a position mapping relationship is established, and triaxial vibration characteristics, attitude characteristics and characteristic frequency band information are extracted from the nodes in or near the candidate anomaly section. The triaxial vibration characteristics, attitude characteristics, and characteristic frequency band information corresponding to the segment nodes are synchronized with the DAS characteristics in time, spatially registered, and fused with confidence to output the fault type, fault location, and fault level.

8. The predictive maintenance method for belt conveyors based on fiber optic silicon photonics triaxial accelerometers as described in claim 6, characterized in that, When issuing early warnings about abnormal fault risks of the belt conveyor based on the changes in vibration state in the three orthogonal directions at various locations along the conveyor line in the time domain, a multi-source fusion artificial intelligence model specifically for the belt conveyor scenario is used. include: The node feature encoding module is used to encode the feature vector F_i(t), attitude angle θ_i(t), φ_i(t), and health index HI of each node; The DAS feature encoding module is used to encode the time-frequency energy distribution, anomalous segment length, spatial continuity, and propagation direction characteristics in a continuous spatial vibration field. The fusion decision module is used to perform time synchronization, spatial registration, noise suppression and weight fusion on the above encoding results, and output the fault type, fault location, fault level, trend warning and remaining life estimation results of idler bearing failure, jamming, uneven wear, belt misalignment or idler frame deformation.

9. A fiber-optic silicon photonic triaxial accelerometer network, characterized in that, include: A multi-core optical cable (1) has an interface reserved at each preset interval for one or more optical fiber cores to be led out. A silicon photonic triaxial accelerometer (2) is connected to the inner core of each interface of a multi-core optical cable (1) via communication. The fiber optic silicon photonics triaxial accelerometer network system receives the detection light sent by the silicon photonics demodulation and control unit (3), and feeds back the reflected light from each silicon photonics triaxial accelerometer (2) to the silicon photonics demodulation and control unit (3). The silicon photonics demodulation and control unit (3) calculates the acceleration information of each silicon photonics triaxial accelerometer (2) at its location based on the reflected light, and obtains the vibration state changes in the three orthogonal directions in the time domain at each position along the belt conveyor, so as to provide early warning of abnormal fault risks of the belt conveyor.

10. A computer storage unit, characterized in that, The computer storage unit stores a program that can be executed by the silicon photonics demodulation and control unit (3). When the program is executed, it triggers the silicon photonics demodulation and control unit (3) to send detection light to the multi-core optical cable (1) connected to it, and causes the silicon photonics demodulation and control unit (3) to calculate the acceleration information of each silicon photonics triaxial accelerometer (2) at its location based on the reflected light of each silicon photonics triaxial accelerometer (2) connected to the multi-core optical cable (1), thereby obtaining the vibration state changes in the three orthogonal directions in the time domain at each location along the belt conveyor, and thus providing early warning of abnormal fault risks of the belt conveyor.

Images