Supply chain management device and supply order processing method

Abstract

The invention discloses an intelligent supply chain management device and a supply order processing method. The device comprises a main conveying line, an intelligent spreading and preprocessing unit, a visual perception system, an edge calculation and control center and an intelligent labeling subsystem. The intelligent spreading and preprocessing unit is used for carrying out personalized spreading processing on the logistics articles through a self-adaptive brush body and a posture adjusting mechanism; the visual perception system comprises a multi-view visual module which is respectively used for initial state acquisition, process monitoring and post-spreading evaluation; the edge calculation and control center generates a spreading strategy with labeling as a target based on the visual data, and outputs a standardized data packet containing accurate positioning and labeling poses; the intelligent labeling subsystem executes high-precision labeling operation and completes visual verification; the full-process intelligent operation from goods flat-spreading preprocessing to automatic labeling is achieved, the labeling accuracy and processing efficiency are greatly improved, and the full-process intelligent labeling system is suitable for modern logistics sorting and supply chain management scenes.

Description

Technical Field

[0001] This invention relates to the field of logistics automation and supply chain management technology, specifically to a supply chain management device and a supply order processing method. Background Technology

[0002] With the rapid development of e-commerce and modern logistics, supply chain management faces increasing pressure to process orders and improve efficiency. At logistics sorting centers, a large number of irregularly shaped and unevenly shaped packages or items need to be processed efficiently. A key step is affixing labels containing order information, delivery addresses, and other details to these items.

[0003] Traditional labeling processes typically rely on manual operation or simple automated equipment. For stacked, wrinkled, or irregularly shaped items, existing equipment often struggles to label them accurately and directly, usually requiring manual sorting and flattening first. This is not only inefficient and labor-intensive, but also prone to errors in high-volume scenarios, leading to inaccurate label placement, unclear information, or even labeling failure, which in turn affects subsequent sorting and delivery processes.

[0004] Currently, some leveling equipment exists on the market, such as mechanical devices based on brushes or rollers. However, these devices are mostly single-function and lack the ability to intelligently sense and adaptively adjust to the condition of the items. They typically use fixed leveling patterns and pressures, failing to provide personalized processing for items of different sizes, materials, shapes, and initial stacking states. This can easily lead to damage to fragile items or poor leveling of flexible items. Furthermore, these devices often operate in isolation from downstream processes such as labeling and sorting, lacking system-level collaboration and data linkage, resulting in efficiency bottlenecks and information gaps throughout the entire processing flow.

[0005] Furthermore, existing technologies for assessing flatness quality largely rely on manual visual inspection or simple photoelectric sensors, which cannot accurately identify and quantify the flatness, surface features, and optimal labeling area of ​​an item, making it difficult to provide reliable data support for subsequent high-precision automated labeling. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an integrated supply chain management device and method capable of intelligently identifying the status of items, adaptively performing pre-processing, and providing precise positioning data for automated labeling, thereby improving the automation level, processing accuracy, and overall efficiency of the entire order processing flow.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] A supply chain management device, comprising:

[0009] The main conveyor line is used to continuously transport logistics goods along the conveying direction.

[0010] The intelligent leveling and pretreatment unit, mounted above the main conveyor line, includes a leveling frame, a rotating shaft, an adaptive leveling brush body, a rotation drive module, and an attitude adjustment mechanism. The adaptive leveling brush body consists of multiple independently controllable flexible bristle modules mounted on the rotating shaft. The rotation drive module drives the rotating shaft to rotate in the opposite direction to the conveying direction of the main conveyor line. The attitude adjustment mechanism is mounted on the leveling frame and drives the rotating shaft to perform vertical lifting and lowering movements.

[0011] The visual perception system includes an entrance visual module located upstream of the intelligent leveling and preprocessing unit, a process monitoring visual module located on its side, and an exit visual module located downstream of it. The entrance visual module is used to acquire the initial three-dimensional information of the logistics items. The process monitoring visual module is used to monitor the shape changes during the leveling process in real time. The exit visual module is used to acquire the flatness, outline, and optimal labeling area information of the logistics items after leveling.

[0012] Preferably, an edge computing and control hub is also included, which is communicatively connected to the main conveyor line, the intelligent leveling and preprocessing unit, and the visual perception system. Its configuration includes: identifying the status of logistics items based on data from the inlet visual module; generating personalized leveling and preprocessing strategies based on the identification results, the goals of which include preparing a flat and accurately positioned item surface for subsequent labeling operations; adjusting the strategy in real time based on feedback from the process monitoring visual module; and evaluating the preprocessing quality based on data from the outlet visual module to generate a standardized item data package containing precise item positioning data and recommended labeling poses.

[0013] Preferably, the system also includes an intelligent labeling subsystem, which includes a label adjustment mechanism and a label supply module. The edge computing and control center sends the item positioning data and recommended labeling pose from the standardized item data package to the intelligent labeling subsystem to control the label adjustment mechanism and the label supply module to perform accurate labeling operations.

[0014] Preferably, the intelligent labeling subsystem also includes a post-labeling visual verification module, which is used to verify the position, angle and content of the label after labeling is completed, and feed the results back to the edge computing and control center to trigger the corresponding processing flow.

[0015] Preferably, the edge computing and control hub is also connected to a supply chain digital twin platform to synchronize and optimize the entire physical processing flow from pre-processing to labeling completion in virtual space.

[0016] Preferably, the entrance vision module and the exit vision module also integrate a deep learning recognition unit, which is used to identify the original markings on the surface of the item and verify them with the order database to determine the labeling strategy.

[0017] Preferably, the edge computing and control center has a pre-built knowledge base of item labeling strategies, which is used to jointly optimize the flat preprocessing strategy by combining the surface characteristics and labeling process requirements of different types of items.

[0018] Another technical problem to be solved by the present invention is to provide a supply order processing method based on the intelligent supply chain management device described in any of the above claims, comprising the following steps:

[0019] Goods are transported via the main conveyor line, and information is collected and identified by the entrance vision module.

[0020] Based on the recognition results, a personalized distribution and preprocessing strategy is generated to ensure labeling quality.

[0021] The intelligent leveling and preprocessing unit executes the strategy, and the process monitoring and adjustment are performed in real time through the process monitoring vision module;

[0022] After preprocessing, the export vision module evaluates the data and generates a standardized item data package containing precise positioning and recommended labeling poses.

[0023] Based on the standardized item data package, the intelligent labeling subsystem is controlled to perform high-precision labeling operations at designated locations on the items.

[0024] Perform visual verification on the labeling results and trigger subsequent processes based on the verification results and order information.

[0025] As a preferred option, the system also includes self-learning and optimization steps: continuously collecting effect data and strategy data during the processing, and iteratively optimizing the strategy knowledge base and generation model through machine learning algorithms to achieve adaptive improvement in system performance.

[0026] The beneficial effects of this invention are as follows:

[0027] By integrating subsystems such as intelligent flattening preprocessing, visual perception, edge computing control, and automated labeling, this invention achieves fully automated closed-loop processing from item identification and sorting to labeling and verification, significantly reducing manual intervention and greatly improving processing efficiency and consistency. The core of this invention lies in providing a systematic solution to a key challenge in automated labeling—uneven item surfaces and inconsistent positioning. Through adaptive intelligent flattening of items based on visual feedback and precise identification of the optimal labeling area, a flat and accurately positioned surface is provided for subsequent labeling operations, fundamentally improving the accuracy, flatness, and success rate of labeling.

[0028] Through multiple independently controllable adaptive flattening brush bodies, posture adjustment mechanisms, and real-time closed-loop control based on vision and pressure sensing, the system can dynamically generate and adjust personalized processing strategies according to the characteristics (such as size, shape, material, and hardness) and initial state of different items. This enables flexible and refined processing of diverse logistics items while effectively preventing damage. A multi-view visual perception system covering the inlet, process, and outlet, along with a post-labeling verification module, enables real-time monitoring and quality assessment of the entire processing process. All perceived data and processing strategies are recorded and can be synchronously analyzed through a digital twin platform, providing a solid foundation for process optimization, problem tracing, and data-driven decision-making.

[0029] This invention's device deeply integrates pre-processing and labeling operations, and achieves data sharing and collaborative control among subsystems through edge computing and a control center. The generated standardized item data package runs throughout the entire process, ensuring the continuity and consistency of information flow. This facilitates seamless integration with upstream order systems and downstream sorting and packaging systems, building a smart supply chain. The system design incorporates self-learning and optimization mechanisms, continuously collecting processing effect data and iteratively optimizing the strategy knowledge base and control model through machine learning algorithms. This enables the system to adaptively evolve, continuously responding to new item types and processing requirements, maintaining long-term technological advancement and processing efficiency. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the supply chain management device of the present invention. Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments are not intended to limit the present invention.

[0032] Example

[0033] See Figure 1 As shown, the intelligent supply chain management device provided by this invention mainly includes the following core subsystems:

[0034] Main conveyor system 1: A roller conveyor driven by a variable frequency motor, the conveying speed can be continuously adjusted within the range of 0.2-1.5m / s, the conveyor belt width is 800mm, and the load capacity is ≤50kg / m;

[0035] Intelligent leveling and pretreatment unit 2: Installed on a gantry frame 2.5m above the conveyor line, including: Leveling frame 21: made of aluminum alloy profile frame, with dimensions of 2000mm×1200mm×800mm; Rotating shaft 22: a hollow steel shaft with a diameter of 80mm, chrome-plated surface, supported at both ends by self-aligning roller bearings; Adaptive leveling brush body 23: composed of 8 independent flexible bristle modules, each module is 150mm wide and the bristle length is 200mm, made of polyurethane material, with a Shore hardness of 65A; Rotation drive module 24: uses a servo motor (rated power 1.5kW) with a precision reducer (reduction ratio 1:50); Attitude adjustment mechanism 25: adopts a lifting structure with a lifting stroke of 600mm;

[0036] Visual Perception System 3: Entrance 3D Vision Module 31: Employs a binocular structured light camera (resolution 1280×960), installed at a height of 3.0m, with a field of view of 1.5m×1.2m; Process Monitoring Vision Module 32: Employs a high-speed industrial camera (frame rate 200fps), installed on the side; Exit 3D Vision Module 33: Employs a line laser 3D scanner with a scanning frequency of 2000Hz and an accuracy of ±0.05mm; Edge Computing and Control Hub: Built on an industrial PC (Intel Core i7 processor, 32GB memory) and PLC (Siemens S7-1500 series), running ROS (Robot Operating System) and a real-time control system.

[0037] When goods enter the main conveyor line, the entrance 3D vision module collects point cloud data of the goods. The system first performs item segmentation and feature extraction: the item segmentation algorithm adopts improved Euclidean clustering segmentation;

[0038] Feature extraction and recognition:

[0039] The system extracts the following feature vectors for each item: Geometric features: bounding box dimensions (L,W,H), volume V, surface area S, centroid position; Morphological features: flatness F=min(L,W) / H, elongation E=max(L,W) / min(L,W); Surface features: rate of curvature change, wrinkle depth distribution.

[0040] For the recognition of the original surface markings, a hybrid network based on deep learning is used:

[0041] Formula for calculating recognition accuracy:

[0042] P_Identification = α·P_Barcode + β·P_OCR + γ·P_Special Mark;

[0043] Among them: α+β+γ=1, α=0.5, β=0.3, γ=0.2;

[0044] P_barcode = (1 - decoding failure rate) × positioning accuracy;

[0045] P_OCR = Character-level accuracy × Word-level accuracy;

[0046] The edge computing and control center generates personalized cost-sharing strategies based on item characteristics and order information. Strategy parameters include: cost-sharing intensity model:

[0047] F_i=k_1·m·g·μ_i+k_2·Δh_i+k_3·(1-s_i);

[0048] Where, F_i: applied force (N) of the i-th bristle module; m: mass of the item (kg), estimated by volume and density; g: gravitational acceleration; μ_i: friction coefficient between the item and the bristles, determined based on material identification results; Δh_i: height difference of the i-th module area (mm); s_i: surface flatness score (0-1); k_1,k_2,k_3: adjustment coefficients, calibrated experimentally.

[0049] Brush motion trajectory planning: Bézier curves are used to plan the brush motion path.

[0050] P(t)=Σ_{i=0}^nB_i^n(t)·P_i;

[0051] B_i^n(t)=C(n,i)·t^i·(1-t)^{ni};

[0052] Where P(t): the position of the brush body at time t; P_i: the coordinates of the control point; n: the order of the Bézier curve, usually taken as 3; C(n,i): the binomial coefficient.

[0053] Amortized quality evaluation function:

[0054] Q=w_1·(1-σ_h)+w_2·S_flat+w_3·(1-θ_max)+w_4·T_efficiency;

[0055] Where: σ_h: standard deviation of surface height; S_flat: percentage of flat area; θ_max: maximum tilt angle; T_efficiency: processing time efficiency; w_i: weighting coefficient, ∑w_i=1.

[0056] The process monitoring vision module acquires images at a frame rate of 200fps, and the system calculates the following feedback parameters in real time: morphological change rate:

[0057] ΔM_t = ||f_t - f_{t-1}|| / Δt;

[0058] Where f_t is the morphological feature vector at time t, containing the position coordinates of 8 key points.

[0059] A method combining PID control and fuzzy logic is adopted:

[0060] Δu(t)=K_p·e(t)+K_i·∫e(τ)dτ+K_d·de(t) / dt;

[0061] Where the error e(t) = Q_target - Q_actual, the controller parameters are adjusted online through fuzzy rules: if e is large and Δe is large, then increase K_p; if e is small and Δe is small, then increase Ki; if Δe changes rapidly, then increase K_d.

[0062] The export 3D vision module acquires the point cloud of the flattened object surface. The system determines the optimal labeling pose using the following algorithm: Surface suitability score:

[0063] S_surface(x,y)=α·F_flat(x,y)+β·F_curvature(x,y)+γ·F_visibility(x,y);

[0064] Where: F_flat(x,y): flatness score, based on the variance of the local surface normal vector; F_curvature(x,y): curvature score, to avoid labeling in raised or recessed areas; F_visibility(x,y): visibility score, to ensure that the label is not obscured.

[0065] The labeling position and orientation are determined by optimizing the following objective function:

[0066] max_{x,y,θ} [S_surface(x,y) - λ·D_edge(x,y) + μ·A_alignment(x,y,θ)];

[0067] Constraints: Distance from the edge of the item ≥ 10mm; Label area ≤ 0.3% of the available flat area; Angle between the long axis of the label and the main direction of the item ≤ 30°.

[0068] The multi-axis labeling robotic arm adopts a 6-DOF serial structure with a repeatability of ±0.05mm. The labeling process control includes: employing time-optimal trajectory planning to meet acceleration and velocity constraints.

[0069] min ∫_0^T [1 + w·||a(t)||^2] dt;

[0070] st v_min ≤ v(t) ≤ v_max;

[0071] a_min ≤ a(t) ≤ a_max;

[0072] j_min ≤ j(t) ≤ j_max;

[0073] The labeling pressure is adaptively adjusted based on the surface material using a force / position hybrid control system.

[0074] F_label = F_base + k·(E_material - E_avg);

[0075] Where F_base=2N, k=0.1N / GPa, and E_material is the elastic modulus of the material.

[0076] After labeling, the visual inspection module captures the label image and performs the following inspections:

[0077] Position deviation detection:

[0078] Δp = ||p_actual - p_target||;

[0079] Δθ = |θ_actual - θ_target|;

[0080] Acceptance criteria: Δp ≤ 2mm, Δθ ≤ 3°.

[0081] Label quality assessment: Label flatness was assessed using the SSIM (Structural Similarity Index).

[0082] SSIM(x,y) = [(2μ_xμ_y + C1)(2σ_xy + C2)] / [(μ_x² + μ_y² + C1)(σ_x² +σ_y² + C2)];

[0083] Acceptance criteria: SSIM ≥ 0.85;

[0084] The system continuously collects and processes data to construct a training sample set D = {(x_i, y_i, r_i)}, where: x_i: item feature vector; y_i: strategy parameters; r_i: processing result score;

[0085] The policy network is updated periodically using deep reinforcement learning (PPO algorithm):

[0086] L(θ) = E_t[min(π_θ(a_t|s_t) / π_θ_old(a_t|s_t)·A_t, clip(π_θ / π_θ_old,1-ε, 1+ε)·A_t)];

[0087] Where A_t is the dominance function and ε=0.2 is the trimming parameter.

[0088] The item labeling strategy knowledge base is stored in a graph database, where nodes represent item types and edges represent the transition probabilities of processing strategies. Weekly automatic knowledge base optimization includes: merging similar item categories (Jaccard similarity > 0.8); eliminating inefficient strategies (success rate < 90%); and adding newly discovered optimization strategies.

[0089] The supply chain digital twin platform receives real-time device operation data, including: equipment status data (updated every 100ms); item processing data (one record generated for each item); and quality control data (including images and measurement results). The twin platform runs simulation models to predict and optimize system performance, including a throughput prediction model.

[0090] TPH = 3600 / [T_process + max(T_pallet, T_label) + σ];

[0091] Where σ is a random disturbance term, representing the uncertainty of the system.

[0092] Taking a corrugated cardboard box with dimensions of 400×300×150mm as an example:

[0093] Item entry: Conveyor speed 0.8m / s, item arrives at the recognition area;

[0094] Feature recognition: The system identifies it as a "medium-sized cardboard box", with an estimated weight of 2.5kg and slight wrinkles on the surface;

[0095] Strategy generation: A moderate force spreading strategy is adopted, with a brush rotation speed of 30 rpm and a descent speed of 50 mm / s;

[0096] Execution with flat brush: The brush touches the surface of the item three times, with a total time of 2.5 seconds;

[0097] Labeling planning: The optimal labeling position is determined to be the upper right corner of the front, with coordinates (320, 220, 150).

[0098] Labeling execution: robotic arm movement time 1.2 seconds, labeling pressure 2.8N;

[0099] Quality verification: Positional deviation 0.8mm, angular deviation 1.5°, SSIM=0.92;

[0100] Data logging: The processing results are stored in the database for subsequent analysis.

[0101] Based on actual testing, the system performance indicators of this invention are as follows:

[0102] Processing capacity: Maximum processing speed 1800 pieces / hour;

[0103] Labeling accuracy: 99.2% (positional deviation ≤ 2mm);

[0104] Success rate of leveling: 98.5% (flatness score ≥ 0.8);

[0105] System availability: 99.8% (mean time between failures > 2000 hours).

[0106] The above embodiments of the present invention are not intended to limit the scope of protection of the present invention. The implementation of the present invention is not limited thereto. All other modifications, substitutions or alterations made to the above structure of the present invention based on the above content of the present invention, in accordance with ordinary technical knowledge and common practice in the field, without departing from the basic technical idea of ​​the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A supply chain management apparatus, characterized by, Comprise: a main conveying line for continuously transporting logistics articles along a conveying direction; a smart flattening and pretreatment unit arranged above the main conveying line, comprising a flattening frame, a rotating shaft, an adaptive flattening brush body, a rotating drive module, and a posture adjusting mechanism; the adaptive flattening brush body is composed of multiple independently controllable flexible bristle modules and is installed on the rotating shaft; the rotating drive module is used to drive the rotating shaft to rotate in the opposite direction of the conveying direction of the main conveying line; the posture adjusting mechanism is installed on the flattening frame and is used to drive the rotating shaft to perform vertical lifting motion; a visual perception system, comprising an entrance visual module arranged upstream of the smart flattening and pretreatment unit, a process monitoring visual module arranged on the side thereof, and an exit visual module arranged downstream thereof; the entrance visual module is used to obtain initial three-dimensional information of the logistics articles; the process monitoring visual module is used to monitor the morphological changes in the flattening process in real time; the exit visual module is used to obtain the flatness, profile, and optimal labeling area information of the logistics articles after flattening.

2. The supply chain management apparatus according to claim 1, characterized by, Further comprise an edge computing and control hub, which is communicatively connected with the main conveying line, the smart flattening and pretreatment unit, and the visual perception system, and is configured to: identify the state of the logistics articles according to the data of the entrance visual module; generate an individualized flattening and pretreatment strategy based on the identification result, the target of the strategy including preparing a flat and accurately positioned article surface for subsequent labeling work; adjust the strategy in real time according to the feedback of the process monitoring visual module; and evaluate the pretreatment quality according to the data of the exit visual module, and generate a standardized article data package containing article positioning data and recommended labeling pose.

3. The supply chain management apparatus according to claim 1, wherein Further comprise a smart labeling subsystem, which comprises a label adjusting mechanism and a label supply module; the edge computing and control hub sends the article positioning data and the recommended labeling pose in the standardized article data package to the smart labeling subsystem to control the label adjusting mechanism and the label supply module to perform accurate labeling work.

4. The supply chain management apparatus according to claim 1, wherein The smart labeling subsystem further comprises a post-labeling visual checking module for checking the position, angle, and content of the label after labeling and feeding back the results to the edge computing and control hub to trigger corresponding processing procedures.

5. The supply chain management apparatus according to claim 3, wherein The edge computing and control hub is further connected with a supply chain digital twin platform for synchronizing and optimizing the entire physical processing procedure from flattening and pretreatment to labeling completion in a virtual space.

6. The supply chain management apparatus according to claim 1, wherein The entrance visual module and the exit visual module further integrate a deep learning identification unit for identifying the original mark on the surface of the article and performing association verification with an order database to determine the labeling strategy.

7. The supply chain management apparatus according to claim 1, wherein The edge computing and control hub has a preinstalled article labeling strategy knowledge base for jointly optimizing the flattening and pretreatment strategy in combination with the surface characteristics of different categories of articles and the labeling process requirements.

8. A supply order processing method based on the intelligent supply chain management apparatus according to any one of claims 1 to 7, characterized by, Comprise the following steps: transporting logistics articles through the main conveying line and collecting and identifying information by the entrance visual module; generating an individualized flattening and pretreatment strategy with the core of ensuring labeling quality based on the identification result; The control intelligent flatting and pre-processing unit executes the strategy and is monitored and adjusted in real time by the process monitoring visual module; After pre-processing, the exit visual module evaluates and generates a standardized product data package containing accurate positioning and recommended labeling pose; Based on the standardized product data package, the control intelligent labeling subsystem performs high-precision labeling work at the specified product location; The labeling result is visually checked, and the subsequent process is triggered according to the checking result and order information. 9.The smart supply chain management device based supply order processing method of claim 8, wherein, It also includes system self-learning and optimization steps: continuously collect effect data and strategy data during processing, iteratively optimize the strategy knowledge base and generation model through machine learning algorithms, to achieve self-adaptive improvement of system performance.

Images