Port and shipping equipment health perception monitoring system based on 5g and lora hybrid transmission

Abstract

This invention provides a health monitoring system for port and shipping equipment based on hybrid 5G and LoRa transmission. A sensor cluster is custom-designed for the harsh operating conditions of port and shipping facilities. Through structural optimization, electromagnetic shielding, and differential output design, the system significantly improves the sensors' anti-interference capabilities and protection performance. An intelligent scheduling algorithm based on fuzzy hierarchical analysis is constructed to achieve dynamic intelligent traffic splitting and seamless switching between LoRa and 5G networks. Link weights are dynamically adjusted for different priority data streams, balancing the low power consumption and wide coverage of LoRa with the high bandwidth and low latency advantages of 5G. Addressing the non-stationary, strongly coupled, and multi-condition operating characteristics of port and shipping equipment, a CNN-LSTM-Attention multi-task fusion model is constructed to simultaneously achieve fault classification and diagnosis and remaining service life prediction. This improves fault diagnosis accuracy and early fault identification rate, reduces RUL prediction error, and enables early fault warning.

Description

Technical Field

[0001] This invention relates to the field of ship operation and maintenance technology, and in particular to a port and shipping equipment health perception and monitoring system based on 5G and LoRa hybrid transmission, as well as a port and shipping equipment health perception and monitoring method and device, electronic equipment and computer-readable storage medium based on 5G and LoRa hybrid transmission. Background Technology

[0002] With the continuous expansion of global trade, the port and shipping industry, as a core hub of the logistics supply chain, is accelerating its transformation towards intelligent and unmanned operations. The operational reliability of port and shipping equipment (quay cranes, gantry cranes, yard cranes, conveying equipment, hydraulic systems, etc.) directly determines port operation efficiency, operating costs, and production safety. Port and shipping equipment operates under complex and harsh conditions of high load, strong corrosion, strong electromagnetic interference, large vibration, and high dust levels. Key components (motors, reducers, bearings, gears, hydraulic systems, steel structures) are prone to wear, fatigue, cracks, and leaks. Traditional equipment management models have the following unavoidable technical defects: Limitations of the sensing layer technology: Traditional sensors generally suffer from insufficient protection levels and weak anti-interference capabilities. In the strong electromagnetic and salt spray corrosive environments of ports and shipping, they are prone to temperature drift, zero-point drift, and signal distortion, making it impossible to achieve high-precision and stable acquisition of equipment operating parameters. Traditional acquisition terminals only have simple data transmission functions and lack multi-protocol compatibility, edge preprocessing, and local inference capabilities. The full raw data transmission results in high bandwidth consumption and high terminal power consumption, making it unsuitable for the large-scale, multi-node monitoring needs of ports.

[0003] Technical bottlenecks in the transmission layer: Existing port and shipping equipment monitoring mostly adopts a single transmission mode: wired transmission is difficult to implement, prone to wear and tear, and cannot be adapted to the highly mobile port and shipping equipment; single 5G transmission has coverage blind spots in enclosed areas such as ship cabins, yard pipe corridors, and internal cavities of equipment, and the terminal power consumption is high, with insufficient battery power supply nodes; single LoRa transmission, although low power consumption and wide coverage, has low transmission rate and high latency, and cannot handle the high-frequency vibration, acoustic emission, and other high-volume data required for fault diagnosis, thus failing to meet real-time requirements. Existing hybrid transmission solutions are mostly simple master-slave switching without dynamic intelligent traffic distribution mechanisms, failing to balance the core contradictions of power consumption, bandwidth, latency, and coverage.

[0004] Shortcomings of diagnostic layer technology: Traditional equipment monitoring relies on manual inspection and simple threshold alarms, which cannot achieve early fault identification; existing intelligent diagnostic solutions mostly use a single CNN or LSTM model, which cannot fully extract the features of non-stationary, strongly coupled, and multi-condition operating data of port and shipping equipment, resulting in low fault identification rate for small samples, lack of a fault feature library specific to port and shipping equipment, poor generalization ability, and can only achieve post-fault diagnosis. They cannot complete accurate prediction of remaining useful life (RUL) and early warning, and cannot support preventive maintenance of equipment.

[0005] Insufficient system integration: Existing monitoring systems are mostly simple splicing of sensing, transmission and analysis modules, without achieving deep integration. Data links are not smooth, response delays are high, and the coordination between modules is poor. It is impossible to form a full life cycle management system of "data collection-analysis and diagnosis-early warning and maintenance-closed-loop optimization", which is difficult to meet the intelligent and refined requirements of modern smart ports for equipment management. Summary of the Invention

[0006] To address the technical problems existing in the prior art, the present invention provides the following technical solution: On the one hand, a health perception and monitoring system for port and shipping equipment based on 5G and LoRa hybrid transmission is provided. The system adopts a five-layer integrated architecture of end-edge-network-cloud-application, including a terminal perception layer, a hybrid transmission layer, a platform service layer, and an application interaction layer, wherein: The terminal perception layer includes a high-precision anti-interference sensor cluster adapted to the harsh working conditions of ports and shipping, and an edge intelligent acquisition terminal that is matched with the sensor cluster. The high-precision anti-interference sensor cluster is used to collect the operating parameters of port and shipping equipment. The edge intelligent acquisition terminal integrates an FPGA preprocessing unit, an edge computing unit with NPU computing power, and a LoRa+5G dual-mode communication module to realize edge preprocessing of collected data, local anomaly identification, and dual-link data transmission and reception. The hybrid transmission layer includes a LoRa low-power wide-area subnet, a 5G transmission subnet, and a hybrid transmission collaborative gateway. The LoRa low-power wide-area subnet is used to achieve low-power wide-coverage data transmission throughout the port. The 5G transmission subnet is configured with a port-specific network slice to achieve high-bandwidth, low-latency data transmission. The hybrid transmission collaborative gateway has a built-in intelligent scheduling engine to achieve dynamic intelligent traffic splitting, seamless switching, and protocol conversion between LoRa and 5G dual links. The platform service layer is deployed on a port private cloud cluster and includes a cloud data platform, an intelligent diagnosis and prediction engine, an equipment lifecycle management module, and a dedicated fault feature library for port and shipping equipment. The intelligent diagnosis and prediction engine is equipped with a multi-task fusion deep learning model to realize fault classification diagnosis and remaining service life prediction of port and shipping equipment. The application interaction layer includes a multi-terminal visual management and control platform, an early warning and emergency response module, and an operation and maintenance management module. These modules are used to achieve visualized display of equipment operation data, tiered early warning issuance, closed-loop management of operation and maintenance work orders, and remote equipment control. On the other hand, a health perception and monitoring method for port and shipping equipment based on 5G and LoRa hybrid transmission is provided, which is applied to the aforementioned port and shipping equipment health perception and monitoring system based on 5G and LoRa hybrid transmission. The method includes the following steps: System deployment and initialization configuration: High-precision anti-interference sensor clusters were deployed and fixed at locations prone to failure in port and shipping equipment. An independent edge intelligent acquisition terminal was deployed for each piece of port and shipping equipment, completing sensor wiring, address configuration, and acquisition strategy settings. A LoRa low-power wide-area subnet, a 5G transmission subnet, and a hybrid transmission collaborative gateway were deployed to achieve full port network coverage and intelligent scheduling rule configuration. A port private cloud cluster was built, deploying the platform service layer and application interaction layer, completing the construction of a dedicated fault feature library for port and shipping equipment, deployment of pre-trained models, and matching of system parameters across the entire chain. The system performs real-time acquisition and edge preprocessing of equipment operation data. During normal operation, the edge intelligent acquisition terminal controls the sensor cluster to acquire port and shipping equipment operation parameters according to the conventional acquisition strategy. The raw data is transmitted to the FPGA preprocessing module to complete hardware filtering, time domain and frequency domain feature extraction. The main control unit inputs the preprocessed feature data into the lightweight anomaly recognition model of the edge computing unit to complete local real-time inference. If an abnormal feature is identified, the acquisition strategy is immediately upgraded to increase the sensor sampling frequency. The main control unit completes standardized encapsulation and local caching of all data, waiting for transmission scheduling. Intelligent scheduling and data transmission for 5G / LoRa hybrid transmission: The dual communication modules of the edge intelligent acquisition terminal collect link communication quality parameters in real time and report them to the hybrid transmission collaborative gateway. The gateway's intelligent scheduling engine is based on the intelligent scheduling algorithm of fuzzy hierarchical analysis. It generates traffic allocation decisions based on data packet priority, data volume, and real-time link status, and schedules data packets of different priorities to LoRa links, 5G links, or dual-link redundant transmission respectively. When the quality of the current working link is detected to be lower than a preset threshold, seamless link switching is triggered. After receiving dual-link data, the gateway completes decapsulation, verification, and protocol conversion, and forwards the valid data to the cloud data platform. Cloud-based data processing and intelligent fault diagnosis and lifespan prediction: After receiving data, the cloud data platform performs data cleaning, missing value completion, normalization, and multi-sensor timestamp synchronization and alignment to generate a standardized feature matrix, which is then pushed to the intelligent diagnosis and prediction engine. The intelligent diagnosis and prediction engine calls the CNN-LSTM-Attention multi-task fusion model to complete deep feature extraction, matches the features with the port and shipping equipment's dedicated fault feature library, and outputs the fault type, fault location, fault severity, and remaining lifespan prediction. When the diagnosis result exceeds the preset warning threshold, a corresponding level of warning / alarm instruction is generated. The platform implements a system of early warning issuance, operation and maintenance coordination, and closed-loop optimization. It issues tiered early warning information based on the severity of faults, notifying relevant operation and maintenance and dispatch personnel through multiple channels. Simultaneously, it triggers emergency plans in conjunction with the port's emergency management system. Operation and maintenance personnel generate and issue work orders through the platform, upload repair results and fault handling information after completing on-site inspections, and update equipment ledgers and operation and maintenance records. The platform adds new fault data and operation and maintenance feedback data to a dedicated fault feature library for port and shipping equipment, regularly retrains and optimizes the diagnostic model, and distributes the optimized lightweight model to edge intelligent acquisition terminals for OTA updates. It also optimizes system acquisition strategies, early warning thresholds, and operation and maintenance plans to achieve refined management of the entire equipment lifecycle.

[0007] On the other hand, an electronic device is provided, comprising: a processor; and a memory storing computer-readable instructions, which, when executed by the processor, implement the method described above.

[0008] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored therein, the at least one instruction being loaded and executed by a processor to implement the method described above.

[0009] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. Perception Level: Deep integration of high-precision anti-interference perception and edge intelligence to overcome the data collection bottleneck in complex port and shipping environments. This solution features a customized sensor cluster designed for harsh working conditions in ports and shipping. Through structural optimization, electromagnetic shielding, and differential output design, it significantly improves the sensor's anti-interference capability and protection performance, increasing data acquisition accuracy by more than 40%. The edge intelligent acquisition terminal integrates multi-protocol compatibility, hardware preprocessing, and edge computing capabilities, completing data noise reduction, feature extraction, and local anomaly identification on the terminal side. The data compression ratio is ≥60%, and the local anomaly response time is ≤100ms. This solution solves the technical defects of traditional acquisition terminals, such as high bandwidth consumption and slow response caused by the lack of preprocessing, edge inference, and full data transmission.

[0010] 2. Transmission Layer: 5G / LoRa intelligent collaborative hybrid transmission network solves the core contradictions of single transmission technologies, achieving highly reliable transmission with no blind spots across the entire network. This solution breaks through the limitations of existing single transmission technologies by constructing an intelligent scheduling algorithm based on fuzzy hierarchical analysis, enabling dynamic intelligent traffic splitting and seamless switching between LoRa and 5G dual networks. It dynamically adjusts link weights for data streams of different priorities, taking into account the advantages of LoRa's low power consumption and wide coverage, as well as 5G's high bandwidth and low latency. The power consumption of regular data transmission is reduced by more than 80%, the latency of fault data transmission is ≤10ms, and the transmission success rate is ≥99.99%. It perfectly solves the industry pain points of coverage blind spots in port closed areas and the contradiction between terminal power consumption and bandwidth, and avoids the technical defects of existing hybrid transmission fixed traffic splitting and inability to dynamically adapt.

[0011] 3. Diagnostic Level: A multi-task fusion deep learning model combined with a port and shipping-specific fault feature library enables accurate fault diagnosis and early warning. This solution addresses the non-stationary, strongly coupled, and multi-condition operating characteristics of port and shipping equipment by constructing a CNN-LSTM-Attention multi-task fusion model. This model simultaneously achieves fault classification and diagnosis as well as remaining service life prediction, with a fault diagnosis accuracy of ≥98.5%, an early fault identification rate of ≥95%, and a RUL prediction error of ≤8%. It can provide early warning of faults 7 to 30 days in advance. Combined with a dedicated fault feature library for port and shipping equipment and a closed-loop iteration mechanism, this solution overcomes the technical shortcomings of traditional single models, such as incomplete feature extraction, low fault identification rate with small samples, and poor generalization ability. This upgrades equipment management from passive maintenance to preventive maintenance.

[0012] 4. System Level: Integrated end-to-end system construction to build a closed-loop management system for the entire equipment lifecycle, comprehensively improving port operational efficiency. This solution deeply integrates high-precision sensing, hybrid transmission, intelligent diagnosis, and operation and maintenance linkage to construct a health monitoring system that integrates the entire "end-edge-network-cloud-application" chain. It realizes closed-loop management of the entire process of equipment status monitoring, data transmission, analysis and diagnosis, and early warning operation and maintenance. Compared with the traditional manual inspection mode, this solution can reduce the sudden failure rate of equipment by more than 70%, shorten the operation and maintenance response time by more than 60%, and extend the service life of equipment by more than 15%. It significantly reduces equipment maintenance costs and downtime losses, helps ports achieve intelligent upgrading, enhances the core competitiveness of the industry, and promotes the high-quality and sustainable development of the port and shipping industry. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the hardware architecture provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the terminal perception and edge preprocessing process provided in an embodiment of the present invention; Figure 3 This is a flowchart of the method provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0015] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0016] I. This solution proposes a port and shipping equipment health perception and monitoring system based on 5G and LoRa hybrid transmission, which realizes full-dimensional, all-time, and high-precision health status monitoring and intelligent fault early warning of port and shipping equipment, and comprehensively improves the level of port equipment management.

[0017] II. System Overall Architecture This system adopts a five-layer integrated architecture of "device-edge-network-cloud-application", using consistent technical terminology throughout, with logical connections and deep collaboration across all layers. The specific architecture is as follows: Terminal perception layer: It consists of a high-precision anti-interference sensor cluster and an edge intelligent acquisition terminal, which is responsible for high-precision acquisition of device operating parameters, edge preprocessing and local anomaly identification. Hybrid transmission layer: Composed of LoRa low-power wide-area subnet, 5G transmission subnet, and hybrid transmission collaborative gateway, the core of which realizes intelligent data offloading, high-reliability transmission and seamless link switching; Platform service layer: It consists of cloud data platform, intelligent diagnosis and prediction engine, equipment life cycle management module, and port and shipping equipment exclusive fault feature library. The core functions are to realize standardized data processing, in-depth mining, fault diagnosis and life prediction. Application Interaction Layer: Composed of a multi-terminal visual management and control platform, an early warning and emergency response module, and an operation and maintenance management module, the core functions of which include data visualization, early warning release, operation and maintenance work order management, and remote control. Security protection system: It runs through all levels and covers terminal access authentication, transmission encryption, platform permission control, data backup and disaster recovery, ensuring data security and operational stability of the entire system chain.

[0018] III. System Hardware Components like Figure 1 As shown, the hardware system consists of the following components: 3.1 Terminal Perception Layer 3.1.1 High-precision anti-interference sensor cluster Customized for the complex operating conditions of port and shipping facilities, all sensors meet the IP68 protection standard, and feature built-in anti-salt spray corrosion coating, electromagnetic shielding cavity, and wide voltage power management module. Operating temperature range is -40℃ to 85℃. Specific classifications and parameters are as follows: Core operating condition monitoring sensors: Vibration acceleration sensor: MEMS type, fully sealed titanium alloy housing, internally potted with high and low temperature resistant epoxy resin, dual differential signal output circuit to suppress common mode interference, measuring range ±50g, frequency response 0.5Hz~10kHz, suitable for vibration characteristic acquisition of rotating equipment. Acoustic emission sensor: Built-in preamplifier circuit, signal-to-noise ratio ≥65dB, response frequency 100kHz~1MHz, used for early fault detection of internal cracks and wear in equipment; Tilt sensor: High-precision MEMS type, measurement accuracy ±0.01°, used for deformation monitoring of quay crane beams and gantry steel structures; Auxiliary parameter monitoring sensors: temperature and humidity sensors, hydraulic system pressure / level sensors, motor current / voltage sensors, and switch status sensors, all of which support industrial standard bus protocols.

[0019] 3.1.2 Edge Intelligent Data Acquisition Terminal As the core edge hardware of the system, one unit is independently deployed for each port and shipping equipment, installed inside the equipment protection and control cabinet. The hardware composition and internal connection examples are shown in Table 1 below:

[0020] Table 1 3.2 Hybrid Transport Layer This layer is the core of the system's transmission, breaking through the limitations of single transmission technologies and constructing a hybrid transmission architecture with dual-network collaboration. The hardware composition and communication connections are as follows: LoRa Low Power Wide Subnet Hardware components: LoRa module (terminal node) of edge intelligent acquisition terminal, LoRa concentrator; Topology: Star topology. Each LoRa concentrator uses an 8-channel SX1302 chip with a maximum receiving sensitivity of -142dBm and a coverage radius of ≥5km. 3~5 concentrators are deployed throughout the port area to achieve blind-spot-free coverage of the yard, wharf, anchorage, and pipe gallery. Communication protocol: Optimized LoRaWAN protocol, custom time-slot ALOHA channel access mechanism to avoid data collisions between multiple nodes; Connection method: LoRa terminal nodes connect to LoRa concentrators via wireless air interface, and concentrators connect to hybrid transmission collaborative gateways via gigabit Ethernet interface.

[0021] 5G transmission subnet Hardware components: 5G module (terminal node) of edge intelligent acquisition terminal, 5G macro base station / indoor base station in port, and 5G core network UPF sinking node; Network configuration: Deploy a port-dedicated 5G network slice to allocate independent bandwidth and QoS priority to the monitoring system, realize local data offloading without leaving the public network, and ensure data security and low-latency transmission; Connection method: 5G terminal nodes connect to the base station via 5G NR air interface, the base station connects to the UPF sink node via optical fiber, and the UPF node accesses the hybrid transmission collaborative gateway via 10 Gigabit optical fiber, while configuring a public network VPN redundant backup link.

[0022] Hybrid Transmission Collaborative Gateway As the core control hardware for dual-network collaboration, it adopts an industrial-grade rack server, equipped with an Intel Xeon D-1718T 8-core 16-thread CPU, 32GB DDR4 memory, 2TB SSD storage, and dual power supply redundancy; it has 4 built-in gigabit Ethernet ports and 2 10-gigabit optical ports, and is equipped with an FPGA hardware acceleration module to realize hardware acceleration of transmission scheduling algorithms, dual protocol stack conversion, and high-speed data processing.

[0023] Uplink connection: Access to the port's industrial Ethernet via 10 Gigabit fiber optic cable, connecting to the cloud data middleware of the platform service layer; Downlink connection: Connects to the LoRa concentrator via Gigabit Ethernet and to the 5G UPF downlink node via 10 Gigabit fiber optic cable; Core functions: real-time link status monitoring, intelligent traffic offloading and scheduling, seamless link switching, protocol conversion, data verification and forwarding, and downlink control command encapsulation and issuance.

[0024] 3.3 Platform Service Layer The system is deployed using a port private cloud cluster to ensure high computing power, high reliability, and high scalability. The hardware components are as follows: Computing server cluster: 2 redundant high-performance servers, each with an Intel Xeon Gold 6330 CPU (32 cores, 64 threads) and 128GB of memory, used for training, inference, and task scheduling of the intelligent diagnostic model. Distributed storage cluster: 4 storage servers with a total capacity of 192TB, deploying the distributed time series database InfluxDB and object storage MinIO to store structured monitoring data and unstructured raw waveforms, model files, and fault characteristic data, respectively; Web and Application Servers: Two redundant servers are used for service publishing, interface management, and business logic processing in the application interaction layer. Security hardware: Industrial-grade firewalls, intrusion detection systems (IDS), and data encryption machines to ensure platform network and data security; Cluster connection: All servers are clustered together through 10 Gigabit Ethernet switches and connected to a hybrid transmission and collaboration gateway through a firewall to achieve end-to-end data communication.

[0025] 3.4 Application Interaction Layer Port dispatch center: LED visualization screen, industrial-grade PC workstation, sound and light alarm equipment; Maintenance terminals: industrial tablets, smartphones, and portable maintenance computers, accessing the platform via the port's intranet / public VPN; Linked hardware: Port emergency broadcasting system and equipment emergency shutdown control module, which are connected to the industrial Ethernet platform to realize early warning and emergency linkage.

[0026] IV. System Interactive Control Process The system's interaction adopts a full-link collaborative mechanism of "downlink command hierarchical control and uplink data hierarchical aggregation," which is divided into four core stages, ensuring the closed-loop and real-time nature of the interaction logic throughout the process.

[0027] 4.1 Terminal Perception and Edge Preprocessing like Figure 2 As shown, the specific implementation steps are as follows: Initialization self-test: After the system is powered on, the power management module of the edge intelligent acquisition terminal supplies power to all mounted sensors. The sensors complete self-test, sampling parameter configuration and protocol initialization, and report the ready status to the main control unit through the bus. The main control unit completes the initialization of the FPGA module, NPU module and dual communication module, registers the terminal node with the hybrid transmission collaborative gateway, and synchronizes the link configuration parameters.

[0028] Data Acquisition and Hardware Preprocessing: The main control unit sends acquisition commands to the sensors according to the preset acquisition strategy. The sensors acquire the equipment operating parameters in real time and convert them into digital signals, which are then transmitted to the interface module via the bus. The interface module forwards the raw data to the FPGA preprocessing module, which sequentially performs hardware second-order Butterworth low-pass filtering, adaptive power frequency interference suppression, and time-domain / frequency-domain feature extraction to complete data compression and noise reduction. The preprocessed feature data is then sent back to the main control unit.

[0029] Edge computing local inference: The main control unit inputs feature data into the lightweight CNN-LSTM anomaly detection model of the NPU module (see “Lightweight CNN-LSTM Anomaly Detection Algorithm on the Edge Side” below) to complete local real-time inference; if anomaly features are detected, the acquisition strategy is immediately upgraded, switching from the normal mode to the warning / fault mode, increasing the sensor sampling frequency, and marking the abnormal data with high priority; the inference results are synchronously sent back to the main control unit for data encapsulation and transmission scheduling.

[0030] Data caching and encapsulation: The main control unit encapsulates the preprocessed data and inference results in a unified format (including device ID, sensor ID, timestamp, data type, priority flag, and CRC check bit) and stores it in the local eMMC module; local caching is implemented in the network outage state, and automatic network outage resume transmission is triggered after the network is restored to ensure data integrity.

[0031] 4.2 Hybrid Transmission Control Link status monitoring: The LoRa and 5G modules of the edge intelligent acquisition terminal collect link communication quality parameters (LoRa: RSSI, SNR, packet loss rate, latency; 5G: RSRP, RSRQ, available bandwidth, latency, packet loss rate) every 100ms and report them to the hybrid transmission collaborative gateway through the control channel; the gateway's transmission scheduling engine receives and updates the link status data of all network nodes in real time.

[0032] Intelligent traffic offloading and scheduling: The scheduling engine uses an intelligent scheduling algorithm based on fuzzy hierarchical analysis to dynamically generate traffic offloading decisions based on data packet priority, data volume, and real-time link status, and then sends these decisions to the dual communication modules of the edge terminals. Low-priority routine data: scheduled for transmission via LoRa link to reduce terminal power consumption and save 5G bandwidth; Medium-priority early warning data: Dual-link redundant transmission ensures reliable data transmission; High-priority faults and high-frequency data: scheduled to 5G link for transmission, allocated with dedicated network slices to ensure low-latency, high-bandwidth transmission.

[0033] Seamless link switching: When the scheduling engine detects that the quality of the current working link is lower than the preset threshold, it immediately triggers a link switching command and sends it to the edge terminal to seamlessly switch the data stream to the backup link with a switching latency of ≤50ms. During the switching process, the data packets are encoded with consecutive sequence numbers to avoid data loss, duplication and out-of-order, and solve the problem of transmission interruption in the port coverage blind spot.

[0034] Data forwarding and command issuance: After receiving dual-link data, the gateway completes decapsulation, CRC verification, and protocol conversion, and converts LoRaWAN / 5G PDU data into MQTT protocol and forwards it to the cloud data platform; at the same time, it receives control commands issued by the platform, converts them into the corresponding link protocol format, and issues them to the edge intelligent acquisition terminal. After the terminal executes the command, it sends the execution result back to the platform.

[0035] 4.3 Platform Service Layer Control Data platform processing: The cloud-based data platform receives data forwarded by the gateway and sequentially completes data cleaning, missing value completion, normalization processing, and multi-sensor timestamp synchronization and alignment. It stores structured data in a time-series database and unstructured raw data in object storage. At the same time, it pushes standardized feature data to the intelligent diagnosis and prediction engine in real time.

[0036] Intelligent diagnostic reasoning: The intelligent diagnostic engine calls the CNN-LSTM-Attention multi-task fusion model to perform deep extraction of spatial, temporal and key features from the input feature matrix, and performs feature matching with the port and shipping equipment-specific fault feature library to output fault type, fault location, fault severity and remaining service life prediction value; the diagnostic results are synchronously stored in the equipment full life cycle management module and fault feature library.

[0037] Warning and control commands: When the diagnostic results exceed the preset warning threshold, the engine immediately generates a warning / alarm command of the corresponding level. On the one hand, it is pushed to the application interaction layer, and on the other hand, it is sent to the hybrid transmission and collaboration gateway through the data platform, and finally delivered to the edge intelligent acquisition terminal to trigger actions such as sampling strategy upgrade and device emergency linkage.

[0038] Model Iteration and Update: The platform regularly retrains and optimizes the diagnostic model based on newly added fault labeling data and operation and maintenance feedback data; the optimized lightweight model is then sent to the NPU module of the edge terminal via the 5G link to complete the remote OTA update of the terminal model, thereby realizing the continuous iterative improvement of the system's diagnostic capabilities.

[0039] 4.4 Application Interaction Layer Control Visualization: PCs and mobile devices can access the web server via HTTP / HTTPS protocol to obtain real-time operating status of the equipment, monitoring data curves, fault warning information, and health assessment reports. The visualization interface enables 3D modeling of the equipment, real-time data display, and visualization of fault location.

[0040] Manual control: Maintenance personnel issue manual control commands (adjust sampling frequency, start remote diagnostics, issue maintenance work orders, remote control of equipment) through the interactive interface. The commands are transmitted to the platform service layer via the web server, and then sent to the edge terminal through the hybrid transmission gateway. After the terminal executes the commands, it sends the results back to the interactive interface for display.

[0041] Early warning and emergency response linkage: The early warning / alarm information generated by the platform is notified to relevant personnel through interactive interface pop-ups, audible and visual alarms, SMS, APP push, etc. At the same time, it is linked with the port emergency management system to trigger emergency plans and realize emergency actions such as equipment shutdown and operation scheduling adjustment to ensure port production safety.

[0042] V. System Application Control Methods like Figure 3 As shown, this method covers the entire process of system deployment and initialization to closed-loop optimization. Each step clearly defines the implementation process, technical principles, and technical effects, ensuring that the system is feasible, replicable, and highly reliable.

[0043] Step 1: System Deployment and Initial Configuration Implementation process: Sensor deployment: Sensors are installed at designated locations in areas of port and shipping equipment prone to failure: Vibration and temperature / humidity sensors are deployed on the hoisting motors, reducers, and wheel boxes of quay cranes / gantry cranes; tilt sensors are deployed on the steel structure beams and gantry frames; pressure and level sensors are deployed on the hydraulic system; and current and voltage sensors are deployed on the electrical cabinets. The sensors are fixed with stainless steel brackets, and waterproof and sealed connectors are used at the wiring points. Shielded cables are kept away from sources of strong electromagnetic interference such as frequency converters and high-voltage cabinets.

[0044] Edge terminal deployment and configuration: Install one edge intelligent acquisition terminal on each device, complete the wiring and address configuration of all sensors, set sensor sampling parameters, communication protocols, acquisition strategy thresholds, and complete the burning and initialization of the terminal's local inference model.

[0045] Hybrid transmission network deployment: Deploy LoRa concentrators at high points in the port, configure frequency bands, spreading factors, and channel parameters to achieve full coverage of the port; complete the deployment of 5G base stations and UPF sinking nodes, configure dedicated network slices and QoS rules for the monitoring system; deploy a hybrid transmission collaborative gateway, complete the configuration of dual-network access parameters, scheduling rules, and protocol conversion rules, and achieve communication docking with edge terminals.

[0046] Platform and application layer deployment: Build a port private cloud cluster, deploy a data platform, intelligent diagnostic engine, and equipment lifecycle management module, construct a dedicated fault feature library for port and shipping equipment, and complete the deployment and parameter configuration of pre-trained models; deploy a multi-terminal visual management and control platform, complete the configuration of user permissions, equipment ledgers, early warning rules, and linkage strategies, and realize access authorization for operation and maintenance terminals.

[0047] By employing a layered and tiered deployment approach, and considering the fault characteristics of port and shipping equipment and the features of the port environment, the layout of sensor points and network coverage are optimized. This achieves seamless integration and parameter matching across the entire chain of perception, transmission, platform, and application, ensuring the comprehensiveness of system data collection and the reliability of transmission. The complete system chain construction achieves 100% coverage of port and shipping equipment monitoring points, eliminates blind spots in the transmission network, and completes system parameter configuration, enabling stable operation and laying the foundation for subsequent equipment health monitoring.

[0048] Step 2: Real-time acquisition and edge preprocessing of equipment operation data Implementation process: Normal state data acquisition: When the system is running normally, the edge intelligent acquisition terminal controls the sensor cluster to acquire data at a low frequency according to the normal acquisition strategy: vibration sensor sampling frequency 1kHz, temperature and humidity, electrical parameters sampling frequency 1 time / minute, and structural tilt angle sampling frequency 1 time / 10 minutes.

[0049] Hardware preprocessing: The raw sensor data is transmitted to the FPGA module. First, adaptive filtering is performed to remove 50Hz power frequency interference and its harmonics. Then, a second-order Butterworth low-pass filter is used to remove high-frequency random noise. Subsequently, time-domain features (mean, peak value, RMS value, kurtosis, margin factor) and frequency-domain features (spectral peak, centroid frequency, and mean square frequency extracted after FFT transformation) are extracted to achieve data noise reduction and compression.

[0050] Local anomaly identification and strategy upgrade: The main control unit inputs feature data into the NPU lightweight model to complete local anomaly inference; if the anomaly probability exceeds the threshold, it immediately switches to the early warning mode and increases the vibration sampling frequency to 5kHz; if a clear fault feature is identified, it switches to the fault mode and increases the sampling frequency to 20kHz to continuously collect raw vibration waveform data.

[0051] Data encapsulation and caching: The main control unit adds device identifiers, timestamps, priority markers and check bits to all data to complete standardized encapsulation, stores it in the local storage module, and waits for transmission scheduling.

[0052] Utilizing edge computing technology, data preprocessing and anomaly identification are completed on the terminal side, reducing the amount of raw data transmission. A hierarchical sampling strategy ensures complete capture of abnormal data. Hardware filtering and feature extraction effectively suppress strong interference in the port and shipping environment, improving the data signal-to-noise ratio. This achieves high-precision real-time acquisition of equipment operation data, with a data preprocessing compression ratio ≥60%, significantly reducing transmission bandwidth usage. Local anomaly identification response time is ≤100ms, enabling complete capture of fault data. Simultaneously, the terminal sleep mode current is ≤10μA, significantly extending battery life.

[0053] Step 3: Intelligent Scheduling and Data Transmission of 5G / LoRa Hybrid Transmission Implementation process: Real-time link status monitoring: The dual communication modules of the edge terminal collect link quality parameters every 100ms and report them to the hybrid transmission collaborative gateway. The scheduling engine updates the entire network link status database in real time.

[0054] Dynamic intelligent traffic offloading and scheduling: The scheduling engine is based on an intelligent scheduling algorithm (see "Intelligent scheduling algorithm for 5G / LoRa hybrid transmission based on fuzzy hierarchical analysis" below) to generate traffic offloading decisions based on data packet priority, data volume, and link status: small data packets with normal priority are scheduled to LoRa links and allocated dedicated channels and time slots; data packets with warning priority are transmitted with dual-link redundancy; and large data packets with fault priority are scheduled to 5G links and allocated dedicated network slices and high QoS priority.

[0055] Seamless link switching: When the current link quality is detected to be below the threshold (LoRa: RSSI<-130dBm, packet loss rate>10%; 5G: RSRP<-110dBm, latency>50ms), link switching is triggered immediately, with a switching latency of ≤50ms, ensuring uninterrupted data flow.

[0056] Data verification and forwarding: After receiving data from both links, the gateway completes decapsulation, verification, deduplication, and protocol conversion, and forwards the valid data to the cloud data platform via the MQTT protocol. At the same time, it completes the encapsulation and issuance of downlink control commands.

[0057] Based on the complementary characteristics of 5G and LoRa technologies, fuzzy hierarchical analysis is used to achieve dynamic data stream splitting and seamless link switching. This fully leverages the advantages of LoRa's low power consumption and wide coverage, and 5G's large bandwidth and low latency, resolving the contradiction between transmission coverage, power consumption, and bandwidth in the complex environment of ports. This enables data transmission without blind spots throughout the port area, achieving a transmission success rate of ≥99.99%, reducing power consumption for regular data transmission by more than 80%, reducing latency for fault data transmission by ≤10ms, and improving bandwidth utilization by more than 70%, thus breaking through the performance bottleneck of single transmission technologies.

[0058] Step 4: Cloud-based data processing and intelligent fault diagnosis and lifespan prediction Implementation process: Cloud-based data standardization processing: After receiving the data, the data platform uses the 3σ criterion to remove outliers, linear interpolation to fill in missing values, and min-max normalization to map the feature data to the [0,1] interval, thereby completing the synchronization and alignment of multi-sensor data timestamps and generating a standardized feature matrix.

[0059] Deep feature extraction and fault diagnosis: The feature matrix is ​​input into the CNN-LSTM-Attention fusion model. Spatial features are extracted through the CNN layer, temporal features are extracted through the Bi-LSTM layer, and the Attention layer assigns high weights to key fault features, outputting fused features. The model is then matched with the fault feature library to output the fault type, location, and severity.

[0060] Remaining useful life prediction: Through the RUL prediction branch of the model, the remaining useful life of key components of the equipment is output, and a health status assessment report is generated.

[0061] Warning Trigger: The diagnostic and prediction results are compared with preset thresholds. If the threshold is exceeded, a warning / alarm message of the corresponding level is generated, triggering the emergency response rules.

[0062] By integrating the spatial feature extraction capabilities of CNNs, the long-term dependency capture capabilities of LSTMs, and the key feature focusing capabilities of Attention, a multi-task deep learning model is constructed to solve the challenge of feature extraction from non-stationary and strongly coupled data of port and shipping equipment. Combined with a dedicated fault feature library, it enables accurate fault diagnosis and early warning. It can provide early warning of equipment faults 7 to 30 days in advance, solving the problems of low fault recognition rate and poor generalization ability of traditional algorithms with small sample sizes.

[0063] Step 5: Early Warning Issuance, Operation and Maintenance Collaboration, and System Closed-Loop Optimization Implementation process: Tiered early warning issuance: The platform generates Level 1 (attention), Level 2 (inspection), and Level 3 (emergency shutdown) early warning information based on the severity of the fault, and notifies the corresponding operation and maintenance and dispatch personnel through visual interfaces, audible and visual alarms, SMS, and APP push notifications.

[0064] Operation and maintenance closed-loop management: Operation and maintenance personnel can view fault details, historical data, and operation and maintenance guidelines through the platform, and generate and issue operation and maintenance work orders; on-site operation and maintenance personnel can receive work orders through industrial tablets, and after completing the repair, upload the repair results and fault handling status to the platform, and update the equipment ledger and operation and maintenance records.

[0065] Model Iteration and Optimization: The platform adds the original data, feature data, diagnostic results, and maintenance feedback of this fault to the fault feature library as new samples for model training. It regularly retrains and optimizes the parameters of the diagnostic model to improve the model's diagnostic accuracy.

[0066] Terminal model update and strategy optimization: The optimized lightweight model is delivered to the edge terminal via the 5G link to complete the local model OTA update; at the same time, based on the device's full life cycle data, the collection strategy, early warning threshold and operation and maintenance plan are optimized to achieve refined management of the device's full life cycle.

[0067] A closed-loop management mechanism of "data collection - analysis and diagnosis - early warning and maintenance - data feedback - model optimization" is constructed to enable the system to self-learn and self-optimize, continuously improving system performance and upgrading the equipment management model from passive maintenance to preventive maintenance. This achieves closed-loop management of equipment failures, reducing maintenance response time by more than 60%, decreasing the rate of sudden equipment failures by more than 70%, extending equipment lifespan by more than 15%, and continuously improving system diagnostic accuracy with operation, fully adapting to the multi-condition operating requirements of port and shipping equipment.

[0068] VI. Detailed Explanation of Core Algorithms The core algorithms of this solution are deeply integrated with hardware and software systems and application steps, ensuring the feasibility and technological innovation of the algorithms.

[0069] 6.1 Intelligent Scheduling Algorithm for 5G / LoRa Hybrid Transmission Based on Fuzzy Hierarchical Analysis This algorithm runs on the FPGA acceleration module and main control CPU of the hybrid transmission collaborative gateway, corresponding to system application step 3. The core implementation is dynamic intelligent data flow splitting and optimal link selection, avoiding the technical defects of fixed splitting in existing hybrid transmission.

[0070] Core objective function The algorithm aims to maximize the overall transmission efficiency of the link, and the objective function is:

[0071] The operating mechanism is as follows: This formula calculates the comprehensive transmission utility value of LoRa and 5G links by weighted summation of four core evaluation indicators of the link, and selects the link with the greater utility value as the transmission link of the current data stream; at the same time, it dynamically adjusts the weight coefficients according to the priority of data packets to achieve the optimal balance of transmission performance, power consumption and latency.

[0072] Definitions of each term in the formula: The maximum value operator aims to find the maximum total value. The largest transmission link; : Link overall transmission utility value, normalized value range [0,1], the larger the value, the better the link overall transmission performance; The weighting coefficients of each evaluation indicator satisfy the constraints. It is calculated using fuzzy hierarchical analysis and dynamically adjusted according to data packet priority. : Link transmission quality index, normalized value [0,1], calculation formula is as follows ,in Real-time packet loss rate of the link; Link power consumption metric, normalized to [0,1], calculated using the following formula: ,in This represents the current transmit power of the communication module. This represents the module's maximum transmit power. : Link transmission rate index, normalized value [0,1], calculation formula is as follows ,in For the real-time available bandwidth of the link, This represents the maximum bandwidth of the link. Link transmission delay metric, normalized value [0,1], calculation formula is as follows: ,in For the real-time end-to-end latency of the link, This represents the maximum acceptable latency for the system.

[0073] Weight dynamic adjustment rules Regular priority data packets: Focus on optimizing power consumption and transmission quality, and weight allocation. , , , LoRa links are preferred. Priority alert data packets: Focus on optimizing transmission reliability and latency, and weight allocation. , , , It employs dual-link redundant transmission; Fault priority data packets: Focus on optimizing transmission rate and latency, and weight allocation. , , , 5G links should be given priority.

[0074] 6.2 Lightweight CNN-LSTM Anomaly Detection Algorithm for Edge Side This algorithm is deployed on the NPU module of the edge intelligent acquisition terminal, corresponding to step 2 of the system application. Its core function is to quickly identify local anomalies on the terminal side. It uses depthwise separable convolution to achieve lightweight model and adapt to the limited computing power on the edge side.

[0075] Algorithm Network Structure The algorithm consists of an input layer, a lightweight CNN convolutional layer, an LSTM layer, a fully connected layer, and an output layer. The input is the preprocessed temporal feature matrix. ,in For timing length, For feature dimension, For the real number field, This indicates a belonging relationship.

[0076] 1. Lightweight CNN Convolutional Layers Depthwise separable convolution is used instead of traditional convolution to reduce computational cost. The convolution operation formula is as follows:

[0077] The operating mechanism is as follows: Depth-separable convolution is divided into two steps: the first step is depthwise convolution (dw), which performs convolution on each input channel individually to extract spatial features within the channel; the second step is pointwise convolution (pw), which achieves channel feature fusion through 1×1 convolution. Compared with traditional convolution, the number of parameters and computation is reduced by more than 80%, which is suitable for edge computing power constraints.

[0078] Definitions of each term in the formula: Feature maps output by convolutional layers; The ReLU activation function has the following formula: It is used to introduce nonlinear transformations to improve the model's feature representation ability; The kernel weight matrix of a depthwise convolution, with a size of [size missing]. Each input channel corresponds to one convolutional kernel; The convolution kernel weight matrix for pointwise convolution, with size . ,in Input the number of channels. Number of output channels; : Two-dimensional convolution operator; : Input feature matrix to convolutional layer; : Convolutional layer bias term, used to adjust the numerical distribution of feature maps.

[0079] 2. LSTM layer Used to extract temporal features, capture the temporal evolution of device states, and solve the gradient vanishing problem in traditional RNNs. Core unit calculation formula: , The operating mechanism is as follows: LSTM uses three gate structures—forget gate, input gate, and output gate—to control the forgetting, updating, and output of time-series information, effectively capturing the dependencies of long-term time-series data and achieving accurate extraction of the time-series evolution patterns of device states.

[0080] Definitions of each term in the formula: At the current moment, The previous moment; The output of the forget gate, after activation of the Sigmoid function, takes a value of [0,1], controlling the state of the cell at the previous time step. The degree of information forgetting is 0 for complete forgetting and 1 for complete retention. The input gate output, after activating the Sigmoid function, takes a value of [0,1], controlling the candidate state at the current time. The degree of information retention; : The current candidate state of the cell, the value of tanh after activation is [-1,1]; The current cell state is the core memory unit of the LSTM, storing temporal feature information; The output gate outputs a value [0,1] after Sigmoid activation, controlling the current cell state. The degree of information output; The output of the hidden layer at the current time, i.e., the temporal features extracted by the LSTM layer; : These are the weight matrices for the forget gate, input gate, candidate state, and output gate, respectively; These are the bias terms for the corresponding gating; The Sigmoid activation function has the following formula: The values ​​are mapped to the [0,1] interval to achieve gate control; Hyperbolic tangent activation function, formula is: This maps the values ​​to the interval [-1, 1]. : Hadamard product operator, which multiplies corresponding elements of two matrices; Tensor concatenation operation concatenates the hidden layer output from the previous time step with the input features from the current time step, and uses the result as the input to the gating unit.

[0081] 3. Output Layer The fully connected layer fuses the temporal features output by the LSTM and then outputs the anomaly probability through the Sigmoid activation function. ,when When the preset threshold is reached, the device is deemed to be in an abnormal state, triggering an upgrade of the data collection strategy.

[0082] 6.3 Cloud-based CNN-LSTM-Attention Fault Diagnosis and RUL Prediction Algorithm This algorithm is deployed on the platform service layer computing server cluster, corresponding to system application step 4. It is a multi-task fusion deep learning model that shares the feature extraction backbone and simultaneously realizes fault classification diagnosis and remaining service life prediction.

[0083] Shared Feature Extraction Backbone The input is the normalized multi-sensor time-series feature matrix. ,in For timing length, The feature dimension is consistent with the edge-side feature dimension to ensure contextual consistency.

[0084] CNN Convolutional Layers: 3 traditional convolutional layers extract local spatial features of device faults and output spatial feature maps. , Number of output channels; Bi-LSTM layer: Two bidirectional LSTM layers simultaneously capture the temporal dependencies of both the forward and backward directions, outputting temporal features. , For the hidden layer dimension; Multi-head attention layer: Weights are assigned to temporal features, focusing on key fault features and suppressing irrelevant interference. Core formula: ， The operating mechanism is as follows: The multi-head attention mechanism uses multiple sets of query (Q), key (K), and value (V) matrices to compute eight attention heads in parallel, capturing fault features of different dimensions and assigning higher weights to key features related to faults, thus significantly improving the model's ability to identify early faults and small-sample faults.

[0085] Definitions of each term in the formula: These are the query matrix, key matrix, and value matrix, respectively, derived from the input features. Obtained through linear transformation; : Key matrix The dimension is used to scale the dot product result to avoid the gradient of the softmax function disappearing due to excessively large values; : Matrix transpose operator; The normalized exponential function maps attention scores to the [0,1] interval, with a total sum of 1, to obtain the weight coefficient of each feature. : No. The output of each attention head, For the total number of attention heads; Tensor concatenation operation concatenates the outputs of multiple attention heads along the feature dimension; : Output layer weight matrix, which maps the concatenated multi-head features to the target output dimension.

[0086] Attention layer outputs fused feature matrix It serves as a shared input for both the fault classification and RUL prediction branches.

[0087] Fault Classification Branches Fusion features After global average pooling, the input consists of two fully connected layers, and the output layer uses the Softmax activation function to output the probability distribution of fault types. , The final diagnosis result is the category corresponding to the highest probability of common fault types (including normal conditions) in port and shipping equipment.

[0088] Softmax function formula: , in For the fully connected layer The output of each node, For the first The probability of a fault class is equal to the sum of the probabilities of all classes, which is 1.

[0089] RUL Prediction Branch Fusion features The input consists of two fully connected layers, and the output layer uses a linear activation function to predict the remaining useful life of key components of the device. The unit is hours.

[0090] Multi-task loss function The joint loss function is used to optimize both tasks simultaneously, as shown in the formula:

[0091] in: Cross-entropy loss function, used to optimize fault classification accuracy; Mean squared error (MSE) loss function, used to optimize RUL prediction accuracy; , Weighting coefficients balance the loss contributions of the two tasks.

[0092] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 4 As shown, optionally, electronic device 410 may include a first processor 2001.

[0093] Optionally, the electronic device 410 may also include a memory 2002 and a transceiver 2003.

[0094] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0095] The following is combined Figure 4 A detailed description of each component of electronic device 410 is provided below: The first processor 2001 is the control center of the electronic device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0096] Optionally, the first processor 2001 can perform various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0097] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0098] In a specific implementation, as one example, the electronic device 410 may also include multiple processors, for example... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0099] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0100] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0101] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0102] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0103] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0104] It should be noted that, Figure 4 The structure of the electronic device 410 shown does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0105] Furthermore, the technical effects of the electronic device 410 can be referred to the technical effects of the methods described in the above method embodiments, and will not be repeated here.

[0106] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0107] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0108] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0109] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0110] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0111] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0112] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0113] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0114] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0117] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A health monitoring system for port and shipping equipment based on a hybrid transmission of 5G and LoRa, characterized in that, The system adopts a five-layer integrated architecture of terminal-edge-network-cloud-application, including a terminal perception layer, a hybrid transmission layer, a platform service layer, and an application interaction layer, wherein: The terminal perception layer includes a high-precision anti-interference sensor cluster adapted to the harsh working conditions of ports and shipping, and an edge intelligent acquisition terminal that is matched with the sensor cluster. The high-precision anti-interference sensor cluster is used to collect the operating parameters of port and shipping equipment. The edge intelligent acquisition terminal integrates an FPGA preprocessing unit, an edge computing unit with NPU computing power, and a LoRa+5G dual-mode communication module to realize edge preprocessing of collected data, local anomaly identification, and dual-link data transmission and reception. The hybrid transmission layer includes a LoRa low-power wide-area subnet, a 5G transmission subnet, and a hybrid transmission collaborative gateway. The LoRa low-power wide-area subnet is used to achieve low-power wide-coverage data transmission throughout the port. The 5G transmission subnet is configured with a port-specific network slice to achieve high-bandwidth, low-latency data transmission. The hybrid transmission collaborative gateway has a built-in intelligent scheduling engine to achieve dynamic intelligent traffic splitting, seamless switching, and protocol conversion between LoRa and 5G dual links. The platform service layer is deployed on a port private cloud cluster and includes a cloud data platform, an intelligent diagnosis and prediction engine, an equipment lifecycle management module, and a dedicated fault feature library for port and shipping equipment. The intelligent diagnosis and prediction engine is equipped with a multi-task fusion deep learning model to realize fault classification diagnosis and remaining service life prediction of port and shipping equipment. The application interaction layer includes a multi-terminal visualization management and control platform, an early warning and emergency response module, and an operation and maintenance management module, which are used to realize the visualization display of equipment operation data, the release of hierarchical early warnings, the closed-loop management of operation and maintenance work orders, and the remote control of equipment.

2. The system according to claim 1, characterized in that, All sensors in the high-precision anti-interference sensor cluster have an IP68 protection rating, and are equipped with a built-in anti-salt spray corrosion coating, an electromagnetic shielding cavity, and a wide-voltage power management module. The operating temperature range is -40℃ to 85℃. The sensor cluster includes core operating condition monitoring sensors and auxiliary parameter monitoring sensors. The core operating condition monitoring sensors include vibration acceleration sensors, acoustic emission sensors, and tilt sensors. The auxiliary parameter monitoring sensors include temperature and humidity sensors, hydraulic system pressure / level sensors, motor current / voltage sensors, and switch status sensors.

3. The system according to claim 1 or 2, characterized in that, The edge intelligent acquisition terminal is deployed independently for each port and shipping equipment and installed in the equipment protection and control cabinet. It includes a main control unit, a multi-protocol interface module, an FPGA preprocessing module, an edge computing unit, a dual-mode communication module, and a power supply and storage module. The main control unit uses an STM32H743IIT6 microcontroller with a main frequency of 400MHz and an FPU floating-point unit, which is used for task scheduling, logic control and data interaction of the entire terminal. The multi-protocol interface module includes 8 RS485 interfaces, 2 CAN bus interfaces, 4 analog input interfaces, and 2 digital input interfaces. It is compatible with Modbus-RTU, CANopen, and Profibus-DP industrial protocols and is used to connect to sensor clusters. The FPGA preprocessing module uses an XC7A35T chip to achieve high-speed real-time data sampling, hardware digital filtering, time-domain / frequency-domain feature extraction, and complete edge data preprocessing. The edge computing unit has a built-in NPU computing power module with a total computing power of 1 TOPS, which is used to run a lightweight fault diagnosis model to achieve rapid identification of anomalies on the terminal side. The dual-mode communication module includes a LoRa module and a 5G module. The LoRa module uses the SX1262 chip and operates in the 470~510MHz unlicensed frequency band. The 5G module uses the RM500U-CN module, which supports SA / NSA dual-mode and is used to realize dual-link data transmission and reception. The power and storage module supports wide voltage DC12~48V input, has a built-in lithium battery backup power supply with a battery life of ≥72h, and is equipped with 8GB eMMC flash memory for local data caching and offline data transfer resumption.

4. The system according to claim 1, characterized in that, The LoRa low-power wide-area subnet adopts a star topology, including LoRa modules built into edge intelligent acquisition terminals as terminal nodes, and multiple LoRa concentrators. Each LoRa concentrator uses an 8-channel SX1302 chip, with a maximum receiving sensitivity of -142dBm and a coverage radius of ≥5km. 3-5 concentrators are deployed throughout the port area to achieve blind-spot-free coverage. The LoRa low-power wide-area subnet uses an optimized LoRaWAN protocol and is configured with a custom time-slot ALOHA channel access mechanism to avoid data collisions between multiple nodes.

5. The system according to claim 1 or 4, characterized in that, The 5G transmission subnet includes 5G modules built into edge intelligent acquisition terminals as terminal nodes, port 5G macro base stations / indoor distributed base stations, and 5G core network UPF sinking nodes. The 5G transmission subnet allocates dedicated network slices and independent QoS priorities to the monitoring system to achieve local data offloading and ensure data security and low-latency transmission. The 5G terminal nodes are connected to the base stations through the 5G NR air interface, the base stations are connected to the UPF sinking nodes through optical fibers, and the UPF nodes are connected to the hybrid transmission collaborative gateway through 10 Gigabit optical fibers, while configuring redundant backup links for the public network VPN.

6. The system according to claim 1, 4, or 5, characterized in that, The hybrid transmission collaborative gateway adopts an industrial-grade rack server, configured with an Intel Xeon D-1718T 8-core 16-thread CPU, 32GB DDR4 memory, 2TB SSD storage, dual power supply redundancy, and built-in 4 gigabit Ethernet ports, 2 10-gigabit optical ports, and an FPGA hardware acceleration module. The uplink of the hybrid transmission collaborative gateway connects to the port industrial Ethernet platform service layer via 10-gigabit fiber optic cable, and the downlink connects to the LoRa concentrator via gigabit Ethernet and to the 5G UPF downlink node via 10-gigabit fiber optic cable. The intelligent scheduling engine runs an intelligent scheduling algorithm based on fuzzy hierarchical analysis, with the objective of maximizing the overall transmission utility of the link. The objective function is: Definitions of each term in the formula: The maximum value operator aims to find the maximum total value. The largest transmission link; : Link overall transmission utility value, normalized value range [0,1], the larger the value, the better the link overall transmission performance; The weighting coefficients of each evaluation indicator satisfy the constraints. It is calculated using fuzzy hierarchical analysis and dynamically adjusted according to data packet priority. : Link transmission quality index, normalized value [0,1], calculation formula is as follows ,in Real-time packet loss rate of the link; Link power consumption metric, normalized to [0,1], calculated using the following formula: ,in This represents the current transmit power of the communication module. This represents the module's maximum transmit power. : Link transmission rate index, normalized value [0,1], calculation formula is as follows ,in For the real-time available bandwidth of the link, This represents the maximum bandwidth of the link. Link transmission delay metric, normalized value [0,1], calculation formula is as follows: ,in For the real-time end-to-end latency of the link, This represents the maximum acceptable latency for the system.

7. The system according to claim 1, characterized in that, The edge computing unit of the edge intelligent acquisition terminal is equipped with a lightweight CNN-LSTM anomaly detection algorithm on the edge side. The algorithm network structure includes, in sequence, an input layer, a lightweight CNN convolutional layer, an LSTM layer, a fully connected layer, and an output layer. The input to the input layer is the preprocessed temporal feature matrix. Where T=64 is the time series length and F=12 is the feature dimension; The lightweight CNN convolutional layer uses depthwise separable convolution instead of traditional convolution, and the convolution operation formula is as follows: Definitions of each term in the formula: Feature maps output by convolutional layers; The ReLU activation function has the following formula: It is used to introduce nonlinear transformations to improve the model's feature representation ability; The kernel weight matrix of a depthwise convolution, with a size of [size missing]. Each input channel corresponds to one convolutional kernel; The convolution kernel weight matrix for pointwise convolution, with size . ,in Input the number of channels. Number of output channels; : Input feature matrix to convolutional layer; : Convolutional layer bias term, used to adjust the numerical distribution of feature maps; The LSTM layer is used to extract temporal features; The output layer outputs anomaly probability through a Sigmoid activation function. ,when When the preset threshold is reached, the device is deemed to be in an abnormal state, triggering an upgrade of the data collection strategy.

8. The system according to claim 1, characterized in that, The intelligent diagnosis and prediction engine is equipped with a multi-task fusion deep learning model, which is a CNN-LSTM-Attention multi-task fusion model. The model includes a shared feature extraction backbone, a fault classification branch, and a RUL prediction branch, and is optimized using a multi-task joint loss function. The shared feature extraction backbone consists of a CNN convolutional layer, a Bi-LSTM layer, and a multi-head attention layer, with the input being a normalized multi-sensor temporal feature matrix. Where T=256 is the temporal length and F=12 is the feature dimension; the CNN convolutional layer consists of 3 traditional convolutional layers, used to extract the spatial local features of device faults; The Bi-LSTM layer is a two-layer bidirectional LSTM layer used to capture the temporal dependencies between the forward and backward directions. The multi-head attention layer is used for weight allocation of temporal features, and the core formula is: ， Definitions of each term in the formula: These are the query matrix, key matrix, and value matrix, respectively, derived from the input features. Obtained through linear transformation; : Key matrix The dimension is used to scale the dot product result to avoid the gradient of the softmax function disappearing due to excessively large values; : Matrix transpose operator; The normalized exponential function maps attention scores to the [0,1] interval, with a total sum of 1, to obtain the weight coefficient of each feature. : No. The output of each attention head, For the total number of attention heads; Tensor concatenation operation concatenates the outputs of multiple attention heads along the feature dimension; : Output layer weight matrix, which maps the concatenated multi-head features to the target output dimension; Attention layer outputs fused feature matrix This serves as a shared input for both the fault classification and RUL prediction branches. The fault classification branch performs global average pooling on the fusion features output by the multi-head attention layer, inputs them into two fully connected layers, and outputs the fault type probability distribution using the Softmax activation function. The category corresponding to the maximum probability is taken as the final diagnosis result. The RUL prediction branch inputs the fused features output from the multi-head attention layer into two fully connected layers. The output layer uses a linear activation function to output the predicted remaining lifespan of key components of the device. The formula for the multi-task joint loss function is as follows: in: Cross-entropy loss function, used to optimize fault classification accuracy; Mean squared error (MSE) loss function, used to optimize RUL prediction accuracy; , Weighting coefficients balance the loss contributions of the two tasks.

9. The system according to claim 1, characterized in that, The port private cloud cluster of the platform service layer includes a computing server cluster, a distributed storage cluster, a web and application server, and security protection hardware; The computing server cluster is configured with two redundant high-performance servers, each with an Intel Xeon Gold 6330 CPU (32 cores, 64 threads) and 128GB of memory, used for training, inference, and task scheduling of the intelligent diagnostic model. The distributed storage cluster is configured with 4 storage servers with a total capacity of 192TB. It deploys the distributed time-series database InfluxDB and the object storage MinIO to store structured monitoring data and unstructured raw waveforms, model files, and fault characteristic data, respectively. The Web and application server is configured with two redundant servers for service publishing, interface management, and business logic processing of the application interaction layer. The security hardware includes an industrial-grade firewall, an intrusion detection system, and a data encryption machine, which are used to ensure the security of the platform network and data. All servers are clustered together via 10 Gigabit Ethernet switches and connected to a hybrid transmission and collaboration gateway via a firewall to achieve end-to-end data communication.

10. A method for health sensing and monitoring of port and shipping equipment based on hybrid transmission of 5G and LoRa, characterized in that, The method for the port and shipping equipment health sensing and monitoring system based on 5G and LoRa hybrid transmission, as described in any one of claims 1-9, comprises the following steps: System deployment and initialization configuration: High-precision anti-interference sensor clusters were deployed and fixed at locations prone to failure in port and shipping equipment. An independent edge intelligent acquisition terminal was deployed for each piece of port and shipping equipment, completing sensor wiring, address configuration, and acquisition strategy settings. A LoRa low-power wide-area subnet, a 5G transmission subnet, and a hybrid transmission collaborative gateway were deployed to achieve full port network coverage and intelligent scheduling rule configuration. A port private cloud cluster was built, deploying the platform service layer and application interaction layer, completing the construction of a dedicated fault feature library for port and shipping equipment, deployment of pre-trained models, and matching of system parameters across the entire chain. The system performs real-time acquisition and edge preprocessing of equipment operation data. During normal operation, the edge intelligent acquisition terminal controls the sensor cluster to acquire port and shipping equipment operation parameters according to the conventional acquisition strategy. The raw data is transmitted to the FPGA preprocessing module to complete hardware filtering, time domain and frequency domain feature extraction. The main control unit inputs the preprocessed feature data into the lightweight anomaly recognition model of the edge computing unit to complete local real-time inference. If an abnormal feature is identified, the acquisition strategy is immediately upgraded to increase the sensor sampling frequency. The main control unit completes standardized encapsulation and local caching of all data, waiting for transmission scheduling. Intelligent scheduling and data transmission for 5G / LoRa hybrid transmission: The dual communication modules of the edge intelligent acquisition terminal collect link communication quality parameters in real time and report them to the hybrid transmission collaborative gateway. The gateway's intelligent scheduling engine is based on the intelligent scheduling algorithm of fuzzy hierarchical analysis. It generates traffic allocation decisions based on data packet priority, data volume, and real-time link status, and schedules data packets of different priorities to LoRa links, 5G links, or dual-link redundant transmission respectively. When the quality of the current working link is detected to be lower than a preset threshold, seamless link switching is triggered. After receiving dual-link data, the gateway completes decapsulation, verification, and protocol conversion, and forwards the valid data to the cloud data platform. Cloud-based data processing and intelligent fault diagnosis and lifespan prediction: After receiving data, the cloud data platform performs data cleaning, missing value completion, normalization, and multi-sensor timestamp synchronization and alignment to generate a standardized feature matrix, which is then pushed to the intelligent diagnosis and prediction engine. The intelligent diagnosis and prediction engine calls the CNN-LSTM-Attention multi-task fusion model to complete deep feature extraction, matches the features with the port and shipping equipment's dedicated fault feature library, and outputs the fault type, fault location, fault severity, and remaining lifespan prediction. When the diagnosis result exceeds the preset warning threshold, a corresponding level of warning / alarm instruction is generated. The platform implements a system of early warning issuance, operation and maintenance coordination, and closed-loop optimization. It issues tiered early warning information based on the severity of faults, notifying relevant operation and maintenance and dispatch personnel through multiple channels. Simultaneously, it triggers emergency plans in conjunction with the port's emergency management system. Operation and maintenance personnel generate and issue work orders through the platform, upload repair results and fault handling information after completing on-site inspections, and update equipment ledgers and operation and maintenance records. The platform adds new fault data and operation and maintenance feedback data to a dedicated fault feature library for port and shipping equipment, regularly retrains and optimizes the diagnostic model, and distributes the optimized lightweight model to edge intelligent acquisition terminals for OTA updates. It also optimizes system acquisition strategies, early warning thresholds, and operation and maintenance plans to achieve refined management of the entire equipment lifecycle.

Images