Automatic packaging and stacking integrated system

Abstract

The invention discloses an automatic packaging and stacking integrated system which comprises a packaging unit, a dynamic marshalling unit, an intelligent stacking unit, a digital twin monitoring unit and an edge decision center. According to the invention, mixed production of bagged, boxed and barreled products is supported through the switchable clamp (vacuum chuck / clamping jaw / holding tool) and the liftable multi-track conveyor; according to the multi-line cooperative efficiency, the dynamic marshalling unit achieves efficient matching of 3-5 packaging lines and 1-2 stacking machines, the stacking efficiency is improved by 83% (for example, 12 boxes per minute of a traditional system is improved to 22 boxes per minute), and the problems of congestion and idling during multi-line convergence are solved; based on multi-robot scheduling of a Hungary algorithm and reinforcement learning, the task completion time is shortened by 20% after 1000 iterations, the replacement frequency of a robot clamp is reduced by 18%, and idle loss of equipment is avoided.

Description

TECHNICAL FIELD

[0001] The application relates to an automatic packaging and stacking integrated system device, in particular to an automatic packaging and stacking integrated system, and belongs to the technical field of automatic production lines. BACKGROUND

[0002] With the progress and development of science and technology, in modern industrialized production of chemicals, food, feed, fertilizer and the like, in order to improve work efficiency and work quality and reduce labor intensity, production enterprises have gradually used automatic packaging and automatic stacking equipment to replace manual packaging and stacking of bagged products.

[0003] According to the search, a kind of automatic packaging and stacking integrated system is disclosed in Chinese patent No.CN103101772B, which is grouped according to the working beat requirement of front-end stacking equipment, and regularly transports packaging to stacking equipment for stacking operation at a certain distance or time interval, to realize the automation integration from packaging to stacking.But in the above-mentioned patent product, it is designed only for bagged products, cannot be compatible with box, barrel and other different forms of packaging, and is difficult to adapt to the packaging and stacking of multi-form products;At present, the detection link and production link data of existing automatic packaging and stacking integrated system in market are cut off, and defect product traceability needs manual inquiry of multi-system records (average traceability time > 4 hours), and the leakage detection rate is as high as 3%. SUMMARY

[0004] The purpose of the present application is to provide an automatic packaging and stacking integrated system to solve the above problems.

[0005] The application achieves the above-mentioned purpose by the following technical scheme, an automatic packaging and stacking integrated system, comprising a packaging unit, a dynamic grouping unit, an intelligent stacking unit, a digital twin monitoring unit and an edge decision hub, characterized in that: the packaging unit mainly completes product metering, packaging and primary detection;The dynamic grouping unit realizes multi-line product confluence, sorting and beat matching;The intelligent stacking unit performs automatic stacking operation and supports multi-robot cooperation;The digital twin monitoring unit realizes real-time monitoring and pre-performance of the whole process through virtual mirroring;The edge decision hub is mainly responsible for data processing, algorithm scheduling and instruction issuing;

[0006] The packaging unit is mainly composed of a stacking robot, a magnetic suspension electric clamp and a vision centering module, the packaging unit and the dynamic grouping unit are connected through a liftable multi-track conveyor, the track of the liftable multi-track conveyor is driven by a servo motor and a ball screw, the dynamic grouping unit includes a six-axis mechanical arm sorting module and an AGV mobile roller group;

[0007] The intelligent stacking unit and the digital twin monitoring unit realize real-time data interaction through a 5G network, the intelligent stacking unit comprises a mixed reality guiding system, the stacking robot is configured with MR glasses, the stacking robot supports gesture human-computer interaction, and the stacking robot has a ''man-machine-environment'' collaborative obstacle avoidance function.

[0008] The edge decision hub comprises an edge computing node and a blockchain storage module, and the blockchain storage module supports product life cycle tracing.

[0009] Preferably, the six-axis mechanical arm sorting module is loaded with a force control end effector, the thrust adjustment range of the six-axis mechanical arm sorting module is 5-35N, the six-axis mechanical arm sorting module can support bagged, boxed and barrel-shaped product sorting, and the AGV mobile roller group supports cross-rail flexible material receiving, and the maximum bearing weight of the AGV mobile roller group is greater than 500kg.

[0010] Preferably, the edge computing node is based on a stacking type planning algorithm and a robot kinematics solving module, and the data processing delay of the edge decision hub is less than 10ms.

[0011] Preferably, the packaging unit is provided with a green packaging module, the green packaging module comprises a waste recycling conveyor and an AI ink optimization algorithm, the waste recycling conveyor automatically winds and weighs the film waste, and the utilization rate of materials is improved.

[0012] The AI ink optimization algorithm dynamically adjusts the ink amount according to the surface area of the packaging material, and saves the ink.

[0013] Preferably, the stacking robot adopts an NSGA-II multi-objective optimization algorithm to realize Pareto optimization of stacking stability, space utilization and energy consumption, and the average energy consumption is reduced by greater than 18%.

[0014] Preferably, the intelligent stacking unit guides the robot stacking by using a ToF depth camera, and the intelligent stacking unit is provided with a six-dimensional force sensor.

[0015] Preferably, the digital twin monitoring unit realizes 1:1 virtual mirroring of equipment based on point cloud modeling and a physical engine, and the digital twin monitoring unit supports process parameter sensitivity analysis and AR assisted maintenance.

[0016] Preferably, the edge decision hub adopts a mixed model of the Hungarian algorithm and reinforcement learning to schedule multi-robot tasks, and the edge decision hub realizes 72-hour early warning of equipment failure through an LSTM network.

[0017] The present application has the following beneficial effects:

[0018] 1. Support mixed production of bagged, boxed, and barrelled products through switchable clamps (vacuum cups / grippers / grippers) and liftable multi-track conveyors;

[0019] 2. Multi-line coordination efficiency: dynamic grouping units achieve efficient matching of 3-5 packaging lines and 1-2 stackers, with stacker efficiency improved by 83% (e.g., from 12 boxes per minute in traditional systems to 22 boxes per minute), solving congestion and idling problems during multi-line convergence;

[0020] 3. Multi-robot scheduling based on Hungarian algorithm and reinforcement learning, with task completion time reduced by 20% after 1000 iterations, and robot clamp replacement frequency reduced by 18%, avoiding equipment idle loss;

[0021] 4. Blockchain traceability: production data (packaging parameters, test results) are real-time chained, with a data tamper-proof rate of 100%, supporting product lifecycle traceability (such as raw material batch, production time, stacking coordinates), meeting ISO22000 and other compliance requirements;

[0022] 5. Virtual-real collaborative optimization: 1:1 virtual model and physical equipment kinematic error <0.5%, supporting process parameter sensitivity analysis (prediction error <5%), new product introduction cycle reduced from 2 weeks to 3 days, saving actual debugging time by more than 90%;

[0023] 6. Waste recycling conveyor (speed adjustable from 0.5-2m / s) automatically rolls up film waste, with material utilization rate statistical error <1%, recycling rate reaching more than 98%, saving raw material cost by 100-150 thousand yuan per year; AI ink optimization dynamically adjusts ink volume according to the surface area of the packaging, saving ink by 15%-20%, reducing ink consumption by more than 3000 liters per year, while avoiding environmental pollution caused by over-spraying. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A whole system flowchart of an automatic packaging and stacking integrated system is proposed for the present invention;

[0025] Figure 2 A system flowchart and algorithm formula fusion architecture of an automatic packaging and stacking integrated system is proposed for the present invention;

[0026] Figure 3 An AI ink optimization control flowchart in an automatic packaging and stacking integrated system is proposed for the present invention. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all the embodiments.

[0028] Embodiment one:

[0029] Referring to Figures 1-2 An automatic packaging and stacking integrated system, comprising a packaging unit, a dynamic grouping unit, an intelligent stacking unit, a digital twin monitoring unit and an edge decision hub, characterized in that: the packaging unit mainly completes product metering, packaging and primary detection; the dynamic grouping unit realizes multi-line product convergence, sorting and beat matching; the intelligent stacking unit performs automatic stacking operation and supports multi-robot collaboration; the digital twin monitoring unit realizes real-time monitoring and pre-performance of the whole process through virtual mirroring; the edge decision hub is mainly responsible for data processing, algorithm scheduling and instruction issuing;

[0030] The packaging unit is mainly composed of a stacking robot, a magnetic suspension electric clamp and a visual centering module. The packaging unit is connected with the dynamic grouping unit through a liftable multi-track conveyor. The track of the liftable multi-track conveyor is driven by a servo motor and a ball screw;

[0031] The intelligent stacking unit and the digital twin monitoring unit realize real-time data interaction through a 5G network. The intelligent stacking unit includes a mixed reality guidance system. The stacking robot is equipped with MR glasses. The stacking robot supports gesture human-computer interaction. The stacking robot has a "human-machine-environment" collaborative obstacle avoidance function. The intelligent stacking unit uses a ToF depth camera to guide the robot stacking. The intelligent stacking unit is equipped with a six-dimensional force sensor;

[0032] The edge decision hub is built-in with a distributed computing architecture. The edge decision hub includes an edge computing node and a blockchain storage module. The blockchain storage module supports product lifecycle traceability. The edge computing node is based on a stacking type planning algorithm and a robot kinematics solving module. The data processing delay of the edge decision hub is less than 10ms. The edge decision hub uses a Hungarian algorithm and a reinforcement learning hybrid model to schedule multi-robot tasks. The edge decision hub realizes 72-hour early warning of equipment failure through an LSTM network.

[0033] In this embodiment, it should be noted that the magnetic suspension electric clamp is a magnetic suspension electric clamp library (containing 5 clamps, positioning accuracy ±0.05mm). The clamp is "plug and play" through a linear motor, and the replacement time is shortened to 8 seconds.

[0034] A visual centering system is added: an industrial camera (resolution 2048x1536) detects the position offset of the packaging in real time, and a servo motor drives the clamp to dynamically compensate (compensation accuracy ±0.2mm).

[0035] The mechanical structure of the liftable multi-track conveyor adopts an aluminum alloy profile frame, is equipped with a servo motor (power 1.5 kW) and a ball screw transmission mechanism, has a track lifting speed of 50 mm / s and a maximum stroke of 500 mm, supports flexible combination of 2-4 input tracks and 1-2 output tracks (such as "3-in-1-out" and "4-in-2-out"), and can electrically adjust the track spacing through an HMI interface (adjustment accuracy ±1 mm).

[0036] The liftable multi-track conveyor is provided with an array of built-in pressure sensors (spacing 100 mm) for real-time calculation of the packing density. When the density is greater than 80 kg / m 3 , the PLC triggers a shunting program to guide the excess products into a buffer shelf. The track surface is made of anti-static rubber material with a friction coefficient of 0.3-0.5, which can adapt to the conveying of packaging materials of different materials (such as plastic packaging to prevent slipping and metal barrels to prevent wear).

[0037] The ToF depth camera (resolution 1024x768, point cloud density 5 million / ㎡) is installed at a 45° position above the stacking robot. An industrial-grade PC controller (CPU: i7-12700, GPU: RTX3060) is used to support real-time point cloud processing and path planning.

[0038] The algorithm flow of the intelligent stacking unit is as follows:

[0039] 1. The camera scans the stacking plate area to generate three-dimensional point cloud data.

[0040] 2. The algorithm identifies the stacking plate reference surface and the profile of the products that have been stacked.

[0041] 3. According to the preset stacking type (such as row-column type and staggered type), the optimal placement position and attitude are calculated.

[0042] 4. The robot motion trajectory is generated (error <2 mm) to avoid collision with existing products.

[0043] The main algorithms and formulas involved in the above process are as follows:

[0044] 1. Point cloud preprocessing - voxel grid downsampling (reduce point cloud density and reduce calculation amount)

[0045] The formula is:

[0046] Where: p i is the point coordinates within the voxel; n is the number of points within the voxel.

[0047] 2. Plane segmentation - RANSAC algorithm (identify the stacking plate reference surface)

[0048] The formula is: ax+by+cz+d=0

[0049] Distance from point (x, y, z) to plane:

[0050] 3. Target contour extraction - Convex hull algorithm (Calculate the minimum convex polygon of product contour)

[0051] Formula: Convex hull volume:

[0052] 4. Optimal placement position calculation

[0053] Algorithm logic:

[0054] Space utilization optimization:

[0055] Stability evaluation:

[0056] Where: (x c ,y c ): current stack center of gravity; (x0, y0): tray center

[0057] 5. Robot motion trajectory planning

[0058] Core formula:

[0059] Joint space interpolation: θ(t) = θ0 + (θ1 - θ0) · f(t)

[0060] Where: f(t): quintic polynomial interpolation function

[0061] Ensure: f(0) = 0, f(1) = 1, f'(0) = f'(1) = f"(0) = f"(1) = 0

[0062] Obstacle avoidance constraint: distance(robot, obstacle) ≥ safety distance

[0063] 6. Precision control - error compensation

[0064] Error model:

[0065] PID controller parameters: K p = 0.8, K i = 0.1, K d = 0.05 (need to be tuned in actual project)

[0066] In the packaging unit, a honeycomb multi-line convergence platform is adopted, integrating an RFID scanner (reading speed 200 times / s) and a laser range finder (accuracy ±1mm), which monitors the distance between adjacent packages in real time. When the distance is <10cm, the conveying speed is dynamically adjusted by a servo motor (speed regulation accuracy ±5%), to avoid collision and congestion.

[0067] Distributed computing architecture: integrate edge computing gateway with blockchain node, support local autonomous operation in offline environment (data cache for 72 hours), automatically synchronize to the cloud after network recovery; use machine learning algorithms (such as XGBoost) to optimize the stacking scheme, generate energy-saving reports every 100,000 operations, and reduce energy consumption by 15% annually.

[0068] The algorithm involved in this embodiment is as follows:

[0069] I. Multi-robot dynamic task allocation algorithm

[0070] 1. Comprehensive cost calculation model: C ij = α · D ij + β · S ij + γ · T ij

[0071] Parameter description:

[0072] C ij : The comprehensive cost of robot i processing product j;

[0073] D ij : Spatial distance cost (Euclidean distance, unit: meters);

[0074] S ij : Task switching cost (fixture replacement time conversion, unit: seconds);

[0075] T ij : Processing time cost (including grabbing, moving, and stacking, unit: seconds);

[0076] α = 0.5, β = 0.3, γ = 0.2: Weighting coefficients (adjustable according to production priority).

[0077] 2. Reinforcement learning reward function:

[0078] Parameter description:

[0079] T 传统 : Completion time of traditional fixed allocation scheme;

[0080] T 当前 : Completion time of current allocation scheme;

[0081] N 切换 : Number of robot fixture replacements;

[0082] λ = 0.1: Switching penalty coefficient.

[0083] II. Intelligent stacking planning algorithm

[0084] 1. Space utilization rate calculation:

[0085] Parameter Description:

[0086] V used : Actual volume occupied by the pallet (sum of all product volumes, unit: mm 3 ) ;

[0087] V pallet : Maximum volume of the pallet (width x length x height, unit: mm 3 ).

[0088] 2. Stability score model:

[0089] Parameter Description:

[0090] G: Pallet type gravity center height (unit: mm) ;

[0091] y0: Critical stability height (take 60% of the pallet height) ;

[0092] k = 0.1: Slope adjustment parameter.

[0093] When the gravity center height is below the critical value, the stability score tends to 1 (stable) ; when it exceeds, it tends to 0 (easy to collapse).

[0094] Three, Predictive maintenance algorithm

[0095] 1. Abnormal score calculation (root mean square error) : Parameter Description:

[0096] x i : Real-time sensor measurement value (vibration acceleration / temperature / current) ;

[0097] Normal value predicted by LSTM model;

[0098] n = 30: Sliding window length.

[0099] 2. Remaining life prediction (linear degradation model) :

[0100]

[0101] Parameter Description:

[0102] Threshold: Failure threshold (such as vibration acceleration threshold 8 mm / s) CurrentScore: Current abnormal score

[0103] Slope: Degradation trend slope (fitted by historical data)

[0104] Four, Whole-process quality control model (Bayesian network)

[0105] 1. Joint probability formula: Parameter explanation:

[0106] A: Packaging parameter anomaly (film thickness / sealing temperature);

[0107] B: Grouping parameter anomaly (pushing package strength / conveying speed);

[0108] C: Palletizing parameter anomaly (grabbing force / placement accuracy);

[0109] D: Final quality defect.

[0110] 2. Defect traceability priority: Parameter explanation: Priority (X): The probability priority of factor X causing defects.

[0111] Five, energy optimization model (dynamic programming)

[0112] 1. Equipment energy consumption function: E (t) = E0 + e start +e run ·t

[0113] Parameter explanation:

[0114] E0: Standby energy consumption (unit: kWh);

[0115] e start : Start-stop energy consumption (unit: kWh / time);

[0116] e run : Start-stop energy consumption (unit: kWh / time);

[0117] t: Running time (unit: hours).

[0118] 2. Total production constraint:

[0119] Parameter explanation: Q0: Maximum theoretical production; T0: Production capacity saturation time.

[0120] Formula derivation basis:

[0121] Multi-robot scheduling: Combine the distance, switching, and time three elements in actual production, and balance efficiency and equipment loss through weighted summation;

[0122] Stack stability: Based on the principle of barycenter, use the logistic function to simulate the gradual change process from "stable" to "unstable";

[0123] Predictive maintenance: Use the root mean square error to quantify the deviation of real-time data from the normal model, and the linear model is suitable for early degradation trend prediction;

[0124] Bayesian Network: Uncertainty reasoning based on probability theory, suitable for multi-factor coupled quality traceability scenarios.

[0125] These formulas are verified by industrial field data and can be directly used for parameter calculation and algorithm implementation in system design.

[0126] Example two:

[0127] The algorithm in the above examples is applied to the specific scene analysis of the automatic packaging and stacking integrated system. The operation method is explained in combination with the equipment operation process and control logic as follows:

[0128] I. Application of multi-robot dynamic task allocation formula

[0129] Scenario: Multi-line product mixed stacking

[0130] Input data: Robot state: RobotA (vacuum suction cup, current position (10, 20), can handle bagged products), RobotB (electromagnetic clamp, position (30, 15), can handle boxed products)

[0131] Product to be stacked: Product1 (bagged, position (12, 22), processing time 10s), Product2 (boxed, position (28, 18), processing time 15s)

[0132] Step 1: Calculate the comprehensive cost matrix:

[0133] (RobotA is incompatible with boxed, set to infinity)

[0134] D B1 = ∞ (RobotB is incompatible with bagged)

[0135]

[0136] Task switching cost S ij : RobotA needs to change the clamp (30s) to handle boxed, but it is directly set to infinity because it is incompatible; RobotB has no switching (0s).

[0137] Processing time cost T ij : Product1 = 10s, Product2 = 15s.

[0138] Comprehensive cost matrix:

[0139]

[0140] Step 2: Assign tasks using the Hungarian algorithm

[0141] The only feasible solution: RobotA→Product1, RobotB→Product2, total cost = 3.915+5.305=9.22.

[0142] Step 3: Reinforcement Learning Optimization

[0143] If historical rewards show that frequent tool changes lead to decreased efficiency, the algorithm adjusts the weighting factor β to 0.4 (increases the weight of switching cost), and the next time the assignment prioritizes tasks with high compatibility.

[0144] II. Application of Intelligent Pile Type Planning Formula

[0145] Scenario: Palletizing of Barreled Products (Tray Size 1200x1000x1500mm)

[0146] Input Data:

[0147] Product Size: Diameter 300mm, Height 500mm (Treated as a Cylinder, Bottom Area πx150 2 = 10686mm 2

[0148] Maximum Number of Layers per Tray:

[0149] Step 1: Calculate Maximum Number of Layers per Tray

[0150]

[0151] Number of Layers = 3x3 = 9 Barrels, Three Layers Total 27 Barrels.

[0152] Step 2: Calculate Space Utilization

[0153] V used = 27x(πx150 2 x500) = 9.54x10 9 mm 3

[0154] V pallet = 1200x1000x1500 = 1.8x10 10 mm 3

[0155]

[0156] Step 3: Genetic Algorithm Optimization of Pile Type

[0157] By rotating the barrel (Lay horizontally: Height 300mm, Diameter 500mm), recalculate:

[0158]

[0159]

[0160] Total quantity = 6 x 2 = 12 barrels (seems to decrease, but actually can be interleaved on the second layer)

[0161] After optimization, the space utilization rate is improved to 68%, and the stability score S is improved from 0.7 (straight stacking) to 0.85 (interleaved horizontal placement).

[0162] III. Application of Predictive Maintenance Formula

[0163] Scenario: Conveyor bearing failure early warning

[0164] Input data:

[0165] Vibration acceleration historical data (unit: mm / s):

[0166] [4.2, 4.5, 4.8, 5.1, 5.4, 5.7] (sampled once a day for 6 days)

[0167] LSTM model predicts normal value on day 7: 5.8 mm / s

[0168] Real-time measurement value: 6.5 mm / s (exceeds threshold value 8 mm / s for the first 3 days)

[0169] Step 1: Abnormal score calculation

[0170] (Current score is lower than threshold 0.8, no early warning is triggered, but the trend shows continuous rise)

[0171] Step 2: Remaining life prediction

[0172] Calculate degradation slope:

[0173] Remaining life:

[0174] The system issues a maintenance warning 5 days in advance, and schedules a backup conveyor to take over the task.

[0175] IV. Application of Whole Process Quality Control Formula

[0176] Scenario: Bagged product sealing defect traceability

[0177] Input data:

[0178] Historical statistics: Probability of packaging parameter abnormality (A) P(A) = 0.05, probability of grouping parameter abnormality (B) P(B) = 0.03, probability of stacking parameter abnormality (C) P(C) = 0.02

[0179] Conditional probability of defect occurrence: P(D|A) = 0.6, P(D|B) = 0.2, P(D|C) = 0.1

[0180] Step 1: Calculate joint probability

[0181] P(D) = P(D|A)P(A) + P(D|B)P(B) + P(D|C)P(C)

[0182] = 0.6 x 0.05 + 0.2 x 0.03 + 0.1 x 0.02 = 0.038

[0183] Step 2: Trace priority ranking

[0184]

[0185] The system prioritizes the sealing temperature and film thickness parameters in the packaging section.

[0186] Five, application of energy optimization formula

[0187] Scenario: Equipment start-stop scheduling in three-shift production

[0188] Input data:

[0189] Packaging machine: E0 = 5kWh, e start = 2kWh / time, e run = 20kW

[0190] Total daily working time T 总 = 24h, target output Q 目标 = 10000 pieces

[0191] Output model:

[0192] Step 1: Establish energy consumption function E(t) = 5 + 2 + 20t = 20t + 7kWh

[0193] Step 2: Dynamic programming to solve the optimal running time

[0194] Constraints:

[0195] Actual optimization: At t = 16h, output = 8000 x (1-e -2 ) = 6800 pieces, energy consumption = 20 x 16 + 7 = 327kWh

[0196] Compared with traditional continuous operation (24h): Energy consumption = 20 x 24 + 7 = 487kWh, energy consumption is saved by 33% after optimization.

[0197] Key points of system integration for formula application

[0198] Data acquisition layer: real-time acquisition of device position, running time, energy consumption, etc. through PLC; sensors (laser range finder, force control sensor, vibration sensor) collect 100-1000 times of data per second.

[0199] Edge computing layer: deploy lightweight algorithm engine (such as Python Flask), process 100+ pieces of data per second; use distributed computing framework (such as Apache Spark) to handle multi-robot task allocation in parallel.

[0200] Execution control layer: servo motor adjusts the speed in real time according to the optimal path output by the algorithm (accuracy ±0.1 rpm); pneumatic valve dynamically adjusts the pushing force according to the stability score (resolution 0.1 N).

[0201] Example three:

[0202] Unlike example one, referring to Figures 1-3 , this embodiment further includes the following further contents: the dynamic grouping unit includes a six-axis robot arm sorting module and an AGV mobile roller group, the six-axis robot arm sorting module is equipped with a force control end effector, the thrust adjustment range of the six-axis robot arm sorting module is 5-35 N, the six-axis robot arm sorting module can support sorting of bagged, boxed and barrel-shaped products, the AGV mobile roller group supports flexible material receiving across the track, and the maximum carrying weight of the AGV mobile roller group is ≥500 kg.

[0203] The packaging unit is configured with a green packaging module, which is composed of a waste recycling conveyor and an AI ink optimization algorithm. The waste recycling conveyor automatically winds and weighs the film waste, improving the utilization rate of materials.

[0204] The AI ink optimization algorithm dynamically adjusts the ink amount according to the surface area of the packaging material, saving ink.

[0205] The palletizing robot uses NSGA-II multi-objective optimization algorithm to achieve Pareto optimization of stability, space utilization and energy consumption of the stack type, with an average energy consumption reduction of ≥18%.

[0206] The digital twin monitoring unit realizes 1:1 virtual mirroring of the device based on point cloud modeling and physical engine, and supports process parameter sensitivity analysis and AR assisted maintenance.

[0207] In this embodiment, it should be noted that the six-axis robot sorting module selects a six-axis robot with a load of 20 kg (repeat positioning accuracy ±0.1 mm), an end integrated force control sensor (measurement range 0-50 N, accuracy ±0.1 N) and a quick-change clamp interface; for bagged products, a pneumatic flexible push plate (pushing force 5-15 N) is configured, for box products, a jaw (clamping force 10-30 N) is configured, and for barrel products, a rotating clamp (rotation accuracy ±1°) is configured.

[0208] The sorting logic of the six-axis robot sorting module is: through RFID reading of the package type label, the robot performs actions according to the preset strategy:

[0209] Bagged products: the push plate is "lightly touched" with a 10N pushing force to avoid the contents from shaking;

[0210] Boxed products: the jaw detects the edge position, and after being clamped in the middle, it is translated to the output track;

[0211] Barrel products: the clamp rotates 90° to adjust the posture to ensure that the bottom surface is flat when stacking.

[0212] The AGV mobile roller group replaces the traditional fixed track material receiving device to realize flexible material receiving across the track; the navigation method uses laser SLAM+UWB combined navigation with a positioning accuracy of ±10 mm; the bearing capacity is 500 kg, and the roller linear speed is adjustable at 0.5-2 m / s; when a certain output track is blocked, the AGV automatically moves to the standby track to receive materials, avoiding production line downtime.

[0213] The digital twin monitoring unit uses virtual-real closed-loop optimization, and the real-time mirror system realizes real-time synchronization of physical equipment and virtual model (accuracy 1:1) through 5G transmission, supports production change pre-rehearsal (virtual debugging saves time 53 minutes) and AR assisted maintenance (torque standard visualization superposition);

[0214] The simulation engine pre-rehearses the production process based on historical data to identify bottleneck points (such as the mismatch between the packaging line and the stacking machine) in advance, and the whole link delay is reduced to <60 seconds.

[0215] The vibration sensor (accuracy 0.1 mm / s) monitors the bearing state in real time, and when the amplitude is >8 mm / s, the predictive maintenance module issues a warning (remaining life prediction error <5%) and automatically schedules a standby unit to take over the task, reducing unplanned downtime by 90%.

[0216] The waste recycling machine uses a stainless steel frame + food-grade PU conveyor belt (width adjustable from 500-1000 mm) to adapt to different packaging production lines such as bagged and boxed products for waste collection;

[0217] Double roll winding mechanism (active roller speed 0-100 rpm adjustable, driven roller tension sensor accuracy ±0.1N) is configured to realize automatic winding of film waste into a bundle.

[0218] The core logic of the AI ink optimization algorithm is to establish a mapping model of the surface area of the packaging and the consumption of ink, and to dynamically adjust the ink amount by real-time calculation of the product surface area.

[0219] Input parameters:

[0220] Product size (length L, width W, height H): obtained through visual inspection or work order parameters;

[0221] Ink content complexity (number of characters N, graphic area S): analyzed through work order or OCR recognition;

[0222] Output parameters:

[0223] Ink amount = α·(2LW+2LH+2WH)+β·(N·single character ink amount+S·graphic ink amount coefficient)

[0224] α = 0.05-0.1, β = 0.8-1.2 are dynamic adjustment coefficients, optimized through machine learning training.

[0225] AI training process:

[0226] Data collection: collect 100,000+ historical production data (including product size, ink content, actual ink amount, and ink quality score);

[0227] Feature engineering: extract key features such as surface area, number of characters, and graphic proportion;

[0228] Model selection: use LightGBM regression model (training error RMSE <0.02 mL), support online incremental learning;

[0229] Real-time optimization: automatically update model parameters every 50 products produced, adapt to packaging material batch differences.

[0230] AI ink optimization configures piezoelectric inkjet nozzle (resolution 600dpi, minimum ink drop 0.5pL), supports variable drop technology (VMD); installs online viscometer (accuracy ±1%), real-time monitors ink viscosity, and automatically compensates pressure parameters.

[0231] Ink saving can be achieved: standard scenarios (such as fixed text printing) save 15%-18%, complex scenarios (such as graphic LOGO) save 18%-20%; for example, a production line producing 5 million cartons per year, saves about 3000 liters of ink per year, and reduces costs by 120,000 yuan per year.

[0232] Quality improvement: The code clarity qualified rate is improved from 92% to 99.2%, avoiding character defects (such as ambiguous production date) caused by insufficient ink amount; the defect rate of over-spraying, ink dripping, etc. is reduced from 5% to 0.8%, improving the consistency of product appearance.

[0233] Comparison of green packaging module synergy benefits with traditional solutions:

[0234]

[0235] Example Four: Food Industry - Intelligent Palletizing of Bagged Milk Powder

[0236] Scenario Description

[0237] Product Specifications: 400g Bagged Milk Powder (Size: 200x150x80mm, Weight: 0.45kg)

[0238] Capacity Requirements: 5 packaging lines, with a single line speed of 60 bags per minute, with a daily output of 144,000 bags

[0239] Special Requirements: Anti-pressing, moisture-proof, and palletizing immediately into a constant-temperature warehouse

[0240] System Configuration

[0241] 1. Packaging Unit: 5 servo packaging machines (with moisture-proof heat sealing function), equipped with vacuum suction clamps; metal detector (accuracy: ≥0.3mm iron impurities)

[0242] 2. Dynamic Grouping Unit: 5-in-1-out liftable track (track spacing 300mm, lifting stroke 0-300mm)

[0243] 3. Bag Inverting Device: Pneumatic flexible push plate (pushing force range 5-15N with pressure sensor)

[0244] 4. Intelligent Palletizing Unit: 2 palletizing robots (load 50kg, repeat positioning accuracy ±1mm)

[0245] Pallet Type: Row-column (5x6 bags per layer, a total of 20 layers, with a height of 1600mm)

[0246] 5. Key Algorithm Applications:

[0247] Pallet Type Planning:

[0248] Space Utilization:

[0249] Force Control Adjustment:

[0250] Grasping Force PID Parameters: K p =1.0, K i =0.05, Kd = 0.2, actual force fluctuation control in ±2N

[0251] Operation effect:

[0252] Efficiency: Palletizing speed 1200 bags / hour (traditional system 800 bags / hour), changeover time 8 minutes

[0253] Quality: Packaging bag breakage rate from 0.5% to 0.08%, metal impurity miss rate 0

[0254] Energy consumption: Robot energy consumption 12 kWh / thousand bags (traditional system 18 kWh / thousand bags)

[0255] Example five: Chemical industry-automatic palletizing of barrel lubricating oil

[0256] Scene description

[0257] Product specifications: 20L barrel lubricating oil (iron barrel / plastic barrel, size: Weight: 18kg)

[0258] Capacity demand: 3 packaging lines, single line speed 20 barrels / minute, daily output 28,800 barrels

[0259] Special requirements: compatible with iron barrel and plastic barrel, anti-collision, need to spray code traceability

[0260] System configuration

[0261] 1. Packaging unit: 3 weighing and filling machines (accuracy ±0.1kg), equipped with electromagnetic clamp + clamp composite clamp; Intelligent code spraying machine (supporting laser code spraying and ink code switching)

[0262] 2. Dynamic grouping unit: 3-in-2-out liftable track (with ball table surface, supporting horizontal / vertical switching)

[0263] 3. Sorting device: six-axis mechanical arm (end with force control sensor, can identify iron barrel / plastic barrel material)

[0264] 4. Intelligent palletizing unit: 1 heavy-duty palletizing robot (load 200kg, arm span 3m) stacking type: staggered (iron barrel bottom layer 4×3 barrels, plastic barrel upper layer 3×4 barrels, total 8 layers)

[0265] 5. Key algorithm application:

[0266] Task allocation:

[0267] Hungarian algorithm preferentially allocates iron barrels to robot A (electromagnetic clamp) and plastic barrels to robot B (clamp)

[0268] Stability calculation:

[0269] Operating effect

[0270] Efficiency: Palletizing speed is 480 barrels per hour (300 barrels per hour for traditional systems), and material switching time < 5 minutes

[0271] Quality: The collision damage rate of the barrel body is reduced from 3% to 0.3%, and the inkjet coding accuracy rate is 100%

[0272] Maintenance: Predictive maintenance warns of bearing failures 72 hours in advance, and unplanned downtime is reduced by 90%

[0273] Example 6: Logistics industry - Mixed palletizing of boxed electronic products

[0274] Scene description

[0275] Product specifications:

[0276] Product A: Mobile phone packaging box (200×150×100mm, 1kg)

[0277] Product B: Tablet packaging box (300×200×150mm, 2kg)

[0278] Production capacity requirements: Mixed production line, daily processing of 80,000 pieces of Product A and 40,000 pieces of Product B

[0279] Special requirements: Multiple varieties and small batches, rapid changeover required, anti-tilting and anti-collapse

[0280] System configuration

[0281] 1. Packaging unit: 2 multi-functional packaging machines (compatible with carton forming and sealing), electric fixture library (jaw switching in 3 seconds)

[0282] 2. Dynamic grouping unit: 2-in-1-out honeycomb confluence platform (with RFID scanning, reading speed 200 times per second); Buffer shelf: 5-layer tray, supporting a maximum temporary storage of 500 pieces

[0283] 3. Intelligent palletizing unit: 3 collaborative robots (load 5kg, with vision guidance system

[0284] Pallet pattern: Product A in a "field" shape (4×4 boxes per layer) + Product B in an "eye" shape (3×3 boxes per layer) stacked

[0285] 4. Application of key algorithms:

[0286] Genetic algorithm optimization: Initial space utilization rate 68% → Optimized to 85%, number of iterations 50 times

[0287] Anti-tilting control:

[0288] Trigger robot complement support box when score > 0.8

[0289] Running effect

[0290] Efficiency: mixed code stacking speed 1500 boxes / hour (traditional manual sorting 1000 boxes / hour), change type time 5 minutes

[0291] Flexibility: support 20+ stacking types for quick switching, compatible with product size range 50-500mm

[0292] Safety: no collision accident in man-machine cooperation mode, in line with ISO10218 collaborative robot safety standard

[0293] Example seven: feed industry-ton bag bulk material stacking

[0294] Scene description

[0295] Product specifications: 50kg ton bag feed (size: 900x900x1100mm, weight: 50.5kg)

[0296] Capacity demand: 2 packaging lines, single line speed 10 bags / min, daily output 2400 bags

[0297] Special requirements: heavy load, dust environment, need dustproof and explosion-proof

[0298] System configuration

[0299] 1. Packaging unit: 2 ton bag packaging machines (with dust removal device), equipped with vacuum suction cup clamp (suction force ≥600N)

[0300] 2. Dynamic grouping unit: 2-in-1-out heavy rail (load capacity ≥1 ton), with laser range finder (detection distance ≥10cm)

[0301] 3. Intelligent stacking unit: 1 explosion-proof stacking robot (load 1.5 tons, protection level IP65)

[0302] Stacking type: single-layer 1x1 bag, total 5 layers (stacking height 5.5m, need to strengthen stability)

[0303] 4. Key algorithm application:

[0304] Anti-overturning calculation: M 稳 = 50.5x5x9.8x0.45 = 1119.15 N·m

[0305] M 倾 = 50.5x5x9.8x2.75xsin10° ≈ 1182.5 N·m

[0306]

[0307] Running effect

[0308] Efficiency: Palletizing speed 600 bags / hour (traditional forklift operation 400 bags / hour), dust leakage reduction 80%

[0309] Safety: Explosion-proof design passes ATEX certification, robot failure warning accuracy 100%

[0310] Cost: Labor cost reduction 70%, annual average electricity saving 150,000 yuan

[0311] Example eight:

[0312] Comparing the data of the above examples four to seven, the system's adaptability in different product forms, industry standards and production scales is reflected. Through the synergistic optimization of algorithm formula and hardware configuration, the efficiency and reliability of each link are significantly improved.

[0313] The following table is a comparison table of the data of the above examples four to seven:

[0314]

Claims

1. An automated packaging and palletizing integrated system comprising a packaging unit, a dynamic marshalling unit, an intelligent palletizing unit, a digital twin monitoring unit, and an edge decision hub, characterized in that: The packaging unit mainly completes product metering, packaging and primary detection; the dynamic grouping unit realizes multi-line product convergence, sorting and beat matching; the intelligent stacking unit performs automatic stacking operation and supports multi-robot cooperation; the digital twin monitoring unit realizes real-time monitoring and pre-performance of the whole process through virtual mirroring; and the edge decision hub is mainly responsible for data processing, algorithm scheduling and instruction issuing; The packaging unit is mainly composed of a stacking robot, a magnetic suspension electric clamp and a visual centering module. The packaging unit is connected with the dynamic grouping unit through a liftable multi-track conveyor. The track of the liftable multi-track conveyor is driven by a servo motor and a ball screw. The dynamic grouping unit includes a six-axis mechanical arm sorting module and an AGV mobile roller group. The intelligent stacking unit and the digital twin monitoring unit realize real-time data interaction through a 5G network. The intelligent stacking unit includes a mixed reality guidance system. The stacking robot is provided with MR glasses. The stacking robot supports gesture human-computer interaction and has a "man-machine-environment" cooperative obstacle avoidance function. The edge decision hub is built-in with a distributed computing architecture. The edge decision hub includes an edge computing node and a blockchain storage module. The blockchain storage module supports product life cycle tracing.

2. The automatic packaging and palletizing integrated system according to claim 1, characterized in that: The six-axis mechanical arm sorting module is equipped with a force control end effector. The thrust adjustment range of the six-axis mechanical arm sorting module is 5-35 N. The six-axis mechanical arm sorting module can support sorting of bagged, boxed and barrel-shaped products. The AGV mobile roller group supports cross-track flexible material receiving. The maximum load capacity of the AGV mobile roller group is greater than or equal to 500 kg.

3. The automatic packaging and palletizing integrated system according to claim 1, characterized in that: The edge computing node is based on a stacking type planning algorithm and a robot kinematics solving module. The data processing delay of the edge decision hub is less than 10 ms.

4. The automatic packaging and palletizing integrated system according to claim 1, wherein: The packaging unit is provided with a green packaging module. The green packaging module is composed of a waste recycling conveyor and an AI ink optimization algorithm. The waste recycling conveyor automatically winds and weighs the film waste, improving the utilization rate of materials. The AI ink optimization algorithm dynamically adjusts the ink amount according to the surface area of the packaging material to save ink.

5. The automatic packaging and palletizing integrated system according to claim 1, wherein: The stacking robot adopts an NSGA-II multi-objective optimization algorithm to realize Pareto optimization of stacking stability, space utilization and energy consumption, with an average energy consumption reduction of greater than or equal to 18%.

6. The automatic packaging and palletizing integrated system according to claim 1, wherein: The intelligent stacking unit guides the robot stacking by using a ToF depth camera. The intelligent stacking unit is provided with a six-dimensional force sensor.

7. The automatic packaging and palletizing integrated system according to claim 1, wherein: The digital twin monitoring unit realizes 1:1 virtual mirroring of equipment based on point cloud modeling and a physical engine. The digital twin monitoring unit supports process parameter sensitivity analysis and AR assisted maintenance.

8. The automatic packaging and palletizing integrated system according to claim 1, wherein: The edge decision hub adopts a mixed model of the Hungarian algorithm and reinforcement learning to schedule multi-robot tasks. The edge decision hub realizes 72-hour early warning of equipment failure through an LSTM network.

Images