Automobile assembly line dynamic periodic material distribution scheduling system

Abstract

This invention relates to the field of intelligent manufacturing logistics technology, specifically a dynamic periodic material delivery scheduling system based on an automotive assembly line. The system includes: a data acquisition module that acquires real-time information on material consumption and warehouse inventory status at production workstations; a demand forecasting module that predicts the material demand sequence for each workstation within a future cycle, including precise demand time windows; a constraint management module that integrates constraints such as vehicle capacity, material attributes, and channel rules; and an optimization modeling module that constructs a multi-objective optimization model based on the demand sequence and constraints, aiming to minimize the total travel distance and satisfy all time windows. A scheduling generation module solves this model to generate a dynamic delivery plan including vehicle departure time, loading list, delivery route, and estimated arrival time. This invention achieves accurate dynamic forecasting of material demand and multi-objective collaborative optimization of delivery cost and timeliness.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing logistics technology, and in particular to a dynamic periodic material distribution scheduling system based on automobile assembly lines. Background Technology

[0002] Current material distribution scheduling on automotive assembly lines primarily employs pull or push strategies based on fixed time intervals or static inventory thresholds. Existing solutions typically rely on production plans or historical consumption experience to establish relatively fixed delivery schedules and routes, or trigger replenishment requests when workstation inventory falls below a safety threshold. The drawback of this model is its relatively rigid scheduling logic, which cannot respond in real-time to dynamic material consumption fluctuations caused by production rhythm adjustments, order insertions, or abnormal downtime. When production rhythms change, fixed delivery plans can easily lead to materials arriving too early, resulting in line-side inventory buildup, or arriving too late, causing production line downtime.

[0003] Existing scheduling methods have relatively singular optimization objectives, making it difficult to achieve an effective balance between delivery costs and delivery timeliness. Common methods either solely pursue the shortest delivery route to reduce transportation costs, or treat the time required to meet workstation needs as an inviolable hard constraint for rigid scheduling. The former may lead to production interruptions due to neglecting time requirements; the latter may fail to find a feasible solution or result in a significant increase in delivery distance when demand is high or transportation is restricted. There is a lack of a scheduling mechanism that can simultaneously quantify and evaluate the degree to which delivery distance and time window satisfaction are met, and intelligently balance and collaboratively optimize accordingly. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a dynamic periodic material delivery scheduling system based on automobile assembly lines.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a dynamic periodic material distribution scheduling system based on an automotive assembly line, comprising: The data acquisition module collects material consumption data from production workstations on the automotive assembly line in real time and receives material inventory status information from the material storage management system. The material inventory status information includes the real-time inventory quantity of each material, its location in the storage area, and the physical specifications of the material. The demand forecasting module predicts the material demand sequence for each workstation in the future production cycle based on the material consumption data of the production workstation. The material demand sequence includes the demand material code, the predicted demand quantity, and the demand time window. The constraint management module uses the single loading capacity of the material delivery vehicle, the physical specifications in the material inventory status information, and the passage rules of the logistics channel within the assembly line as constraints. The optimization modeling module constructs a multi-objective optimization model based on the material demand sequence and the constraints, with the goal of minimizing the total travel distance of delivery vehicles and satisfying all demand time windows. The scheduling generation module solves the multi-objective optimization model to generate an initial dynamic delivery plan, which includes the departure time of each delivery vehicle, the list of loaded materials, the delivery route sequence, and the expected arrival time at each workstation.

[0006] As a further aspect of the present invention, the step of predicting the material demand sequence of each workstation in the future production cycle based on the material consumption data of the production workstation includes: The material consumption data includes the actual usage quantity and timestamp of each workstation for different material codes; Time series analysis is performed on the actual usage quantity and corresponding timestamp in the material consumption data to identify the periodic patterns and trend components of each workstation's consumption of each material. Based on the identified periodic patterns and trend components, an exponential smoothing model is used to make rolling predictions of material consumption at workstations, thereby obtaining the basic predicted demand at multiple consecutive time points in the future. Obtain the future production plan for the automotive assembly line, which includes the planned production models, the assembly sequence of each model, and the planned production takt time. Based on the bill of materials for the planned production models, the basic forecast demand is decomposed and revised by model to generate a refined forecast demand driven by the model. By combining the assembly sequence with the production cycle, the refined forecast demand is mapped onto a specific timeline to form the material demand sequence containing a clear demand time window.

[0007] As a further aspect of the present invention, the step of constructing a multi-objective optimization model based on the material demand sequence and the constraints includes: Each demand in the material demand sequence is transformed into a demand node in the optimization model, and the demand node is associated with its demand material code, predicted demand quantity and demand time window; Define decision variables to represent whether a delivery vehicle departs from a storage point and whether it serves a pair of demand nodes in a specific order; Establish an objective function, which is in the form of a weighted summation. Its first part is the sum of the travel distances of all delivery vehicles, and its second part is the penalty cost incurred by all demand nodes that do not receive service within the demand time window. Establish a first set of constraints that ensure that the total volume and weight of the materials loaded on each delivery vehicle do not exceed its single loading capacity. Establish a second set of constraints that ensure that the predicted demand at each demand node must be fully met and completed by the same delivery vehicle in a single service. A third set of constraints is established, which is a time window constraint, to ensure that the delivery vehicle arrives at each demand node no earlier than the start time of its demand time window and no later than the end time of its demand time window. A fourth set of constraints is established, which is based on the traffic rules of the logistics channel and restricts the driving speed of vehicles in a specific area, prohibits turning, or prohibits simultaneous entry.

[0008] As a further aspect of the present invention, solving the multi-objective optimization model to generate an initial dynamic delivery plan includes: An improved genetic algorithm is used to solve the multi-objective optimization model. The improved genetic algorithm uses a mixed encoding method of real numbers and integers to encode the path of the delivery vehicle. Initialize the population, which contains multiple randomly generated initial delivery schemes, each of which satisfies the single-load capacity constraint of the material delivery vehicle; Calculate the fitness value of each individual in the population, where the fitness value is the reciprocal of the objective function value; Perform a selection operation, and select individuals from the current population to enter the mating pool based on the fitness value using a roulette wheel strategy; Perform a crossover operation, using the sequential crossover operator to generate new offspring individuals from individuals in the mating pool; Perform mutation operations to swap two points within a path or reverse a path segment in offspring individuals with a certain probability; For the population after selection, crossover, and mutation operations, feasibility repair is performed to ensure that newly generated individuals satisfy the second set of constraints and the third set of constraints; The process iteratively executes steps from calculating the fitness value to feasibility repair until the preset maximum number of iterations is reached or the objective function value converges. The individual with the smallest objective function value is selected from the final population, and the initial dynamic delivery plan containing vehicle departure time, loading list, and route sequence is generated by decoding.

[0009] As a further aspect of the present invention, it also includes: The dynamic adjustment module distributes the initial dynamic delivery plan to the delivery vehicles for execution, and continuously collects the actual location information of the vehicles and the actual material consumption rate of each workstation during the execution process. The actual material consumption rate is compared with the predicted rate to calculate the material demand prediction deviation rate. When the material demand prediction deviation rate exceeds a preset threshold, the dynamic adjustment process of the delivery plan is triggered. In the dynamic adjustment process of the delivery plan, the initial dynamic delivery plan is rescheduled online by combining the latest material inventory status information and the actual location information of the vehicles. An adjusted dynamic delivery plan is generated and issued, which is used to update the remaining tasks of delivery vehicles to ensure the continuity of material supply. The step of comparing the actual material consumption rate with the predicted rate and calculating the material demand forecast deviation rate includes: Set a fixed monitoring cycle, and at the end of each monitoring cycle, calculate the total actual consumption of all material codes at each workstation during the just-ended cycle. Extract the corresponding workstation, material code, and predicted demand for the same time period from the material demand sequence; For each type of material at each workstation, calculate the absolute difference between the actual total consumption and the predicted demand. The absolute difference is compared with the predicted demand to obtain the material demand prediction deviation rate of the material in the monitoring period of the work site. A weighted average of the material demand forecast deviation rates for all materials at all workstations is used to obtain a global material demand forecast deviation level. When the material demand forecast deviation rate of any material exceeds its independent preset threshold, or the global material demand forecast deviation level exceeds the global preset threshold, the trigger condition is determined to be met, and the dynamic adjustment process of the delivery plan is initiated.

[0010] As a further aspect of the present invention, the online rescheduling of the initial dynamic delivery plan includes: Freeze all service tasks completed by delivery vehicles, i.e., workstation nodes that have arrived and unloaded. Obtain the real-time location information, current cargo status, and remaining route sequence of all delivery vehicles; Obtain the latest material inventory status information and the updated material demand sequence at the time of triggering rescheduling; the updated material demand sequence has been corrected according to the latest actual material consumption rate. The current time is taken as the new scheduling start time, the real-time location of all vehicles is taken as the new starting point, and the current cargo status of the vehicles is taken as the new initial load. Based on the latest material inventory status information, the updated material demand sequence, and the constraints, a local rescheduling optimization model is constructed. The local rescheduling optimization model only optimizes demand nodes that have not yet been served. Solve the local rescheduling optimization model to generate a new remaining path sequence and timetable for each delivery vehicle starting from its current location; The newly generated remaining path sequence and timetable are integrated with the frozen completed tasks to form a new draft delivery plan covering the entire scheduling cycle.

[0011] As a further aspect of the present invention, the construction of a local rescheduling optimization model includes: Determine the optimization scope, which includes all unserved demand nodes, the real-time location of all delivery vehicles, and storage points; Define new decision variables to represent whether the delivery vehicle departs from its current location and whether it serves a pair of unserved demand nodes in a new order; A new objective function is established, which, while minimizing the penalty cost of additional travel distance and time window, adds a penalty term for the magnitude of plan changes in order to maintain the relative stability of the scheduling scheme. Establish loading constraints to ensure that the cumulative load of each delivery vehicle when serving a new route never exceeds its single loading capacity, taking into account its current cargo status. Establish demand satisfaction constraints to ensure that the demand of each unserved node is met in the rescheduled plan; Establish a new time window constraint, taking the current moment as the zero point, recalculate the time for vehicles to arrive at each unserved demand node, and ensure that the time falls within its demand time window. The vehicle paths obtained by solving the local rescheduling optimization model are used as the new remaining path sequence.

[0012] As a further aspect of the present invention, it also includes: The evaluation and parameter adjustment module summarizes the delivery task execution data for the entire day after the end of each complete production day or shift. The delivery task execution data includes total mileage, total number of delivery trips, number of time window violations, and number of rescheduling triggers. Extract material availability rate data for the same period from the warehouse management system. The material availability rate data reflects the frequency and duration of production line stoppages caused by material shortages at workstations. Based on the delivery task execution data and the material availability rate data, calculate the comprehensive efficiency index of the delivery system; The comprehensive performance indicators are compared and analyzed with historical data or preset performance benchmarks. Based on the results of the comparative analysis, the weight coefficients of the objective function in the multi-objective optimization model, the preset threshold of the material demand prediction deviation rate, or the trigger sensitivity parameters of the dynamic adjustment process of the delivery plan are adaptively adjusted. The adjusted parameters will be applied to the scheduling process of the next production cycle.

[0013] As a further aspect of the present invention, the step of calculating the comprehensive performance index of the delivery system based on the delivery task execution data and the material availability rate data includes: The total mileage is standardized and divided by the total weight of delivered materials to obtain the delivery mileage per unit weight of materials. The total number of delivery trips is standardized and divided by the total number of material types to obtain the average delivery frequency per material type. The time window violation rate is obtained by calculating the ratio of the number of violations of the time window to the total number of service nodes. The logistics efficiency sub-indicator is obtained by weighted summing of the delivery mileage per unit weight of material, the average delivery frequency per unit type of material, and the violation rate of the time window. The material availability rate data is converted into the proportion of workstation downtime to total production time, which is used as a sub-indicator for material availability. The logistics efficiency sub-indicator and the material support sub-indicator are weighted and summed again to obtain the comprehensive performance index of the distribution system. The lower the comprehensive performance index value, the better the comprehensive performance.

[0014] As a further aspect of the present invention, it also includes: The emergency delivery dispatch module receives emergency order insertion production instructions, which include the vehicle model to be produced, the production quantity, and the required online time. Based on the bill of materials for the urgently produced vehicle model, immediately calculate the additional urgent material requirements for each workstation caused by the emergency order. Check the current inventory status of the materials to confirm whether the available inventory of the emergency materials meets the demand for the new emergency materials; If the inventory is sufficient, immediately lock the corresponding quantity of materials in the available inventory and create the highest priority emergency demand node for these new urgent material needs; The urgent demand node is inserted into the current dynamic delivery plan, and the demand time window of the urgent demand node is calculated backward based on the required online time. Immediately trigger a local dynamic adjustment process for the aforementioned delivery plan, using the demand time window of the urgent demand node as a hard constraint, and replan the routes of some or all delivery vehicles to ensure that urgent materials are delivered to the destination on time with priority. In subsequent regular scheduling cycles, the emergency demand nodes that have completed delivery will be removed from the demand list, and the system will return to normal scheduling mode.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on real-time material consumption data from production workstations, a time-series forecasting algorithm dynamically generates material demand sequences for each workstation within the future production cycle. These sequences not only include material types and quantities but also provide precise demand time windows. This solution elevates demand forecasting from rough estimates based on historical statistics or fixed timeframes to refined, time-series projections of material consumption at specific future points in time for each workstation. It drives material delivery from a passive response or fixed-timeframe model to a proactive service model guided by forward-looking, precise time windows, providing crucial and accurate input for developing highly timely delivery plans.

[0016] A multi-objective mathematical model is constructed with the common optimization objectives of minimizing the total travel distance of delivery vehicles and maximizing the time window for meeting all demands. During the solution process, these two objectives are collaboratively optimized and balanced. In modeling, the time window requirement is treated as a flexible objective function term that needs to be optimized for its fulfillment, rather than an inviolable absolute constraint. This allows the scheduling solution process to automatically calculate and output a comprehensively optimal balance between reducing overall transportation costs and ensuring timely delivery to each workstation. The generated delivery plan simultaneously considers the economy of the routes and the fit of the time windows, achieving comprehensive optimization of logistics operating costs and production line continuity stability. Attached Figure Description

[0017] Figure 1 This is a timing diagram of the dynamic periodic material delivery scheduling system based on an automobile assembly line as described in this invention. Figure 2 A flowchart generated for the initial dynamic delivery plan. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0020] See Figure 1This invention relates to the field of automotive manufacturing logistics technology, specifically providing an implementation method for a dynamic periodic material distribution scheduling system based on an automotive assembly line. The overall system implementation scheme is as follows: The system collects material consumption data from each production station on the automotive assembly line in real time through a data acquisition module, and receives material inventory status information from a material storage management system. This information includes the real-time inventory quantity of each material, its location in the storage area, and its physical specifications. The demand forecasting module predicts the material demand sequence for each station within a future production cycle based on the collected material consumption data from the production stations. This sequence explicitly provides the required material code, predicted demand quantity, and demand time window. The constraint management module manages constraints such as the single loading capacity of material delivery vehicles, the physical specifications of materials, and the passage rules of logistics channels within the assembly line. The optimization modeling module constructs a multi-objective optimization model based on the aforementioned material demand sequence and various constraints, with the core objectives of minimizing the total travel distance of delivery vehicles and satisfying all demand time windows. Finally, the scheduling generation module is responsible for solving the multi-objective optimization model to generate an initial dynamic delivery plan, which specifies in detail the departure time, loading list, delivery route sequence, and expected arrival time at each workstation for each delivery vehicle.

[0021] In one embodiment of the present invention, after acquiring material consumption data containing the actual usage quantity and timestamps of different material codes at each workstation, the demand forecasting module performs time series analysis on the data to identify the periodic patterns and trend components of each workstation's consumption of each material. Based on the identified patterns and components, the module uses an exponential smoothing model to perform rolling forecasts of workstation material consumption, obtaining the basic forecasted demand for multiple consecutive future time points. Simultaneously, the module acquires the future production plan for the automotive assembly line, which includes the planned production models, the assembly sequence for each model, and the planned production takt time. Based on the bill of materials for the planned production models, the basic forecasted demand is decomposed and corrected by model, generating a refined forecasted demand driven by the model. Combining the assembly sequence and production takt time, the refined forecasted demand is mapped onto a specific timeline, forming a material demand sequence containing clearly defined demand time windows.

[0022] The optimization modeling module transforms each demand in the material demand sequence into a demand node in the optimization model, associated with its demand material code, predicted demand quantity, and demand time window. The module defines decision variables to indicate whether delivery vehicles depart from the storage point and whether they serve a pair of demand nodes in a specific order. The module establishes an objective function, a weighted summation, whose first part is the sum of the distances traveled by all delivery vehicles, and the second part is the penalty cost incurred for all demand nodes not served within the demand time window. The module establishes a first set of constraints to ensure that the total volume and weight of materials loaded by each delivery vehicle does not exceed its single-load capacity. The module establishes a second set of constraints to ensure that the predicted demand quantity of each demand node must be fully satisfied, and that this is accomplished by the same delivery vehicle in a single service. The module establishes a third set of constraints, which are time window constraints, ensuring that the arrival time of delivery vehicles at each demand node is no earlier than the start time and no later than the end time of its demand time window. The module establishes a fourth set of constraints, which are based on the traffic rules of the logistics channel and restrict the speed of vehicles in a specific area, prohibit turning, or prohibit simultaneous entry.

[0023] In practical implementation, the demand forecasting module acquires material consumption data for production workstations from the data acquisition module. This data includes workstation identifiers, material codes, actual usage quantities, and timestamps accurate to the second. The demand forecasting module groups the historical production workstation material consumption data by workstation and material code. For each group, it applies time series decomposition to the actual usage quantity and corresponding timestamp sequence, separating the daily recurring periodic components and the trend components reflecting long-term changes. Based on the identified periodic and trend components, the demand forecasting module uses a Holt-Winters exponential smoothing model to perform forward multi-step rolling forecasts for the consumption of each material at each workstation, obtaining the basic forecasted demand for multiple consecutive future time points. Simultaneously, the demand forecasting module receives future production plans for the automotive assembly line from the upper-level production execution system. These plans clearly list the sequence of planned production models, the assembly order for each model, and fixed production cycle times. The demand forecasting module decomposes the basic forecasted demand according to the material composition ratio of the planned production models based on the bill of materials for each model. It then weights and adjusts the decomposed demand based on the model's assembly sequence and planned production quantity, generating a refined forecasted demand driven by the specific model. Combining the specific assembly sequence and production cycle time, the demand forecasting module precisely maps the refined forecasted demand onto a timeline with production cycle time intervals, specifying the start and end times of the demand time window for each demand, ultimately forming a structured material requirement sequence.

[0024] In some embodiments, the optimization modeling module receives the material demand sequence output by the demand forecasting module, and transforms each independent demand instance in the material demand sequence into a demand node in the optimization model. The demand node attributes include a unique identifier, associated workstation location coordinates, a set of demand material codes, the predicted demand quantity for each material, and the demand time window. The optimization modeling module defines two sets of decision variables. The first set of decision variables is a binary variable, indicating whether a delivery vehicle departs from the warehouse to serve a certain demand node. The second set of decision variables is also a binary variable, indicating whether a delivery vehicle immediately proceeds to another specific demand node after completing service at one demand node. The optimization modeling module establishes the objective function of the multi-objective optimization model. The objective function uses a weighted summation method to transform the two objectives into a single objective, and its mathematical expression is: ; Where: character F represents the total value of the objective function, character K represents the set of delivery vehicles, character N represents the set of all nodes (including storage points and demand nodes), and character... Represents the distance traveled from node i to node j, character The second set of decision variables represents whether vehicle k travels from node i to node j. and These are preset weighting coefficients. The character N' represents the set of all requirement nodes. This represents the unit time penalty cost for demand node i not being served within the time window. Represents the actual time when the vehicle arrives at demand node i, character and These represent the start and end times of the demand time window for demand node i, respectively. The optimization modeling module establishes the first set of constraints, ensuring that for any delivery vehicle, the total volume of all materials loaded on it does not exceed the vehicle's maximum volumetric load capacity and maximum payload. The optimization modeling module establishes the second set of constraints, ensuring that the entire predicted demand for each demand node in the material demand sequence must be delivered by a uniquely designated delivery vehicle in a single stop service. The optimization modeling module establishes the third set of constraints, which are time window constraints. These constraints are calculated using the time flow in the vehicle routing model, forcing the delivery vehicle to arrive at any demand node within the start and end times of that node's demand time window. The optimization modeling module establishes the fourth set of constraints, set based on a digital map rule base for the logistics channels within the assembly line. This rule base defines speed limits for specific road segments, rules prohibiting turning at intersections, and one-way traffic rules for narrow passages. These rules are translated into direct restrictions on vehicle travel paths and travel times in the model.

[0025] See Figure 2 In one embodiment of the present invention, the scheduling generation module uses an improved genetic algorithm to solve a multi-objective optimization model. This improved genetic algorithm encodes the delivery vehicle paths using a mixed real-number and integer encoding method. The module initializes a population containing multiple randomly generated initial delivery plans, each satisfying the single-load capacity constraint of the material delivery vehicle. The module calculates the fitness value of each individual in the population, which is the reciprocal of the objective function value. The module performs a selection operation, selecting individuals from the current population to enter the mating pool based on their fitness values ​​using a roulette wheel strategy. The module performs a crossover operation, generating new offspring individuals from the individuals in the mating pool using a sequential crossover operator. The module performs a mutation operation, exchanging two points within the path or reversing path segments in the offspring individuals with a certain probability. The module performs feasibility repair on the population after selection, crossover, and mutation operations, ensuring that the newly generated individuals satisfy the second and third sets of constraints. The module iteratively executes the steps from calculating the fitness value to feasibility repair until a preset maximum number of iterations is reached or the objective function value converges. The module selects the individual with the smallest objective function value from the final population and decodes it to generate an initial dynamic delivery plan that includes vehicle departure time, loading list, and route sequence.

[0026] In its implementation, the scheduling generation module employs an improved genetic algorithm to solve the multi-objective optimization model. This improved algorithm uses a mixed real-number and integer encoding method to encode the delivery vehicle paths. The real-number part represents the sequential order of vehicle visits to nodes, while the integer part identifies the specific delivery vehicle number serving each node. The scheduling generation module initializes the population, which contains multiple randomly generated initial delivery plans. Each initial delivery plan is constructed by randomly arranging demand nodes and allocating vehicles during the generation process, and each initial delivery plan undergoes loading verification to ensure it meets the single-load capacity constraint of the material delivery vehicles. The scheduling generation module calculates the fitness value of each individual in the population. The fitness value is the reciprocal of the output value of the multi-objective optimization model's objective function. The output value of the multi-objective optimization model's objective function is calculated by decoding the individual into a specific delivery path and then inputting it into the objective function of the multi-objective optimization model. The scheduling generation module performs a selection operation, using a roulette wheel selection strategy based on the fitness value to select individuals from the current population to enter the mating pool. In the roulette wheel selection strategy, the probability of each individual being selected is proportional to its fitness value. The scheduling generation module performs a crossover operation, using a sequential crossover operator to generate new offspring individuals from individuals in the mating pool. The sequential crossover operator randomly selects a path subsequence from the parent individuals and directly stores it in the corresponding position of the offspring individual. Missing nodes are selected sequentially from another parent individual. The scheduling generation module then performs a mutation operation, swapping two points within a path or reversing a path segment for offspring individuals with a preset mutation probability. The two-point swap operation randomly selects two nodes in the path and swaps their positions. The path segment reversal operation randomly selects a continuous segment in the path and completely reverses the order of nodes within that segment. The scheduling generation module then performs feasibility repair on the population after selection, crossover, and mutation operations. This feasibility repair process checks whether each newly generated individual satisfies the second and third sets of constraints. Individuals violating the second set of constraints are corrected by splitting requirements or reallocating vehicles. Individuals violating the third set of constraints are corrected by locally adjusting the node access order or inserting waiting times. The scheduling generation module iteratively executes steps from calculating fitness values ​​to feasibility repair until the algorithm reaches the preset maximum number of iterations or the change in the output value of the multi-objective optimization model's objective function of the best individual in the population across multiple generations is less than the convergence threshold. The scheduling generation module then selects the individual with the smallest output value of the multi-objective optimization model from the final population, decodes the individual's code to extract the node sequence list served by each delivery vehicle, and combines this with the material demand sequence and warehouse information to generate an initial dynamic delivery plan containing vehicle departure time, loaded material list, delivery route sequence, and estimated arrival time at each workstation.

[0027] In some embodiments, the improved genetic algorithm employs a selection operation based on the following formula to calculate the probability of each individual being selected into the mating pool: ; Where: characters The first in the representative population The probability of an individual being selected by the roulette strategy, character Representing the Fitness value of an individual, character This represents the total number of individuals in the current population. The roulette wheel strategy randomly selects individuals based on a calculated probability distribution; individuals with higher fitness values ​​have a higher probability of being selected and entering the mating pool to participate in subsequent crossover operations. In some embodiments, the improved genetic algorithm uses a sequential crossover operator involving two parent individuals. The sequential crossover operator first randomly defines the same crossover interval on the encoded strings of the two parent individuals. The encoded fragment of parent individual one within this interval is directly copied to the same position in the offspring individual. The remaining empty slots in the offspring individual are filled with the encoded elements according to the order in which they appeared in parent individual two, but elements already copied from parent individual one must be excluded.

[0028] Optionally, the improved genetic algorithm uses a random generation method in the initialization phase to generate schemes that not only satisfy the single-load capacity constraint of the material delivery vehicle, but also generates some high-quality initial individuals through fast heuristic rules such as nearest neighbor insertion to accelerate population convergence. Optionally, the improved genetic algorithm sets a dynamically adjusted mutation probability in the mutation operation, which decreases linearly with the number of iterations, thereby maintaining population diversity in the early stages of the search and enhancing local mining capabilities in the later stages.

[0029] It is understandable that the feasibility repair step in the improved genetic algorithm is a crucial process to ensure that the algorithm always searches within the feasible solution space, and this step is forcibly implemented after each generation of new population. It is also understandable that the mixed real-integer and integer encoding method used in the improved genetic algorithm can naturally express the joint information of node access order and vehicle allocation in the vehicle routing problem. This encoding method allows crossover and mutation operations to be performed directly on the encoded string without complex mapping transformations.

[0030] In one embodiment of the present invention, the system includes a dynamic plan adjustment module. This module issues an initial dynamic delivery plan to delivery vehicles for execution and continuously collects the actual location information of the vehicles and the actual material consumption rate of each workstation during execution. The module compares the actual material consumption rate with the predicted rate, calculates the material demand prediction deviation rate, and triggers a dynamic adjustment process for the delivery plan when the material demand prediction deviation rate exceeds a preset threshold. In the dynamic adjustment process, the initial dynamic delivery plan is rescheduled online by combining the latest material inventory status information and the actual location information of the vehicles. The module generates and issues an adjusted dynamic delivery plan, which is used to update the remaining tasks of the delivery vehicles.

[0031] In calculating the material demand forecast deviation rate, the module sets a fixed monitoring period. At the end of each monitoring period, it calculates the total actual consumption of all material codes at each workstation within the just-ended period. The module extracts the predicted demand for the corresponding workstation, material code, and period of the same length from the material demand sequence. For each material at each workstation, the module calculates the absolute difference between the actual consumption and the predicted demand. The module compares this absolute difference with the predicted demand to obtain the material demand forecast deviation rate for that material at that workstation within that monitoring period. The module performs a weighted average calculation of the material demand forecast deviation rates for all materials at all workstations to obtain a global material demand forecast deviation level. When the material demand forecast deviation rate of any material exceeds its independent preset threshold, or when the global material demand forecast deviation level exceeds the global preset threshold, the trigger condition is determined to be met, and the dynamic adjustment process of the delivery plan is initiated.

[0032] In implementation, the dynamic planning adjustment module receives the initial dynamic delivery plan from the scheduling generation module and distributes it to the onboard terminals of the delivery vehicles for execution. During execution, the module continuously collects the actual location information of the delivery vehicles through onboard GPS and RFID devices, and continuously collects the actual material consumption rate of each workstation through material consumption sensors at the assembly line workstations. The module compares the actual material consumption rate with the predicted rate provided by the demand forecasting module, calculates the material demand forecast deviation rate, and triggers the dynamic adjustment process for the delivery plan when the deviation rate exceeds a preset threshold. In the dynamic adjustment process, the module obtains the latest material inventory status information from the data acquisition module and, combined with the actual location information of the delivery vehicles, performs online rescheduling of the initial dynamic delivery plan. The module then generates and distributes the adjusted dynamic delivery plan, which is used to update the remaining task instructions displayed on the onboard terminals of the delivery vehicles, ensuring the continuity of material supply.

[0033] The dynamic planning adjustment module calculates the material demand forecast deviation rate by setting a fixed monitoring period, the length of which is related to the production cycle and delivery cycle. At the end of each monitoring period, the dynamic planning adjustment module calculates the actual total consumption of all material codes at each workstation within the just-ended period. The actual consumption data is stored using the workstation and material code as a joint primary key. The dynamic planning adjustment module extracts the corresponding workstation, corresponding material code, and predicted demand within the same monitoring period from the material demand sequence maintained by the demand forecasting module. For each material at each workstation, the dynamic planning adjustment module calculates the absolute difference between the actual total consumption and the predicted demand. The absolute difference is the absolute value of the actual total consumption minus the predicted demand. The dynamic planning adjustment module compares the calculated absolute difference with the predicted demand to obtain the material demand forecast deviation rate for that material at that workstation within that monitoring period. The dynamic planning adjustment module performs a weighted average calculation of the material demand forecast deviation rates for all materials at all workstations to obtain a global material demand forecast deviation level. When the material demand forecast deviation rate of any material exceeds its independent preset threshold, or the overall material demand forecast deviation level exceeds the overall preset threshold, the dynamic adjustment module determines that the trigger condition has been met and starts the dynamic adjustment process of the delivery plan.

[0034] In some embodiments, the formula for calculating the material demand forecast deviation rate is as follows: ; Where: characters Represents the monitoring period Inside, workstation For materials Material demand forecast deviation rate. Represents the monitoring period Inside, workstation For materials The actual total consumption. (Character) This represents data extracted from the material demand sequence, corresponding to the monitoring period. workstation and materials The predicted demand. (Character) Represents a specific monitoring period. Global material demand forecast deviation level. The calculation method is for all To correspond to the predicted demand We calculate the weighted sum and average of the weights.

[0035] In some embodiments, the independent preset threshold and the global preset threshold are set based on the importance of the material and its historical consumption volatility. A relatively lenient independent preset threshold is set for critical materials or materials with high consumption volatility, while a relatively strict independent preset threshold is set for regular materials. The global preset threshold serves as a warning line for overall system stability, and its value is lower than the independent preset threshold for most materials.

[0036] Optionally, the monitoring cycle length can be dynamically adjusted based on the actual operating status of the production line. When production pace accelerates or order fluctuations are significant, the monitoring cycle can be automatically shortened to improve response sensitivity; when production is stable, the monitoring cycle can be automatically extended to reduce the system's computational load. Optionally, when calculating the global material demand forecast deviation level, the dynamic planning adjustment module can assign different weights to different workstations, giving higher weights to bottleneck workstations or critical assembly workstations to highlight their importance.

[0037] It is understandable that the dynamic adjustment process of the delivery plan triggered by the dynamic adjustment module is an online rescheduling process, which works in conjunction with the offline global scheduling of the scheduling generation module. Similarly, it is understandable that the calculation and monitoring of the material demand forecast deviation rate is a continuous, closed-loop process that provides the entire dynamic periodic material delivery scheduling system with the ability to perceive and respond to changes in the production environment.

[0038] In one embodiment of the present invention, when the dynamic adjustment module reschedules the initial dynamic delivery plan online, it first freezes all completed service tasks of delivery vehicles, i.e., workstation nodes that have arrived and unloaded. The module obtains the real-time location information, current loading status, and remaining path sequence of all delivery vehicles. The module obtains the latest material inventory status information and the updated material demand sequence at the time the rescheduling is triggered; this updated material demand sequence has been corrected according to the latest actual material consumption rate. The module uses the current time as the new scheduling start time, the real-time location of all vehicles as the new starting point, and the current loading status of the vehicles as the new initial loading amount. Based on the latest material inventory status information, the updated material demand sequence, and constraints, the module constructs a local rescheduling optimization model, which optimizes only the demand nodes that have not yet been served. The module solves this local rescheduling optimization model, generating a new remaining path sequence and timetable for each delivery vehicle starting from its current location. The module integrates the newly generated remaining path sequence and timetable with the frozen completed tasks to form a new draft delivery plan covering the entire scheduling cycle.

[0039] When constructing the local rescheduling optimization model, the module determines the optimization scope, which includes all unserved demand nodes, the real-time locations of all delivery vehicles, and storage points. The module defines new decision variables to indicate whether a delivery vehicle departs from its current location and whether it serves a pair of unserved demand nodes in the new order. The module establishes a new objective function that minimizes the additional travel distance and time window penalty costs, while adding a penalty term for the magnitude of plan changes to maintain the relative stability of the scheduling scheme. The module establishes loading constraints to ensure that the cumulative load of each delivery vehicle serving the new route never exceeds its single-load capacity, taking into account its current loading status. The module establishes demand satisfaction constraints to ensure that the demand of each unserved demand node is satisfied in the rescheduled plan. The module establishes new time window constraints, recalculating the arrival time of vehicles at each unserved demand node with the current time as the zero point, ensuring that this time falls within its demand time window. The module uses the vehicle paths obtained from solving the local rescheduling optimization model as the new sequence of remaining paths.

[0040] In practical implementation, the dynamic planning adjustment module performs online rescheduling of the initial dynamic delivery plan. First, the module freezes all completed service tasks for all delivery vehicles. Completed service tasks refer to workstation nodes where vehicles have arrived, unloaded, and been confirmed and signed for at the workstations. These nodes are marked as completed and removed from subsequent optimization decision variables. The module obtains the real-time location coordinates, current loading status, and remaining path sequences not yet executed according to the previous plan from the wireless communication network for all delivery vehicles. The module obtains the latest material inventory status information at the time of triggering rescheduling from the data acquisition module and the updated material demand sequence from the demand forecasting module. The updated material demand sequence has been rolled over and remapped according to the latest actual material consumption rate. The module uses the current system time as the new scheduling start time, the real-time location of all delivery vehicles as the new path planning starting point, and the current loading status of the delivery vehicles as the new initial load. The dynamic planning adjustment module constructs a partial rescheduling optimization model based on the latest material inventory status information, the updated material demand sequence, and the constraints defined by the constraint management module. This model optimizes only all unserved demand nodes; frozen nodes are not included. The module solves the partial rescheduling optimization model, generating a new remaining path sequence and an estimated timetable for each unserved node for each delivery vehicle, starting from its current location. The module then integrates the newly generated remaining path sequence and timetable with the frozen completed task history to form a new draft delivery plan covering the period from the initial scheduling time to the end of the complete scheduling cycle, which is then issued for execution.

[0041] In some embodiments, constructing a local rescheduling optimization model first requires determining the optimization scope, which includes the set of all unserved demand nodes, the set of virtual starting points composed of the real-time locations of all delivery vehicles, and the storage points. The dynamic planning adjustment module defines new decision variables to indicate whether a delivery vehicle departs from its current location and whether it serves a pair of unserved demand nodes in the new order. The dynamic planning adjustment module establishes a new objective function for the local rescheduling optimization model. This new objective function, while minimizing the additional travel distance and time window penalty costs, adds a penalty term for the magnitude of plan changes. The mathematical expression of the new objective function is as follows: ; Where: characters Represents the objective value of the local rescheduling optimization model, character This represents the set of delivery vehicles participating in the rescheduling. (Character) Representative vehicle The virtual starting point is formed by the real-time position of the character. Represents the set of all unserved demand nodes. (Character) Representative vehicle From position Drive to the node The distance, if If it's a virtual starting point, then it's the actual distance. (Character) It is a new decision variable, representing the vehicle. Whether from Drive to .character , , These are weighting coefficients. (Character) Representative node Time window violation penalty coefficient, character Represents the vehicle's arrival at the node The deviation between the time and its required time window. (Character) Representatives targeting vehicles The penalty term for the difference between the old and new planned routes is used to measure the magnitude of the plan change. The dynamic plan adjustment module establishes loading constraints to ensure that the sum of the current load and the cumulative newly added load in the route for each delivery vehicle when serving a new route never exceeds its single loading capacity. The dynamic plan adjustment module also establishes demand satisfaction constraints to ensure that the demand of each unserved demand node is satisfied in the rescheduled plan and served only once by one vehicle. The dynamic plan adjustment module establishes new time window constraints, which use the current rescheduling trigger time as the time zero point, recalculate the time for a vehicle to arrive at each unserved demand node from its current location, and force that this time fall within its demand time window. Refer to Table 1, which shows an example of a system state snapshot at a rescheduling trigger time. This snapshot provides input data for constructing the local rescheduling optimization model.

[0042] Table 1: System State Snapshot Table at Rescheduling Time

[0043] Optional, penalty for plan changes The specific calculation can be based on the vehicle The Jaccard difference or sequence edit distance between the new path sequence and the old remaining path sequence on the set of service nodes. Optionally, the local rescheduling optimization model can be solved using a fast heuristic algorithm based on tabu search to meet the high real-time requirements of online rescheduling.

[0044] It is understandable that the local rescheduling optimization model is a scaled-down vehicle routing problem model that considers the new starting point. It is also understandable that the mechanism of freezing completed tasks ensures that the rescheduling process does not change the past, allowing optimization to focus on future delivery activities that have not yet occurred, thus guaranteeing the consistency and executability of the scheduling logic.

[0045] In one embodiment of the invention, the system includes an evaluation and parameter adjustment module. This module summarizes the day's delivery task execution data at the end of each complete production day or shift, including total mileage, total delivery trips, number of time window violations, and number of rescheduling triggers. The module extracts material availability rate data from the warehouse management system for the same period, reflecting the frequency and duration of production line stoppages due to material shortages. Based on the delivery task execution data and material availability rate data, the module calculates the comprehensive performance index of the delivery system. The module compares this comprehensive performance index with historical data or a preset performance benchmark. Based on the results of the comparison analysis, the module adaptively adjusts the weight coefficients of the objective function in the multi-objective optimization model, the preset threshold for the material demand forecast deviation rate, or the trigger sensitivity parameters of the dynamic adjustment process for the delivery plan. The module applies the adjusted parameters to the scheduling process of the next production cycle.

[0046] When calculating the overall performance index of the delivery system, the module standardizes the total mileage by dividing it by the total weight of delivered materials to obtain the delivery mileage per unit weight of material. The module also standardizes the total number of delivery trips by dividing it by the total number of material types delivered to obtain the average delivery frequency per material type. The module calculates the ratio of the number of time window violations to the total number of service nodes to obtain the time window violation rate. The module then performs a weighted sum of the delivery mileage per unit weight of material, the average delivery frequency per material type, and the time window violation rate to obtain the logistics efficiency sub-index. The module converts the material availability rate data into the proportion of workstation downtime to total production time, serving as the material support sub-index. Finally, the module performs a weighted sum of the logistics efficiency sub-index and the material support sub-index to obtain the overall performance index of the delivery system; a lower index value indicates better overall performance.

[0047] The system also includes an emergency delivery scheduling module. This module receives emergency order insertion instructions, which include the vehicle model, production quantity, and required online time. Based on the bill of materials for the emergency production vehicle model, the module immediately calculates the new emergency material requirements for each workstation caused by the emergency order insertion. The module checks the current material inventory status to confirm whether the available inventory of emergency materials can meet the new emergency material requirements. If the inventory is sufficient, the module immediately locks the corresponding quantity of materials in the available inventory and creates the highest priority emergency demand node for these new emergency material requirements. The module inserts the emergency demand node into the current dynamic delivery plan, with the demand time window of the emergency demand node calculated backwards from the required online time. The module immediately triggers a local dynamic adjustment process for the delivery plan, using the demand time window of the emergency demand node as a hard constraint, and replans the routes of some or all delivery vehicles to ensure that emergency materials are delivered on time with priority. In subsequent regular scheduling cycles, the module removes the emergency demand nodes that have already been delivered from the demand list, and the system returns to normal scheduling mode.

[0048] In practice, the evaluation and parameter adjustment module is activated after a complete production day or shift. This module aggregates the day's delivery task execution data, including the total mileage of all delivery vehicles, the total number of delivery trips, the number of time window violations for all demand nodes, and the number of rescheduling triggered by the dynamic planning adjustment module. The module also extracts material availability rate data from the material warehousing management system database for the same period. This data records the specific start and end times and cumulative duration of production line stoppages due to material shortages at each workstation. Based on the delivery task execution data and material availability rate data, the module calculates the overall performance index of the delivery system. It then compares the calculated overall performance index with historical data from the same period in the previous workday or shift, or with a preset performance baseline. Based on the results of this comparison, the module adaptively adjusts the weight coefficients of each part of the objective function in the multi-objective optimization model, the independent preset threshold and global preset threshold for the material demand forecast deviation rate in the dynamic planning adjustment module, or the trigger sensitivity parameters of the dynamic adjustment process for the delivery plan. The evaluation and parameter adjustment module applies the adjusted parameter configuration file to the scheduling generation module and the dynamic planning adjustment module during the next production cycle.

[0049] When calculating the comprehensive performance index of the delivery system, the evaluation and parameter adjustment module standardizes the total mileage by dividing it by the total weight of all delivered materials to obtain the delivery mileage per unit weight of material. It also standardizes the total number of delivery trips by dividing it by the total number of material types to obtain the average delivery frequency per material type. The module calculates the ratio of time window violations to the total number of service nodes to obtain the time window violation rate. Finally, it weights and sums the delivery mileage per unit weight of material, the average delivery frequency per material type, and the time window violation rate to obtain a logistics efficiency sub-index. The module converts the material availability rate data into the proportion of workstation downtime to total production time, serving as a material support sub-index. Finally, it weights and sums the logistics efficiency sub-index and the material support sub-index again to obtain the final comprehensive performance index of the delivery system. A lower comprehensive performance index value indicates better comprehensive performance. The formula for calculating the comprehensive performance index is: ; Where: characters Represents a comprehensive performance indicator. (Character) Represents total mileage, character This represents the total weight of the delivered materials, therefore Delivery distance per unit weight of material. Represents the total number of delivery trips, characters This represents the total number of types of materials delivered, therefore Average delivery frequency per unit material type. Represents the number of violations of the time window, character This represents the total number of service nodes, therefore For the time window violation rate. (Character) , , These are the weighting coefficients for the logistics efficiency sub-indicator. (Character) Represents the total production line downtime due to material shortages, character Represents the total production time, therefore This is a sub-indicator for material support. (Character) The weighting coefficients for the material support sub-indicators.

[0050] In some embodiments, the emergency delivery scheduling module receives an emergency order insertion production instruction from the production management system. This instruction is transmitted in the form of a structured data message, containing the emergency production vehicle model code, production quantity, and latest online time. The emergency delivery scheduling module retrieves the vehicle model's bill of materials based on the emergency production vehicle model code and immediately calculates the new emergency material requirements for each relevant workstation caused by the emergency order insertion. The module queries the current material inventory status information provided by the data acquisition module to confirm whether the available inventory quantity of emergency materials meets the new emergency material requirements. If the inventory is sufficient, the module immediately sends an instruction to the material storage management system to lock the corresponding quantity of materials in the available inventory and creates emergency demand nodes with the highest priority for these new emergency material requirements. The module inserts the emergency demand nodes into the demand list of the current dynamic delivery plan maintained by the dynamic planning adjustment module. The demand time window for the emergency demand node is calculated by subtracting the material delivery and preparation time from the latest online time. The module immediately sends a signal to the dynamic planning adjustment module to trigger a partial dynamic adjustment process for the delivery plan, using the demand time window of the emergency demand node as an inviolable hard constraint, requiring the replanning of some or all delivery vehicle routes. In subsequent regular scheduling cycles, the emergency delivery scheduling module monitors the completion status of emergency demand nodes, removes emergency demand nodes that have been delivered from the demand list, and restores the system to normal scheduling mode.

[0051] Optionally, the parameter adjustment strategy of the evaluation and parameter adjustment module can adopt a strategy based on reinforcement learning, adjusting specific parameters upwards or downwards with a fixed step size or an adaptive step size according to the direction and magnitude of the change of the comprehensive performance index relative to historical or benchmark values. Optionally, when creating an emergency demand node, the emergency delivery scheduling module will assign it a unique identifier prefix that distinguishes it from regular demand nodes, and highlight it in the interfaces and logs of all relevant scheduling modules.

[0052] It is understandable that the evaluation and parameter adjustment module achieves self-optimization of system parameters through periodic evaluation and feedback adjustment, enabling the scheduling system to adapt to long-term changes in production patterns. It is also understandable that the emergency delivery scheduling module, through a standardized interruption handling process, enables the conventional dynamic periodic scheduling system to cope with sudden changes in production plans.

[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A dynamic periodic material distribution scheduling system based on an automotive assembly line, characterized in that: The system includes: The data acquisition module collects material consumption data from production workstations on the automotive assembly line in real time and receives material inventory status information from the material storage management system. The material inventory status information includes the real-time inventory quantity of each material, its location in the storage area, and the physical specifications of the material. The demand forecasting module predicts the material demand sequence for each workstation in the future production cycle based on the material consumption data of the production workstation. The material demand sequence includes the demand material code, the predicted demand quantity, and the demand time window. The constraint management module uses the single loading capacity of the material delivery vehicle, the physical specifications in the material inventory status information, and the passage rules of the logistics channel within the assembly line as constraints. The optimization modeling module constructs a multi-objective optimization model based on the material demand sequence and the constraints, with the goal of minimizing the total travel distance of delivery vehicles and satisfying all demand time windows. The scheduling generation module solves the multi-objective optimization model to generate an initial dynamic delivery plan, which includes the departure time of each delivery vehicle, the list of loaded materials, the delivery route sequence, and the expected arrival time at each workstation.

2. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 1, characterized in that, The step of predicting the material demand sequence for each workstation in the future production cycle based on the material consumption data of the production workstations includes: The material consumption data includes the actual usage quantity and timestamp of each workstation for different material codes; Time series analysis was performed on the actual usage quantity and corresponding timestamp in the material consumption data to identify the periodic patterns and trend components of each workstation's consumption of each material. Based on the identified periodic patterns and trend components, an exponential smoothing model is used to make rolling predictions of material consumption at workstations, thereby obtaining the basic predicted demand at multiple consecutive time points in the future. Obtain the future production plan for the automotive assembly line, which includes the planned production models, the assembly sequence of each model, and the planned production cycle time. Based on the bill of materials for the planned production models, the basic forecast demand is decomposed and revised by model to generate a refined forecast demand driven by the model. By combining the assembly sequence with the production cycle, the refined forecast demand is mapped onto a specific timeline to form the material demand sequence containing a clear demand time window.

3. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 1, characterized in that, The construction of a multi-objective optimization model based on the material demand sequence and the constraints includes: Each demand in the material demand sequence is transformed into a demand node in the optimization model, and the demand node is associated with its demand material code, predicted demand quantity and demand time window; Define decision variables to represent whether a delivery vehicle departs from a storage point and whether it serves a pair of demand nodes in a specific order; Establish an objective function, which is in the form of a weighted summation. Its first part is the sum of the travel distances of all delivery vehicles, and its second part is the penalty cost incurred by all demand nodes that do not receive service within the demand time window. Establish a first set of constraints that ensure that the total volume and weight of the materials loaded on each delivery vehicle do not exceed its single loading capacity. Establish a second set of constraints that ensure that the predicted demand at each demand node must be fully met and completed by the same delivery vehicle in a single service. A third set of constraints is established, which is a time window constraint, to ensure that the delivery vehicle arrives at each demand node no earlier than the start time of its demand time window and no later than the end time of its demand time window. A fourth set of constraints is established, which is based on the traffic rules of the logistics channel and restricts the driving speed of vehicles in a specific area, prohibits turning, or prohibits simultaneous entry.

4. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 1, characterized in that, Solving the multi-objective optimization model to generate an initial dynamic delivery plan includes: An improved genetic algorithm is used to solve the multi-objective optimization model. The improved genetic algorithm uses a mixed encoding method of real numbers and integers to encode the path of the delivery vehicle. Initialize the population, which contains multiple randomly generated initial delivery schemes, each of which satisfies the single-load capacity constraint of the material delivery vehicle; Calculate the fitness value of each individual in the population, where the fitness value is the reciprocal of the objective function value; Perform a selection operation, and select individuals from the current population to enter the mating pool based on the fitness value using a roulette wheel strategy; Perform a crossover operation, using the sequential crossover operator to generate new offspring individuals from individuals in the mating pool; Perform mutation operations to swap two points within a path or reverse a path segment in offspring individuals with a certain probability; For the population after selection, crossover, and mutation operations, feasibility repair is performed to ensure that newly generated individuals satisfy the second set of constraints and the third set of constraints; The process iteratively executes steps from calculating the fitness value to feasibility repair until the preset maximum number of iterations is reached or the objective function value converges. The individual with the smallest objective function value is selected from the final population, and the initial dynamic delivery plan containing vehicle departure time, loading list, and route sequence is generated by decoding.

5. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 1, characterized in that, Also includes: The dynamic adjustment module distributes the initial dynamic delivery plan to the delivery vehicles for execution, and continuously collects the actual location information of the vehicles and the actual material consumption rate of each workstation during the execution process. The actual material consumption rate is compared with the predicted rate to calculate the material demand prediction deviation rate. When the material demand prediction deviation rate exceeds a preset threshold, the dynamic adjustment process of the delivery plan is triggered. In the dynamic adjustment process of the delivery plan, the initial dynamic delivery plan is rescheduled online by combining the latest material inventory status information and the actual location information of the vehicles. An adjusted dynamic delivery plan is generated and issued, which is used to update the remaining tasks of delivery vehicles to ensure the continuity of material supply. The step of comparing the actual material consumption rate with the predicted rate and calculating the material demand forecast deviation rate includes: Set a fixed monitoring cycle, and at the end of each monitoring cycle, calculate the total actual consumption of all material codes at each workstation during the just-ended cycle. Extract the corresponding workstation, material code, and predicted demand for the same time period from the material demand sequence; For each type of material at each workstation, calculate the absolute difference between the actual total consumption and the predicted demand. The absolute difference is compared with the predicted demand to obtain the material demand prediction deviation rate of the material in the monitoring period of the work site. A weighted average of the material demand forecast deviation rates for all materials at all workstations is used to obtain a global material demand forecast deviation level. When the material demand forecast deviation rate of any material exceeds its independent preset threshold, or the global material demand forecast deviation level exceeds the global preset threshold, the trigger condition is determined to be met, and the dynamic adjustment process of the delivery plan is initiated.

6. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 5, characterized in that, The online rescheduling of the initial dynamic delivery plan includes: Freeze all service tasks completed by delivery vehicles, i.e., workstation nodes that have arrived and unloaded. Obtain the real-time location information, current cargo status, and remaining route sequence of all delivery vehicles; Obtain the latest material inventory status information and the updated material demand sequence at the time of triggering rescheduling; the updated material demand sequence has been corrected according to the latest actual material consumption rate. The current time is taken as the new scheduling start time, the real-time location of all vehicles is taken as the new starting point, and the current cargo status of the vehicles is taken as the new initial load. Based on the latest material inventory status information, the updated material demand sequence, and the constraints, a local rescheduling optimization model is constructed. The local rescheduling optimization model only optimizes demand nodes that have not yet been served. Solve the local rescheduling optimization model to generate a new remaining path sequence and timetable for each delivery vehicle starting from its current location; The newly generated remaining path sequence and timetable are integrated with the frozen completed tasks to form a new draft delivery plan covering the entire scheduling cycle.

7. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 6, characterized in that, The construction of a local rescheduling optimization model includes: Determine the optimization scope, which includes all unserved demand nodes, the real-time location of all delivery vehicles, and storage points; Define new decision variables to represent whether the delivery vehicle departs from its current location and whether it serves a pair of unserved demand nodes in a new order; A new objective function is established, which, while minimizing the penalty cost of additional travel distance and time window, adds a penalty term for the magnitude of plan changes in order to maintain the relative stability of the scheduling scheme. Establish loading constraints to ensure that the cumulative load of each delivery vehicle when serving a new route never exceeds its single loading capacity, taking into account its current cargo status. Establish demand satisfaction constraints to ensure that the demand of each unserved node is met in the rescheduled plan; Establish a new time window constraint, taking the current moment as the zero point, recalculate the time for vehicles to arrive at each unserved demand node, and ensure that the time falls within its demand time window. The vehicle paths obtained by solving the local rescheduling optimization model are used as the new remaining path sequence.

8. The dynamic periodic material distribution scheduling system based on an automobile assembly line according to claim 1, characterized in that, Also includes: The evaluation and parameter adjustment module summarizes the delivery task execution data for the entire day after the end of each complete production day or shift. The delivery task execution data includes total mileage, total number of delivery trips, number of time window violations, and number of rescheduling triggers. Extract material availability rate data for the same period from the warehouse management system. The material availability rate data reflects the frequency and duration of production line stoppages caused by material shortages at workstations. Based on the delivery task execution data and the material availability rate data, calculate the comprehensive efficiency index of the delivery system; The comprehensive performance indicators are compared and analyzed with historical data or preset performance benchmarks. Based on the results of the comparative analysis, the weight coefficients of the objective function in the multi-objective optimization model, the preset threshold of the material demand prediction deviation rate, or the trigger sensitivity parameters of the dynamic adjustment process of the delivery plan are adaptively adjusted. The adjusted parameters will be applied to the scheduling process of the next production cycle.

9. The dynamic periodic material distribution scheduling system based on an automotive assembly line according to claim 8, characterized in that, The calculation of the comprehensive performance index of the delivery system based on the delivery task execution data and the material availability rate data includes: The total mileage is standardized and divided by the total weight of delivered materials to obtain the delivery mileage per unit weight of materials. The total number of delivery trips is standardized and divided by the total number of material types to obtain the average delivery frequency per material type. The time window violation rate is obtained by calculating the ratio of the number of violations of the time window to the total number of service nodes. The logistics efficiency sub-indicator is obtained by weighted summing of the delivery mileage per unit weight of material, the average delivery frequency per unit type of material, and the violation rate of the time window. The material availability rate data is converted into the proportion of workstation downtime to total production time, which is used as a sub-indicator for material availability. The logistics efficiency sub-indicator and the material support sub-indicator are weighted and summed again to obtain the comprehensive performance index of the distribution system. The lower the comprehensive performance index value, the better the comprehensive performance.

10. The dynamic periodic material distribution scheduling system based on an automobile assembly line according to claim 1, characterized in that, Also includes: The emergency delivery dispatch module receives emergency order insertion production instructions, which include the vehicle model to be produced, the production quantity, and the required online time. Based on the bill of materials for the urgently produced vehicle model, immediately calculate the additional urgent material requirements for each workstation caused by the emergency order. Check the current inventory status of the materials to confirm whether the available inventory of the emergency materials meets the demand for the new emergency materials; If the inventory is sufficient, immediately lock the corresponding quantity of materials in the available inventory and create the highest priority emergency demand node for these new urgent material needs; The urgent demand node is inserted into the current dynamic delivery plan, and the demand time window of the urgent demand node is calculated backward based on the required online time. Immediately trigger a local dynamic adjustment process for the aforementioned delivery plan, using the demand time window of the urgent demand node as a hard constraint, and replan the routes of some or all delivery vehicles to ensure that urgent materials are delivered to the destination on time with priority. In subsequent regular scheduling cycles, the emergency demand nodes that have completed delivery will be removed from the demand list, and the system will return to normal scheduling mode.

Images