Industrial equipment intelligent operation and maintenance management system and method based on 5G-MOM

Abstract

The invention discloses an industrial equipment intelligent operation and maintenance management system and method based on 5G-MOM, and belongs to the technical field of industrial internet and intelligent manufacturing. The system comprises a multi-source heterogeneous data acquisition layer deployed in industrial equipment, an edge computing node cluster based on 5G, a cloud intelligent analysis platform and a man-machine collaborative operation and maintenance terminal. The method comprises the following steps of collecting equipment vibration, temperature and current multi-dimensional working condition data in real time through a 5G network; performing data cleaning and feature extraction by using edge computing nodes, and constructing an equipment operation digital twin model; a cloud deep neural network is adopted to carry out fusion analysis on the multi-dimensional time series data, and self-adaptive diagnosis and residual life prediction of a fault mode are realized; a dynamic maintenance strategy is generated based on an MOM system, and field personnel are guided to execute precise maintenance through an AR terminal. According to the invention, 5G ultra-low time delay communication and an industrial mechanism model are creatively combined, and real-time visual management and predictive maintenance decision optimization of the equipment health state are realized.

Description

Technical Field

[0001] The present invention relates to the field of industrial Internet and intelligent manufacturing technology, and in particular to an intelligent operation and maintenance management system and method for industrial equipment based on 5G-MOM. Background Art

[0002] As industrial equipment becomes increasingly complex, traditional operations and maintenance systems are plagued by issues such as limited data collection, insufficient real-time performance, and reliance on manual experience for fault diagnosis. Existing technologies typically employ single-sensor monitoring coupled with threshold alarm mechanisms (e.g., shutdown due to excessive vibration), lacking global awareness and predictive capabilities for equipment health. This leads to frequent unplanned downtime and high operations and maintenance costs. Furthermore, most systems utilize wired networks or 4G communications, which struggle to meet the high-concurrency, low-latency data transmission requirements of industrial scenarios, hindering the timeliness of operations and maintenance decisions.

[0003] Chinese invention patent publication number CN113420132A discloses an "Internet of Things-based remote monitoring system for industrial equipment." It collects equipment data by deploying temperature and humidity sensors and current detection modules, and uses cloud servers to detect anomalies and generate maintenance recommendations. Although this solution achieves basic remote data monitoring, it has significant drawbacks: (1) Data collection is limited to a single physical quantity (such as temperature or current) and does not integrate multi-dimensional signals such as vibration, resulting in incomplete fault feature extraction; (2) It adopts a centralized cloud processing mode and does not introduce edge computing nodes. Direct uploading of raw data causes high network bandwidth pressure and diagnostic response delays exceeding 200ms; (3) Maintenance strategy generation relies on a fixed rule base and cannot be combined with real-time dynamic optimization of working conditions, resulting in a low match between maintenance instructions and actual equipment status.

[0004] Based on the above existing technologies, the following technical problems urgently need to be solved in the field of intelligent operation and maintenance of industrial equipment:

[0005] Efficient fusion of multi-source heterogeneous data: How to achieve real-time synchronous collection and transmission of multi-dimensional data such as vibration, temperature, and current through 5G networks, and resolve heterogeneous compatibility issues among different industrial equipment communication protocols (such as Modbus and Profinet);

[0006] The collaborative optimization problem of low latency and high precision: How to build an edge-cloud collaborative computing architecture to achieve accurate identification of failure modes and remaining life prediction through deep learning models while reducing network transmission load;

[0007] The problem of adaptive generation of dynamic maintenance strategies: How to deeply integrate the production scheduling data of the MOM system with the health status of the equipment to generate maintenance strategies that can be dynamically adjusted according to the production plan and spare parts inventory, and achieve precise guidance of on-site operations through augmented reality (AR) technology.

[0008] In response to the above problems, the present invention proposes an intelligent operation and maintenance management system and method for industrial equipment based on 5G-MOM, aiming to break through the technical bottleneck of traditional operation and maintenance systems. Summary of the Invention

[0009] In view of the above existing problems, the present invention is proposed.

[0010] Therefore, the present invention provides an intelligent operation and maintenance management system and method for industrial equipment based on 5G-MOM to solve the technical problems of delayed response and high fault misjudgment rate of traditional operation and maintenance systems.

[0011] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0012] In a first aspect, the present invention provides an intelligent operation and maintenance management system for industrial equipment based on 5G-MOM, including:

[0013] The multi-source heterogeneous data acquisition layer is deployed on the industrial equipment itself and integrates vibration sensors, temperature sensors, current monitoring modules, and industrial protocol adapters. It is used to collect multi-dimensional operating status data of the equipment in real time and achieve millisecond-level data transmission through the 5G network.

[0014] A 5G-based edge computing node cluster, consisting of distributed edge servers and embedded data processing modules, is used to filter noise, remove outliers, and standardize raw data. It also generates device operation feature vectors through time-frequency domain feature extraction algorithms, and simultaneously builds a dynamically updated device digital twin model.

[0015] The cloud-based intelligent analysis platform deploys a fault diagnosis model based on a deep neural network and an LSTM life prediction model. It is used to integrate and analyze the multi-dimensional time series feature data uploaded by edge nodes, and output equipment health status scores, fault type probability distributions, and remaining service life prediction results.

[0016] The human-machine collaborative operation and maintenance terminal integrates the MOM system maintenance strategy engine and the AR augmented reality interactive interface. It is used to receive cloud analysis results and generate a dynamic maintenance plan that includes maintenance priorities, spare parts requirement lists and operation instructions. The AR terminal visually guides on-site personnel to perform maintenance operations.

[0017] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management system described in the present invention, the feature extraction algorithm of the edge computing node cluster satisfies the following relationship:

[0018]

[0019] Among them, E k Indicates the energy value of the kth frequency band, Wk,i is the coefficient of the i-th node in the k-th frequency band after the vibration signal is decomposed by wavelet packet, which is used to construct the frequency domain energy entropy feature vector.

[0020] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management system described in the present invention, the fault diagnosis model of the cloud-based intelligent analysis platform adopts a multi-scale convolutional neural network, and the convolutional layer output satisfies:

[0021]

[0022] in, is the convolution kernel weight, b (l) is the bias term, is the original or front-layer feature data input to the l-th layer of the convolutional neural network, P×Q is the convolution kernel size, σ is the ReLU activation function, and the fault features of different frequency ranges are captured through multi-scale convolution kernels.

[0023] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management system of the present invention, the hidden state update of the LSTM life prediction model satisfies:

[0024] h t =o t ⊙tanh(c t )

[0025] Among them, t is the output gating signal, c t The cell state is used to model the long-term temporal dependencies during device performance degradation.

[0026] As a preferred solution of the 5G-MOM-based intelligent operation and maintenance management system for industrial equipment described in the present invention, the human-machine collaborative operation and maintenance terminal integrates the equipment maintenance knowledge graph module, and its construction satisfies the following relationship:

[0027]

[0028] Among them, e i ,e j The smallest semantic unit for device maintenance knowledge, r ij It is a rule engine that connects semantic units and drives automated reasoning from fault diagnosis to maintenance decision-making. The entity set ε contains three types of nodes: equipment components, failure modes, and maintenance operations. The relationship set Define component-fault association and fault-maintenance mapping rules to automatically generate maintenance path reasoning chains based on cloud-based diagnosis results.

[0029] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management method of the present invention, the following steps are included:

[0030] Step S1: Vibration signals, temperature distribution, three-phase current waveforms, and equipment control parameters are synchronously collected through a sensor array deployed on industrial equipment, and data packets are transmitted to an edge computing node using a 5G network.

[0031] Step S2: Perform sliding window filtering on the raw data at the edge node to extract the time-domain statistical features of mean, variance, and kurtosis and the frequency-domain energy entropy features, and build a real-time state mapping of the digital twin model in combination with the physical parameters of the equipment;

[0032] Step S3: Input the feature data into the cloud-based deep neural network, extract fault-sensitive features through multi-scale convolutional layers, combine the attention mechanism to model the temporal dependency, output the fault location result and remaining life interval, and trigger the MOM system to generate a maintenance work order;

[0033] Step S4: Based on the maintenance work order priority, the disassembly process 3D animation, torque calibration parameters, and safety operation prompts are pushed to the AR terminal. After the on-site personnel complete the maintenance operation through the AR interface, the equipment health status is updated to the digital twin model, forming a closed-loop optimization.

[0034] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management method of the present invention, in step S2, the time domain kurtosis feature calculation satisfies:

[0035]

[0036] Where μ is the signal mean and σ is the standard deviation, which are used to characterize the intensity of the impact component of the vibration signal.

[0037] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management method of the present invention, in step S3, the attention mechanism weight distribution satisfies:

[0038]

[0039] Among them, Q is the query vector, K is the key vector, and d k It is a vector dimension, which is used to enhance the feature weight of the fault-sensitive time point.

[0040] As a preferred solution of the 5G-MOM-based intelligent operation and maintenance management method for industrial equipment described in the present invention, in step S3, the fault location result is generated by a multi-model voting mechanism, and the final fault type determination satisfies:

[0041]

[0042] Among them, y k is the predicted category of the k-th base model, w k is its F1-score weight on the validation set, and δ is the indicator function used to improve the diagnostic robustness under complex working conditions.

[0043] As a preferred solution of the 5G-MOM-based industrial equipment intelligent operation and maintenance management method of the present invention, in step S4, the maintenance priority calculation satisfies:

[0044] P=w1·S health +w2·C cost +w3·T deadline

[0045] Among them, S health Score the device health, C cost is the downtime cost coefficient, T deadline is the time urgency of the current production task, w1, w2, and w3 are weight coefficients, which are used to quantify the influence of different decision factors on maintenance priority. w1 is the equipment health score weight, w2 is the downtime cost coefficient weight, and w3 is the production plan urgency weight.

[0046] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the 5G-MOM-based industrial equipment intelligent operation and maintenance management method as described in the first aspect of the present invention.

[0047] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent operation and maintenance management method for industrial equipment based on 5G-MOM as described in the first aspect of the present invention.

[0048] The beneficial effects of the present invention are:

[0049] The present invention deeply integrates the multi-source heterogeneous data acquisition layer with the high bandwidth and low latency characteristics of the 5G network to achieve millisecond-level synchronous acquisition and stable transmission of multi-dimensional working condition data such as vibration, temperature, and current of industrial equipment, breaking through the limitations of incomplete data dimensions and incomplete feature information extraction in the traditional single sensor monitoring mode, and significantly improving the comprehensiveness and real-time performance of equipment status perception; based on the localized data cleaning and feature modeling technology of the edge computing node cluster, while effectively reducing the cloud processing load, a dynamically updated equipment digital twin model is constructed through joint analysis of the time and frequency domains, which improves the conversion efficiency of raw data to feature vectors by more than 60%, providing high signal-to-noise ratio input data for subsequent intelligent analysis; the cloud-based intelligent analysis platform adopts a collaborative architecture of multi-scale convolutional neural networks and LSTM life prediction models, combined with transfer learning mechanisms and incremental training of historical fault case libraries, which can not only adaptively identify early weak fault characteristics under complex working conditions, but also accurately predict the remaining service life of the equipment. The confidence interval reduces the fault misjudgment rate from 15% in the traditional threshold alarm mode to less than 3%; through the deep integration of the MOM system and AR terminals, dynamic optimization of maintenance strategies and virtual-real fusion guidance of on-site operations are achieved. Prioritized maintenance work orders can be automatically generated based on real-time production plans, spare parts inventory, and equipment health scores. The AR interface overlays a three-dimensional model of the equipment's internal structure and operation guidance animations, which increases on-site personnel maintenance efficiency by more than 50% and reduces equipment unplanned downtime by more than 40%. In addition, the system's built-in maintenance knowledge graph and multi-model voting mechanism build a closed-loop reasoning chain from fault diagnosis to maintenance decision-making. By continuously comparing the data differences between predicted results and actual maintenance effects, the weight parameters of the cloud model are reversely optimized to form an intelligent operation and maintenance system with self-evolution capabilities, ultimately achieving technical results of saving more than 30% in operation and maintenance costs and improving the overall equipment efficiency (OEE) by 25%, providing a highly reliable and adaptable solution for the full life cycle management of industrial equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 Schematic diagram of the 5G-MOM-based intelligent operation and maintenance management system for industrial equipment in Example 1;

[0052] Figure 2 This is a flow chart of the intelligent operation and maintenance management method for industrial equipment based on 5G-MOM in Example 2. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0056] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an intelligent operation and maintenance management system for industrial equipment based on 5G-MOM, including:

[0057] The multi-source heterogeneous data acquisition layer is deployed on the industrial equipment itself and integrates vibration sensors, temperature sensors, current monitoring modules, and industrial protocol adapters. It is used to collect multi-dimensional operating status data of the equipment in real time and achieve millisecond-level data transmission through the 5G network.

[0058] A 5G-based edge computing node cluster, consisting of distributed edge servers and embedded data processing modules, is used to filter noise, remove outliers, and standardize raw data. It also generates device operation feature vectors through time-frequency domain feature extraction algorithms, and simultaneously builds a dynamically updated device digital twin model.

[0059] The cloud-based intelligent analysis platform deploys a fault diagnosis model based on a deep neural network and an LSTM life prediction model. It is used to integrate and analyze the multi-dimensional time series feature data uploaded by edge nodes, and output equipment health status scores, fault type probability distributions, and remaining service life prediction results.

[0060] The human-machine collaborative operation and maintenance terminal integrates the MOM system maintenance strategy engine and the AR augmented reality interactive interface. It is used to receive cloud analysis results and generate a dynamic maintenance plan that includes maintenance priorities, spare parts requirement lists and operation instructions. The AR terminal visually guides on-site personnel to perform maintenance operations.

[0061] The feature extraction algorithm of the edge computing node cluster satisfies the following relationship:

[0062]

[0063] Among them, E k Indicates the energy value of the kth frequency band, W k,i is the coefficient of the i-th node in the k-th frequency band after the vibration signal is decomposed by wavelet packet, which is used to construct the frequency domain energy entropy feature vector.

[0064] The fault diagnosis model of the cloud-based intelligent analysis platform uses a multi-scale convolutional neural network, and the output of its convolutional layer satisfies the following requirements:

[0065]

[0066] in, is the convolution kernel weight, b (l) is the bias term, is the original or front-layer feature data input to the l-th layer of the convolutional neural network, P×Q is the convolution kernel size, σ is the ReLU activation function, and the fault features of different frequency ranges are captured through multi-scale convolution kernels.

[0067] The hidden state update of the LSTM life prediction model satisfies:

[0068] h t =o t ⊙tanh(c t )

[0069] Among them, t is the output gating signal, c t The cell state is used to model the long-term temporal dependencies during device performance degradation.

[0070] The human-machine collaborative operation and maintenance terminal integrates the equipment maintenance knowledge graph module, and its construction satisfies the following relationships:

[0071]

[0072] Among them, e i ,e j The smallest semantic unit for device maintenance knowledge, r ij It is a rule engine that connects semantic units and drives automated reasoning from fault diagnosis to maintenance decision-making. The entity set ε contains three types of nodes: equipment components, failure modes, and maintenance operations. The relationship set Define component-fault association and fault-maintenance mapping rules to automatically generate maintenance path reasoning chains based on cloud-based diagnosis results.

[0073] After the industrial equipment is started, the multi-source heterogeneous data acquisition layer uses vibration sensors tightly integrated in key parts of the equipment to capture the impact and wear characteristics of mechanical transmission components in real time. The temperature sensor array monitors the temperature gradient distribution in heat-sensitive areas such as motor windings and bearing seats at a sampling frequency of 50 times per second. The current monitoring module synchronously records the current waveform phase and harmonic distortion rate of the three-phase motor. At the same time, the industrial protocol adapter automatically parses the operating parameters of different protocols such as Modbus and Profinet in the equipment control system, encapsulates the multi-dimensional working condition data into standardized data packets, and completes data aggregation across regional equipment clusters within 5 milliseconds through the high-priority transmission channel allocated by 5G network slicing technology. After the edge computing node cluster receives the raw data stream, the embedded data processing module immediately starts a three-level processing pipeline: first, a sliding window filtering algorithm is used to suppress high-frequency noise in the vibration signal, and a box plot analysis method is used to eliminate outliers in temperature monitoring; then the current waveform data is normalized to eliminate dimensional differences caused by grid voltage fluctuations; finally, a time-frequency domain joint analysis engine is used to extract time domain statistical features (such as the peak-to-peak value and waveform factor of the vibration signal) and frequency domain energy distribution features (such as the spectral center of gravity offset of the current harmonic component) from the preprocessed data, and a dynamic digital twin is constructed based on the equipment's three-dimensional CAD model and real-time operating parameters, and lightweight feature vectors and twin state snapshots are uploaded to the cloud every 30 seconds. After the cloud-based intelligent analysis platform receives the feature data, the deep neural network fault diagnosis module first initiates multi-scale convolution kernel scanning. Through convolution layers of different sizes, it captures local abnormal patterns in the vibration spectrum and long-period trend changes in the current waveform in parallel. Combined with the attention mechanism, it dynamically weights the feature contribution of key time points and outputs a probability distribution map of typical faults such as bearing peeling and gear tooth breakage. At the same time, the LSTM life prediction model analyzes the temporal correlation of performance degradation layer by layer based on the equipment's historical degradation curve and current feature vector, and generates a confidence interval prediction report for the remaining service life. All analysis results are cross-validated by the MOM system maintenance strategy engine and the production planning database to automatically generate prioritized maintenance work orders. After the human-machine collaborative operation and maintenance terminal receives the work order instruction, the AR interactive interface immediately activates the equipment positioning and navigation function, and accurately superimposes the virtual maintenance instructions on the physical equipment through SLAM technology: After wearing AR glasses, on-site personnel can observe the meshing status of the internal gears of the equipment through perspective, and follow the three-dimensional animation demonstration to step by step complete operations such as bearing disassembly and clearance calibration. After each maintenance step is completed, the system automatically verifies the matching degree of the torque sensor feedback data with the standard parameters, and abnormal operations will trigger real-time alarms; after the maintenance is completed, the updated equipment status data is transmitted back to the digital twin model, forming a two-way mapping closed loop from the physical entity to the virtual model. At the same time, the maintenance process video and effect evaluation data are entered into the cloud knowledge base to provide incremental training samples for subsequent model iterations.The entire process uses the 5G-MOM architecture to achieve seamless integration of data collection, analysis and decision-making, and execution feedback, enabling equipment health management to transition from traditional post-maintenance to predictive maintenance.

[0074] Example 2, reference Figure 2 , which is the second embodiment of the present invention, provides an intelligent operation and maintenance management method for industrial equipment based on 5G-MOM, including the following steps:

[0075] Step S1: Vibration signals, temperature distribution, three-phase current waveforms, and equipment control parameters are synchronously collected through a sensor array deployed on industrial equipment, and data packets are transmitted to an edge computing node using a 5G network.

[0076] At key monitoring points on industrial equipment, a modular approach is adopted to deploy multi-type sensor arrays based on the structural characteristics of the equipment and the distribution of fault-sensitive areas. High-precision piezoelectric vibration sensors are installed at the meshing point between the high-speed rotating bearing seat and the gearbox. A dual fixation method, magnetic base and threaded fastening, ensures the signal-to-noise ratio of the collected signal. The vibration signal is sampled 100,000 times per second to capture wide-band (0-20kHz) time-domain waveforms and impact events. Miniaturized infrared temperature sensor arrays are deployed in temperature-sensitive areas such as the motor stator windings and hydraulic system oil circuit interfaces. An equidistant annular distribution strategy is used to reconstruct the three-dimensional temperature field, updating the temperature gradient data every millisecond. At the output end of the drive cabinet of the three-phase asynchronous motor, a non-intrusive current transformer and a high-speed AD sampling module are integrated to synchronously capture the instantaneous amplitude, phase angle, and harmonic distortion rate of the A / B / C three-phase currents, with a sampling accuracy of 0.5. At the same time, a multi-protocol industrial gateway is installed in the equipment PLC control cabinet to read the speed setpoint and OPC data in the Modbus TCP register in real time through hardware-level signal isolation technology. Pressure control parameters in the UA server and position feedback signals in the Profinet IO module. All sensor nodes achieve microsecond-level time alignment through a precision clock synchronization module (using the IEEE 1588 Precision Time Protocol), ensuring strict time-scale alignment of multi-dimensional data including vibration, temperature, current, and control parameters. During data acquisition, the vibration sensor's IEPE power supply circuit and anti-electromagnetic interference shielding work together to effectively suppress high-frequency noise introduced by the on-site inverter and high-voltage cables. The temperature sensor has a built-in self-calibration algorithm that dynamically compensates for measurement deviation based on ambient temperature and humidity. The current monitoring module uses differential input and digital filtering to eliminate common-mode interference. After local pre-processing, the collected multi-source data is encapsulated into a unified data packet format according to a preset frame structure. This data packet contains a timestamp, device ID, sensor type code, raw value, and CRC checksum. It is transmitted over the 5G industrial gateway's URLLC (Ultra-Reliable Low Latency Communication) channel. Network slicing technology is used to assign the highest priority QoS policy to the vibration signal, ensuring a stable end-to-end transmission latency of less than 3ms. Temperature and current data are transmitted in parallel via eMBB (enhanced Mobile Broadband) slices. When encountering multipath interference caused by metal structures within the factory, adaptive modulation and coding technology dynamically adjusts the signal modulation method and error correction code rate, keeping packet loss and retransmission times to less than two per second while maintaining a transmission success rate of over 95%. During transmission, the edge computing node continuously monitors the channel status. If it detects that the 5G signal strength is below -90dBm and has lasted for 200ms, it automatically switches to the local LoRa wireless link for data caching. Once the 5G connection is restored, it resumes the transmission, ensuring data flow continuity.

[0077] Step S2: Perform sliding window filtering on the raw data at the edge node to extract the time-domain statistical features of mean, variance, and kurtosis and the frequency-domain energy entropy features, and build a real-time state mapping of the digital twin model in combination with the physical parameters of the equipment;

[0078] In step S2, the temporal kurtosis feature calculation satisfies:

[0079]

[0080] Where μ is the signal mean and σ is the standard deviation, which are used to characterize the intensity of the impact component of the vibration signal.

[0081] After receiving the raw data stream from multi-source sensors, the edge node first initiates the sliding window filtering mechanism: for vibration signals, a Hanning window function with a width of 512 sampling points and a step size of 256 points is used for segmentation. Within each window, a finite impulse response (FIR) low-pass filter is used to filter out high-frequency noise above 20kHz. At the same time, a sliding window based on median filtering is used for temperature data. The window length is set to a continuous sampling sequence of 10 seconds. By replacing abnormal fluctuation values point by point, spike interference caused by instantaneous poor sensor contact is eliminated. The current waveform data is divided into windows according to the power frequency period (50Hz corresponds to 20ms), and the adaptive Kalman filter algorithm is used to separate the fundamental and harmonic components. After filtering, the system extracts features from the time-domain signal once per second. For vibration signals, the mean is calculated within each sliding window to reflect the overall vibration energy level. The variance is used to quantify the intensity of signal fluctuations, and kurtosis is used to identify the density of transient impact events using fourth-order moment statistics. For temperature data, the gradient variance is calculated based on the spatial distribution of each measurement point within the window to characterize the diffusion trend of local overheating areas in the equipment. The current harmonic distortion rate is dynamically updated based on the effective value ratio of the fundamental wave to each harmonic within the window. In the frequency-domain feature extraction stage, the vibration signal is divided into eight equal-width frequency bands using a fast Fourier transform (FFT). The percentage of energy in each frequency band to the total energy is calculated to form a spectral energy distribution vector. Wavelet packet decomposition is also used to extract the energy entropy of 16 sub-bands at a three-level decomposition depth to characterize the frequency-domain information chaos within complex vibration patterns. For the current signal, a short-time Fourier transform is used to generate a time-spectrum diagram, extracting the amplitude variation trajectory of specific harmonics (such as the 5th and 7th harmonics) as electrical features. All time-frequency domain feature parameters interact in real time with the device's physical parameter library. Structural parameters such as the number of bearing rollers and gear module are retrieved from the device's 3D CAD model. Combined with operating parameters such as current speed and load, the feature vectors are mapped to corresponding nodes in the digital twin model. For example, if a 20% energy spike in the frequency band corresponding to a particular gear's meshing frequency is detected, the twin model automatically correlates the gear's material fatigue coefficient with historical stress distribution data, highlighting areas of excessive tooth contact stress in the virtual model. Simultaneously, the bearing's remaining life prediction weights are dynamically adjusted based on a correlation curve between bearing raceway surface roughness parameters and vibration kurtosis. A data validation module runs simultaneously to check the integrity of the feature vectors. If the variance of a sensor channel falls below a threshold for three consecutive windows, a sensor health alert is triggered. If the frequency-domain energy entropy calculation result exceeds the material's physical property limit, a manual review flag is automatically inserted. The processed feature dataset and twin status snapshot are compressed and transmitted via the 5G uplink. A rolling cache of the last 24 hours is retained on the node for online model calibration and retrospective analysis.

[0082] Step S3: Input the feature data into the cloud-based deep neural network, extract fault-sensitive features through multi-scale convolutional layers, combine the attention mechanism to model the temporal dependency, output the fault location result and remaining life interval, and trigger the MOM system to generate a maintenance work order;

[0083] In step S3, the attention mechanism weight distribution satisfies:

[0084]

[0085] Among them, Q is the query vector, K is the key vector, and d k It is a vector dimension, which is used to enhance the feature weight of the fault-sensitive time point.

[0086] In step S3, the fault location result is generated through a multi-model voting mechanism, and the final fault type determination satisfies:

[0087]

[0088] Among them, y k is the predicted category of the k-th base model, w k is its F1-score weight on the validation set, and δ is the indicator function used to improve the diagnostic robustness under complex working conditions.

[0089] After receiving the feature dataset uploaded by the edge node, the cloud-based intelligent analysis platform first activates the data alignment and contextual association engine. This platform compares the real-time feature vectors with historical operating condition records, a fault case database, and data from groups of similar devices in the device's full lifecycle database. It automatically completes missing feature fields due to network jitter and adds metadata tags such as device model, cumulative operating hours, and recent maintenance history to each data record. The preprocessed data stream is then fed into a multi-scale convolutional neural network module, which utilizes a parallelized convolution kernel design. Large-scale convolution kernels (e.g., with a time window span of up to 10 seconds) capture the slow drift of vibration energy in the low-frequency band, reflecting progressive faults such as bearing wear. Medium-scale convolution kernels (with a 1-second window) focus on the periodic fluctuations of current harmonic amplitudes, identifying medium-frequency faults such as motor rotor eccentricity. Small-scale convolution kernels (with a 0.1-second window) perform high-resolution scanning of transient impact components in the vibration signal, accurately locating sudden anomalies such as gear tooth breakage or ball bearing spalling. After each convolutional layer output, a temporal attention mechanism is dynamically activated. By analyzing the correlation strength between adjacent time steps in the feature sequence, it assigns higher weights to feature vectors at key nodes, such as the equipment startup and shutdown phases and load sudden changes, thereby suppressing redundant information interference during steady-state operation. The fault diagnosis branch maps the weighted features into a decision space encompassing 12 typical fault categories. A hierarchical classification strategy is employed: first, the main frequency band offset of the vibration spectrum is used to identify the general fault category (e.g., bearing fault, gear fault), then the phase correlation of the current harmonic components is combined to refine the specific fault mode (e.g., differentiating between bearing outer ring spalling and inner ring cracking). Ultimately, the probability distribution and confidence level assessment for each fault type are output. Meanwhile, the life prediction branch feeds the time series features into a bidirectional LSTM network, which uses memory cells to store long-term dependencies during device performance degradation. During the model training phase, a mixed dataset of accelerated life test data and actual field failure cases is injected, enabling the prediction engine to adapt to differences in degradation rates under different operating conditions. The prediction results are presented as remaining useful life intervals (e.g., "72-96 hours") and are dynamically calibrated based on the equipment's real-time load factor and historical maintenance effectiveness data. After the diagnosis and prediction results are integrated through credibility, the maintenance strategy engine of the MOM system is triggered: the order schedule, spare parts inventory status and personnel schedule in the production planning database are called, and the maintenance priority index of each device is calculated by combining the equipment health score, downtime cost coefficient and production urgency weight. For example, for equipment with a remaining life of less than 48 hours and located on a critical production line, a red emergency work order is automatically generated and spare machine resources are reserved; for non-core equipment with a wider prediction interval, it is recommended to process it in the next planned maintenance window. During the work order generation process, the knowledge graph module initiates multi-path reasoning: based on the failure probability distribution, maintenance records of similar cases are retrieved, the required spare parts models, special tool lists and safety operating specifications are associated, and the availability of spare parts in the inventory system and the minimum replenishment cycle are verified.The final maintenance work order includes a three-dimensional schematic diagram of the fault location, an animation link to the disassembly and assembly process, standard torque parameters, and a safety warning label, which is pushed to the designated human-machine collaboration terminal through an encrypted data channel.

[0090] Step S4: Based on the maintenance work order priority, the disassembly process 3D animation, torque calibration parameters, and safety operation prompts are pushed to the AR terminal. After the on-site personnel complete the maintenance operation through the AR interface, the equipment health status is updated to the digital twin model, forming a closed-loop optimization.

[0091] In step S4, the maintenance priority calculation satisfies:

[0092] P=w1·S health +w2·C cost +w3·T deadline

[0093] Among them, S health Score the device health, C cost is the downtime cost coefficient, T deadline is the time urgency of the current production task, w1, w2, and w3 are weight coefficients, which are used to quantify the influence of different decision factors on maintenance priority. w1 is the equipment health score weight, w2 is the downtime cost coefficient weight, and w3 is the production plan urgency weight.

[0094] After a maintenance work order is generated, the MOM system intelligently dispatches the work order based on the equipment's geographic location, the maintenance personnel's skill level, and the status of their tool equipment. Upon receiving the task instruction, the AR terminal immediately activates the spatial positioning module: Using binocular cameras and LiDAR sensors, it performs millisecond-level 3D reconstruction of the on-site environment. Combined with pre-loaded equipment BIM model data, it uses an improved semantic SLAM algorithm to precisely match virtual operation instructions with the physical spatial coordinates of the physical equipment, with an error within ±0.5mm. When on-site personnel put on AR glasses, their line of sight automatically locks onto the target equipment. A translucent 3D exploded view of the faulty component is superimposed on the interface, with abnormal areas highlighted with pulsing red light. By pinching their fingers in the air, they can dissect the equipment casing layer by layer, observing the meshing state of internal gear sets or the wear of bearing raceways through perspective. During the disassembly process guidance phase, the AR interface presents a split-screen display: a floating window on the left plays a 3D animation of the bearing disassembly process, demonstrating the puller installation angle and hydraulic pressure application curve; the right side displays the current bolt preload value fed back by the digital twin model in real time. When the wrench torque sensor detects that the actual torque value deviates by more than ±5% from the standard parameter, the interface immediately pops up an orange warning box and freezes the animation playback until the operation is corrected and guidance is resumed. The safety operation reminder system is deeply integrated with the enterprise EHS specification library: when a person enters within 1 meter of a high-voltage electrical cabinet, the AR glasses automatically overlay an insulating glove wearing detection icon. The camera AI recognizes that the person is not wearing compliant protective equipment, triggering an audible and visual alarm and locking the subsequent operation steps. During the hydraulic system maintenance phase, the interface dynamically displays the residual pressure value in the pipeline. If it detects that the pressure relief is not complete and attempts to disassemble the joint, the system will force a 20-second safety confirmation countdown and jointly close the upstream valve. After completing each maintenance sub-step, the operator confirms it by voice commanding "Step Complete" or by swiping a check mark with their hand. The system automatically verifies the associated sensor data. For example, after a bearing is installed, the real-time vibration characteristic analysis module immediately initiates a 10-second short-term monitoring. If the newly collected vibration RMS value drops by more than 30% compared to the historical baseline, the installation is deemed qualified. Otherwise, the instructions for that step are retraced and a recalibration prompt is prompted. After maintenance is completed, the AR terminal enters acceptance mode: it retrieves the device's standard health parameter curve from the digital twin model and compares it with the 24-hour post-maintenance test data (including vibration spectrum, temperature gradient, and current harmonic distortion). If key indicators return to the green safety zone, the work order is marked as closed in the MOM system. If any residual anomalies remain, a secondary maintenance plan is generated. The entire operation video (including gesture interaction traces, tool usage sequence, and environmental conditions) is desensitized and encrypted and uploaded to a cloud-based knowledge base. Video semantic segmentation technology is used to extract typical maintenance action clips for updating the AR guidance animation library and the operation specification knowledge graph.At the same time, maintenance effect data (such as fault elimination rate, time deviation value, and spare parts consumption) are back-injected into the self-learning engine of the digital twin model, dynamically adjusting the life prediction weight coefficient of similar faults and optimizing the feature extraction threshold of the edge computing node, forming a full-link adaptive optimization closed loop from physical maintenance to virtual model and then to algorithm iteration.

[0095] This embodiment also provides a computer device, which is suitable for the intelligent operation and maintenance management method of industrial equipment based on 5G-MOM, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent operation and maintenance management method of industrial equipment based on 5G-MOM as proposed in the above embodiment.

[0096] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0097] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent operation and maintenance management method for industrial equipment based on 5G-MOM as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0098] In summary, the present invention deeply integrates the multi-source heterogeneous data acquisition layer with the high bandwidth and low latency characteristics of the 5G network to achieve millisecond-level synchronous acquisition and stable transmission of multi-dimensional working condition data such as vibration, temperature, and current of industrial equipment, breaking through the limitations of incomplete data dimensions and incomplete feature information extraction in the traditional single sensor monitoring mode, and significantly improving the comprehensiveness and real-time performance of equipment status perception; based on the localized data cleaning and feature modeling technology of the edge computing node cluster, while effectively reducing the cloud processing load, a dynamically updated equipment digital twin model is constructed through joint analysis of the time and frequency domains, which improves the conversion efficiency of raw data to feature vectors by more than 60%, providing high signal-to-noise ratio input data for subsequent intelligent analysis; the cloud-based intelligent analysis platform adopts a collaborative architecture of multi-scale convolutional neural networks and LSTM life prediction models, combined with transfer learning mechanisms and incremental training of historical fault case libraries, which can not only adaptively identify early weak fault characteristics under complex working conditions, but also accurately predict the remaining service life of equipment. The confidence interval reduces the fault misjudgment rate from 15% in the traditional threshold alarm mode to less than 3%; through the deep integration of the MOM system and AR terminals, dynamic optimization of maintenance strategies and virtual-real fusion guidance of on-site operations are achieved. Prioritized maintenance work orders can be automatically generated based on real-time production plans, spare parts inventory, and equipment health scores. The AR interface overlays a three-dimensional model of the equipment's internal structure and operation guidance animations, which increases on-site personnel maintenance efficiency by more than 50% and reduces equipment unplanned downtime by more than 40%. In addition, the system's built-in maintenance knowledge graph and multi-model voting mechanism construct a closed-loop reasoning chain from fault diagnosis to maintenance decision-making. By continuously comparing the data differences between the predicted results and the actual maintenance effects, the weight parameters of the cloud model are reversely optimized to form an intelligent operation and maintenance system with self-evolution capabilities, ultimately achieving technical results of saving more than 30% in operation and maintenance costs and improving the overall equipment efficiency (OEE) by 25%, providing a highly reliable and adaptable solution for the full life cycle management of industrial equipment.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. The intelligent operation and maintenance management system for industrial equipment based on 5G-MOM is characterized by: include: The multi-source heterogeneous data acquisition layer is deployed on the industrial equipment itself and integrates vibration sensors, temperature sensors, current monitoring modules, and industrial protocol adapters. It is used to collect multi-dimensional operating status data of the equipment in real time and achieve millisecond-level data transmission through the 5G network. A 5G-based edge computing node cluster, consisting of distributed edge servers and embedded data processing modules, is used to filter noise, remove outliers, and standardize raw data. It also generates device operation feature vectors through time-frequency domain feature extraction algorithms, and simultaneously builds a dynamically updated device digital twin model. The cloud-based intelligent analysis platform deploys a fault diagnosis model based on a deep neural network and an LSTM life prediction model. It is used to integrate and analyze the multi-dimensional time series feature data uploaded by edge nodes, and output equipment health status scores, fault type probability distributions, and remaining service life prediction results. The human-machine collaborative operation and maintenance terminal integrates the MOM system maintenance strategy engine and the AR augmented reality interactive interface. It is used to receive cloud analysis results and generate a dynamic maintenance plan that includes maintenance priorities, spare parts requirement lists and operation instructions. The AR terminal visually guides on-site personnel to perform maintenance operations.

2. The 5G-MOM-based intelligent operation and maintenance management system for industrial equipment according to claim 1, characterized in that: The feature extraction algorithm of the edge computing node cluster satisfies the following relationship: Among them, E k Indicates the energy value of the kth frequency band, W k,i is the coefficient of the i-th node in the k-th frequency band after the vibration signal is decomposed by wavelet packet, which is used to construct the frequency domain energy entropy feature vector.

3. The 5G-MOM-based intelligent operation and maintenance management system for industrial equipment according to claim 1, characterized in that: The fault diagnosis model of the cloud-based intelligent analysis platform adopts a multi-scale convolutional neural network, and the output of its convolutional layer satisfies: in, is the convolution kernel weight, b (l) is the bias term, is the original or front-layer feature data input to the l-th layer of the convolutional neural network, P×Q is the convolution kernel size, σ is the ReLU activation function, and the fault features of different frequency ranges are captured through multi-scale convolution kernels.

4. The 5G-MOM-based intelligent operation and maintenance management system for industrial equipment according to claim 1, characterized in that: The hidden state update of the LSTM life prediction model satisfies: h t =o t ⊙tanh(c t ) Among them, t is the output gating signal, c t The cell state is used to model the long-term temporal dependencies during device performance degradation.

5. The 5G-MOM-based intelligent operation and maintenance management system for industrial equipment according to claim 1, characterized in that: The human-machine collaborative operation and maintenance terminal integrated equipment maintenance knowledge graph module is constructed to meet the following relationship: Among them, e i ,e j The smallest semantic unit for device maintenance knowledge, r ij It is a rule engine that connects semantic units and drives automated reasoning from fault diagnosis to maintenance decision-making. The entity set ε contains three types of nodes: equipment components, failure modes, and maintenance operations. The relationship set Define component-fault association and fault-maintenance mapping rules to automatically generate maintenance path reasoning chains based on cloud-based diagnosis results.

6. A 5G-MOM-based industrial equipment intelligent operation and maintenance management method, which is implemented based on the 5G-MOM-based industrial equipment intelligent operation and maintenance management system according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step S1: Vibration signals, temperature distribution, three-phase current waveforms, and equipment control parameters are synchronously collected through a sensor array deployed on industrial equipment, and data packets are transmitted to an edge computing node using a 5G network. Step S2: Perform sliding window filtering on the raw data at the edge node to extract the time-domain statistical features of mean, variance, and kurtosis and the frequency-domain energy entropy features, and build a real-time state mapping of the digital twin model in combination with the physical parameters of the equipment; Step S3: Input the feature data into the cloud-based deep neural network, extract fault-sensitive features through multi-scale convolutional layers, combine the attention mechanism to model the temporal dependency, output the fault location result and remaining life interval, and trigger the MOM system to generate a maintenance work order; Step S4: Based on the maintenance work order priority, the disassembly process 3D animation, torque calibration parameters, and safety operation prompts are pushed to the AR terminal. After the on-site personnel complete the maintenance operation through the AR interface, the equipment health status is updated to the digital twin model, forming a closed-loop optimization.

7. The intelligent operation and maintenance management method for industrial equipment based on 5G-MOM according to claim 6, characterized in that: In step S2, the temporal kurtosis feature calculation satisfies: Where μ is the signal mean and σ is the standard deviation, which are used to characterize the intensity of the impact component of the vibration signal.

8. The intelligent operation and maintenance management method for industrial equipment based on 5G-MOM according to claim 6, characterized in that: In step S3, the attention mechanism weight distribution satisfies: Among them, Q is the query vector, K is the key vector, and d k It is a vector dimension, which is used to enhance the feature weight of the fault-sensitive time point.

9. The intelligent operation and maintenance management method for industrial equipment based on 5G-MOM according to claim 6, characterized in that: In step S3, the fault location result is generated through a multi-model voting mechanism, and the final fault type determination satisfies: Among them, y k is the predicted category of the k-th base model, w k is its F1-score weight on the validation set, and δ is the indicator function used to improve the diagnostic robustness under complex working conditions.

10. The intelligent operation and maintenance management method for industrial equipment based on 5G-MOM according to claim 6, characterized in that: In step S4, the maintenance priority calculation satisfies: P=w1·S health +w2·C cost +w3·T deadline Among them, S health Score the device health, C cost is the downtime cost coefficient, T deadline is the time urgency of the current production task, w1, w2, and w3 are weight coefficients, which are used to quantify the influence of different decision factors on maintenance priority. w1 is the equipment health score weight, w2 is the downtime cost coefficient weight, and w3 is the production plan urgency weight.

Images