L-shaped conveying belt path dynamic programming and congestion prediction system

Abstract

The application discloses an L-shaped conveying belt path dynamic programming and congestion prediction system, relates to the technical field of automatic control, and collects the material weight, speed and spacing information of each node on the conveying path, constructs a conveying path node graph, and calculates the congestion index of each path section; in combination with a light mass material jumping sensitivity model, a high-risk node is identified and marked as a congestion prediction area, a dynamic optimization model is constructed based on path risk distribution and material flow direction, a plurality of path adjustment strategies are generated and evaluated, the optimal strategy is executed according to priority, path speed adjustment, standby path switching and feeding rhythm control are realized; the strategy effect is evaluated in real time through a feedback mechanism, the path state is updated, and running data is synchronized to a historical database, forming a learnable conveying operation file, and the application can effectively improve the stability and intelligent response capability of the conveying system and is suitable for high-speed sorting and automatic packaging scenes.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, specifically to a dynamic planning and congestion prediction system for an L-shaped conveyor belt path. Background Technology

[0002] In modern production lines, L-shaped conveyor belts are widely used to transfer materials between vertical and horizontal directions. However, most existing L-shaped conveyor belt systems use fixed paths and constant speeds, and are often used in conjunction with apron conveyor structures for high-speed operation. While aprons made of low-friction materials such as PP plastic can reduce energy consumption at high speeds, they also cause lightweight items (such as packaged goods weighing less than 30g) to bounce due to inertia, leading to misaligned, incorrect, or stacked boxes. These phenomena often occur concentratedly at conveyor bends, ascending sections, or drop sections, easily triggering misidentification by misjudgment equipment (such as checkweighers and barcode scanners), resulting in incorrect rejection, incorrect sorting, and other malfunctions, severely impacting production line cycle time and product qualification rates.

[0003] Especially in automated packaging lines with diverse product types, small batches, and high-speed cycles, erroneous rejections are becoming increasingly frequent, requiring additional manual monitoring and verification of rejection results, resulting in a significant waste of human and material resources. Furthermore, current systems generally lack the ability to dynamically predict and warn of potential congestion points and bottleneck sections within the conveying path, making it impossible to achieve real-time adjustments based on the movement trends of lightweight products. This has become a major technical challenge restricting the development of intelligent conveying systems. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic planning and congestion prediction system for L-shaped conveyor belt paths to address the shortcomings of the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a dynamic planning and congestion prediction system for an L-shaped conveyor belt path, comprising:

[0006] Material information acquisition module: Collects real-time material information of each node on the L-shaped conveyor belt path and constructs a conveyor path node diagram, in which each node includes the current material weight, speed, spacing and the path segment to which it belongs;

[0007] Congestion Calculation Module: Based on the changes in material spacing and speed at each node, calculate the congestion index between nodes, which is the probability value of material stacking, jumping, or box-connecting trends occurring in each path segment.

[0008] Risk node judgment module: Based on the preset light mass material bounce sensitivity model and combined with the congestion index, it judges the set of high-risk nodes in the current path and marks the corresponding path segment as the congestion prediction area;

[0009] Model building module: Based on the distribution results of the congestion prediction area and the material flow map, a dynamic optimization model of the current path is built, and a set of optional path adjustment strategies is generated, including adjusting the path speed, switching to the backup path and adjusting the feeding rhythm.

[0010] Evaluation module: Performs simulation calculations on each strategy in the path adjustment strategy set, evaluates its impact on the target indicators, and obtains a strategy priority list;

[0011] Feedback and Adjustment Module: Based on the priority list, it executes the optimal strategy and adjusts the path adaptively and actively avoids high-risk sections based on the real-time feedback and adjusted path information.

[0012] Operation file generation module: Synchronizes the adjusted route information to the historical database to form a transport operation file that includes congestion points, bounce frequency and strategy response effects.

[0013] Preferably, the real-time material information collection at each node along the L-shaped conveyor belt path includes:

[0014] In the critical path section of the L-shaped conveyor belt, the material's transit time, pressure per unit area, and state of existence were measured.

[0015] Non-contact measurement of the front-to-back distance between adjacent materials is performed, and the material velocity vector is calculated in combination with the conveyor belt running speed.

[0016] Real-time acquisition of the weight of a single material passing through a node;

[0017] The collected data undergoes preliminary fusion processing to generate a node data stream containing four-dimensional features: weight, speed, spacing, and segment labels. This data is then updated in real time to the transport path node map.

[0018] Preferably, the step of calculating the congestion index between nodes based on the changes in material spacing and speed at each node includes:

[0019] Obtain the average material velocity and average distance between any two adjacent nodes, and calculate the rate of change of velocity. The expression is: ;in, and denoted as the average velocity at the two nodes, respectively, and d is the distance between the two nodes;

[0020] Combining the spacing threshold and the speed change rate threshold, the path is divided into "smooth flow", "congested flow" or "high-risk flow" using segment feature classification;

[0021] The frequency of material bouncing events and stacking events in each section, as well as the current average distance between materials, are collected, and a weighted average sum is calculated to obtain the congestion index CI.

[0022] Preferably, the step of determining the set of high-risk nodes in the current path and marking the corresponding path segment as a congestion prediction area includes:

[0023] A vibration sensitivity model based on material weight and velocity was constructed, and materials with a weight of less than 30 grams and a velocity of more than 500 millimeters per second were defined as highly vibration-sensitive materials.

[0024] Filter all nodes in the path and extract nodes that meet the bounce sensitivity condition and have a congestion index greater than 0.6;

[0025] The selected nodes are clustered into a set of high-risk nodes, and risk segment labels are generated based on their path distribution locations;

[0026] The route segment is marked as a congestion prediction zone, and an early warning signal is generated.

[0027] Preferably, the step of constructing a dynamic optimization model for the current path based on the distribution results of the congestion prediction area and the material flow map includes:

[0028] Obtain all marked congestion prediction sections in the path node diagram, and combine them with the material flow direction diagram to determine the main logistics channel, material density gradient, and alternative path distribution structure of the current path;

[0029] A graph search algorithm is used to search for a set of feasible paths in the path graph, sort them by congestion index weight, remove path segments with CI values ​​exceeding a set threshold, and construct a dynamic optimization path model.

[0030] Based on the load conditions of different path segments and historical operation data, three types of path adjustment strategies are generated: speed reduction for high CI sections, switching the conveying direction when there is a low-congestion alternative path, and reducing the feeding frequency when the overall load of the path is higher than the set threshold.

[0031] Combine the policy set with the objective optimization function, calculate the policy priorities, and sort and output them.

[0032] Preferably, the simulation calculation for each strategy in the path adjustment strategy set includes:

[0033] Construct a multi-objective evaluation model that includes false rejection rate, delivery stability and unit throughput, and set the expected weight value for each objective;

[0034] For each strategy in the strategy set, simulate the state evolution within 1 minute after the strategy is executed, including changes in material distribution, path load, and congestion index;

[0035] Based on the simulation output, calculate the comprehensive score for each strategy. The scoring function is the weighted sum of the scores and weights of each target indicator.

[0036] Strategies are ranked from highest to lowest based on their overall scores, and a strategy priority list is generated.

[0037] Preferably, the implementation of adaptive path adjustment and active avoidance of high-risk sections includes:

[0038] Extract the target policy with the highest score from the policy priority list and send the execution command, while simultaneously locking the current path status parameters;

[0039] During strategy execution, real-time data collection is conducted on material speed, spacing, and congestion index changes at path nodes to establish a feedback data stream.

[0040] The dynamic feedback algorithm is used to compare the rate of change of the congestion index before and after the execution. If the rate of change is lower than the preset improvement threshold, the strategy will automatically switch to the suboptimal one.

[0041] The adjusted path information is written back to the path node graph to update the risk status of the sections, enabling proactive avoidance of high-risk sections and adaptive path adjustment.

[0042] Preferably, the step of synchronizing the adjusted path information to the historical database includes:

[0043] Extract key operational data after path adjustment, including real-time congestion index, number of jump events, and material spacing distribution at each node;

[0044] Record the type of strategy executed, parameter settings, start and end times, and post-execution feedback improvement rate;

[0045] The information is structured and encapsulated into standard delivery and operation file units, including timestamps, path numbers, risk levels, and response strategy identifiers.

[0046] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0047] 1. This invention utilizes techniques such as congestion index modeling, bounce sensitivity identification, path dynamic planning, strategy priority evaluation, and feedback closed-loop control to achieve intelligent perception and dynamic response to issues such as bounce, stacking, and box-to-box formation of lightweight materials in an L-shaped conveyor belt system. The system can automatically identify high-risk path nodes, predict potential abnormal sections, and implement path adjustment, speed adjustment, or material cycle control through strategy simulation and optimization, significantly improving the stability and error rejection control capabilities of the conveyor system.

[0048] 2. This invention has significant advantages in real-time response, precise adjustment, data closed-loop, and strategy self-evolution. By accumulating historical databases and operational records, it can further drive machine learning models to optimize strategy generation and risk prediction, constructing an intelligent conveying platform with adaptive, self-diagnostic, and self-optimizing capabilities, suitable for high-cycle, high-precision, and complex intelligent packaging and sorting production lines. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0050] Figure 1 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0052] For examples, please refer to Figure 1 As shown in this embodiment, an L-shaped conveyor belt path dynamic planning and congestion prediction system includes:

[0053] Material information acquisition module: Collects real-time material information of each node on the L-shaped conveyor belt path and constructs a conveyor path node diagram, in which each node includes the current material weight, speed, spacing and the path segment to which it belongs;

[0054] Congestion Calculation Module: Based on the changes in material spacing and speed at each node, calculate the congestion index between nodes, which is the probability value of material stacking, jumping, or box-connecting trends occurring in each path segment.

[0055] Risk node judgment module: Based on the preset light mass material bounce sensitivity model and combined with the congestion index, it judges the set of high-risk nodes in the current path and marks the corresponding path segment as the congestion prediction area;

[0056] Model building module: Based on the distribution results of the congestion prediction area and the material flow map, a dynamic optimization model of the current path is built, and a set of optional path adjustment strategies is generated, including adjusting the path speed, switching to the backup path and adjusting the feeding rhythm.

[0057] Evaluation module: Performs simulation calculations on each strategy in the path adjustment strategy set, evaluates its impact on the target indicators, and obtains a strategy priority list;

[0058] Feedback and Adjustment Module: Based on the priority list, it executes the optimal strategy and adjusts the path adaptively and actively avoids high-risk sections based on the real-time feedback and adjusted path information.

[0059] Operation file generation module: Synchronizes the adjusted route information to the historical database to form a transport operation file that includes congestion points, bounce frequency and strategy response effects.

[0060] The present invention deploys the following data collection devices in multiple key sections of the L-shaped conveyor belt (including horizontal sections, corner sections, ascending and climbing sections, descending sections, and path intersection sections):

[0061] Infrared sensor array: used to acquire time information of materials entering and leaving specific nodes in order to calculate the speed of materials passing through the nodes;

[0062] Laser ranging module: installed above or to the side of the conveying path, used for non-contact detection of the spatial distance between adjacent materials;

[0063] High-frequency load cell: installed at a specific node at the bottom of the conveyor belt to obtain the weight of a single piece of material passing through that point in real time;

[0064] Pressure sensor array: used to detect the unit pressure distribution of materials during the conveying process, and to help determine the stacking or box-linking trend;

[0065] Industrial vision recognition unit: including high-speed camera and edge image processing chip, installed above some key nodes, to collect material shape features, color and distribution images for structured analysis of non-standard behaviors such as material stacking and deformation.

[0066] Each acquisition device is connected to the edge computing module via Ethernet or industrial CAN bus. The edge computing module integrates a set of data fusion and feature extraction programs to synchronously correct and time-series match multidimensional sensing data and output standardized node feature vectors.

[0067] In this invention, the transport path node graph is modeled using graph theory, wherein:

[0068] Each node represents a monitoring location on the conveyor belt with data acquisition capabilities;

[0069] Node features are defined as quadruples Ni={Wi,Vi,Di,Zi}, where:

[0070] Wi represents the weight of the material at the current node (in grams).

[0071] Vi represents the velocity of the material at this node (unit: mm / s);

[0072] Di represents the distance between the material and the previous material (unit: mm).

[0073] Zi represents the path segment number to which the node belongs (e.g., horizontal segment number 1, corner segment number 2, uphill segment number 3, etc.).

[0074] Nodes are connected by directed edges, and edge weights can represent parameters such as speed changes and congestion risks, which facilitates subsequent path analysis algorithms.

[0075] The node graph uses a sliding window update mechanism, which means that a batch of new data is collected every preset time interval (such as 200ms) to cover the old data window, so as to achieve dynamic updates.

[0076] To assist in identifying stacking, bouncing, or box-linking phenomena of lightweight materials during high-speed conveying, this invention introduces an industrial visual recognition method, specifically including the following steps:

[0077] A high-speed industrial camera captures key sections of the conveyor belt at a frame rate of no less than 200fps. The image data is then fed into an edge processing chip for grayscale conversion, background subtraction, and edge detection to extract the material's outer contour.

[0078] By using target detection algorithms (such as the YOLOv5-tiny lightweight model) to identify individual materials in an image and assign them numbers, the time-series tracking of the same material can be achieved.

[0079] The vertical coordinate difference of the same material with the same number in two adjacent frames is calculated. If the Y-axis displacement of the material fluctuates more than the set threshold Tjump=5mm in three consecutive frames, it is marked as "jumping".

[0080] The centroid overlap matching method is used to determine whether the spatial overlap of multiple numbered materials is greater than the threshold Toverlap=80%. If the condition is met, it is determined to be a "stacked" phenomenon and the node diagram is reported.

[0081] It should be noted that the jump judgment threshold Tjump=5mm: determined through experimental calibration based on the actual conveyor belt thickness and vibration characteristics; the stacking overlap rate threshold Toverlap=80%: set according to the visual recognition error and material size fluctuation tolerance; and the minimum spacing tolerance between nodes Dmin=10mm.

[0082] Data collected by multiple sensors is processed uniformly in the edge computing module using the following fusion algorithm:

[0083] All sensor data are timestamped and standardized, and the sampling frequency is uniformly set to 100Hz.

[0084] Linear interpolation is used for missing data points, and median filtering is used to smooth out fluctuation outliers.

[0085] By integrating multi-sensor data from each node with visual recognition results, a standardized four-element feature vector is generated.

[0086] All node vectors are sorted along the time axis and mapped to a graph structure to construct a real-time path graph structure with topological order.

[0087] In the completed transport path node diagram, each node contains the following basic data fields:

[0088] Current average material velocity, in millimeters per second (mm / s);

[0089] Current material front-to-back spacing, in millimeters (mm);

[0090] The frequency of material jitter events and stacking events, expressed in events per second;

[0091] The route segment number is used to identify the route's structural features (such as turning sections, uphill sections, etc.).

[0092] The material movement state between any two adjacent nodes reflects the smoothness of the path in that segment. Therefore, the congestion index is calculated from two dimensions: the "rate of change of speed" and the "material spacing" between adjacent nodes, and the frequency of historical events is introduced as a disturbance indicator to construct an overall evaluation model.

[0093] The rate of change of speed is an important parameter for judging the stability of material flow. Especially in the corner section of the conveyor or the load switching section, if the material speed changes drastically, it will often cause jumping or box-blocking.

[0094] In this invention, the rate of change of velocity (denoted as ΔV) between any two adjacent nodes is defined as: ;in, and denoted as the average velocity at the two nodes, respectively, and d is the distance between the two nodes (in millimeters).

[0095] To simplify calculation and judgment, this invention sets the empirical threshold TV to 0.2, with the unit being millimeters per second (mm / s·mm). That is, when ΔV is greater than 0.2, the segment is marked as a "speed change zone" and is considered to have potential congestion signs.

[0096] In continuous logistics systems, the average spacing between materials is a key parameter for judging flow density and smoothness. When the spacing is consistently less than the standard value, materials are very likely to stack or come into contact with each other.

[0097] Therefore, this invention sets the lower limit threshold TD for spacing to 15 mm. When the spacing D of any node is detected to be less than TD, the path segment is considered a "high-density area".

[0098] In addition, to quantify the space utilization of the path, a relative spacing index is introduced, which is the ratio of the current node spacing D to the reference spacing Dref. The reference spacing Dref is the average value of the system under long-term operation without stacking or bouncing, and is set to 40 mm.

[0099] During node graph operation, the system will count the number of jump events and stack events occurring per unit time for each path segment in real time, denoted as N_jump and N_stack, respectively. Jump events are identified by the industrial vision unit; that is, when the vertical position of the same material with the same number fluctuates by more than 5 mm in three consecutive image frames, a jump event is recorded. Stack events are determined by image overlap; when the overlap area of ​​two objects in an image exceeds 80%, it is recorded as a stack event. The event statistics period is calculated every 1 second by default, that is, the total number of events is counted per unit time.

[0100] This invention employs a weighted fusion algorithm to construct the congestion index CI, the mathematical expression of which is as follows:

[0101] The Congestion Index (CI) is equal to the weighted sum of its three components: Where: N_stack: number of stacked events per unit time; N_jump: number of jump events per unit time; T: statistical time window (default is 1 second); D: current average spacing of materials; D_ref: reference spacing (set to 40 mm); α, β, γ: weight coefficients of each factor, with preferred values ​​of 0.4, 0.4 and 0.2 respectively.

[0102] This model can flexibly adapt to different path structures and logistics types, and it has a stronger ability to identify bounce and stacking problems, especially in the transportation of lightweight products.

[0103] The calculated congestion index (CI) will be directly mapped to the weights of the corresponding edges in the transport path node graph, forming a graph structure with congestion risk as the core indicator, which can be used by subsequent path planning algorithms.

[0104] This invention sets the threshold CI_threshold to 0.6. When the CI value of any path segment exceeds this threshold, the system marks that path segment as a "potential congestion segment" and triggers the following response mechanism:

[0105] The system calls up alternative paths, dynamically adjusts the feeding rhythm, reduces the conveying speed of this section, and initiates visual tracking enhancement analysis to verify the causes of risks.

[0106] This mechanism can effectively avoid phenomena such as accidental rejection, skipping, and overlapping boxes caused by congestion, thereby improving the stability and intelligent response capability of the entire conveying system.

[0107] A bounce sensitivity model is a discriminant model used to assess the likelihood of material bounce during a conveying path. In this invention, by using material weight and conveying speed as two core indicators, a simple and effective empirical model is constructed to quickly identify highly bounce-sensitive materials.

[0108] Input parameter definition:

[0109] Material weight W: acquired in real time by a high-frequency weighing sensor, in grams (g).

[0110] Material velocity V: Calculated by laser rangefinder combined with timestamp difference, in millimeters per second (mm / s).

[0111] Spacing D: Distance from the previous material, in millimeters, used for the jump trend enhancement item.

[0112] Sensitivity rule model setting: Based on a large amount of experimental data and transport observations, the frequency of fluctuation phenomenon increases significantly under the following material characteristics:

[0113] Weight less than 30 grams;

[0114] The speed is greater than 500 millimeters per second.

[0115] Therefore, the present invention preferably employs the following rule-based judgment method:

[0116] If the material weight in a node is less than 30 grams and the material velocity is greater than 500 millimeters per second, the material is determined to be a highly vibration-sensitive material, and the node is marked as a "vibration-sensitive node".

[0117] To improve model stability, the following supplementary items can be introduced in engineering applications:

[0118] If the spacing D is less than 75% of the reference spacing D_ref, the risk of jitter increases by one level;

[0119] Detecting two consecutive jumping behaviors in visual recognition can correct the model's sensitivity threshold.

[0120] To further ensure the accuracy of path segment judgment results, this invention combines the output of the bounce sensitivity model with the congestion index as the basis for judging high-risk nodes.

[0121] As defined in the aforementioned specification, the congestion index is a path congestion level indicator calculated by combining the frequency of stacking events, the frequency of bouncing events, and the current material spacing. Its value ranges from 0 to 1, indicating the likelihood of dense logistics or stacking behavior occurring on the path segment.

[0122] This invention sets the congestion index threshold CI_threshold to 0.6. That is, a node is only identified as a high-risk node when it is a volatile sensitive node and its CI value is greater than 0.6.

[0123] This joint condition ensures that the algorithm considers both the material's own bouncing characteristics and the actual running state of the path, avoiding misjudgment or overjudgment.

[0124] In the path node graph, the system performs the following filtering and clustering steps on all nodes:

[0125] Iterate through all path nodes and select nodes that simultaneously meet the following conditions:

[0126] The material at this node is a vibration-sensitive material;

[0127] The congestion index CI calculated for the current node is ≥ 0.6.

[0128] The set of nodes that have been filtered out is defined as the high-risk candidate node set, denoted as R.

[0129] Considering that jumping or stacking behaviors often occur continuously along a path rather than in isolated nodes, path segment-level clustering is required for the candidate node set R. The specific steps are as follows:

[0130] Sort the nodes in the node set R according to their path location information;

[0131] Nodes with consecutive node numbers less than or equal to 2 are grouped into the same cluster unit;

[0132] Each clustering unit is labeled as a subset of high-risk nodes, denoted as R_i;

[0133] Each path segment covered by R_i is marked as a risk path segment and numbered Z_i.

[0134] This clustering strategy is based on path continuity and data topology, adapts to the actual layout of the conveyor belt, and is especially suitable for continuous risk identification in fault-prone areas such as corner sections and climbing sections.

[0135] For each set of high-risk nodes R_i, the corresponding path segment Z_i will be marked as a congestion prediction segment in real time, and the following operations will be initiated:

[0136] In the transport path node graph, update the attribute status of path segment Z_i to "predicted congestion";

[0137] Introduce a label weight (e.g., increase the weight by 0.2) into the edge weights to influence subsequent path selection.

[0138] Package the path segment to which Z_i belongs and its high-risk node numbers into early warning information;

[0139] The warning information is transmitted to the path optimization and control module via the internal bus;

[0140] The control module responds by generating corresponding adjustment strategies, such as deceleration, diversion, and temporary cycle adjustment.

[0141] All congestion prediction sections and their triggering conditions are stored in a historical database; the model can be self-learned and the hopping sensitivity parameters can be updated based on historical records.

[0142] The dynamic optimization path model constructed in this invention is based on the transportation path node graph. It introduces CI values ​​as edge weights for modeling and uses a heuristic search algorithm to achieve real-time path substitution and load prediction analysis.

[0143] In the path graph, the weight W of each path edge is given by the following formula: Where d is the physical distance of the path segment (in millimeters), CI is the congestion index of the segment (ranging from 0 to 1), and K is an adjustment factor, set to 100 by default. By introducing a CI weighting mechanism, the search algorithm automatically avoids high-risk paths.

[0144] This invention employs a weighted A* algorithm for path selection, with the heuristic function being the sum of the weights already traversed on the current path and the straight-line distance to the target node. The algorithm flow is as follows:

[0145] Set the starting point and target discharge port;

[0146] Traverse the path nodes, ignoring path segments with a CI value greater than 0.8;

[0147] Dynamically adjust search priority, giving preference to path segments with low CI values ​​and short distances;

[0148] Returns a set of feasible paths P, each path with its path cost and average CI.

[0149] This algorithm can calculate a set of alternative paths in milliseconds for subsequent strategy decisions.

[0150] Based on the completed dynamic path model, the system generates the following three types of path adjustment strategies to cope with operational anomalies in different scenarios:

[0151] This is suitable for situations where the CI value of a path is high but there is no alternative path. By reducing the conveyor speed of this path segment (e.g., by 20%), the material spacing can be artificially increased, reducing the probability of bouncing and stacking. The adjustment range is dynamically mapped according to the CI value; no adjustment is made when CI is 0.6, and the speed is reduced by 30% when CI is 0.9.

[0152] When there are path segments with a CI higher than 0.8 in the primary path, and the average CI value of the alternative paths is less than 0.6, the system performs a path switch. Alternative paths must meet the following conditions:

[0153] Parallel to the current path;

[0154] Having the same target export;

[0155] The current no-load or low-load condition is (i.e., the number of materials passing through per unit time is less than 10 pieces / minute).

[0156] The system issues a scheduling command to redirect part or all of the feed flow to an alternative path, achieving temporary decoupling.

[0157] If there is a widespread increase in CI values ​​across the entire path network (e.g., more than 60% of path segments have CI values ​​higher than 0.6) and there is no effective path switching solution, the system will automatically reduce the feeding frequency. By default, the cycle time is reduced by 10% each time (e.g., from 60 pieces per minute to 54 pieces per minute), and the path status is reassessed every 30 seconds to restore the feeding rhythm.

[0158] The above strategies are prioritized to select the most suitable control method for the current operating state. This invention defines a set of strategy evaluation functions with the goal of minimizing the expected false rejection rate E while maintaining a high logistics throughput rate T.

[0159] The following definitions apply to each strategy S_i: the expected false rejection rate E_i, the path stability index St_i (0–1), and the throughput T_i (pieces / minute) per unit time.

[0160] The overall score, Score_i, is calculated using the following weighting function: Score_i equals weight A multiplied by 1 minus E_i, plus weight B multiplied by St_i, plus weight C multiplied by T_i. Recommended weight settings are: A=0.5, B=0.3, C=0.2. The strategy with the highest score is executed first.

[0161] After scoring and ranking the strategies, the top two strategies with the highest scores are sent in parallel to the candidate pool of the control module, and the scheduling engine selects the best one to execute based on the real-time load and cycle time.

[0162] In this embodiment, some core parameters are adaptively set based on the material type, structural layout, and production cycle of the conveying system, mainly including:

[0163]

[0164] The above parameters can be dynamically adjusted based on historical data and production targets to ensure that the system always maintains optimal operating conditions.

[0165] To enable effective comparison of adjustment strategies for different paths, this invention defines the following three core evaluation objectives, which form the basis of the multi-objective evaluation model:

[0166] Objective 1: False Reject Rate: The false rejection rate refers to the ratio of the number of non-compliant rejections caused by fluctuations, stacking, or path anomalies to the total number of shipments per unit time, expressed as a percentage. This indicator reflects system stability and identification accuracy, and is preferably the primary evaluation objective, with a weight set at 0.5.

[0167] Objective 2: Stability Index: This index quantifies the change in the range of material spacing fluctuations in the path after strategy implementation. It is defined as the standard deviation of the average material spacing per unit time in the path; a lower value indicates more stable material flow. The preferred weight for this index is 0.3.

[0168] Objective 3: Throughput: The number of materials successfully delivered from the starting point to the target inspection area per unit time, measured in pieces per minute. This is used to measure the impact of the strategy on overall production line efficiency, with a preferred weight of 0.2.

[0169] All three indicators are based on simulation output data and are incorporated into the comprehensive score after standardization.

[0170] Each strategy evaluation simulation lasts 60 seconds and is used to predict the impact of the strategy on the logistics system within 1 minute. The duration can be dynamically adjusted.

[0171] Simulation input parameters include, but are not limited to:

[0172] Current path node graph structure;

[0173] Distribution of congestion prediction zones;

[0174] Current material flow rate, weight, and cycle time;

[0175] The path adjustment strategies to be evaluated (such as deceleration, switching paths, adjusting pace, etc.).

[0176] An event-driven discrete simulation model is used, with material units as the basic granularity, to simulate their movement, waiting, and switching states along a path according to strategic behaviors. Key material behaviors during the simulation (such as bouncing, stacking, box concatenation, and erroneous rejection) are recorded as state event triggers in the simulation log.

[0177] By setting the starting material queue, simulation path nodes, and strategy behavior response rules, the complete behavioral evolution process of the target path under the action of the strategy can be modeled.

[0178] After the simulation is complete, the following numerical output results will be collected:

[0179] E: False rejection rate (unit: %)

[0180] S: Conveying stability index (unit: mm);

[0181] T: Throughput (unit: pieces / minute);

[0182] To standardize dimensions and facilitate scoring and ranking, the above indicators will undergo min-max normalization. The normalization formula is as follows: For an indicator value X of a certain strategy, the normalized score X_n is: , where X_max and X_min are the maximum and minimum values ​​in the historical evaluation window.

[0183] For the false rejection rate E and the spacing fluctuation S, smaller values ​​are better, and a normalized result of 1 indicates optimal performance. For the throughput T, larger values ​​are better, and a normalized result closer to 1 indicates optimal performance.

[0184] Finally, for each strategy Si, calculate its comprehensive score Score_i: Where E_n, S_n, and T_n are the normalized scores for false rejection rate, stability, and throughput, respectively.

[0185] After all path adjustment strategies have completed simulation evaluation and score calculation, the system will execute the following sorting process:

[0186] Sort all strategies from highest to lowest using the overall score Score_i as the key;

[0187] For strategies with the same score, the one with lower execution logic complexity will be given priority (e.g., speed adjustment is preferred over path switching).

[0188] Generate a strategy priority list L, which includes strategy number, strategy description, score, and expected improvement value for incorrect removals.

[0189] The following is an example of a list L structure:

[0190]

[0191] In the aforementioned Strategy Priority List (SPL), each candidate strategy has been evaluated through multi-objective simulation and assigned a corresponding score. The system prioritizes the strategy with the highest score for execution.

[0192] Extract the highest-scoring strategy S_opt from SPL. Strategy information includes:

[0193] Strategy types (speed adjustment, path switching, beat adjustment, etc.);

[0194] Execute the target path segment;

[0195] Parameter values ​​(such as speed reduction, target path ID switching, and beat adjustment ratio);

[0196] Expected scope of impact and expected rate of improvement.

[0197] Policy control commands are issued to the transport control layer via the path control module. Depending on the policy type, one of the following behaviors is triggered:

[0198] Adjust the frequency of the drive motor in the target path segment to control the conveying speed;

[0199] Switch the direction of the electric sorting device to guide the materials into the backup path;

[0200] Adjust the control signal of the feeding device to change the feeding frequency or interval.

[0201] Meanwhile, the system saves a snapshot of the current path status parameters for subsequent feedback comparison and analysis.

[0202] After the strategy is executed, the system immediately enters the dynamic feedback collection phase to evaluate whether the strategy has achieved the expected control effect. The collected feedback data includes:

[0203] The average velocity of materials in the current path segment, in millimeters per second;

[0204] Average distance between adjacent materials, in millimeters;

[0205] The updated Congestion Index (CI) value for the route segment;

[0206] Frequency of bouncing and stacking events.

[0207] The preferred data sampling frequency is 5 times per second, with a continuous acquisition window of 10 seconds to form a stable data baseline.

[0208] The system uses a sliding window averaging method and a median filtering algorithm to process the feedback data, avoiding abnormal peak values ​​from misleading the determination of the strategy's effectiveness.

[0209] This invention defines the Improvement Rate (IR) to determine whether the control effect of the current strategy on the target path segment meets the target.

[0210] The feedback improvement rate (IR_CI) is defined as the change in the CI value of a path segment before and after the strategy is implemented. It is calculated as follows: IR_CI equals (CI_before − CI_after) divided by the previous value CI_before. Where: CI_before is the congestion index of the path segment before the strategy is implemented; CI_after is the average CI value over a 10-second window after the strategy is implemented.

[0211] The present invention sets the minimum improvement threshold T_IR for judging the effectiveness of the strategy to be 0.1, that is, the improvement rate must reach more than 10% to be considered as the strategy producing a substantial control effect.

[0212] If IR_CI ≥ 0.1, the current strategy is effective, and the system will maintain the current state and wait for the next cycle evaluation.

[0213] If IR_CI < 0.1, it is considered that the current strategy has not met expectations, and the system will automatically call the second-ranked strategy S_alt in the priority list to enter a new round of execution and evaluation.

[0214] This judgment mechanism has self-correcting capabilities, ensuring that path control does not fall into inefficient strategies.

[0215] Once the strategy is confirmed to be effective, the system will perform the following two updates to the path node graph structure:

[0216] Write the CI value, material speed, and spacing of the current path segment into the node graph database;

[0217] Update the "Status Label" of this path segment to "Adjusted";

[0218] The marking strategy affects the execution time window of the segment, which is used for training subsequent learning modules.

[0219] The system will reanalyze path segments whose current CI value is still higher than 0.6. If the path segment remains in a high-risk state 10 seconds after adjustment, the system will execute the following avoidance procedure:

[0220] Search for alternative paths that run parallel to high-risk segments and determine if their mean CI is less than 0.5;

[0221] If an available path exists, a new material flow diagram is generated, and the main logistics path is replanned.

[0222] Adjust the electric sorting device to prioritize guiding the feed to the alternative path, thereby reducing the load on the original path;

[0223] Set the shortest recovery assessment period (preferably 60 seconds) and reassess whether the high-risk path segment has recovered.

[0224] After the system executes the path adjustment strategy, the control module will continuously monitor and record several key operating parameters of the target path segment. To ensure the integrity and consistency of the feedback data, this invention employs the following data extraction method:

[0225] Based on the path node graph, within a 10-second evaluation window after policy execution, the latest Congestion Index (CI) for all path segments is calculated, and nodes with a CI value greater than or equal to 0.6 are selected as the "Congested Points" set. Each point records the following:

[0226] Node number;

[0227] Section type (straight line, corner, uphill, etc.);

[0228] CI value;

[0229] Current path load (number of materials passing through per unit time);

[0230] Current path speed and spacing.

[0231] The jolt event detection results are obtained from the industrial vision subsystem. Each jolt event is determined by a sequence of frames where the vertical displacement of the material exceeds the jolt threshold of 5 mm, as identified by the high-speed camera. Within the window after the strategy is executed, the number of jolt events on the target path segment is counted, and the frequency is recorded (number of events divided by the time window length, in events / second).

[0232] After each strategy execution, the system also needs to record the strategy's execution parameters and actual response results to evaluate the strategy's effectiveness and subsequent reuse. Specifically, this includes the following fields:

[0233] Strategy types (such as speed adjustment, path switching, beat adjustment, etc.);

[0234] Execution start and end times (timestamp format);

[0235] Strategy parameter values ​​(such as speed regulation ratio, path ID, cycle time reduction, etc.);

[0236] Post-implementation feedback improvement rate (IR value, in percentage);

[0237] Successful (True / False, depending on whether the IR is greater than the set threshold of 0.1).

[0238] To ensure data standardization and traceability, this invention encapsulates operational data into Transport Log Units and stores them using a unified data structure format. This structure includes, but is not limited to, the following fields:

[0239] File number (unique identifier, format: path ID + timestamp);

[0240] Path segment number;

[0241] Material type;

[0242] List of congested locations (including node number and CI value);

[0243] Frequency of fluctuation;

[0244] Execution strategy number and parameters;

[0245] Policy response improvement rate (IR);

[0246] Status labels (e.g., "effective", "partially mitigated", "ineffective");

[0247] Timestamps and execution cycles.

[0248] All fields are encapsulated in JSON format and can be uploaded to the cloud or local area network database via the edge gateway interface. Alternatively, a CSV structure can be used for lightweight local storage.

[0249] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A dynamic planning and congestion prediction system for an L-shaped conveyor belt path, characterized in that: include: Material information acquisition module: Collects real-time material information of each node on the L-shaped conveyor belt path and constructs a conveyor path node diagram, in which each node includes the current material weight, speed, spacing and the path segment to which it belongs; Congestion Calculation Module: Based on the changes in material spacing and speed at each node, calculate the congestion index between nodes, which is the probability value of material stacking, jumping, or box-connecting trends occurring in each path segment. Risk node judgment module: Based on the preset light mass material bounce sensitivity model and combined with the congestion index, it judges the set of high-risk nodes in the current path and marks the corresponding path segment as the congestion prediction area; Model building module: Based on the distribution results of the congestion prediction area and the material flow map, a dynamic optimization model of the current path is built, and a set of optional path adjustment strategies is generated, including adjusting the path speed, switching to the backup path and adjusting the feeding rhythm. Evaluation module: Performs simulation calculations on each strategy in the path adjustment strategy set, evaluates its impact on the target indicators, and obtains a strategy priority list; Feedback and Adjustment Module: Based on the priority list, it executes the optimal strategy and adjusts the path adaptively and actively avoids high-risk sections based on the real-time feedback and adjusted path information. Operation file generation module: Synchronizes the adjusted route information to the historical database to form a transport operation file that includes congestion points, bounce frequency and strategy response effects.

2. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 1, characterized in that: in, The collection of real-time material information at each node along the L-shaped conveyor belt path includes: In the critical path section of the L-shaped conveyor belt, the material's transit time, pressure per unit area, and state of existence were measured. Non-contact measurement of the front-to-back distance between adjacent materials is performed, and the material velocity vector is calculated in combination with the conveyor belt running speed. Real-time acquisition of the weight of a single material passing through a node; The collected data undergoes preliminary fusion processing to generate a node data stream containing four-dimensional features: weight, speed, spacing, and segment labels, which is then updated in real time to the transport path node map.

3. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 1, characterized in that: The calculation of the congestion index between nodes based on the changes in material spacing and speed at each node includes: Obtain the average material velocity and average distance between any two adjacent nodes, and calculate the rate of change of velocity. The expression is: ;in, and denoted as the average velocity at the two nodes, respectively, and d is the distance between the two nodes; Combining the spacing threshold and the speed change rate threshold, the path is divided into "smooth flow", "congested flow" or "high-risk flow" using segment feature classification; The frequency of material bouncing events and stacking events in each section, as well as the current average distance between materials, are collected, and a weighted average sum is calculated to obtain the congestion index CI.

4. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 1, characterized in that: in, The step of determining the set of high-risk nodes in the current path and marking the corresponding path segment as a congestion prediction area includes: A vibration sensitivity model based on material weight and velocity was constructed, and materials with a weight of less than 30 grams and a velocity of more than 500 millimeters per second were defined as highly vibration-sensitive materials. Filter all nodes in the path and extract nodes that meet the bounce sensitivity condition and have a congestion index greater than 0.6; The selected nodes are clustered into a set of high-risk nodes, and risk segment labels are generated based on their path distribution locations; The route segment is marked as a congestion prediction zone, and an early warning signal is generated.

5. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 1, characterized in that: in, The dynamic optimization model for the current path, based on the distribution results of the congestion prediction zone and the material flow map, includes: Obtain all marked congestion prediction sections in the path node diagram, and combine them with the material flow direction diagram to determine the main logistics channel, material density gradient, and alternative path distribution structure of the current path; A graph search algorithm is used to search for a set of feasible paths in the path graph, sort them by congestion index weight, remove path segments with CI values ​​exceeding a set threshold, and construct a dynamic optimization path model. Based on the load conditions of different path segments and historical operation data, three types of path adjustment strategies are generated: speed reduction for high CI sections, switching the conveying direction when there is a low-congestion alternative path, and reducing the feeding frequency when the overall load of the path is higher than the set threshold. Combine the policy set with the objective optimization function, calculate the policy priorities, and sort and output them.

6. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 5, characterized in that: in, The simulation calculation for each strategy in the path adjustment strategy set includes: Construct a multi-objective evaluation model that includes false rejection rate, delivery stability and unit throughput, and set the expected weight value for each objective; For each strategy in the strategy set, simulate the state evolution within 1 minute after the strategy is executed, including changes in material distribution, path load, and congestion index; Based on the simulation output, calculate the comprehensive score for each strategy. The scoring function is the weighted sum of the scores and weights of each target indicator. Strategies are ranked from highest to lowest based on their overall scores, and a strategy priority list is generated.

7. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 6, characterized in that: in, The implementation of adaptive path adjustment and active avoidance of high-risk sections includes: Extract the target policy with the highest score from the policy priority list and send the execution command, while simultaneously locking the current path status parameters; During strategy execution, real-time data collection is conducted on material speed, spacing, and congestion index changes at path nodes to establish a feedback data stream. The dynamic feedback algorithm is used to compare the rate of change of the congestion index before and after the execution. If the rate of change is lower than the preset improvement threshold, the strategy will automatically switch to the suboptimal one. The adjusted path information is written back to the path node graph to update the risk status of the sections, enabling proactive avoidance of high-risk sections and adaptive path adjustment.

8. The L-shaped conveyor belt path dynamic planning and congestion prediction system according to claim 1, characterized in that: in, The step of synchronizing the adjusted path information to the historical database includes: Extract key operational data after path adjustment, including real-time congestion index, number of jump events, and material spacing distribution at each node; Record the type of strategy executed, parameter settings, start and end times, and post-execution feedback improvement rate; The information is structured and encapsulated into standard delivery and operation file units, including timestamps, path numbers, risk levels, and response strategy identifiers.

Images