Multi-robot control method and system based on industrial internet of things, and medium

Abstract

The invention relates to the technical field of industrial automation, in particular to a multi-robot control method and system based on the industrial Internet of Things and a medium. The method comprises the following steps: acquiring industrial warehouse channel data; generating a warehouse digital twin model based on the industrial warehouse channel data; performing multi-robot path conflict detection according to the warehouse digital twin model to obtain path conflict data; determining a robot number according to the path conflict data; performing motor winding damage detection based on the robot number to obtain motor winding damage data; performing iron core loosening analysis based on the motor winding damage data to obtain iron core loosening data; performing vibration suppression control optimization according to the iron core loosening data to obtain vibration suppression control data; and robot scheduling is carried out according to the path conflict data to obtain robot scheduling data. According to the invention, the path planning precision and fault response efficiency of the multi-robot system are improved based on the industrial automation technology, and the operation efficiency of industrial warehouse operation is improved.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to a multi-robot control method, system and medium based on the Industrial Internet of Things. Background Technology

[0002] Traditional multi-robot control methods suffer from coarse-grained path conflict detection, often relying solely on static maps or single-task path planning. This lack of comprehensive analysis of actual warehouse dynamics, aisle load, and real-time robot behavior leads to delayed and inaccurate path conflict detection, easily causing robot collisions or task congestion. Robot status monitoring relies too heavily on simple operational data statistics, such as task completion rate and current load changes, making it difficult to capture potential faults at the underlying hardware level, such as microscopic damage like motor winding burnout, core loosening, or bearing surface peeling. This prevents early fault prediction and intervention. Scheduling optimization and fault control logic are disconnected, failing to form a collaborative control mechanism with a closed loop of perception-evaluation-feedback. In dense multi-robot operation, it is difficult to coordinate the exit of faulty robots and the rescheduling of healthy robots. Traditional systems lack deep perception and modeling methods for changes in multiple physical fields such as electromagnetic, thermal, and mechanical fields, making it difficult to achieve refined diagnosis and control strategy optimization for core robot components such as motors and bearings. Ultimately, this affects the stability and task execution efficiency of the entire industrial system. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a multi-robot control method based on the Industrial Internet of Things to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a multi-robot control method based on the Industrial Internet of Things (IIoT) includes the following steps: Step S1: Acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data; Step S2: Determine the robot number based on the path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. Step S3: Perform robot scheduling based on path conflict data to obtain robot scheduling data; perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. Step S4: Perform emergency braking control based on bearing fault data to obtain emergency braking control data; transmit the emergency braking control data and vibration suppression control data to the robot industrial IoT management platform to execute multi-robot control tasks.

[0005] This invention constructs a digital twin model based on industrial warehouse aisle data, enabling real-time reflection of dynamic changes in the warehouse environment and complex aisle topology. Overcoming the limitations of traditional static maps, it achieves accurate detection of multi-robot path conflicts, effectively avoiding conflict omissions and delays in path planning, and improving robot safety and task continuity. Combined with robot identification numbers, it performs in-depth damage detection on motor windings, down to the microscopic level, enabling early identification of potential faults such as winding erosion and core loosening, providing a scientific basis for equipment maintenance. Based on core loosening data, it optimizes vibration suppression control, reducing the impact of equipment vibration on performance and extending the lifespan of core components by specifically adjusting control parameters, while ensuring stable robot operation. Robot scheduling fully utilizes path conflict information to rationally arrange task execution order, dynamically simulating the robotic arm's handling process, accurately assessing the impact of load on mechanical components, and achieving comprehensive monitoring of bearing faults such as surface peeling and eccentricity. Emergency braking control, combined with bearing fault data, enables rapid response and safe isolation of faulty robots, reducing accident risks and ensuring the overall system's operational safety. Vibration suppression and emergency braking control data are uniformly transmitted to an industrial IoT management platform, enabling closed-loop feedback and collaborative optimization of multi-robot control tasks. This promotes deep integration of perception, evaluation, and execution, improving the intelligence level and task execution efficiency of multi-robot systems. The overall system integrates electromagnetic, thermal, and mechanical multi-physical field information, establishing a closed loop from the hardware level to the control strategy. This significantly enhances the health management capabilities and dynamic scheduling adaptability of robot equipment, effectively improving the stability, safety, and production efficiency of multi-robot systems in industrial warehouses.

[0006] Preferably, this specification also provides a multi-robot control system based on the Industrial Internet of Things (IIoT) for executing the multi-robot control method based on the IIoT described above. The multi-robot control system based on the IIoT includes: A multi-robot path conflict detection module is used to acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; and perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data. The vibration suppression control optimization module is used to determine the robot number based on path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; and optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. The bearing fault detection module is used to perform robot scheduling based on path conflict data to obtain robot scheduling data; to perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; and to perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. The multi-robot control module is used to perform emergency braking control based on bearing fault data to obtain emergency braking control data; the emergency braking control data and vibration suppression control data are transmitted to the robot industrial IoT management platform to execute multi-robot control tasks.

[0007] The present invention relates to a multi-robot control system based on the Industrial Internet of Things (IIoT). This system can implement any of the multi-robot control methods based on the IIoT of the present invention. It serves as a medium for coordinating the operation and signal transmission between various modules to complete the multi-robot control method based on the IIoT of the Industrial Internet of Things. The internal modules of the system cooperate with each other, improving the path planning accuracy and fault response efficiency of the multi-robot system, and significantly enhancing the safety and operational efficiency of industrial warehouse operations.

[0008] Optionally, this specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements any one of the multi-robot control methods based on the Industrial Internet of Things. Attached Figure Description

[0009] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the steps of the multi-robot control method based on the Industrial Internet of Things of the present invention. Figure 2 This is a detailed flowchart of step S1 in the present invention; Figure 3 This is a schematic diagram of the robotic arm handling process in this invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0010] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0011] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0012] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0013] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a multi-robot control method based on the Industrial Internet of Things (IIoT), the method comprising the following steps: Step S1: Acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data; In this embodiment, a LiDAR system (Velodyne VLP-16, 360° omnidirectional scanning, ranging accuracy ±3cm) and a high-definition camera (1920×1080 resolution, 30fps frame rate) are used to jointly acquire spatial point cloud data and image information of all passages in the industrial warehouse. The radar-acquired data undergoes point cloud preprocessing, including ground filtering, noise removal, and edge sharpening. After processing, the passage cross-section is divided into slices at 2m intervals to form a two-dimensional cross-sectional view. Boundary fitting operations are then performed on each cross-sectional view to extract geometric parameters such as passage width, height, and curvature. The passage width is limited to between 1.2m and 3.5m, and areas with a turning radius less than 1.8m are recorded as curve points. These parameters constitute the passage geometric structure set. A topology graph is constructed based on the geometric structure set. Nodes in the topology graph represent intersections or transformation points, and edges represent passage segments. The attributes of each edge include length (in meters), width (in meters), maximum number of robots it can accommodate (in units), and passage time (calculated based on a robot movement speed of 1.5m / s). Based on this topology, path simulation calculations are performed using a rule-driven approach. A FIFO queuing strategy is employed to simulate the trajectory of each robot from the starting point to the end point, recording the expected timestamp of each robot at each node. Based on path timestamp comparison, if the expected entry time difference between two robots on the same channel segment is less than 3 seconds and the distance between them is less than 1 meter, a path conflict is identified. The conflict data structure includes the conflicting robot number, conflicting channel number, conflict timestamp, and position coordinates. All conflict data is aggregated and output as path conflict data for subsequent control analysis.

[0014] Step S2: Determine the robot number based on the path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. In this embodiment, the conflict robot numbers included in the acquired path conflict data are used to extract the corresponding robot control board IDs and internal state storage units in the multi-robot control system. Using the control board (with an STM32F407 main controller) as the interface, the binding position data of the motor interface module is read to obtain the physical installation position coordinates of the motor (unit: mm). A thermal imager (Fluke Ti401 PRO, temperature measurement accuracy ±2°C) is used to perform infrared scanning on the motor end to identify overheated winding areas. Areas with abnormal temperatures exceeding 20°C from room temperature (with a set temperature rise threshold of 75°C) are identified as key detection targets. The motor winding wiring method is confirmed through disassembly and inspection. A digital multimeter (e.g., FLUKE 87V) is used to test the wiring resistance; a resistance below 0.3Ω indicates a delta connection, while a resistance above 0.8Ω indicates a star connection. For star connections, a portable metallographic microscope (1000× magnification) is used to inspect the copper wire surface. If charring, bubbles, or cracks are observed, and the measured ablation length exceeds 5mm, it is recorded as copper wire ablation. Copper busbar cross-sections are treated with chemical etching (using a 10% nitric acid alcohol solution). If a significant oxide layer thickness exceeding 0.02 mm is observed, it is recorded as copper busbar oxidation. For delta connections, residual arc burn marks are detected using a portable spectrometer (such as the Rigaku KT-100S). Burn mark areas larger than 5 mm² are considered to be present. 2 This indicates the presence of arc burn-off. By combining the copper busbar oxidation data and the arc burn-off data, a motor winding damage data structure is formed, which includes damage type, location, and damage level (classified as mild, moderate, and severe, corresponding to damage lengths of <10mm, 10-20mm, and >20mm, respectively).

[0015] Step S3: Perform robot scheduling based on path conflict data to obtain robot scheduling data; perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. In this embodiment, conflict locations recorded in the path conflict data are used to match specific coordinates in the warehouse topology map, and the robot numbers occupying the node simultaneously are analyzed. Combining the current task priority of each robot (set to levels 1-5, with 5 being the highest) and the remaining task distance (unit: m), task priority is adjusted to ensure high-priority robots have priority passage. After the scheduling sequence is output, it is transmitted to the robot navigation module (embedded ARM Cortex-A72) of the scheduling control system, where navigation commands are assigned sequentially. Using a dynamic simulation toolbox (e.g., MATLAB Simulink Robotics System Toolbox), the scheduled robot path and task information are input, the robotic arm load range is set to 2kg to 50kg, the grasping cycle to 3s-8s, and the running speed to 1.2m / s, simulating the entire handling process of the robotic arm grasping and placing, and outputting the robotic arm handling data for each task cycle, including end-effector position deviation (unit: mm), running speed, and acceleration (unit: m / s²). 2 The simulation results included the load curve of the rolling bearing region. Transport cycles with a curvature greater than the threshold of 0.05 / s were marked. Abnormal segments were recorded based on rotational errors (angle offset > 0.3°) collected by displacement sensors. A magnetic eddy current detector was used to detect spalling on the bearing surface. Areas with reflected signal amplitudes below 30dB and echo time delays exceeding 0.2ms were marked as spalling regions. The location, area, and depth of all spalling regions were recorded as bearing fault data.

[0016] Step S4: Perform emergency braking control based on bearing fault data to obtain emergency braking control data; transmit the emergency braking control data and vibration suppression control data to the robot industrial IoT management platform to execute multi-robot control tasks.

[0017] In this embodiment, targets with a spalling level of "severe" (i.e., spalling area exceeding 15% of the bearing surface area or spalling depth exceeding 0.3mm) are selected from the bearing fault data to determine that the robot has a critical hardware risk. The control system sends a braking command to the robot control board via the CAN bus. The control board controls the MOSFET switch to cut off the main drive power and triggers the electromagnetic brake (voltage 24V, delay <10ms) to perform an emergency braking operation. After braking, the system simultaneously determines the vibration source frequency band (main frequency band range between 120Hz and 280Hz) based on the core loosening data obtained in step S2, finds the bandpass damper parameters in the control strategy (center frequency ±10Hz, damping ratio set to 0.3), calls the structural response frequency domain simulation subroutine, inputs the control parameters and combines them with the motor structure stiffness matrix and impedance matrix, analyzes the system response through Discrete Fourier Transform (DFT), and generates vibration suppression control data. This data includes recommended mechanical fastening strategies, pressure plate adjustment direction, and specific locking force values ​​(unit: N), which are then transmitted to the main control server of the industrial IoT platform (configured as NVIDIA Jetson AGX Orin) for execution instruction scheduling.

[0018] Of particular importance, step S4 includes the following steps: Step S41: Assess the fault level based on the bearing fault data; In this embodiment, the collected bearing fault data undergoes signal preprocessing. A bandpass filter is used to filter out power supply interference at 50Hz and its harmonics, with the filter bandwidth set to 500Hz to 20kHz to ensure the extraction of valid fault signals. Subsequently, time-domain statistical methods are used to calculate the root mean square (RMS), peak power factor (PKF), kurtosis, and impulsiveness. The thresholds for RMS, PKF, kurtosis, and impulsiveness are set to 0.3g, PKF, 3, and 0.6, respectively. Frequency domain analysis employs Fast Fourier Transform (FFT) to calculate the amplitudes and peak frequencies of major harmonics within the frequency range, focusing on anomalous frequency peaks within the 1kHz to 10kHz range, with a threshold set to 0.2g. Based on a comprehensive weighted evaluation of time-domain and frequency-domain parameters, the weight allocation is: RMS 30%, PKF 25%, kurtosis 20%, impulsiveness 15%, and peak frequency 10%. A fault score is obtained through weighted calculation, ranging from 0 to 100. 0-20 represents normal faults, 21-40 represents minor faults, 41-60 represents moderate faults, 61-80 represents severe faults, and 81-100 represents extremely severe faults. Finally, the comprehensive score is mapped to a 0-5 level system, with level 1 corresponding to 0-20 points, level 2 to 21-40 points, level 3 to 41-60 points, level 4 to 61-80 points, and level 5 to 81-100 points. The fault level data is then output for subsequent use.

[0019] Step S42: Develop an emergency braking control strategy based on the fault level; In this embodiment, a mapping relationship is established between fault levels and braking control parameters. The initiation delay, braking force, and duration of emergency braking are adjusted according to the fault level. The initiation delay is the time from system detection of a fault to execution of the braking command: 500 milliseconds for level 1, 400 milliseconds for level 2, 250 milliseconds for level 3, 100 milliseconds for level 4, and 50 milliseconds for level 5. The braking force is defined as a percentage of the maximum braking force: 20% for level 1, 40% for level 2, 60% for level 3, 80% for level 4, and 100% for level 5. The braking duration is defined as the shortest time to execute the braking action: 200 milliseconds for level 1, 400 milliseconds for level 2, 600 milliseconds for level 3, 800 milliseconds for level 4, and 1000 milliseconds for level 5. Combining the current speed, acceleration, and load status of the robotic arm, the braking parameters are adjusted using a linear interpolation algorithm to ensure a continuous and smooth braking process. The control strategy data format includes a braking initiation timestamp (milliseconds), braking force percentage (integer 0-100), and duration (milliseconds), which are output as emergency braking control command parameters.

[0020] Step S43: Generate an emergency braking control command signal according to the emergency braking control strategy to obtain emergency braking control data; In this embodiment, the braking parameters generated in step S42 are encapsulated according to an industrial communication protocol, using a frame structure definition based on the CAN bus. The command frame length is fixed at 8 bytes, where 1 byte is the command identifier, 1 byte is the braking force percentage, 2 bytes are the start delay (in milliseconds, big-endian format), 2 bytes are the duration (in milliseconds, big-endian format), and 2 bytes are the CRC checksum. The generated control commands are sent to the robotic arm execution unit cyclically at a period of 10 milliseconds. The sending interface uses a CAN transceiver conforming to the ISO11898 standard at the physical layer to ensure real-time communication and stability. All commands are checked using CRC16-CCITT before transmission to ensure data integrity. The emergency braking control data includes the command frame sequence and timestamp, stored in binary file format, and supports log recording and traceability.

[0021] Step S44: Transmit emergency braking control data and vibration suppression control data to the robot industrial IoT management platform to execute multi-robot control tasks; In this embodiment, data transmission is achieved through an industrial Ethernet interface, and network communication is conducted using a TCP / IP protocol stack. Data transmission encapsulates emergency braking control data and vibration suppression control data in JSON format. Fields include robot ID (string, 16 characters long), control type (enumerated value, 0 for emergency braking, 1 for vibration suppression), control parameters (containing specific values), and timestamp (ISO8601 format). The transmission channel uses the TLS 1.3 encryption protocol to ensure data security. The transmission rate is set to 1 Mbps, with a latency not exceeding 20 milliseconds to ensure real-time uploading of control data. The management platform performs format and timing verification on the received data, and data validity is confirmed using a SHA-256 digest. The platform uses a message queue service (such as RabbitMQ) to process and distribute control tasks, ensuring the orderliness and consistency of data in a multi-robot environment and coordinating the execution of multi-robot control tasks.

[0022] Preferably, step S1 includes the following steps: Step S11: Obtain industrial warehouse aisle data and extract aisle geometry; In this embodiment, the acquisition of industrial warehouse aisle data is accomplished through a coordinated effort of a fixed-deployment LiDAR system and a stereo camera system. The LiDAR system is a Velodyne VLP-32C, operating at 10Hz, with a 360° scanning range, a maximum detection distance of 200 meters, and a ranging error not exceeding ±3cm. The stereo camera system uses an Intel RealSense D435i with a viewing angle of 85.2°×58° and a depth resolution of 1280×720, working in conjunction with an IMU module to record spatial structure and visual texture data in real time. After acquiring the raw data, a raster map generation tool is used to spatially mesh the laser point cloud, setting the grid size to 10cm×10cm. The laser reflection value and aisle height variation are recorded for each grid point. Based on the grid reflection values, fixed obstacles such as metal pillars and shelves are removed, and the remaining continuous area is extracted as the aisle data region. The length, width, and cross-sectional area of ​​each aisle are calculated, with the width limited to between 1.2m and 3.0m and the cross-sectional area less than 1.5m². 2 The area is marked as a non-traffic zone. The final geometric data includes: coordinates of the start and end points of the passage (labeled in the global coordinate system, with an accuracy of ±5cm), width (accurate to 0.1m), length (unit: m), boundary slope (representing the curvature change of the passage, unit: ° / m), and obstacle spacing (unit: m), outputting a structured data table to provide a foundation for topology modeling.

[0023] Step S12: Model the channel topology based on the channel geometry to obtain warehouse channel topology layer data; In this embodiment, based on the channel geometric structure data extracted in step S11, a channel topology graph is generated using a regular graph construction method. Each channel intersection or turning point is defined as a topology graph node, the channel body is an edge, and the node numbering adopts a sequential encoding method, numbered clockwise from the entrance. Each edge in the topology graph includes the following attributes: start point number, end point number, channel number, length (unit: m), maximum number of passages in the channel (rounded down by the channel width divided by the robot width 1.0m), and passage time (calculated based on a set robot speed of 1.5m / s). If there is a bidirectional channel between two nodes, the topology graph is constructed as a bidirectional edge graph structure. Dijkstra's algorithm is used for reachability verification, finding the shortest path between all node pairs and recording the path distance and hop count. The completed topology layer data is stored in graph structure JSON format, including complete topology node coordinates, connection relationships, attribute sets, and dynamic label fields, supporting subsequent behavior layer binding and digital twin synchronous updates.

[0024] Step S13: Simulate the channel transportation behavior based on the industrial warehouse channel data to obtain channel transportation behavior data, and construct the warehouse channel behavior layer based on the channel transportation behavior data to obtain warehouse channel behavior layer data; In this embodiment, the trajectory data collected in real time by the LiDAR and stereo camera system deployed in step S11 is combined with RFID access record data (reading frequency 13.56MHz, antenna spacing 2m) to establish channel transportation behavior data. Information such as the timestamp, speed, number of stops, stop time, and acceleration of each robot in each channel is recorded. The transportation behavior data divides the travel speed into five levels (0-0.5m / s, 0.5-1.0m / s, 1.0-1.5m / s, 1.5-2.0m / s, >2.0m / s), and the number of stops is measured in times per minute. For each channel, the number of robots passing through that channel, the average travel speed, and the total number of stops within a certain time period (set as 30 minutes) are statistically analyzed to form a behavior index matrix. This matrix is ​​then bound to the edge attributes in the topology graph using logical mapping to form "warehouse channel behavior layer data." For example, during the period from 9:00 AM to 9:30 AM, 15 robots passed through channel AB, with an average speed of 1.2 m / s and 18 stops. The final output behavioral layer data is expressed in a two-dimensional table structure, including fields such as channel number, time period, speed distribution, passage frequency, and stopping probability, which serve as dynamic behavioral parameters to support subsequent twin construction.

[0025] Step S14: Construct a warehouse digital twin model based on warehouse aisle topology layer data and warehouse aisle behavior layer data; In this embodiment, the warehouse channel topology layer data constructed in step S12 is fused with the channel behavior layer data formed in step S13. The fusion process is based on the one-to-one correspondence principle of channel numbers, binding the static attributes (geometric data, connectivity structure) of each channel edge with dynamic behavior data (speed, frequency, blocking rate). After fusion, a digital twin object set is formed, driving the evolution of the twin model with a unified time step (updated every 5 seconds). The digital twin modeling platform accesses the data streams of all sensors and execution controllers based on the Industrial Internet of Things OPC UA protocol to achieve real-time dynamic synchronization. The twin scene uses a 3D graphics engine (Unity engine or WebGL rendering engine) to render the warehouse map in real time, and pushes key fields such as the position, running status, current channel number, and estimated arrival time of each robot via the WebSocket protocol. The digital twin model has functional components such as channel entities, edge weight mapping, and time evolution, supporting simulation of running path conflicts and task scheduling analysis. The final digital twin model is marked with a unified timestamp to indicate the current state. It includes the complete warehouse layout, channel topology, dynamic information on transportation behavior, and real-time robot location and scheduling strategy, providing a physical-information dual foundation for the next step of path conflict detection.

[0026] Step S15: Perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data; In this embodiment, multi-robot path conflict detection is based on spatiotemporal comparison of data such as the current position, target point, current path, and estimated arrival time of all robots recorded in the digital twin model. A "channel occupancy time window list" is established for each channel edge, with each time window including the start time, end time, and occupying robot number. If any two robot occupancy windows overlap and the spatial distance is less than the robot's body length (set to 1.0m), it is considered a path conflict event. The conflict detection time precision is set to 100ms, and a conflict traversal is performed on the global path graph in each calculation cycle. Conflict information is represented in the form of a quadruple: <conflict channel number, conflict timestamp, robot number A, robot number B>. Additional attribute fields include conflict area length, estimated occupancy duration, corresponding task number, and task priority. The detection process calls a timed scheduling thread, running on an industrial edge computing unit (based on NVIDIA Jetson AGX Orin) deployed on the server side, using a memory-level data sharing cache library to improve read efficiency, with an average processing latency of no more than 50ms. The resulting path conflict data is pushed to the task scheduling system and robot control system in real time as a constraint.

[0027] Preferably, step S2, which involves detecting motor winding damage based on the robot number, includes: The motor installation area is determined based on the robot's serial number; In this embodiment, each robot is recorded with a unique serial number (e.g., RBT-0001) upon leaving the factory. This serial number is written into the database through the industrial IoT platform's device registration module. The platform retrieves the mechanical structure drawing corresponding to this serial number via the OPC UA protocol. The drawing clearly marks the motor installation location, including installation coordinates (X, Y, Z, in millimeters), mounting base plate number, and motor specifications (e.g., three-phase asynchronous motor Y2-132M-4). After retrieving the installation location information for this serial number, the platform uses the robot's own IMU (Inertial Measurement Unit) and an industrial vision positioning system to jointly calibrate the motor area. The IMU positioning error is less than 2cm. Combined with the vision recognition system (resolution not less than 1920×1080, recognition accuracy ±1.5mm), the specific locations of the motor housing, junction box, and cable bundle are identified, completing the calibration of the motor installation area. The positioning result is expressed in CAD layer coordinates and imported into the subsequent wiring recognition module.

[0028] The winding wiring is identified based on the motor installation area to obtain the winding wiring data; In this embodiment, a machine vision system is used to acquire images of the inside of the motor junction box, employing a combined binocular imaging and infrared imaging detection method. The image acquisition device is installed near the robot's inspection port, using a close-range focusing method (focal length set to 15cm, aperture F / 2.8, exposure time 10ms), with image clarity controlled at 1080p or higher to ensure clear and distinguishable winding connections. Image preprocessing uses the Canny edge extraction and color block segmentation algorithm, extracting connection paths for winding wires using color recognition (common industrial standards include yellow, green, red, and black), and combining this with physical measurements of wire exit angles and routing radii to identify their routing methods. The winding connection data includes: the number of wire leads, connection method, connection point location (two-dimensional pixel coordinates + three-dimensional spatial coordinates), wire exit color identification, wire exit sequence, and grounding wire identification information, outputting JSON structured data to the wiring pattern classification module.

[0029] Based on the winding connection configuration, star connection data and delta connection data are obtained; In this embodiment, the acquired winding connection data is subjected to three-phase lead-out path topology analysis. According to national standards GB755-2008 and IEC 60034-1, star connections exhibit a common neutral point connection characteristic, where one end of each of the three-phase winding terminals is connected to a common neutral point, and the other ends are connected to power supplies. If three conductors are identified in the image as being connected to the same point, and that point is not connected to the main power supply, it is marked as a star connection; conversely, if the three-phase terminals directly form a closed-loop interconnection, it is a delta connection. A judgment rule is set for connection pattern identification: connection point spacing less than 10mm and connection angle error not exceeding 10° are considered to belong to the same connection terminal. Connection pattern clustering is achieved using the spatial point clustering method (DBSCAN, radius ε set to 8mm, minimum number of points MinPts=3). The classification results are written as structured data, including connection method, lead number, phase information, and three-dimensional coordinates of the connection structure, divided into "star connection data" and "delta connection data" for subsequent use by their respective fault detection modules.

[0030] Copper wire erosion detection was performed based on star connection data to obtain copper wire erosion data. In this embodiment, copper wire ablation detection is conducted using a combination of infrared thermography and optical surface feature detection. The infrared camera, model FLIR A615, has a wavelength response range of 7.5μm–14μm and a thermal sensitivity below 50mK. Temperature measurement of the windings is performed after the motor has been running stably for 30 minutes. A temperature measurement area is set corresponding to the location of the star-connected copper wire. If the temperature exceeds 95°C and the duration exceeds 180 seconds, an initial ablation detection marker is triggered. Simultaneously, surface optical detection is performed, using an industrial macro camera to identify changes in copper wire surface color (judgment criterion: change from bright metallic to dark brown) and local deformation (scorch marks, cracks). The image analysis algorithm uses HSV color space segmentation + morphological operations to identify scorch marks. The detection threshold is set at a reduction in image grayscale value exceeding 15% and an area greater than 5mm². 2 This is recorded as the ablation area. The ablation data includes: the three-dimensional coordinates of the ablation area, the peak temperature of the thermal image, the ablation duration, the ablation area, the conductor number, and the ablation level (mild, moderate, severe), forming a copper wire ablation dataset for subsequent copper busbar oxidation analysis.

[0031] Based on the copper wire ablation data, copper busbar oxidation analysis was performed to obtain copper busbar oxidation data; In this embodiment, historical operating current data (acquired via an industrial acquisition module at a sampling frequency of 100Hz) of the copper wire ablation area is matched with thermal imaging data to analyze the overlap between current overcurrent waveform and ablation temperature rise periods. Oxidation analysis employs contact resistance detection technology, using a four-wire resistance measurement method to measure the contact resistance value at the copper busbar connection points. The standard resistance should be within 0.5mΩ; a resistance greater than 2.0mΩ is considered severe oxidation, while a resistance greater than 1.0mΩ but less than 2.0mΩ is considered mild to moderate oxidation. Simultaneously, Raman spectroscopy analysis is performed on the oxidized area surface, with a peak shift range of 210cm. -1 Up to 230cm -1 If a Cu2O oxidation peak is identified, and the corresponding peak intensity is greater than the threshold intensity of 500 cps, it is determined to be copper busbar oxidation. The generated data includes: oxidation level, oxidation area, resistance change, oxidation phase characteristics, corresponding copper wire number, and historical current data summary, and the output is "copper busbar oxidation data".

[0032] Arc burn-off detection is performed based on the triangular wiring data to obtain arc burn-off data; In this embodiment, arc burn detection at the delta connection point employs a three-pronged approach: voltage waveform acquisition, high-speed imaging, and acoustic emission. Real-time monitoring of winding phase and line voltage waveform changes is performed at a sampling frequency of 20kHz. If a voltage spike (amplitude change exceeding 30% and duration less than 2ms) is detected, it is considered a suspected arc phenomenon. Simultaneously, a high-speed imaging system is activated for optical flash detection (exposure frame rate not less than 500 frames / second). If a discharge spot is detected at a location consistent with the delta connection point, it is recorded as an arc burn event. The acoustic emission sensor, with a frequency range of 20kHz to 200kHz, is positioned near the motor wiring cavity. It detects typical "sharp discharge sound waves" through event triggering, and the results are fused with the imaging and voltage waveform data for verification, ultimately locating the arc event. Data records include: burn time, voltage spike value, arc location, acoustic emission amplitude, arc discharge frequency, and connection number, forming "arc burn data."

[0033] By integrating copper busbar oxidation data and arc burnout data, motor winding damage data is obtained.

[0034] In this embodiment, copper busbar oxidation data and arc burnout data are aligned according to wiring number and timestamp, and damage level fusion is performed. A unified damage level scoring mechanism is set up: copper busbar oxidation is scored in 5 levels based on resistance value and oxidation peak intensity, and arc burnout is scored in 5 levels based on voltage change amplitude and arc frequency. Finally, a weighted score is calculated according to the weight ratio (oxidation score weight 0.6, arc score weight 0.4). A complete damage record is generated for each wiring point, including damage level, damage type proportion, damage cycle, cumulative score value, and risk marker bit (divided into low, medium, and high), and the output forms "motor winding damage data" for subsequent core loosening analysis. The data format uses structured JSON and binary backup format, stored locally and in the cloud industrial IoT platform, respectively.

[0035] Preferably, step S2, which involves performing core loosening analysis based on motor winding damage data, includes: Electromagnetic stress data is obtained by performing electromagnetic stress detection based on motor winding damage data. In this embodiment, the marked ablation points, abnormal contact resistance points, and arc burn locations in the motor winding damage data are classified according to the winding phase. Based on this, the motor operating current and voltage sampling data (sampling frequency set to 20kHz) are retrieved, and the Lorentz force formula in Maxwell's equations is applied. The current density J can be determined by the winding cross-sectional area (measured value, such as 22.3 mm²). 2 The magnetic flux density B is calculated from the current amplitude (e.g., 45A) and obtained from the measured data of the Hall flux sensor, model LEM HLSR 50-P / SP3, with a measurement accuracy of ±1%. The magnetic flux density acquisition interval is set to 10ms. The distribution of the main magnetic poles and leakage magnetic field inside the motor is measured. If the magnetic flux density at a certain point exceeds the design standard value (e.g., 2.0T) and the current density at the corresponding winding damage point exceeds 50A / mm², the magnetic flux density will be determined. 2 This area is then marked as the point of electromagnetic stress increase. Based on the electromagnetic force density calculation formula σ=μ0JB, the electromagnetic stress value at each winding node can be derived. If the stress calculation result exceeds 80MPa (approaching the yield limit of silicon steel sheet), it is identified as a high electromagnetic stress zone. The output format includes location coordinates, electromagnetic stress magnitude, stress direction vector, associated winding number, and other information, forming "electromagnetic stress data".

[0036] Identify areas of concentrated electromagnetic stress based on electromagnetic stress data; In this embodiment, electromagnetic stress data points are clustered using three-dimensional spatial coordinates. The MeanShift algorithm is used with a clustering radius of 10mm. Regions with more than 10 points and an average stress value greater than 60MPa are identified as "electromagnetic stress concentration regions." Based on the three-dimensional mesh generation results of the motor stator model (each unit mesh has a side length of 5mm), the volume of continuous meshes with stress exceeding 50MPa is accumulated and statistically analyzed. If the volume exceeds 500mm... 3 The regions are continuously connected and marked as first-level concentrated regions. The concentrated region marker records the following data items: grid index number, center coordinates, total volume, average electromagnetic stress value, location of the maximum stress point, and the winding number involved. This result is written into the subsequent thermal expansion detection module for use.

[0037] Thermal expansion data of the stator inner wall was obtained by detecting the thermal expansion of the stator inner wall in the electromagnetic stress concentration area. In this embodiment, thermal expansion detection employs infrared thermal imaging temperature measurement. The imaging device is an FLIRA655sc infrared camera with a thermal sensitivity of 30mK and a resolution of 640×480. The camera is mounted at the observation port of the motor running platform, with the measurement angle set perpendicular to the radial direction of the motor. Dynamic imaging is performed on the electromagnetic stress concentration area, with a sampling time of no less than 30 minutes. After collecting temperature data, a thermal distribution map is established, and the temperature contour lines are divided into segments of 0.5°C to match the temperature rise changes at the stress concentration locations. If a temperature rise exceeding 10°C is detected in the corresponding area (starting from the ambient temperature baseline before motor startup), and the area of ​​this area is no less than 100mm², the thermal expansion is considered complete. 2 These are marked as areas of significant thermal expansion. The amount of thermal expansion is determined by the thermal expansion formula. Calculations were performed, and the coefficient of thermal expansion of the motor stator material was set to be... With a reference length L0 = 50 mm, the expansion amount of the stator inner wall is obtained (for example, an expansion of approximately 6 μm). The thermal expansion data includes the center coordinates, expansion direction, temperature change, expansion value, and corresponding grid number, forming the "stator inner wall thermal expansion data".

[0038] The direction of thermal expansion is determined based on the thermal expansion data of the stator inner wall. In this embodiment, multiple expansion displacement vectors within the thermal expansion region are directionally fitted, and principal component analysis (PCA) is used to extract the direction with the maximum variance as the principal direction of thermal expansion. The input is a multi-point three-dimensional displacement vector matrix, and the output is the first principal component vector. The directional error angle tolerance is set to no more than ±5°. If the angle between the expansion direction and the electromagnetic stress direction is less than 15°, they are merged and marked as "composite concentrated expansion zone". The thermal expansion direction data is expressed in vector format, with fields including center point coordinates, expansion angle (with the motor center as a reference), expansion unit vector length (normalized to 1), and actual expansion amount (μm level).

[0039] Warpage detection of the iron core pressure plate is performed based on the direction of thermal expansion to obtain pressure plate warpage data; In this embodiment, a laser displacement sensor (such as the KEYENCE LK-G5000 series, with a resolution of 0.01 μm) is installed at the inspection port on the top of the motor stator, directly opposite the pressure plate, to detect the change in the vertical distance from the edge of the pressure plate to the reference plane. The scanning frequency is 10 kHz, the scanning line width is 5 mm, the number of scanning lines is 50, and the total coverage width is 250 mm. After the motor is heated and running, warpage height data is collected in the pressure plate area corresponding to the main direction of thermal expansion. A warpage height greater than 30 μm is considered mild warpage, and greater than 80 μm is considered severe warpage. The detection data includes the scanning position coordinates of each point, warpage height, maximum value, minimum value, and average value of the area, and the warpage area is calculated. In addition, if warpage causes the gap between the edge of the pressure plate and the inner wall of the stator to increase beyond the specified limit (e.g., the original design gap is 0.2 mm, but the actual measurement is 0.8 mm), it is marked as a severe structural offset. The final dataset is written into "Pressure Plate Warpage Data" for core loosening assessment.

[0040] Of particular importance is the detection of core plate warpage based on the direction of thermal expansion, which includes: Solve the nodal displacement field based on the direction of thermal expansion; In this embodiment, a high-precision temperature sensor array deployed on-site collects temperature data of the pressure plate surface and internal structure. The sensor accuracy is ±0.1 degrees Celsius, the sampling frequency is 1Hz, and the acquisition time is no less than 30 minutes to cover typical operating thermal cycles. Based on the obtained temperature distribution, using a coefficient of thermal expansion of 1.2 × 10⁻⁶... -5 Based on the core material parameters of / K and combined with heat conduction theory, the theoretical linear expansion amount in the direction of thermal expansion at each monitoring point is calculated. This thermal expansion data is mapped to a discrete nodal coordinate system, with a node spacing of 5 mm to ensure structural detail capture. Using rigid body motion analysis methods from structural mechanics, the corresponding three-dimensional displacement is calculated based on the thermal expansion length change of each node, with the displacement unit being millimeters. The nodal displacement field includes the X, Y, and Z direction offsets of all nodes, with the data format being node number and corresponding three-dimensional displacement value. This nodal displacement field reflects the deformation trend of the pressure plate after thermal expansion.

[0041] The warping amplitude is calculated based on the nodal displacement field. In this embodiment, nodal displacement field data is used to extract the normal displacement component of the nodes perpendicular to the pressure plate plane as the main indicator of warping deformation. The pressure plate area is divided into equally spaced grid cells with a grid side length of 10 mm. The maximum and minimum values ​​of the normal displacement of all nodes within each cell are counted, and the local warping amplitude is calculated as the difference between the maximum and minimum values, in millimeters. The statistical results of the warping amplitude of the entire pressure plate include the maximum local warping amplitude, the average warping amplitude, and the standard deviation, which are used for subsequent deformation level determination. The threshold for the maximum local warping amplitude is set to 0.3 mm; areas exceeding this threshold are marked as key areas of concern. The warping amplitude statistical process is implemented through data traversal and numerical calculation, and the results are stored in the form of a two-dimensional array, where each array element corresponds to the warping amplitude of a grid cell.

[0042] Determine the deformation level of the pressure plate based on the warpage amplitude; In this embodiment, local warpage amplitudes are classified and deformation levels are determined according to a set threshold range. Specifically, warpage amplitudes less than or equal to 0.1 mm are defined as "slight deformation"; those between 0.1 mm and 0.3 mm are defined as "moderate deformation"; those exceeding 0.3 mm but less than 0.6 mm are defined as "significant deformation"; and those greater than 0.6 mm are defined as "severe deformation". The warpage amplitudes of all grid cells on the pressure plate surface are assigned corresponding deformation level labels according to this standard, generating a two-dimensional deformation level matrix. The proportion of each level is statistically analyzed and expressed as a percentage for overall deformation level determination. This classification operation utilizes standard thresholds for direct numerical comparison, eliminating the need for complex calculations. The classification results serve as crucial input for subsequent reconstruction steps.

[0043] The warped three-dimensional pressure plate is reconstructed based on the deformation level of the pressure plate to obtain the warped three-dimensional pressure plate data. In this embodiment, the deformation level of each mesh element is mapped back to three-dimensional node coordinates. Combined with node displacement field data, the three-dimensional shape of the warped pressure plate is reconstructed. Specifically, based on the normal displacement of each node, its spatial coordinate position is adjusted to form a three-dimensional pressure plate geometric model with realistic deformation characteristics. The adjusted nodes are connected into patches using a standard triangular meshing method to construct a complete three-dimensional mesh structure. During the reconstruction process, it is ensured that the connection relationship between nodes remains consistent with the original structure to avoid mesh distortion and topological errors. The generated three-dimensional mesh model file is stored in a common format, such as STL or OBJ, and contains node coordinates and triangular patch index information. This three-dimensional pressure plate data accurately reflects the spatial deformation of regions with different deformation levels, and is used for subsequent curvature calculations.

[0044] Warpage curvature is calculated based on warped three-dimensional pressure plate data; In this embodiment, based on the obtained three-dimensional pressure plate mesh model, the local curvature index is calculated for each mesh node. The normal variation rate method based on the spatial position of adjacent nodes is used, calculating the curvature by measuring the change in the angle between the normal vectors of a node and its neighboring nodes. The neighborhood of a node is set to a local mesh with a radius of 15 mm to ensure smooth calculation and highlight local features. The calculated curvature values ​​are expressed in millimeters. -1 The curvature is measured in units, reflecting the degree of curvature on the pressure plate surface. All node curvature data is stored as an array structure, containing node numbers and corresponding curvature values. A curvature heatmap is plotted based on the curvature distribution, with high curvature areas displayed in red and low curvature areas in blue, aiding in the determination of pressure plate deformation distribution.

[0045] The severity of warping is assessed based on the warping curvature, and the warping data of the pressure plate is obtained. In this embodiment, the severity of the pressure plate area is divided according to a curvature threshold. The curvature threshold is set to less than 0.2 mm. -1 Slight warping, ranging from 0.2 to 0.5 mm. -1 Moderate warping, greater than 0.5 mm. -1 Severe warping is identified. The number of nodes and corresponding spatial areas for each severity level are statistically analyzed to generate a warping severity distribution report. Pressure plate warping data includes the maximum curvature value, average curvature value, area proportions of slight, moderate, and severe warping regions, and coordinates of major warping locations. The data is stored in JSON or CSV format with precisely defined fields. This data serves as input for the core loosening assessment and vibration control modules, supporting subsequent automated analysis and decision-making.

[0046] The degree of core loosening is determined based on the pressure plate warping data, and the core loosening data is obtained. In this embodiment, the gap increment in the measured warping data is included in the evaluation by comparing the original design value of the gap between the assembly pressure plate and the iron core (set to 0.2mm ± 0.05mm according to manufacturing specifications). If the pressure plate is found to have dislodged from its original position, causing the gap to expand to more than 0.5mm and lasting for more than 30 minutes, it is defined as "slight loosening"; if the gap is greater than 1mm or the detected warping area exceeds 20cm², it is considered "slightly loose". 2 If the core loosening is not severe, it is defined as "severely loose". The degree of core loosening is scored using a five-level system: Level 0 (no loosening), Level 1 (slight), Level 2 (mild), Level 3 (moderate), and Level 4 (severe). The scoring is based on a weighted calculation of five indicators: maximum warpage value, average warpage value, warpage area, thermal expansion direction offset, and stress concentration degree. The weights are set as follows: maximum warpage value 0.3, average value 0.2, warpage area 0.2, direction offset 0.1, and stress value 0.2. Core loosening data includes: loosening level, score, corresponding pressure plate number, location coordinates, historical temperature rise changes, and stress data reference index, all of which are stored in the industrial IoT platform database as the input source for subsequent vibration suppression control modules.

[0047] Preferably, step S2, which optimizes vibration suppression control based on core loosening data, includes: The main vibration frequency band was determined based on the data of core loosening. In this embodiment, data including timestamps, loosening levels, pressure plate warping values, thermal expansion directions, and stress concentration area numbers are extracted from the core loosening data. Using a triaxial accelerometer (model PCB356A32, range ±500g, sensitivity 10mV / g) operating in motor mode, three sets of orthogonal sensors are installed on the stator shell corresponding to the pressure plate area at a sampling frequency of 5kHz to collect acceleration time-series data in the X, Y, and Z directions. Each set of data is sampled for 10 seconds. The collected acceleration signals are analyzed using FFT (Fast Fourier Transform) with a resolution set to 0.5Hz. The peak amplitude points of each frequency band in the spectrum are statistically analyzed. A detection threshold is set as the baseline mean plus three times the standard deviation. If the peak value of a certain frequency band exceeds this threshold, it is defined as the dominant vibration frequency band. Typically, the main vibration frequency caused by a loose iron core is concentrated in the 60–120Hz range. If a peak amplitude of 0.93g is detected at 95Hz (higher than the threshold of 0.61g), then 95Hz is recorded as the current main vibration frequency band. The final output data structure of the main vibration frequency band includes the center frequency, bandwidth, peak amplitude, vibration direction vector, and corresponding iron core number.

[0048] Determine the target frequency band suppression index based on the main vibration frequency band; In this embodiment, the target frequency range is constructed as 90–100Hz, centered on the main vibration frequency band (e.g., 95Hz) and with a bandwidth of ±5Hz. Based on the vibration damping design requirements of the motor housing's inherent frequency, the peak amplitude of the target frequency band is used as a benchmark, and the suppression index is set as "suppressing the peak amplitude of this frequency band to below the baseline mean plus one standard deviation." Specifically, the target is to reduce the vibration amplitude at 95Hz from 0.93g to below 0.41g. This target frequency band suppression index includes five key parameters: center frequency (f0=95Hz), bandwidth (Δf=10Hz), initial peak value (A0=0.93g), target limit (A1=0.41g), and suppression ratio (A1 / A0=0.44). These indicators form the vibration control optimization input standard, used for the next step of damper type selection and control parameter construction.

[0049] Select a bandpass damper based on the target frequency band suppression index, and perform frequency domain simulation of the structural response based on the bandpass damper to obtain the frequency domain data of the structural response. In this embodiment, a bandpass damper based on viscoelastic materials is selected. Its structure consists of a steel plate sandwich layer and high-damping rubber (e.g., 3MISD112 material, loss factor η=0.45, suitable frequency band 50–150Hz). The damper is 60mm long and 5mm thick, with the rubber layer accounting for 70% of the length. It is installed in the area directly opposite the motor housing pressure plate. A structural response simulation platform is established using the finite element tool ANSYS Mechanical APDL. The modeling method is the equivalent element method, constructing the motor housing, pressure plate, and damper using 20-node SOLID226 elements. Material parameter inputs include the housing material (Q235 steel, elastic modulus E=2.1×10⁻⁶). 11 Pa, Poisson's ratio μ=0.3), the elastic modulus of the damping layer material is 1.2×10⁻⁶. 6 Pa, density is 1100 kg / m³ 3 The boundary conditions are set as fixed cantilevered pressure plate and shell constraint, with simulated load applied from below. The frequency sweep range is 50Hz–200Hz, with a sweep interval of 0.5Hz. The simulation outputs the structural response acceleration amplitude curve, recording the peak value changes in the target frequency band (90–100Hz) before and after the damper installation, and deriving the frequency-response amplitude correspondence table. The response amplitude unit is m / s². 2 The frequency accuracy error is less than 0.25Hz, forming the frequency domain data of the structural response.

[0050] Vibration suppression control optimization is performed based on the frequency domain data of the structural response to obtain vibration suppression control data.

[0051] In this embodiment, the amplitude reduction curve of the target frequency band response is extracted from the structural response frequency domain data, and the amplitude reduction rate at each frequency point before and after the bandpass damper installation is calculated. The number of frequency points with a reduction rate greater than 60% is compared with the target frequency band coverage. If the ratio is greater than 80%, the damper structure and installation parameters are considered to have met the optimization conditions. Based on this analysis result, the installation fastening force of the bandpass damper is reverse-corrected (controlled between 12–15 N·m), the material compressive prestress is 0.35 MPa, and the rubber layer thickness correction value is calculated based on the optimal response point. For example, if the original thickness is 5 mm, the optimized thickness is 5.5 mm to meet a wider frequency band response range. Finally, vibration suppression control data is generated, including the following fields: target frequency band center value, control bandwidth, damper material parameters, installation torque, vibration reduction ratio, compressive prestress value, control strategy number, and verification timestamp. This control data will be synchronously uploaded to the industrial IoT control platform and written into the vibration management module in the robot execution unit for closed-loop control.

[0052] Preferably, step S3 includes the following steps: Step S31: Determine the conflict location based on the path conflict data to obtain conflict location data; In this embodiment, after performing path conflict detection, the collected data includes robot number, motion path coordinates (in millimeters, using a three-dimensional Cartesian x, y, z format), timestamp (accurate to 0.01 seconds), and path intersection marker. Conflict determination is based on the minimum spatial distance d_min between two robots, with a threshold of 300mm. If the spatial distance between the end effectors of two robots at the same travel time is less than this threshold, a path conflict is determined. The conflict location is determined based on points in the three-dimensional path trajectory where d_min ≤ 300mm. The spatial coordinate difference is calculated using the Euclidean distance formula, and the center point within the conflict area is selected as the conflict coordinate record, with an accuracy error not exceeding ±10mm. The conflict location data structure includes conflict time, robot number pair, conflict coordinate points (x, y, z), and minimum distance value d_min. All fields are stored in CSV format and uploaded in real time to the path scheduling submodule in the industrial IoT platform.

[0053] Step S32: Adjust the task allocation order based on conflict location data; In this embodiment, after the conflict location data is determined, the original task queue information of the corresponding conflicting robots is obtained. This information includes each robot's number, task ID, estimated start and end times of the task, coordinates of the task's location, priority (1–10), and task category label. The task reordering strategy is a dual-objective optimization problem of maximizing task priority and minimizing total task time. A heuristic task reordering algorithm (such as one based on the AntColonyOptimization scheduler) is adopted, fixing high-priority tasks and reordering low-priority tasks in the conflict area. The algorithm parameters include the number of iterations N=100, pheromone decay rate ρ=0.8, and maximum allowable task delay time ΔT_max=60s. After the conflicting robot tasks are reordered, their task start times are updated, and their path generation modules are synchronized. Finally, a new task allocation order table is output, with fields including robot number, task number, reordered start time, end time, scheduling order value, etc., and saved as a scheduling configuration JSON structure.

[0054] Step S33: Perform robot scheduling according to the task allocation order to obtain robot scheduling data; In this embodiment, the scheduling table after task rearrangement for each robot is obtained, and trajectory generation is performed sequentially according to the timeline. An inverse kinematics solver is used, taking into account the target task point coordinates, movement speed (0.1–1.5 m / s), load weight (2–50 kg), and obstacle avoidance parameters (safe distance of 500 mm), and outputting the sequence of joint angles. Based on the range of joint angle variations, a maximum acceleration limit (≤1.2 rad / s²) is set. 2The maximum rotational speed (≤2.0 rad / s) was set, and the scheduling trajectory was verified through offline simulation with a simulation step size of 0.01 seconds. The final scheduling data format was a list of time-series joint commands for each robot, including timestamps, 6-axis angle values, trajectory point numbers, speeds, accelerations, and other information.

[0055] Step S34: Simulate robotic arm handling based on robot scheduling data, where the single load mass is set to 2kg-50kg, and obtain robotic arm handling data; In this embodiment, robot scheduling data is input into the handling task simulation module, and objects of known mass are loaded. The mass values ​​are taken from the mass configuration table of real handling workpieces, such as motor housing: 16.5kg, reducer components: 37kg, bearing components: 5kg, etc. The load mass range is set to 2–50kg, and the simulation environment is a dynamic physics simulation environment integrating the ROS (Robot Operating System) platform and the Gazebo simulator. During the simulation, the joint torque, joint position, and three-axis forces (in N) and acceleration (m / s²) of the end effector in each handling task are collected. 2 The system records the stress concentration area, dynamic response changes, and periodic disturbances of the robotic arm under different loads. The sampling frequency is set to 1kHz. The output data includes the robotic arm handling data, with fields such as task number, load mass, timestamp, axis number, drive current, angular velocity, and end force.

[0056] Step S35: Detect surface spalling of bearing rolling elements based on robotic arm handling data to obtain surface spalling data; In this embodiment, the rolling bearings on the drive and driven shafts are first located based on the joint torque fluctuation data recorded in the robotic arm's handling data. An electromagnetic vibration sensor (model: Kistler8763B, frequency response range: 0–10kHz) is installed on the bearing housing to continuously collect vibration signals. Short-time Fourier Transform (STFT) is used to extract high-frequency pulse features, with a sliding window width of 128ms and a frequency resolution of 10Hz, extracting abnormal waveform regions in the mid-to-high frequency range (3–6kHz). When the detected pulse peak amplitude exceeds a set threshold (three times the average background noise level; for example, if the average is 0.3g, the threshold is 0.9g), a peeling event is identified. Further, combining the peak spacing and amplitude variation trend, a spectral entropy algorithm is used to determine whether it is surface peeling. Surface peeling data is output, including the peeling location number, timestamp, amplitude, and estimated peeling area (in mm). 2 ( ) and vibration frequency response.

[0057] Step S36: Detect eccentric rolling elements based on the robotic arm handling data to obtain eccentric rolling element data; In this embodiment, the end effector disturbance record from the transport data is used to analyze its coupling characteristics with the periodic dynamic response of the rolling bearing. The Hilbert-Huang transform (HHT) is used to perform envelope demodulation on the acquired vibration signal to extract the asymmetric amplitude components during the rolling element rotation process. The eccentricity detection threshold is set to a deviation factor of >1.8 (the ratio of upper and lower peak values) for the envelope signal waveform, and a period repeatability standard deviation of less than 0.15. When detecting eccentricity, the rotation time per revolution of the rolling element (e.g., 250ms) and the abnormal peak angle offset (e.g., 45°) are simultaneously acquired. If the offset angle and amplitude period occur synchronously, it is recorded as an eccentric rolling element event. The output data includes the rolling element number, eccentricity angle, rotation period, eccentricity amplitude, and number of abnormal periods.

[0058] Step S37: Integrate surface spalling data and eccentric rolling element data to obtain bearing fault data.

[0059] In this embodiment, the detection results of steps S35 and S36 are integrated to construct a comprehensive bearing fault diagnosis database. Using the bearing number as the primary key, and combining fields such as spalling area value, spalling frequency, eccentricity angle, and eccentricity cycle repetition rate, the current comprehensive fault state of the bearing is determined. A fault state mapping table is used for judgment; if the spalling area > 1.0 mm... 2 If the eccentricity angle is greater than 30° and both occur within the same time window (±2 seconds), it is classified as a Level II fault; if the overlap between the eccentricity and spalling frequencies exceeds 70%, it is classified as a Level III severe fault. A bearing fault data record is generated, including the following fields: bearing number, spalling index, eccentricity index, detection time, fault level, and waveform analysis summary. This record is then transmitted to the host computer system in a standard fault log format (such as the MODBUS protocol).

[0060] Preferably, step S35 includes the following steps: Step S351: Locate the bearing rolling element area based on the robotic arm handling data; In this embodiment, joint torque fluctuations and vibration sensor data are collected during the robotic arm's handling process. Vibration sensors are deployed on the drive shaft and the supporting bearing housing, with a sampling frequency set to 10kHz to ensure the capture of subtle vibration characteristics. Utilizing the periodic changes in joint torque during the handling task, combined with bearing rotational speed data, characteristic frequency bands in the vibration signal are extracted based on time-frequency analysis methods (such as short-time Fourier transform, with a window length of 256 points and an overlap rate of 50%) to locate the typical vibration frequency range of the rolling element (e.g., 500Hz to 2500Hz). The moving area of ​​the rolling element inside the bearing is located by identifying the vibration amplitude cluster area. Furthermore, an accelerometer (range ±50g, sensitivity 10mV / g) mounted on the bearing housing surface collects triaxial acceleration data. Combined with a spatial coordinate positioning system, the relative position of the robotic arm's end effector and the bearing housing is obtained. Multi-sensor fusion technology is used to locate the precise area of ​​the rolling element in three-dimensional space, achieving a positioning accuracy of ±0.5mm. The rolling element region coordinate data is stored in a three-dimensional Cartesian coordinate system (x, y, z, unit: millimeters), providing a foundation for subsequent trajectory determination.

[0061] Step S352: Determine the rolling element contact trajectory based on the rolling element area of ​​the bearing; deploy a particle collector according to the rolling element contact trajectory and collect the detached material to obtain detached material data; In this embodiment, the rolling element contact trajectory is determined based on the rolling element spatial activity area located in step S351, combined with the rolling element size parameters provided by the bearing manufacturer (rolling element diameter range of 5mm to 15mm) and the bearing's rated speed (5000rpm to 15000rpm) to calculate its theoretical motion path. The trajectory calculation uses a geometric constraint method, combined with dynamic position compensation from the robotic arm, to calculate the relative motion trajectory of the rolling element between the inner and outer rings of the bearing. Based on the trajectory, a particle collector is set up. The particle collector uses electrostatic adsorption technology to capture particles detached from the rolling element. The collector area covers 120% of the trajectory path width to ensure all detached particles are collected. The installation position is no more than 1mm from the trajectory center, and the collection cycle is set to sample once per hour, with continuous collection for no less than 24 hours. The shedding material was collected by scanning the surface of the particle collector with a high-precision microscope (0.5 μm resolution) to capture the particle morphology and distribution. The particle size, quantity and morphology were classified using image processing software. The collected shedding material data included particle size distribution (particle diameter range from 0.5 μm to 100 μm), particle density (number of particles per unit area), particle morphological characteristic parameters (aspect ratio, roundness) and collection timestamp. All data were stored in a database for fracture morphology analysis.

[0062] Step S353: Perform fracture morphology detection based on the data of the detached material to obtain fracture morphology data; In this embodiment, fracture morphology detection utilizes scanning electron microscopy (SEM) to perform high-magnification imaging of the detached sample, with the magnification set between 1000x and 10000x to observe the detailed fracture characteristics of the particles. During sample preparation, the detached particles are ultrasonically cleaned and then subjected to metal sputtering, with the sputtering thickness controlled within 10nm to ensure surface conductivity. Image processing of the fracture morphology employs an edge detection algorithm (Canny operator, threshold low 30, high 90) to extract the fracture edges, combined with morphological operations to analyze the fracture shape. Quantitative parameters include fracture crack length (in micrometers, range 50–500μm), fracture roughness (Ra, range 0.2–1.5μm), and fracture initiation angle (10°–80°). Simultaneously, fracture mode classification identifies types such as peeling, fatigue, and brittleness. Fracture morphology data is recorded in a structured data format, including fracture feature images, various fracture parameter indicators, and corresponding time and sample numbers, providing a foundation for subsequent slip fringe analysis.

[0063] Step S354: Perform slip fringe analysis based on the fracture morphology data to obtain slip fringe data; In this embodiment, the slip fringe analysis is based on the high-resolution fracture image obtained in step S353. Texture features are extracted using two-dimensional wavelet transform, employing the Daubechies wavelet basis with a decomposition layer of 5. Fringe direction, spacing, and contrast are extracted. The fringe spacing is determined through image spatial calibration, ranging from 1 μm to 10 μm. The fringe direction is expressed as an angle, accurate to 1° relative to the principal direction of the fracture crack. The contrast of the slip fringe is calculated from the gray-level difference, ranging from 0.1 to 0.8. The uniformity of the fringe arrangement is analyzed by combining fringe density and distribution statistics. The slip fringe data is stored in the form of a numerical matrix, including fringe direction angle, spacing, contrast value, and spatial coordinates. All data are associated with the corresponding fracture sample number for subsequent particle stripping detection.

[0064] Step S355: Perform particle stripping detection based on the slip stripe data to obtain particle stripping data; In this embodiment, particle ablation detection is based on the spatial distribution and density parameters of slip fringes, and an image segmentation algorithm (based on a threshold method, with a grayscale threshold set to 0.3) is used to identify the particle ablation region. A three-dimensional laser scanning microscope (LCSM) is used to measure the depth of the sample surface to obtain the volume and area parameters of the particle ablation region. The minimum ablation particle size is set to 0.5 μm during the detection process. 3 The maximum detection range is controlled at 500μm. 2Within the surface area, data on the quantity, size distribution, and volume of delamination particles were collected in CSV format, with fields including particle ID, volume, surface area, delamination depth, and corresponding slip stripe number. Particle delamination data was cross-referenced with slip stripe data for statistical analysis of the delamination mechanism and its correlation with fracture morphology.

[0065] Step S356: Calculate the surface peeling amount based on the particle peeling data to obtain surface peeling data.

[0066] In this embodiment, the surface peeling amount is statistically analyzed based on the particle peeling volume and surface area data collected in step S355, and the total peeling amount is determined through weighted calculation. The peeling amount calculation adopts a volume-weighted method, which accumulates all particle volumes to obtain the total peeling volume (unit: μm). 3 ), and convert it into the peeled surface area (unit: mm). 2 The conversion factor is calculated based on material density and morphology. A statistical time window of one hour is set, and the volume of all peeled particles within the statistical period is cumulatively calculated. The number and average size trends of the peeled particles are also statistically analyzed. The statistical results are output in a structured report format, including timestamps, cumulative peeling area, number of peeled particles, and average size. Surface peeling data is transmitted to a central monitoring system via an Industrial Internet of Things (IIoT) protocol (such as MQTT) to enable real-time monitoring and support for subsequent maintenance strategies.

[0067] Preferably, step S36 includes the following steps: Step S361: Determine the load disturbance of the end effector based on the robotic arm handling data, and obtain the load disturbance data; In this embodiment, force sensors (such as a six-degree-of-freedom force / torque sensor with a measurement range of 0–200N and an accuracy of 0.1N) mounted on the end effector of the robotic arm are used to collect the mechanical data of the load during the handling process in real time. The acquisition frequency is set to 1kHz to ensure the capture of dynamic disturbance details. The motion trajectory data of the robotic arm (including position, velocity, and acceleration) is synchronized with the load force data in time, and the load disturbance, i.e., the dynamic change of the load during the handling process, is calculated using a numerical differentiation method. The load disturbance is defined as the difference between the current sampled load and the load at the previous time point, in Newtons (N), and the data range is typically between 0–50N. By performing time-series statistics on the load disturbance at all sampling time points, a complete load disturbance dataset is formed. The data format includes timestamps, load disturbance values, and corresponding robotic arm pose information for subsequent drive current statistics.

[0068] Step S362: Calculate the joint drive current based on the load disturbance data, where the standard deviation of the current fluctuation is set to 0.1-3.0; In this embodiment, current sensor data from the actuators of each joint of the robotic arm is collected. The current sampling frequency is 2kHz, the current measurement range is 0–20A, and the accuracy is 0.01A. Combined with the load disturbance data obtained in step S361, synchronization matching is performed according to the corresponding timestamps. For each joint, the standard deviation of the current fluctuation is calculated. The standard deviation is calculated based on the sliding time window method, with a window length of 1 second and a step interval of 0.1 seconds. The statistical range is limited to a standard deviation of 0.1A to 3.0A; fluctuations below 0.1A are considered noise and ignored, while fluctuations above 3.0A are considered outliers. The calculation results form a joint current fluctuation data matrix, including the joint number, the start and end times of the time window, the current standard deviation, and the corresponding load disturbance. The data storage format is a time-series database, supporting subsequent identification of periodic abnormal loads.

[0069] Step S363: Identify the periodic abnormal load of the drive shaft based on the joint drive current and obtain the abnormal load data of the drive shaft; In this embodiment, the joint drive current data collected in step S362 is processed using a spectrum analysis method. The Fast Fourier Transform (FFT) algorithm is used, with an FFT length of 2048 points, a sampling rate of 2kHz, and a frequency resolution of approximately 1Hz. The drive current signal is windowed (Hanning window, 2048 points long, 50% overlap) to identify periodic peaks in the current signal, focusing on the main frequency range of the robotic arm (1Hz to 100Hz). Frequency points in the spectrum whose amplitude exceeds a set threshold (the threshold is the average amplitude plus three times the standard deviation) are identified as abnormal load frequencies. Combined with the time period corresponding to the abnormal frequency, the duration, peak amplitude, and frequency of the abnormal load are statistically analyzed to generate abnormal load data for the drive shaft. The data fields include joint number, abnormal frequency, abnormal amplitude, start and end times of the abnormal period, and amplitude change trend.

[0070] Step S364: Perform dynamic imbalance analysis of the rolling bearing based on the abnormal load data of the drive shaft to obtain dynamic imbalance data; In this embodiment, based on the abnormal frequency identified in step S363 and corresponding to the drive shaft rotation frequency, a method combining time-domain vibration signal analysis and frequency-domain signal power spectrum estimation is used. Power spectral density (PSD) estimation is employed to calculate the vibration energy concentration within ±5Hz of the bearing's rated speed frequency. Vibration data is collected by an accelerometer installed in the bearing housing, with a sampling rate of 10kHz, a range of ±50g, and a data recording duration of at least 60 seconds. The dynamic response function (FRF) is used to analyze the rolling bearing's dynamic response at the abnormal load frequency, calculating the vibration energy ratio at that frequency. The dynamic imbalance coefficient is defined as the ratio of vibration energy at the abnormal frequency to the total vibration energy. The dynamic imbalance data includes frequency values, energy ratios, bearing temperature (measured via thermocouple, range 0–150℃, accuracy ±0.1℃), and the corresponding sampling time.

[0071] Step S365: Determine the asymmetry of the rolling elements based on the dynamic imbalance data to obtain the eccentric rolling element data.

[0072] In this embodiment, the vibration energy ratio and vibration frequency characteristics in the dynamic imbalance data are used in conjunction with the rolling element geometric parameters (such as rolling element diameter and spacing, both determined by the equipment installation manual) to perform asymmetry analysis. The presence of an eccentric rolling element is determined by calculating the proportion of harmonic components and the phase difference of the vibration signal. Specifically, harmonic decomposition technology is used to decompose the vibration signal into its fundamental frequency and its harmonics. An eccentric rolling element typically exhibits a significantly higher fundamental frequency vibration amplitude and a phase lag exceeding 30°. The degree of eccentricity is determined by calculating the rolling element center offset distance, based on the vibration amplitude and phase data. The offset range is limited to 0.1mm to 1.5mm; an offset exceeding 1.5mm is considered severe eccentricity. The eccentric rolling element data structure includes the eccentricity distance, phase difference, fundamental frequency vibration amplitude, and detection time, for subsequent alarm and maintenance reference by the maintenance system.

[0073] Preferably, this specification also provides a multi-robot control system based on the Industrial Internet of Things (IIoT) for executing the multi-robot control method based on the IIoT described above. The multi-robot control system based on the IIoT includes: A multi-robot path conflict detection module is used to acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; and perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data. The vibration suppression control optimization module is used to determine the robot number based on path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; and optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. The bearing fault detection module is used to perform robot scheduling based on path conflict data to obtain robot scheduling data; to perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; and to perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. The multi-robot control module is used to perform emergency braking control based on bearing fault data to obtain emergency braking control data; the emergency braking control data and vibration suppression control data are transmitted to the robot industrial IoT management platform to execute multi-robot control tasks.

[0074] Optionally, this specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements any one of the multi-robot control methods based on the Industrial Internet of Things.

[0075] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0076] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A multi-robot control method based on the Industrial Internet of Things, characterized in that, Includes the following steps: Step S1: Acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data; Step S2: Determine the robot number based on the path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. Step S3: Perform robot scheduling based on path conflict data to obtain robot scheduling data; perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. Step S4: Perform emergency braking control based on bearing fault data to obtain emergency braking control data; transmit the emergency braking control data and vibration suppression control data to the robot industrial IoT management platform to execute multi-robot control tasks.

2. The multi-robot control method based on the Industrial Internet of Things according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain industrial warehouse aisle data and extract aisle geometry; Step S12: Model the channel topology based on the channel geometry to obtain warehouse channel topology layer data; Step S13: Simulate the channel transportation behavior based on the industrial warehouse channel data to obtain channel transportation behavior data, and construct the warehouse channel behavior layer based on the channel transportation behavior data to obtain warehouse channel behavior layer data; Step S14: Construct a warehouse digital twin model based on warehouse aisle topology layer data and warehouse aisle behavior layer data; Step S15: Perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data.

3. The multi-robot control method based on the Industrial Internet of Things according to claim 1, characterized in that, Step S2, which involves detecting motor winding damage based on the robot's identification number, includes: The motor installation area is determined based on the robot's serial number; The winding wiring is identified based on the motor installation area to obtain the winding wiring data; Based on the winding connection configuration, star connection data and delta connection data are obtained; Copper wire erosion detection was performed based on star connection data to obtain copper wire erosion data. Based on the copper wire ablation data, copper busbar oxidation analysis was performed to obtain copper busbar oxidation data; Arc burn-off detection is performed based on the triangular wiring data to obtain arc burn-off data; By integrating copper busbar oxidation data and arc burnout data, motor winding damage data is obtained.

4. The multi-robot control method based on the Industrial Internet of Things according to claim 1, characterized in that, Step S2, which involves core loosening analysis based on motor winding damage data, includes: Electromagnetic stress data is obtained by performing electromagnetic stress detection based on motor winding damage data. Identify areas of concentrated electromagnetic stress based on electromagnetic stress data; Thermal expansion data of the stator inner wall was obtained by detecting the thermal expansion of the stator inner wall in the electromagnetic stress concentration area. The direction of thermal expansion is determined based on the thermal expansion data of the stator inner wall. Warpage detection of the iron core pressure plate is performed based on the direction of thermal expansion to obtain pressure plate warpage data; The degree of core loosening is determined based on the warping data of the pressure plate, and the core loosening data is obtained.

5. The multi-robot control method based on the Industrial Internet of Things according to claim 1, characterized in that, Step S2, which optimizes vibration suppression control based on core loosening data, includes: The main vibration frequency band was determined based on the data of core loosening. Determine the target frequency band suppression index based on the main vibration frequency band; Select a bandpass damper based on the target frequency band suppression index, and perform frequency domain simulation of the structural response based on the bandpass damper to obtain the frequency domain data of the structural response. Vibration suppression control optimization is performed based on the frequency domain data of the structural response to obtain vibration suppression control data.

6. The multi-robot control method based on the Industrial Internet of Things according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Determine the conflict location based on the path conflict data to obtain conflict location data; Step S32: Adjust the task allocation order based on conflict location data; Step S33: Perform robot scheduling according to the task allocation order to obtain robot scheduling data; Step S34: Simulate robotic arm handling based on robot scheduling data, where the single load mass is set to 2kg-50kg, and obtain robotic arm handling data; Step S35: Detect surface spalling of bearing rolling elements based on robotic arm handling data to obtain surface spalling data; Step S36: Detect eccentric rolling elements based on the robotic arm handling data to obtain eccentric rolling element data; Step S37: Integrate surface spalling data and eccentric rolling element data to obtain bearing fault data.

7. The multi-robot control method based on the Industrial Internet of Things according to claim 6, characterized in that, Step S35 includes the following steps: Step S351: Locate the bearing rolling element area based on the robotic arm handling data; Step S352: Determine the rolling element contact trajectory based on the rolling element area of ​​the bearing; deploy a particle collector according to the rolling element contact trajectory and collect the detached material to obtain detached material data; Step S353: Perform fracture morphology detection based on the data of the detached material to obtain fracture morphology data; Step S354: Perform slip fringe analysis based on the fracture morphology data to obtain slip fringe data; Step S355: Perform particle stripping detection based on the slip stripe data to obtain particle stripping data; Step S356: Calculate the surface peeling amount based on the particle peeling data to obtain surface peeling data.

8. The multi-robot control method based on the Industrial Internet of Things according to claim 6, characterized in that, Step S36 includes the following steps: Step S361: Determine the load disturbance of the end effector based on the robotic arm handling data, and obtain the load disturbance data; Step S362: Calculate the joint drive current based on the load disturbance data, where the standard deviation of the current fluctuation is set to 0.1-3.0; Step S363: Identify the periodic abnormal load of the drive shaft based on the joint drive current and obtain the abnormal load data of the drive shaft; Step S364: Perform dynamic imbalance analysis of the rolling bearing based on the abnormal load data of the drive shaft to obtain dynamic imbalance data; Step S365: Determine the asymmetry of the rolling elements based on the dynamic imbalance data to obtain the eccentric rolling element data.

9. A multi-robot control system based on the Industrial Internet of Things, characterized in that, For executing the multi-robot control method based on the Industrial Internet of Things as described in claim 1, the multi-robot control system based on the Industrial Internet of Things includes: A multi-robot path conflict detection module is used to acquire industrial warehouse aisle data; generate a warehouse digital twin model based on the industrial warehouse aisle data; and perform multi-robot path conflict detection based on the warehouse digital twin model to obtain path conflict data. The vibration suppression control optimization module is used to determine the robot number based on path conflict data; perform motor winding damage detection based on the robot number to obtain motor winding damage data; perform core loosening analysis based on the motor winding damage data to obtain core loosening data; and optimize vibration suppression control based on the core loosening data to obtain vibration suppression control data. The bearing fault detection module is used to perform robot scheduling based on path conflict data to obtain robot scheduling data; to perform robotic arm handling simulation based on robot scheduling data to obtain robotic arm handling data; and to perform bearing fault detection based on robotic arm handling data to obtain bearing fault data. The multi-robot control module is used to perform emergency braking control based on bearing fault data to obtain emergency braking control data; the emergency braking control data and vibration suppression control data are transmitted to the robot industrial IoT management platform to execute multi-robot control tasks.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-robot control method based on the Industrial Internet of Things as described in any one of claims 1 to 9.

Images