Parallel jacking, transplanting and conveying equipment and method thereof

Abstract

The invention discloses parallel jacking, transplanting and conveying equipment which comprises a frame, a jacking mechanism used for jacking a chassis and a horizontal control mechanism are arranged on the inner side of the frame, a plurality of parallel belt conveying mechanisms driven by rollers are arranged on the chassis, and the parallel jacking, transplanting and conveying equipment further comprises an intelligent tension cooperative control system. The system comprises a distributed tension monitoring network, a self-calibration sensing module, a layered data processing module, a tension abnormal mode recognition module, a digital twin feedback optimization module, a multi-belt synchronous control module and a fault prediction and maintenance decision module. The stability and reliability of equipment operation are improved, the accuracy of the material conveying process is guaranteed, and the precision and stability of material conveying are improved; according to the invention, high-precision control of belt tension and cooperative operation of each mechanism are realized, and the stability and reliability of the system are ensured; and the service life of the belt is prolonged, the equipment failure rate is reduced, and the maintenance cost is reduced.

Description

A parallel lifting transplanting and conveying device and method thereof Technical Field

[0001] This invention relates to the field of conveying equipment and system control technology, specifically to a parallel lifting transplanting conveying equipment and method thereof. Background Technology

[0002] Lifting and transferring machines are mainly used to change the direction of goods transport, moving items from forklifts into or out of the main conveyor line. They are characterized by high load capacity, simple structure, and stable reliability. However, existing lifting and transferring machines still have some problems. Lifting and transferring machines are required to achieve fast and efficient lifting operations, ensure the stability and reliability of the transport process, and simultaneously consider the equipment's service life and maintenance costs.

[0003] With the rapid development of industrial production, mechanical equipment is developing towards intelligence, complexity, automation, precision, and high speed. The connections between various components are becoming closer, and coupled with their complex working environment, mechanical equipment is prone to failure and is difficult to operate and maintain.

[0004] Traditional fault diagnosis techniques include expert systems, model analysis, and signal processing. However, these methods suffer from drawbacks such as poor versatility, complex modeling, weak robustness, and poor representation capabilities, severely limiting their application. In fault diagnosis, sensor data is singular, and the converted image data is also acquired separately by each sensor, thus failing to reflect all the characteristics of the fault from multiple perspectives. Data-level information fusion ensures the integrity of the original information. Therefore, data fusion from multiple sensors is of great significance for fault diagnosis.

[0005] Currently, some methods based on digital twin technology have been applied to equipment fault diagnosis and prediction. However, traditional methods for creating data twins lack the ability to deeply analyze complex data and environmental changes, making it difficult to accurately capture and reflect performance changes under complex environmental conditions, thus affecting prediction accuracy and system responsiveness. Furthermore, existing predictive maintenance systems still need improvement in classification accuracy and processing efficiency when handling large amounts of unlabeled fault data. Therefore, there is an urgent need for an intelligent monitoring and control system that can integrate multi-source sensor information, improve fault diagnosis accuracy, and achieve predictive maintenance.

[0006] Traditional fault diagnosis techniques have poor versatility, complex modeling, weak robustness, and poor characterization capabilities, making it difficult to accurately reflect performance changes under complex environmental conditions, which affects prediction accuracy and system response capabilities.

[0007] The existing distributed tension monitoring network still has issues with sensor layout and data transmission methods that need further optimization, affecting the accuracy and real-time performance of the data.

[0008] Current predictive maintenance systems still need to improve classification accuracy and processing efficiency when dealing with large amounts of unlabeled fault data, and struggle to achieve in-depth analysis capabilities for complex data and environmental changes.

[0009] Existing intelligent monitoring and control systems lack an effective fusion mechanism for multi-source heterogeneous data, and cannot fully utilize the advantages of each sensor, resulting in insufficient accuracy and comprehensiveness in fault diagnosis.

[0010] Traditional fault prediction models lack the ability to dynamically adjust the operating status of equipment, making it difficult to adapt to complex and ever-changing operating environments and unable to effectively extend equipment lifespan and reduce maintenance frequency. Summary of the Invention

[0011] To solve the above-mentioned technical problems, the present invention provides a parallel lifting transplanting conveyor, including a frame, a chassis at the top of the frame, a lifting mechanism and a horizontal control mechanism for lifting the chassis on the inner side of the frame, and a plurality of parallel belt conveyor mechanisms driven by rollers on the chassis, wherein the belt conveyor mechanisms are equipped with tensioning devices; the lifting transplanting conveyor also includes an intelligent tension coordination control system, which includes:

[0012] Distributed tension monitoring network: used to monitor belt tension distribution. It determines the key stress points of the belt system through finite element analysis. 3-5 piezoresistive tension sensors are arranged on each belt. The sensors are wirelessly powered. A complete tension field is constructed through mathematical interpolation algorithm to achieve high-precision monitoring of belt tension distribution.

[0013] The self-calibration sensing module employs a Kalman filter-based sensor fusion algorithm, combined with a belt dynamic model, to achieve mutual calibration between sensors, eliminating temperature drift and long-term usage errors. The system processes the calibration results through an edge computing unit, and the central controller performs global optimization to ensure an overall system measurement accuracy of ±0.3%.

[0014] Hierarchical Data Processing Module: A three-tiered data processing structure is designed, including local processing at sensor nodes, regional integration at edge computing units, and global optimization by the central controller. Sensor nodes utilize ARM Cortex-M4 processors with a sampling rate of 10kHz. Edge computing units employ ARM Cortex-A72 architecture, equipped with quad-core 2GHz processors, achieving processing latency within 5ms. The central controller utilizes a high-performance server with an Intel Core i7 processor, supporting 100Gbps bandwidth. This architecture significantly reduces communication bandwidth requirements and enables real-time fault warning.

[0015] Tension anomaly pattern recognition module: Based on wavelet transform and deep learning, the tension fluctuation feature extraction algorithm is deployed through a central controller and trained using the Adam optimization algorithm to provide early warning of potential faults;

[0016] Digital twin feedback optimization module: Constructs a high-precision digital model that incorporates the nonlinear characteristics of the belt material, and uses real-time data to drive simulation and predict changes in tension distribution; it uses a GPU accelerator for simulation calculations, and compares the real-time monitoring data with the central controller to guide the tensioning mechanism to make pre-adjustments;

[0017] Multi-belt synchronization control module: For multi-belt systems, vision-based real-time monitoring of belt running trajectory is adopted, combined with differential speed compensation algorithm to improve the synchronization accuracy of multi-belt systems when running at high speed; real-time data processing and optimization are realized through central controller;

[0018] Fault prediction and maintenance decision-making module: Based on deep learning algorithms, it analyzes equipment vibration, temperature, and tension data to predict potential faults and automatically adjust operating parameters; it achieves high-speed data transmission through industrial-grade 5G networks and uses digital twin technology to build a virtual image, enabling equipment status visualization and parameter optimization; it deploys lightweight neural networks through edge computing units for local prediction, combined with cloud-based deep reinforcement learning models, to predict the remaining service life of the equipment; it uses Bayesian optimization algorithms to dynamically adjust maintenance thresholds and automatically generates optimal maintenance time windows and specific maintenance item lists based on equipment operating history and current production plans.

[0019] Preferably, the belt conveyor mechanism further includes a real-time tension distribution mapping module, which includes a torque sensor located on the roller bearing seat and infrared thermal imaging monitoring located on each section of the belt, constructs a tension-temperature-deformation relationship model, and generates a belt tension distribution map in real time.

[0020] Preferably, the lifting mechanism includes an eccentric cam, the frame is rectangular and welded from channel steel on four sides, including two longitudinal beams and two transverse beams, a first rotating shaft is provided between the two opposite transverse beams, bearings are provided on both sides of the first rotating shaft, the bearings are located on a first bearing seat, and the first bearing seat is connected to the transverse beam by bolts; eccentric cams are provided at both ends of the first rotating shaft inside the first bearing seat; the lifting mechanism also includes a rotation drive structure.

[0021] Preferably, the rotation drive structure includes a servo motor, a bottom beam is provided between the two longitudinal beams, the servo motor and the gearbox are fixedly installed on the bottom beam, the output shaft of the gearbox is provided with a drive pulley, the first rotating shaft is provided with a concentric driven pulley, the drive pulley and the driven pulley are opposite to each other and are connected by a bottom belt; the intermediate roller is rolledly connected to the eccentric cam to form a cam mechanism.

[0022] Preferably, the horizontal control mechanism includes a second rotating shaft located between the two ends of two opposing crossbeams. Bearings are provided on both sides of the second rotating shaft, and the bearings are located on second bearing seats. The second bearing seats are bolted to the crossbeams. Connecting members are connected to both ends of the second rotating shaft via pins inside the second bearing seats. The connecting members are curved, and the upper end of the connecting member is rotatably connected to a fixing member. The fixing member is fixedly connected to the chassis and can be fastened to the chassis with a nut. The fixing member is perpendicular to the chassis. The lower end of the connecting member is rotatably connected to a connecting rod, which is connected to the lower end of the connecting member on the other side of the second rotating shaft. The connecting members on both sides of the second rotating shaft have the same shape and direction, and the fixing members are also the same. The installation height of the second rotating shafts at both ends is the same. Thus, the bottom surface of the chassis, the fixing members on both sides, the connecting members, and the connecting rod form a parallelogram, and two sets of synchronous parallelogram linkage mechanisms are provided under the chassis.

[0023] Preferably, the chassis is provided with multiple belt conveyor mechanisms arranged in parallel. Each belt conveyor mechanism is driven by a longitudinal roller and includes a first roller at each of the upper ends, a second roller inside the first roller, and subsequently, a third, fourth, and fifth roller inside the second roller, with a sixth roller in the middle at the top. The belt passes sequentially through the upper rollers into the lower section, then sequentially through the seventh, eighth, and ninth rollers, before entering the upper first roller. The roller shafts are all fixed to a vertical plate, which is fixed to the chassis and perpendicular to it. The upper belts are located on the same plane. The fourth roller is a tensioning roller, and its shaft is located on the vertical plate, allowing it to move up and down.

[0024] Preferably, a set of diffuse reflection photoelectric sensors for detecting materials is provided at both the inlet and outlet of the belt conveyor mechanism to realize the linkage control of belt conveying and lifting actions.

[0025] A parallel lifting transplanting and conveying method includes the following steps: Initial state: the chassis is in the lowest position, the belt of the belt conveyor mechanism is kept horizontal, and the photoelectric sensor is in standby state;

[0026] Material conveying: When the material passes through the inlet photoelectric sensor, the sensor sends a signal to the control system, and the roller drives the belt to move, conveying the material to the designated transplanting position;

[0027] Positioning detection: When the material reaches the position of the photoelectric sensor at the outlet, the sensor sends a signal, the belt conveyor stops running, and the material is accurately positioned;

[0028] Parallel lifting: The control system commands the servo motor to reverse, and the power drives the first rotating shaft to rotate through the transmission mechanism. The bias cam on the first rotating shaft rotates synchronously, and the cam profile pushes the intermediate roller to move upward. The intermediate roller drives the chassis to rise through the horizontal fixed plate. During this process, two sets of synchronous parallelogram linkage mechanisms move synchronously under the action of the synchronous shaft to ensure that the chassis always remains horizontal and realizes the vertical parallel lifting of materials.

[0029] Transplanting and docking: When the chassis rises to the preset height, the external transplanting equipment will remove or transfer the material to the target location;

[0030] Reset process: After the material transfer is completed, the control system commands the servo motor to rotate forward, the cam mechanism drives the chassis to descend smoothly to the initial position, the belt conveyor mechanism resumes operation, and enters the next work cycle.

[0031] This invention presents a tension fluctuation feature extraction algorithm based on wavelet transform and deep learning, capable of identifying 12 typical belt anomaly patterns. The system deploys a deep learning model through a central controller, trained using the Adam optimization algorithm, achieving an accuracy rate exceeding 98% and providing a 200ms early warning of potential faults.

[0032] This invention, an intelligent tension collaborative control system, solves the problems of poor universality of fault diagnosis technology, insufficient fusion of multi-source heterogeneous data, and lack of dynamic adjustment capability in existing technologies, providing an intelligent tension collaborative control system. This invention achieves high-precision control of belt tension and coordinated operation of various mechanisms, while ensuring system stability and reliability; it extends belt life, reduces equipment failure rate, and lowers maintenance costs.

[0033] This invention employs a cam mechanism to lift the conveyor. Through the synergistic effect of an eccentric cam and a synchronous parallelogram linkage mechanism, the material is lifted smoothly, avoiding the impact acceleration problem of traditional cylinder or cam lifting methods, and significantly improving the stability and reliability of the equipment operation.

[0034] This invention, through the design of two sets of synchronous parallelogram linkage mechanisms, ensures the consistency and levelness of the conveyor, effectively solving the problem of levelness error caused by low installation accuracy in the prior art, and ensuring the accuracy of the material conveying process;

[0035] This invention employs a belt conveyor mechanism with a multi-stage roller design, which, in conjunction with tensioning rollers, achieves precise belt tensioning, avoids belt misalignment, and further improves the accuracy and stability of material conveying.

[0036] By arranging multiple micro-MEMS strain sensors at key stress points in the belt system and constructing a complete tension field using mathematical interpolation algorithms, high-precision tension distribution monitoring was achieved, overcoming the limitations of traditional single-point monitoring and significantly improving the measurement accuracy and reliability of the system.

[0037] By employing a sensor fusion algorithm based on Kalman filtering and combining it with a belt dynamic model, mutual calibration between sensors was achieved, effectively eliminating temperature drift and long-term usage errors, improving the stability and reliability of the system, and increasing the overall measurement accuracy of the system to ±0.3%.

[0038] A three-level data processing structure was designed, including local processing at sensor nodes, regional integration at edge computing units, and global optimization by the central controller. This significantly reduced communication bandwidth requirements, improved data processing efficiency, and enabled real-time fault warning.

[0039] The tension anomaly pattern recognition algorithm based on wavelet transform and deep learning can identify 12 typical belt anomaly patterns and provide early warning of potential faults 200ms in advance, effectively improving the accuracy and timeliness of fault prediction.

[0040] By constructing a high-precision digital model that incorporates the nonlinear characteristics of belt materials, and combining it with real-time data-driven simulation to predict tension distribution changes, precise monitoring and pre-adjustment of the belt system were achieved, effectively controlling tension fluctuations within ±1.5%, and improving the system's operational stability and controllability.

[0041] The system adopts a low-power design, supports wireless power supply, simplifies installation and maintenance, and achieves self-learning through edge AI algorithms. After 3 months of operation, the fault prediction accuracy has been improved to over 95%, effectively extending the equipment's lifespan by 40%, reducing maintenance frequency, extending the maintenance cycle by 35%, improving the overall reliability of the equipment by 45%, and improving the product delivery and positioning accuracy to ±0.2mm, significantly improving the system's economic benefits and practical value. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 is a perspective view of the parallel lifting transplanting and conveying equipment of the present invention.

[0044] Figure 2 is a schematic diagram of the belt conveyor mechanism.

[0045] Figure 3 is a schematic diagram of the bottom structure of the present invention.

[0046] Figure 4 is a schematic diagram of the present invention with the transverse fixing plate removed.

[0047] Figure 5 is a bottom view of the present invention.

[0048] Figure 6 is a schematic diagram of the intelligent tension collaborative control system of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] The digital model is represented as

[0051] Example 1

[0052] As shown in Figures 1-6: This invention discloses a parallel lifting transplanting conveyor, including a frame, a chassis 12 at the top of the frame, a lifting mechanism and a horizontal control mechanism for lifting the chassis 12 on the inner side of the frame, and multiple parallel belt conveying mechanisms driven by rollers 9 on the chassis 12, each belt conveying mechanism being equipped with a tensioning device; the lifting transplanting conveyor also includes an intelligent tension coordination control system, which includes:

[0053] Distributed tension monitoring network: used to monitor belt tension distribution, determine key stress points of the belt system through finite element analysis, and deploy 3-5 piezoresistive tension sensors 32 on each belt. The piezoresistive tension sensors are miniature MEMS strain sensors, which are wirelessly powered. A complete tension field is constructed through mathematical interpolation algorithms.

[0054] The key stress points of the belt were identified using the finite element analysis (FEA) method, including: the roller envelope section, the section with abrupt change in idler spacing, the belt overlap section, and the transition zone between the heavy-load and unloaded sections. The finite element analysis (FEA) method will not be detailed here. Sensor placement principles:

[0055] The sensor evenly covers the width of the belt to avoid edge effect interference;

[0056] The spacing between adjacent sensors shall not exceed the stress influence radius obtained from the finite element analysis;

[0057] The sensor is installed on the non-working surface of the belt to reduce direct friction with materials and idlers.

[0058] Based on mathematical interpolation algorithms, a continuous tension field is constructed over the entire belt using discrete tension data from a finite number of sensors. The specific steps are as follows:

[0059] Data preprocessing: Filter and reduce noise in the raw data collected by the sensor (e.g., using Kalman filtering) to remove outliers.

[0060] Interpolation calculation: Kriging interpolation (suitable for tension data with strong spatial correlation) or cubic spline interpolation (suitable for fitting smooth tension fields) is used to establish a functional relationship between tension values ​​and spatial location. ,in Let's say the coordinates along the length of the belt. The coordinates are in the width direction of the belt.

[0061] Kriging interpolation formula:

[0062]

[0063] In the formula:

[0064] The tension estimate for the point to be interpolated;

[0065] For the first The measured tension values ​​of each sensor;

[0066] Let be the weighting coefficient, satisfying ;

[0067] The interpolation results are mapped to two-dimensional / three-dimensional tension cloud maps to intuitively display the tension distribution of the belt.

[0068] Self-calibration sensing module: Employs a sensor fusion algorithm based on Kalman filtering, combined with a belt dynamic model, to achieve mutual calibration between sensors, eliminating temperature drift and long-term usage errors.

[0069] The hierarchical data processing module is designed with a three-level data processing structure, including local processing at sensor nodes, regional integration at edge computing units, and global optimization by the central controller. The sensor nodes use ARM Cortex-M4 architecture processors, and the edge computing units use ARM Cortex-A72 architecture processors to achieve real-time fault early warning. The local edge processing layer uses an ARM Cortex-M4 32-bit microcontroller for local preprocessing and initial anomaly screening. The edge computing units are the regional integration and analysis layer, using ARM Cortex-A72 multi-core processors. The central controller is the global optimization decision layer, used for global collaborative optimization and fault prediction.

[0070] Tension anomaly pattern recognition module: Based on wavelet transform and deep learning, the tension fluctuation feature extraction algorithm is deployed through a central controller and trained using the Adam optimization algorithm to provide early warning of potential faults.

[0071] The tension anomaly pattern recognition module includes a signal preprocessing layer, a feature extraction layer, a model training and deployment layer, and a fault early warning layer. Considering the temporal and high-dimensional nature of tension fluctuation features, a CNN-LSTM hybrid model is selected as the core recognition model.

[0072] CNN submodule: Utilizes the local feature extraction capabilities of convolutional and pooling layers to model the spatial correlation of wavelet feature vectors and automatically extract deep local features;

[0073] LSTM submodule: Captures long-term dependencies of feature sequences, adapts to the temporal evolution of tension fluctuations, and effectively identifies the gradual characteristics of fault precursors;

[0074] Fully connected layer and output layer: The deep features extracted by CNN-LSTM are mapped to the fault category space. The output layer uses the Sigmoid function (binary classification: normal / fault precursor) or the Softmax function (multi-classification: different fault types).

[0075] Model training based on Adam optimization algorithm: Collect wavelet feature vectors of normal state and typical fault precursor state of tension system (such as sudden tension change, abnormal periodic fluctuation, trend drift, etc.), and divide them into training set, validation set and test set in a ratio of 7:2:1.

[0076] Digital twin feedback optimization module: Constructs a high-precision digital model that includes the nonlinear characteristics of belt materials, and uses real-time data to drive simulation and predict changes in tension distribution; uses a GPU accelerator for simulation calculations, and compares the real-time monitoring data with the central controller to guide the tensioning mechanism to make pre-adjustments.

[0077] This solution focuses on modeling the nonlinear characteristics of belt materials, GPU-accelerated real-time simulation, data-driven tension prediction, and closed-loop pre-adjustment of the tensioning mechanism to build a high-precision digital twin control system.

[0078] I. Construction of a High-Precision Digital Model for the Nonlinear Characteristics of Belt Materials

[0079] This invention employs a hyperelastic constitutive model to describe the large deformation nonlinear characteristics of belt materials, using the Mooney-Rivlin model, whose strain energy density function is:

[0080]

[0081] in:

[0082] —Material property parameters were obtained through uniaxial tensile testing and least squares fitting.

[0083] —The first and second strain invariants;

[0084] —Volume strain, when the belt is approximately incompressible ;

[0085] — Volume compressibility factor.

[0086] If the belt exhibits significant creep (plastic deformation under long-term load), a Burger creep model correction is applied:

[0087]

[0088] in For the elastic modulus component, The relaxation time constant is calibrated through long-term creep tests under different loads.

[0089] Introducing the temperature-load-wear multi-field coupling factors, a correction coefficient matrix is ​​established:

[0090] Temperature correction: Real-time belt temperature is collected via thermocouples, and the elastic modulus is established based on experimental data. The functional relationship with temperature is embedded in the constitutive model;

[0091] Wear correction: By combining belt mileage and surface wear monitoring data, the belt cross-sectional area and friction coefficient are corrected, which affects the tension transmission efficiency;

[0092] Load distribution correction: Real-time loads in different sections of the belt are collected by distributed tension sensors and used as boundary conditions for the model.

[0093] The belt is discretized into n infinitesimal segments along its length. Each infinitesimal segment is considered as an independent nonlinear spring-damped element. The overall dynamic equations are constructed using the finite element method (FEM).

[0094]

[0095] in It is a nonlinear stiffness matrix that changes in real time with the strain of the micro-element. This is the external excitation force for the drive wheel and tensioner wheel.

[0096] A parameter identification and optimization model driven by real-time data is used: tension and strain monitoring data during belt operation are used to iteratively correct the model using a particle swarm optimization (PSO) algorithm. By controlling core parameters such as these, the model can achieve online self-calibration.

[0097] II. GPU-accelerated real-time tension distribution simulation calculation

[0098] Utilizing a CPU-GPU heterogeneous computing model, acceleration strategies are optimized:

[0099] Mesh partitioning parallelism: The discretized belt micro-segments are distributed to multiple thread blocks of the GPU, with each thread responsible for solving the dynamic equations of one micro-segment, utilizing the parallel computing capabilities of the GPU's massive cores;

[0100] Sparse matrix optimization: Nonlinear stiffness matrix As a sparse matrix, it adopts the CSR (Compressed Sparse Row) storage format to reduce memory usage and improve computational efficiency;

[0101] Iterative solver selection: The conjugate gradient method (CG) or GMRES algorithm is selected to solve the nonlinear equation system. Compared with direct solvers (such as LU decomposition), the computational efficiency is improved by 1 to 2 orders of magnitude, which meets the real-time requirements (single solution time <10ms).

[0102] Real-time simulation process;

[0103] 1. The CPU receives data such as belt speed, temperature, and initial tension from the sensors, which are used as input boundary conditions for the GPU simulation.

[0104] 2. The GPU loads the latest nonlinear constitutive model parameters, solves the dynamic equations in parallel, and outputs the predicted value of the tension distribution in the entire segment;

[0105] 3. The GPU sends the prediction results back to the CPU to complete one simulation cycle. The cycle can be set according to actual needs (50~100ms recommended).

[0106] III. Real-time data comparison of the central controller and pre-adjustment of the tensioning mechanism

[0107] Employing industrial-grade PLCs or embedded controllers, the architecture is divided into three layers:

[0108] Sensing layer: Receives real-time data from tension sensors, temperature sensors, and encoders;

[0109] Decision-making level: Compare the model-predicted tension with the measured tension, and calculate the deviation value. ;

[0110] Execution layer: Outputs tensioning mechanism adjustment commands based on deviation values.

[0111] Feedforward pre-adjustment: Based on the tension change trend predicted by the GPU (such as a sudden increase in tension during the startup phase or a decrease in tension before a sudden load change), the tension command is output in advance to suppress the source of tension fluctuations;

[0112] Compare the deviations between predicted and measured values ,like Initiate PID feedback regulation to correct the error of feedforward control;

[0113] PID control law: ,in This is the adjustment amount for the tensioning mechanism.

[0114] This technical solution has small tension prediction error, high real-time performance and adaptability. Data-driven parameter identification enables the model to automatically optimize as the belt ages and working conditions change, maintaining high accuracy over a long period.

[0115] Multi-belt synchronization control module: For multi-belt systems, vision-based real-time monitoring of belt running trajectory is adopted, combined with differential speed compensation algorithm to ensure that the multi-belt system maintains synchronization accuracy of ±0.5mm when running at high speed; real-time data processing and optimization are realized through central controller.

[0116] This solution uses machine vision technology to collect and analyze the running trajectory parameters of each belt in real time, providing a precise data source for subsequent compensation. Addressing the speed coupling characteristics of multi-belt systems, a differential speed compensation algorithm is employed to offset synchronization errors caused by trajectory deviations through real-time speed adjustment of each belt drive motor.

[0117] Let the first belt in a multi-belt system be... The real-time speed of the belt is The base speed is The trajectory offset is Then the compensation speed The differential equation is:

[0118]

[0119] in:

[0120] This is a scaling factor used for fast response to static offset;

[0121] These are differential coefficients used to suppress abrupt changes in offset and avoid overcompensation.

[0122] The central controller receives offset data from the vision module. Substitute into the differential equation to solve for the real-time compensation speed Output speed control command to the first The variable frequency drive device for the belt adjusts the motor speed to bring the belt track back to the baseline.

[0123] Central controller data processing and optimization:

[0124] Data acquisition layer: Receives trajectory data from each edge node and speed / tension sensor data from each belt in real time via industrial Ethernet (such as Profinet);

[0125] Operational Analysis Layer:

[0126] Run the differential speed compensation algorithm to generate speed control commands;

[0127] Establish a multi-belt synchronization accuracy evaluation model and calculate the overall system synchronization error. ;

[0128] Train machine learning on historical data and optimize , Parameters enable the algorithm to be adaptive.

[0129] Command sending layer: Sends speed control commands and abnormal alarm signals to the corresponding execution units (frequency converters, audible and visual alarms).

[0130] Data storage layer: Stores real-time operational data and historical fault information, supporting traceability analysis and report generation.

[0131] The Time-Sensitive Networking (TSN) protocol is adopted to ensure low latency (≤10ms) and high reliability of data transmission, meeting the real-time requirements of high-speed operation scenarios.

[0132] This solution can improve synchronization accuracy during high-speed operation, with strong real-time performance and good adaptability.

[0133] Fault prediction and maintenance decision-making module: Based on deep learning algorithms, it analyzes equipment vibration, temperature, and tension data to predict potential faults and automatically adjust operating parameters; it achieves high-speed data transmission through industrial-grade 5G networks and uses digital twin technology to build a virtual image, enabling equipment status visualization and parameter optimization; it deploys lightweight neural networks through edge computing units for local prediction, combined with cloud-based deep reinforcement learning models, to predict the remaining service life of the equipment; it uses Bayesian optimization algorithms to dynamically adjust maintenance thresholds and automatically generates optimal maintenance time windows and specific maintenance item lists based on equipment operating history and current production plans.

[0134] This solution, based on deep learning algorithms, constructs a feature extraction and correlation analysis model for multi-source sensor data such as equipment vibration, temperature, and tension. This model accurately identifies weak characteristic signals of early equipment failures, enabling early warning of potential malfunctions. Simultaneously, based on the fault warning level, the model automatically outputs optimized adjustment schemes for equipment operating parameters (such as speed, load, and pressure), and corrects the equipment's operating status in real time through a closed-loop control link, reducing the probability of failure.

[0135] Leveraging the high bandwidth and low latency of industrial-grade 5G networks, high-speed and stable transmission of multi-source sensor data is achieved. Based on digital twin technology, a virtual mirror image mapped 1:1 to the physical device is constructed, integrating the device's geometric model, operational mechanism model, and data-driven model, synchronizing the device's physical status and operating parameters in real time. Through the digital twin visualization platform, it supports 3D panoramic display of the device's operating status, trend analysis of key parameters, and fault tracing simulation. Simultaneously, parameter simulation experiments are conducted using the virtual mirror image to quickly find the optimal operating parameter range for the device and distribute it to the physical device.

[0136] Employing an edge-cloud collaborative computing architecture, we can achieve tiered and accurate prediction of the remaining lifespan of devices.

[0137] Edge side: Deploy lightweight neural networks (such as MobileNet and lightweight Transformer) within the edge computing unit to complete short-term health status assessment and preliminary RUL prediction of the device based on locally collected real-time sensor data, meeting the requirements of low latency and localized decision-making;

[0138] On the cloud side: Deploy a deep reinforcement learning (DRL) model, gather full lifecycle operation data and historical fault records from multiple similar devices, and build a global, high-precision RUL prediction model through trial and error and iterative optimization of reinforcement learning; the edge side and the cloud side achieve complementary integration of short-term and long-term predictions through lightweight updates and interactions of model parameters.

[0139] Based on Bayesian optimization algorithms, and combining historical equipment failure data, maintenance records, and current production schedules, the system dynamically adjusts maintenance thresholds (such as vibration and temperature thresholds) for equipment health status to avoid over- or under-maintenance. Simultaneously, using equipment RUL prediction results as the core input, and comprehensively considering constraints such as production task priority, maintenance resource scheduling costs, and downtime losses, the system automatically generates optimal maintenance time windows (such as production gaps and low-load periods) and a detailed maintenance item list (including spare parts to be replaced, maintenance procedures, and technical standards), achieving coordinated optimization of maintenance strategies and production plans.

[0140] The real-time tension distribution mapping module includes a torque sensor (accuracy ±0.2%) installed in the roller bearing housing and infrared thermal imaging monitoring (temperature resolution 0.05℃) installed in each section of the belt. It constructs a tension-temperature-deformation relationship model, generates a belt tension distribution map in real time, and identifies areas with abnormal tension.

[0141] The lifting mechanism includes an eccentric cam 23. The frame is rectangular and includes two longitudinal beams 1 and two transverse beams 2. A first rotating shaft 21 is connected between the two opposing transverse beams 2. Bearings are provided on both sides of the first rotating shaft 21. The bearings are located on a first bearing seat 22. The first bearing seat 22 is connected to the transverse beams 2 by bolts. Eccentric cams 23 are provided at both ends of the first rotating shaft 21 inside the first bearing seat 22. The lifting mechanism also includes a rotation drive structure.

[0142] The rotation drive structure includes a motor 30, a bottom beam 24 connected between the two longitudinal beams 1, a motor 30 and a gearbox fixedly mounted on the bottom beam 24, a drive pulley 25 on the output shaft of the gearbox, a concentric driven pulley 26 on the first rotating shaft 21, the drive pulley 25 and the driven pulley 26 being opposite to each other and connected by a bottom belt 27; the chassis 12 has transverse fixing plates 28 at both ends, and an intermediate roller 29 on the outer side of the transverse fixing plate 28, the intermediate roller 29 being rolledly connected to the eccentric cam 23 to form a cam mechanism.

[0143] The horizontal control mechanism includes a second rotating shaft 16 located between the two ends of two opposing crossbeams 2. Bearings are provided on both sides of the second rotating shaft 16, and the bearings are located on the second bearing seats 17. The second bearing seats 17 are connected to the crossbeams 2 by bolts. Connecting members 18 are connected to the two ends of the second rotating shaft 16 inside the second bearing seats 17 by pins. The connecting members 18 are curved. The upper end of the connecting member 18 is rotatably connected to the upper fixing member 19. The upper fixing member 19 is fixedly connected to the chassis 12 and can be fastened to the chassis 12 by nuts. The upper fixing member 19 is perpendicular to the chassis 12. The lower end of the connecting member 18 is rotatably connected to the connecting rod 20. The connecting rod 20 is connected to the lower end of the connecting member 18 on the other side of the second rotating shaft. The connecting members 18 on both sides of the second rotating shaft 16 have the same shape and direction, and the upper fixing members 19 are also the same. The installation height of the second rotating shafts at both ends is the same. In this way, the bottom surface of the chassis 12, the upper fixing parts 19 on both sides, the connecting parts 18 and the connecting rods 20 form a parallelogram, and two sets of synchronous parallelogram linkage mechanisms are provided under the chassis 12.

[0144] The roller 9 is vertically arranged, and the belt conveyor mechanism includes multiple rotating rollers with upright plates 6 and belt 3. The upright plates 6 are fixed on the chassis 12 and are perpendicular to the chassis 12.

[0145] The tensioning device includes a fourth roller 10, the shaft of which is mounted on the vertical plate 6 and can move up and down.

[0146] The belt conveyor mechanism is equipped with a set of diffuse reflection photoelectric sensors at both the inlet and outlet for detecting materials. The detection distance is 0-500mm and the response time is ≤1ms. These sensors are used to detect the presence or absence of materials and to achieve linkage control between belt conveying and lifting actions (e.g., when the material reaches the designated position, the belt stops and the lifting mechanism starts).

[0147] Diffuse reflection photoelectric sensors enable the linkage control of belt conveying and lifting operations.

[0148] Eccentric cam 23 profile design: adopts a sinusoidal curve profile, with a push angle of 120°, a far rest angle of 30°, a return angle of 180°, and a near rest angle of 30°, to ensure a smooth and impact-free lifting process, with a maximum acceleration ≤0.5g (g is the acceleration due to gravity).

[0149] Multiple belt conveyor mechanisms are arranged in parallel on the chassis 12. The belt conveyor mechanism is driven by a longitudinal roller 9. The belt conveyor mechanism includes a first roller 4 at both ends of the upper part, a second roller 5 inside the first roller 4, a third roller 11, a fourth roller 10, and a fifth roller 13 inside the second roller 5 in sequence, and a sixth roller 14 in the middle of the upper part. The belt 3 passes through the upper rollers in sequence to enter the lower part, passes through the seventh roller 7, the eighth roller 8, the roller 9, and the ninth roller 15 in sequence, and then enters the upper first roller 4. The rotation shafts of the rollers are all fixed on the vertical plate 6, which is fixed on the chassis 12 and perpendicular to the chassis 12. The upper belt 3 is located on the same plane.

[0150] The connecting rod 20 is designed as a telescopic and adjustable structure, including an inner rod and an outer rod. The inner rod has a scale on its surface, and the outer rod has a locking bolt. By adjusting the length of the inner rod extending into the outer rod, the side length accuracy of the parallelogram can be calibrated. The adjustment range is 0-30mm, ensuring that the chassis 12 remains level during the lifting process.

[0151] This invention uses a single roller to drive multiple belt conveyors for product transport, and includes a tensioning mechanism. The lower part employs two sets of synchronous parallelogram linkage mechanisms to ensure the conveyors remain consistently horizontal, and a cam mechanism is used to lift the conveyors (highly efficient cycle time).

[0152] Initial state: Chassis 12 is in the lowest position, belt 3 of the belt conveyor mechanism is horizontal, and photoelectric sensor is in standby state;

[0153] Material conveying: When the material passes through the inlet photoelectric sensor, the sensor sends a signal to the control system, and the roller 9 drives the belt 3 to run, conveying the material to the designated transplanting position;

[0154] Positioning detection: When the material reaches the position of the photoelectric sensor at the outlet, the sensor sends a signal, the belt conveyor stops running, and the material is accurately positioned;

[0155] Parallel lifting: The control system commands the servo motor 30 to reverse, and the power drives the first rotating shaft 21 to rotate through the transmission mechanism. The bias cam 23 on the first rotating shaft 21 rotates synchronously, and the cam profile pushes the intermediate roller 29 to move upward. The intermediate roller 29 drives the chassis 12 to rise through the transverse fixed plate 28. During this process, two sets of synchronous parallelogram linkage mechanisms move synchronously under the action of the synchronous shaft to ensure that the chassis 12 always remains horizontal, realizing the vertical parallel lifting of materials.

[0156] Transplanting and docking: When the chassis 12 rises to the preset height (positioned by feedback from the servo motor encoder), external transplanting equipment (such as robotic arms or conveyor belts) will pick up or transfer the material to the target location.

[0157] Reset process: After the material transfer is completed, the control system commands the servo motor 30 to rotate forward, the cam mechanism drives the chassis 12 to descend smoothly to the initial position, the belt conveyor mechanism resumes operation, and enters the next work cycle.

[0158] Example 2:

[0159] This invention addresses the problems of poor universality of fault diagnosis technology, insufficient fusion of multi-source heterogeneous data, and lack of dynamic adjustment capability in existing technologies by providing an intelligent tension collaborative control system.

[0160] An intelligent tension collaborative control system includes a distributed tension monitoring network, a self-calibrating sensing module, a hierarchical data processing module, a tension anomaly pattern recognition module, a digital twin feedback optimization module, a multi-belt synchronous control module, and a fault prediction and maintenance decision module.

[0161] The distributed tension monitoring network comprises multiple belt systems, each including an input shaft, intermediate shaft, output shaft, and several planetary gears and bearings. The system uses finite element analysis to determine the key stress points of the belt systems. Three to five miniature MEMS strain sensors (5mm × 5mm in size) are deployed on each belt, achieving an accuracy of ±0.2%. The sensors are wirelessly powered, and a complete tension field is constructed using mathematical interpolation algorithms to achieve high-precision monitoring of belt tension distribution.

[0162] The self-calibration sensing module employs a sensor fusion algorithm based on Kalman filtering, combined with a belt dynamic model, to achieve mutual calibration between sensors, eliminating temperature drift and long-term usage errors. The system processes the calibration results through an edge computing unit, and the central controller performs global optimization to ensure that the overall system measurement accuracy reaches ±0.3%.

[0163] The hierarchical data processing module comprises local processing at sensor nodes (10kHz sampling rate), regional integration at edge computing units (processing latency <5ms), and global optimization by the central controller, reducing communication bandwidth requirements by 80%. The sensor nodes utilize ARM Cortex-M4 processors with a clock speed of 120MHz and 256MB of memory, achieving a sampling rate of 10kHz. The edge computing units employ ARM Cortex-A72 architecture, equipped with a quad-core 2GHz processor and 2GB of memory, with processing latency controlled within 5ms. The central controller utilizes a high-performance server, configured with an Intel Core i7 processor, 16GB of memory, and supports 100Gbps bandwidth. This architecture significantly reduces communication bandwidth requirements, enabling real-time fault warning.

[0164] The tension anomaly pattern recognition module, based on a tension fluctuation feature extraction algorithm using wavelet transform and deep learning, can identify 12 typical belt anomaly patterns. The system deploys a deep learning model through a central controller, trained using the Adam optimization algorithm, achieving an accuracy rate of over 98% and providing a 200ms early warning of potential faults.

[0165] The digital twin feedback optimization module constructs a high-precision digital model incorporating the nonlinear characteristics of the belt material. Through real-time data-driven simulation, it predicts changes in tension distribution and guides the pre-adjustment of the tensioning mechanism, controlling tension fluctuations within ±1.5%. The system uses a GPU accelerator for simulation calculations, achieving a simulation accuracy of ±0.1%. The central controller compares the simulation with real-time monitoring data, guiding the tensioning mechanism to make pre-adjustments and effectively controlling tension fluctuations within ±1.5%.

[0166] The multi-belt synchronization control module is designed for multi-belt systems. It employs vision-based real-time monitoring of the belt's running trajectory, combined with a differential speed compensation algorithm, to ensure a synchronization accuracy of ±0.5mm even at high speeds (linear speeds up to 2m / s). The system uses a central controller for real-time data processing and optimization, keeping control latency within 8ms. This solution achieves precise synchronization control of the belts under high-speed operating conditions through a differential speed compensation algorithm and vision-based real-time monitoring technology. The system will utilize an edge computing architecture and an ARM processor to ensure real-time control commands and low system response latency, meeting the stringent requirements of modern industrial automated production lines for material conveying accuracy and efficiency.

[0167] The fault prediction and maintenance decision-making module analyzes equipment vibration, temperature, and tension data using deep learning algorithms to predict potential faults and automatically adjust operating parameters. High-speed data transmission is achieved through an industrial-grade 5G network, and a virtual image is constructed using digital twin technology to visualize equipment status and optimize parameters. The system deploys lightweight neural networks on edge computing units for local prediction, combined with a cloud-based deep reinforcement learning model, to accurately predict the remaining service life of the equipment.

[0168] This invention employs a hierarchical deep learning architecture and an edge-cloud collaborative computing framework. The core system includes: 1) Multimodal sensor data fusion preprocessing: integrating data from vibration sensors (sampling rate 20kHz), temperature sensors (accuracy ±0.1℃), tension sensors (accuracy ±0.5%), and current sensors (accuracy ±0.2%), extracting features through wavelet transform and principal component analysis, reducing data dimensionality by 85%; 2) Lightweight edge neural network: deploying an optimized hybrid model of convolutional neural networks (CNN) and long short-term memory networks (LSTM) on a field-programmable gate array (FPGA), with the number of parameters controlled within 500K and inference latency <5ms, achieving [the desired result]. 3) Cloud-based deep reinforcement learning model: A time-series prediction model based on the Transformer architecture (12M parameters) is adopted, combined with the Monte Carlo Tree Search (MCTS) algorithm, to build a prediction model for the remaining service life of the equipment, improving the prediction accuracy to 97% and extending the prediction time span to 3 months; 4) Adaptive maintenance decision system: Based on the Bayesian optimization algorithm, the maintenance threshold is dynamically adjusted. Combined with the equipment operating history and current production plan, the optimal maintenance time window and specific maintenance item list are automatically generated, reducing unplanned downtime by 80%. The system continuously optimizes the model through an incremental learning mechanism, and the prediction accuracy can be improved by 0.5% for every 100 hours of new data accumulated. This solution increases the mean time between failures (MTBF) of the equipment by 45%, reduces maintenance costs by 35%, and reduces system resource consumption by 40%, achieving a balance between high-precision prediction and lightweight deployment.

[0169] Example 3:

[0170] An intelligent tension collaborative control system. This system includes a distributed tension monitoring network, a self-calibrating sensing module, a hierarchical data processing module, a tension anomaly pattern recognition module, a digital twin feedback optimization module, a multi-belt synchronous control module, and a fault prediction and maintenance decision-making module.

[0171] The distributed tension monitoring network comprises multiple belt systems, each including an input shaft, intermediate shaft, output shaft, and several planetary gears and bearings. The system uses finite element analysis to determine the key stress points of the belt systems. Three to five miniature MEMS strain sensors (4mm × 4mm in size) are deployed on each belt, achieving an accuracy of ±0.1%. The sensors are wirelessly powered, and a complete tension field is constructed using mathematical interpolation algorithms to achieve high-precision monitoring of belt tension distribution.

[0172] The self-calibration sensing module employs a sensor fusion algorithm based on Kalman filtering, combined with a belt dynamic model, to achieve mutual calibration between sensors, eliminating temperature drift and long-term usage errors. The system processes the calibration results through an edge computing unit, and the central controller performs global optimization to ensure that the overall system measurement accuracy reaches ±0.2%.

[0173] The hierarchical data processing module includes local processing at sensor nodes, regional integration at edge computing units, and global optimization by the central controller. The sensor nodes utilize ARM Cortex-M4 processors with a clock speed of 200MHz, 512MB of memory, and a sampling rate of 15kHz. The edge computing units employ ARM Cortex-A72 architecture, equipped with a 6-core 2.5GHz processor, 4GB of memory, and a processing latency controlled within 3ms. The central controller uses a high-performance server, configured with an Intel Core i9 processor, 32GB of memory, and a bandwidth support of 200Gbps. This architecture significantly reduces communication bandwidth requirements and enables real-time fault warning.

[0174] The tension anomaly pattern recognition module, based on wavelet transform and deep learning-based tension fluctuation feature extraction algorithms, can identify 12 typical belt anomaly patterns. The system deploys a deep learning model through a central controller, trained using the Adam optimization algorithm, achieving an accuracy rate of over 99% and providing a 300ms early warning of potential faults.

[0175] The digital twin feedback optimization module constructs a high-precision digital model incorporating the nonlinear characteristics of the belt material, and uses real-time data to drive simulation and predict changes in tension distribution. The system employs a GPU accelerator for simulation calculations, achieving a simulation accuracy of ±0.05%. By comparing real-time monitoring data with the central controller, it guides the tensioning mechanism to make pre-adjustments, effectively controlling tension fluctuations within ±1.0%.

[0176] This invention provides an adaptive tension control mechanism, comprising: 1) Multi-zone differential tension control: The roller surface is divided into multiple independent control zones (configured as 6-8 zones), each zone is equipped with an independent tension adjustment mechanism, and a combination structure of precision electric push rod (resolution 0.01mm) and torsion spring is adopted to achieve independent and precise control of the tension of each belt; 2) Real-time tension distribution mapping: By installing a high-precision torque sensor (accuracy ±0.2%) on the roller bearing seat and infrared thermal imaging monitoring of each section of the belt (temperature resolution 0.05℃), a tension-temperature-deformation relationship model is constructed to generate a belt tension distribution map in real time and identify areas of abnormal tension; 3) Intelligent wear prediction and compensation: Based on the belt material characteristic curve and cumulative working time, a belt elastic modulus change model is established, and the optimal tension parameters are automatically adjusted according to the belt service life stage (initial, middle, and late stages) to extend the belt service life by up to 40%; 4) Multi-belt synchronous control system: Real-time monitoring of belt running trajectory based on vision (sampling rate 120fps) is adopted, combined with a differential speed compensation algorithm, to ensure that the multi-belt system maintains a synchronization accuracy of ±0.3mm when running at high speed (linear speed up to 2.5m / s). The system utilizes an edge computing unit (ARM Cortex-A72 architecture, 4 cores, 2GHz) for real-time data processing, keeping control latency within 6ms. This solution significantly improves the operational stability of multi-belt systems, enhances product conveying positioning accuracy to ±0.2mm, expands system adaptability to products of different sizes and weights (width range 50-500mm), reduces maintenance frequency, extends maintenance cycles by 35%, and improves overall equipment reliability by 45%.

[0177] This invention provides a high-precision, low-latency multi-belt synchronous control system. Through a differential speed compensation algorithm and vision-based real-time monitoring technology, it achieves precise synchronous control of the belts under high-speed operating conditions. The system employs an edge computing architecture and an ARM processor to ensure real-time control commands and low system response latency, meeting the stringent requirements of modern industrial automated production lines for material conveying accuracy and efficiency.

[0178] The system uses a high-speed linear scan camera to capture the belt edge and marker points, and combines this with a sub-pixel edge detection algorithm to achieve accurate measurement of the belt position.

[0179] A Baslerra L4096-24gm line array camera can be used, which has sufficient resolution and sampling rate to meet the needs of high-speed belt monitoring. With a 16mm focal length industrial lens, a measurement accuracy of 0.05mm can be achieved at a working distance of 500mm. The camera can be mounted and fixed on a frame.

[0180] Each belt is equipped with a set of visual monitoring units, including:

[0181] Linear scan camera: Installation height 500mm, field of view 300mm

[0182] LED linear light source: 350mm in length, 15° installation angle, providing uniform illumination.

[0183] Protective housing: made of aluminum alloy, dimensions 400mm×150mm×100mm, IP65 protection rating.

[0184] A set of high-precision encoders can be installed at each end of the belt.

[0185] The fault prediction and maintenance decision-making module analyzes equipment vibration, temperature, and tension data using deep learning algorithms to predict potential faults and automatically adjust operating parameters. High-speed data transmission is achieved through an industrial-grade 5G network, and a virtual image is constructed using digital twin technology to visualize equipment status and optimize parameters. The system deploys lightweight neural networks on edge computing units for local prediction, combined with a cloud-based deep reinforcement learning model, to accurately predict the remaining service life of the equipment. The system uses a Bayesian optimization algorithm to dynamically adjust maintenance thresholds, automatically generating optimal maintenance time windows and specific maintenance item lists based on equipment operating history and current production plans, reducing unplanned downtime by 80%. The system continuously optimizes the model through an incremental learning mechanism; for every 150 hours of new data accumulated, the prediction accuracy improves by 0.6%.

[0186] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A parallel lifting transplanting and conveying device, characterized in that, The system includes a frame, with a chassis (12) on the upper part of the frame. Inside the frame, there is a lifting mechanism and a horizontal control mechanism for lifting the chassis (12). The chassis (12) is equipped with multiple parallel belt conveyor mechanisms driven by rollers (9). The belt conveyor mechanisms are equipped with tensioning devices. The lifting and transplanting conveying equipment also includes an intelligent tension collaborative control system, which includes: a distributed tension monitoring network for monitoring belt tension distribution, with several tension sensors arranged at key stress points of each belt, and a complete tension field constructed through mathematical interpolation algorithms; a hierarchical data processing module for using a three-level data processing structure, including local processing of sensor nodes, regional integration of edge computing units, and global optimization by the central controller to achieve real-time fault warning; a digital twin feedback optimization module for constructing a high-precision digital model containing the nonlinear characteristics of belt materials, and using real-time data to drive simulation and predict changes in tension distribution; and a GPU accelerator for simulation calculation, with the central controller comparing real-time monitoring data to guide the tensioning mechanism to make pre-adjustments.

2. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, The intelligent tension collaborative control system further includes: a self-calibration sensing module and a tension anomaly pattern recognition module. The self-calibration sensing module is connected to a distributed tension monitoring network. The self-calibration sensing module uses a sensor fusion algorithm based on Kalman filtering, combined with a belt dynamic model, to achieve mutual calibration between sensors. The tension anomaly pattern recognition module uses a tension fluctuation feature extraction algorithm based on wavelet transform and deep learning. A deep learning model is deployed through a central controller and trained using the Adam optimization algorithm to provide early warning of potential faults.

3. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, The intelligent tension collaborative control system also includes a multi-belt synchronization control module; the multi-belt synchronization control module is used to improve the synchronization accuracy of the multi-belt system at high speed by using vision-based real-time monitoring of the belt running trajectory and combining it with a differential speed compensation algorithm; and to realize real-time data processing and optimization through a central controller.

4. The parallel lifting transplanting and conveying equipment according to claim 3, characterized in that, The intelligent tension collaborative control system also includes a fault prediction and maintenance decision module; the fault prediction and maintenance decision module: analyzes equipment vibration, temperature and tension data based on deep learning algorithms, predicts potential faults and automatically adjusts operating parameters; achieves high-speed data transmission through an industrial-grade 5G network, and uses digital twin technology to build a virtual image to realize equipment status visualization and parameter optimization.

5. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, The logical representation of the mathematical interpolation algorithm is as follows: In the formula: The tension estimate for the interpolation point. For the first The measured tension values ​​of each sensor, Let be the weighting coefficient, satisfying The interpolation results are mapped to two-dimensional / three-dimensional tension cloud maps to intuitively display the tension distribution of the belt.

6. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, The high-precision digital model incorporating the nonlinear properties of the belt material is represented as follows: ;in: Strain energy density; Material property parameters were obtained through uniaxial tensile testing and least squares fitting. First and second strain invariants; Volumetric strain, when the belt is approximately incompressible ； : Volumetric compressibility coefficient; if the belt exhibits significant creep, a Burger creep model correction is applied: ;in For the elastic modulus component, The relaxation time constant is calibrated through long-term creep tests under different loads.

7. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, It also includes a real-time tension distribution mapping module, which includes a torque sensor located on the bearing seat of the roller (9) and an infrared thermal imaging monitor located on each section of the belt. It constructs a tension-temperature-deformation relationship model and generates a belt tension distribution map in real time.

8. The parallel lifting transplanting and conveying equipment according to claim 1, characterized in that, The lifting mechanism includes an eccentric cam (23). The frame is rectangular and includes two longitudinal beams (1) and two transverse beams (2). A first rotating shaft (21) is connected between the two opposing transverse beams (2). Bearings are provided on both sides of the first rotating shaft (21). The bearings are located on a first bearing seat (22). The first bearing seat (22) is connected to the transverse beams (2) by bolts. Eccentric cams (23) are provided at both ends of the first rotating shaft (21) inside the first bearing seat (22). The lifting mechanism also includes a rotation drive structure. The rotation drive structure includes a servo motor (30). A bottom beam (24) is connected between the longitudinal beams (1). A servo motor (30) and a gearbox are fixedly installed on the bottom beam (24). An active pulley (25) is provided on the output shaft of the gearbox. A concentric driven pulley (26) is provided on the first rotating shaft (21). The active pulley (25) and the driven pulley (26) are opposite to each other and connected by a bottom belt (27). A transverse fixing plate (28) is provided at both ends of the chassis (12). An intermediate roller (29) is provided on the outer side of the transverse fixing plate (28). The intermediate roller (29) is rolled and connected to the eccentric cam (23) to form a cam mechanism.

9. The parallel lifting transplanting and conveying equipment according to claim 8, characterized in that, The roller (9) is vertically arranged. The belt conveyor mechanism includes multiple rotating rollers with a vertical plate (6) and a belt (3). The vertical plate (6) is fixed on the chassis (12) and is perpendicular to the chassis (12). The tensioning device includes a fourth roller (10). The shaft of the fourth roller (10) is located on the vertical plate (6) and can move up and down. The inlet and outlet of the belt conveyor mechanism are each equipped with a set of diffuse reflection photoelectric sensors for detecting materials, so as to realize the linkage control of belt conveying and lifting action.

10. A conveying method for the parallel lifting transplanting conveyor according to any one of claims 1-9, characterized in that, The process includes the following steps: Initial state: The chassis (12) is at its lowest position, the belt (3) of the belt conveyor is kept horizontal, and the photoelectric sensor is in standby mode; Material conveying: When the material passes through the inlet photoelectric sensor, the sensor sends a signal to the control system, the roller (9) drives the belt (3) to run, and conveys the material to the designated transplanting position; Positioning detection: When the material reaches the position of the outlet photoelectric sensor, the sensor sends a signal, the belt conveyor stops running, and the material is accurately positioned; Parallel lifting: The control system instructs the servo motor (30) to reverse, and the power drives the first rotating shaft (21) to rotate through the transmission mechanism, and the bias cam (23) on the first rotating shaft (21) rotates synchronously. Rotation, the cam profile pushes the intermediate roller (29) upward, the intermediate roller (29) drives the chassis (12) to rise through the transverse fixed plate (28); during this process, two sets of synchronous parallelogram linkage mechanisms move synchronously under the action of the synchronous shaft to ensure that the chassis (12) always remains horizontal and realizes the vertical parallel lifting of the material; transfer docking: when the chassis (12) rises to the preset height, the external transfer equipment takes away or transfers the material to the target position; reset process: after the material transfer is completed, the control system commands the servo motor (30) to rotate forward, the cam mechanism drives the chassis (12) to descend smoothly to the initial position, the belt conveyor mechanism resumes operation and enters the next working cycle.

Images