Factory material distribution optimization method

Abstract

The invention relates to the technical field of factory material distribution, in particular to a factory material distribution optimization method, which comprises the following steps: S1, path state prediction: acquiring path traffic data, predicting the future congestion level of each road section, and generating path traffic information; s2, equipment idle degree evaluation: acquiring equipment state and queuing information, constructing an equipment time window map, and calculating and predicting the idle degree; s3, path scheduling optimization: constructing a task-path-equipment scheduling model, and performing weighted summation according to driving and waiting time to optimize the path and time; and S4, priority fusion evaluation: constructing a task priority scoring model, fusing scheduling optimization, and improving the high-priority task completion efficiency. According to the invention, the scheduling targets of blockage-avoiding driving, high-efficiency loading and unloading and priority completion of high-priority tasks in factory material distribution are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of factory material distribution, and in particular to a factory material distribution optimization method. Background Art

[0002] With the development of intelligent manufacturing and industrial Internet of Things technologies, the material distribution process within the factory is evolving from traditional manual scheduling to automation and intelligence. In the complex factory environment with multiple concurrent tasks, interlaced paths, and shared loading and unloading resources, how to achieve efficient material distribution while ensuring timeliness and resource coordination has become a key issue in production organization. In recent years, path optimization and equipment scheduling have gradually introduced predictive algorithms and data modeling technologies, but there are still challenges such as insufficient integration and untimely response.

[0003] Existing factory logistics scheduling schemes usually take the shortest path or equipment idleness as the main optimization goals. They lack the ability to dynamically predict path congestion trends, which can easily lead to vehicles encountering sudden congestion and delays during mission execution. At the same time, the allocation strategy of loading and unloading equipment mostly adopts static polling or passive queuing mechanism, failing to make forward-looking adjustments based on future load changes of equipment, resulting in low resource utilization efficiency during mission peak periods. More importantly, existing schemes generally do not distinguish the importance of tasks. The scheduling system cannot identify which tasks are more urgent or more critical, making it difficult to ensure the priority completion of critical tasks when resources are limited, affecting production stability and management responsiveness. Summary of the Invention

[0004] The present invention provides a method for optimizing material distribution in a factory area, which can take into account both scheduling efficiency and task importance in global optimization, thereby outputting a more reasonable and business-value-oriented material distribution plan.

[0005] A method for optimizing material distribution in a factory area comprises the following steps:

[0006] S1, route status prediction: Obtain traffic data for each section of the delivery route, and predict the congestion level of each section in the future time period based on the time series prediction model to generate predicted route congestion information;

[0007] S2, Equipment Idleness Assessment: Collect the status data of loading and unloading equipment in each delivery target area, including the current task status, estimated completion time, and queue information, build an equipment usage time window map, and calculate the predicted idleness of each loading and unloading equipment in the specified future time period;

[0008] S3, distribution task path scheduling optimization: based on the predicted path congestion information and the predicted vacancy, a three-dimensional scheduling model of task-path-device is constructed, the vehicle travel time and waiting time are weighted and summed by using the objective function, under the condition of meeting the device available window and path smoothness constraint, the distribution path and arrival time are dynamically optimized, and the optimal path scheduling scheme is output;

[0009] S4, task priority fusion evaluation: based on the time requirement, material type and regional level of the distribution task, a priority scoring model is constructed, and the scoring result is introduced into the scheduling optimization, so that the optimal path scheduling scheme tends to timely complete the high priority task.

[0010] Optionally, the path state prediction in S1 comprises:

[0011] S11, road section traffic data integration: historical operation data, current traffic state data and vehicle scheduling plan of each distribution path section are collected, combined with the factory road network structure, the traffic data is time-aligned and format-unified, and a standardized traffic state data set is constructed;

[0012] S12, congestion level prediction: based on the constructed traffic state data set, a time series prediction model is used to predict the congestion level of each road section in the future time period, and the traffic condition of the predicted path is output.

[0013] Optionally, the road section traffic data integration in S11 comprises:

[0014] S111, road section division and coding: the internal distribution path of the factory is divided into , wherein is the total number of path sections, is the th path section, and a unique code is assigned to each path section , and a mapping relationship with the factory geographical position information is established to form a path section coding table ;

[0015] S112, data collection: for each path section and time step , the traffic data of the road section is obtained from the factory data system, specifically including:

[0016] historical operation data: including the vehicle passing time of each path section in the past time window ;

[0017] current traffic state data: including current speed and current density ;

[0018] Vehicle dispatch plan: including the expected vehicle travel time under the assigned task ;

[0019] S113, time alignment: all traffic data are resampled to a uniform time step ;

[0020] S114, normalization processing: the minimum-maximum normalization processing is uniformly performed on the speed, density, passing time, and travel time

[0021] S115, standardized dataset construction: the traffic data after time alignment and normalization processing are organized into a standardized traffic state dataset .

[0022] Optionally, the congestion level prediction in S12 comprises:

[0023] S121, road segment historical sequence construction: based on the standardized traffic state dataset, the path segment state sequence in a continuous time window is extracted to form an input tensor ;

[0024] S122, congestion state convolution prediction: a time convolution network (TCN) constructed using causal convolution is used to perform multi-layer convolution calculation on the input tensor to predict the passing state of the future time step

[0025] S123, congestion level division: the predicted value is compared with a set of threshold values , and the predicted congestion level is output , including level 0 (smooth), level 1 (light congestion), level 2 (moderate congestion), and level 3 (serious congestion)

[0026] S124, path passing state output: the congestion levels of all path segments in the future time are grouped into a set .

[0027] Optionally, the device idle degree evaluation in S2 comprises:

[0028] S21, loading and unloading device state data acquisition: the state data of loading and unloading devices in each distribution target area are acquired, including the current task state of the device, the expected completion time, and the queuing queue information

[0029] S22, use time window map construction: based on the acquired loading and unloading device state data, the task scheduling plan is combined to construct a use time window map of each device within a certain time range (1 hour in the future) in the future

[0030] S23, device idle degree calculation: according to the time window graph, the total available duration of the device in the predicted time period is counted, and the predicted idle degree of the device is quantified in combination with the average processing time of the task, the number of waiting tasks and other parameters.

[0031] Optionally, the S21 includes the following steps of collecting the state data of the loading and unloading device:

[0032] S211, association of the loading and unloading device identification and the target area: according to the target area number of the distribution task, the unique device number of all loading and unloading devices in the area is obtained by calling the interface between the factory area GIS system and the device management system and the current binding position information, and a target area-device mapping table is established .

[0033] S212, current task state collection: whether each device is in idle, working, maintenance or failure is queried through the interface of the device execution system (MES or scheduling control platform) ;

[0034] S213, predicted completion time acquisition: for the device in the working state , the scheduling task information thereof is read to obtain the working number and the predicted end time of the execution , if it is a multi-task queue, only the predicted completion time of the current task is recorded;

[0035] S214, queue information collection: the task queue of each device arranged in the scheduling system but not yet executed is queried, the number of tasks is counted , and the predicted working duration of each task is recorded , to form a task queue information set .

[0036] Optionally, the S22 includes the following steps of constructing the time window graph:

[0037] S221, time axis division and window initialization: the future 1-hour prediction time range is evenly divided into fixed-length time windows;

[0038] S222, mapping of the working task to the time window: for each device , all tasks in the current task and the queue are mapped to the window number interval according to the task start time and the end time , indicating the occupied section of the task in the window graph;

[0039] ​S223, Using Time Window Graph Generation: Defining Devices The usage time window map is , where each element Indicates that the device is The state within a time window, that is, , the final map Representation device The occupancy and idleness of each time period in the next hour.

[0040] Optionally, the device idleness calculation in S23 includes:

[0041] S231, Idle Window Statistics: Based on the Device Usage Time Window Graph , count the total number of idle windows in the future prediction time period ;

[0042] S232, Estimation of the number of serviceable tasks: Calculate the average task processing time based on the historical operation duration of the equipment in the previous scheduling records , then the number of tasks that the equipment can theoretically complete within the forecast period is ;

[0043] S233, predict idle quantification: define the number of tasks currently waiting for the device as , then the device The idleness index in the forecast period is defined as .

[0044] Optionally, the delivery task path scheduling optimization in S3 includes:

[0045] S31, 3D Scheduling Model Construction: Constructing Triple Combinations in the Scheduling Domain ,in, For the A delivery task, For the feasible paths, For the loading and unloading equipment, in the three-dimensional scheduling model, each triple combination Corresponding to a set of scheduling decision variables, including the estimated travel time of the path , device predicted waiting time , and task assignment Boolean variables ;

[0046] S32, multi-objective weighted scheduling optimization: with the goal of minimizing the total vehicle travel time and the waiting time of loading and unloading equipment, construct the objective function, which is expressed as:

[0047] ;

[0048] wherein, , is a time weight coefficient, , , are respectively the number of tasks, the number of paths, the number of devices;

[0049] S33, constraint condition setting and optimal solution output: the three-dimensional scheduling model meets the device availability constraint and the path fluency constraint, the optimal solution of the objective function is solved and the optimal combination scheme of each task is output, expressed as:

[0050] Device availability constraint: ;

[0051] wherein, is the departure time of the task , is the idle time window set of the device ;

[0052] Path fluency constraint: ;

[0053] wherein, is the maximum path travel time tolerated by the scheduling;

[0054] Optimal combination scheme: ;

[0055] wherein, is the optimal solution combination, indicating the selected path and device combination scheme, so that the corresponding , , respectively indicate the optimal path and the optimal loading and unloading device number index allocated for the distribution task after scheduling optimization solution.

[0056] Optionally, the task priority fusion evaluation in S4 includes:

[0057] S41, task attribute quantification: obtain the attribute index of each distribution task, including task timeliness requirement, material type level and target area level, and respectively map into a standardized score vector, including task timeliness score , task material type score , and task area level score ;

[0058] S42, priority score calculation: based on the task timeliness score , the task material type score , and the task area level score ​, construct a comprehensive priority score function to calculate the comprehensive priority score of the task ;

[0059] S43, scheduling model fusion optimization: introduce priority weighting coefficient in path scheduling, and adjust the scheduling weight by taking the task priority score as a penalty term.

[0060] The beneficial effects of the present application are:

[0061] The present application, by constructing a standardized traffic state data set, fusing historical operation data, current traffic state and scheduling plan information, using a congestion prediction model based on time convolution network to dynamically predict the future traffic conditions of the path segment, realizes the prospective control of the path traffic capacity, compared with the traditional scheme relying on the static shortest path algorithm, can effectively avoid the path that will appear congestion, improve the vehicle traffic efficiency, and reduce the frequency of rescheduling caused by sudden congestion.

[0062] The present application, by collecting the task state and queue information of the loading and unloading equipment in real time, constructing the equipment usage time window atlas, and dynamically quantifying the equipment availability by combining the predicted idle degree, makes the scheduling system complete resource pre-matching in the task path planning stage, avoids the traditional to-point queuing type inefficient distribution mode, and synchronously optimizes the path and equipment resources under the premise of meeting the path smoothness and resource idle constraint, improves the overall stability of task processing and the efficiency of internal material flow in the factory.

[0063] The present application, by introducing a task priority scoring mechanism based on the three-dimensional scheduling model, comprehensively considering the time efficiency requirement of the task, the material type grade and the target area grade, constructing a multi-factor task scoring model, and taking the score value as a scheduling weighting factor in the scheduling objective function, thus prioritizing the rapid completion of key tasks under the condition of overall resource limitation, makes the scheduling system have a business-oriented awareness, realizes the intelligent evolution from task equal processing to value-sensitive scheduling, and significantly improves the response ability and business matching degree of the scheduling system to core production tasks. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only illustrate the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0065] Fig. 1 The flowchart of the distribution optimization method of the embodiment of the present application is shown.

[0066] Fig. 2A schematic diagram for evaluating the device idle degree of the embodiment of the present application. DETAILED DESCRIPTION

[0067] The present application will be described in detail below with reference to the drawings and specific embodiments. For some known technologies, other alternative ways can also be implemented by those skilled in the art; and the drawings are only used to more specifically describe the embodiments and are not intended to specifically limit the present application.

[0068] As shown in Figs. 1-2 A plant material distribution optimization method, comprising the following steps:

[0069] S1, path state prediction: acquiring traffic data of each road section in the distribution path, and predicting the congestion level of each road section in the future time period based on a time series prediction model to generate predicted path congestion information;

[0070] S2, device idle degree evaluation: collecting the state data of the loading and unloading devices in each distribution target area, including the current task state, the estimated completion time and the queuing queue information, constructing a device use time window atlas, and calculating the predicted idle degree of each loading and unloading device in the future specified time period;

[0071] S3, distribution task path scheduling optimization: based on the predicted path congestion information and the predicted idle degree, a task-path-device three-dimensional scheduling model is constructed, a target function is used to weight and sum the vehicle travel time and waiting time, under the condition of meeting the device available window and path smoothness constraint, the distribution path and arrival time are dynamically optimized, and the optimal path scheduling scheme is output;

[0072] S4, task priority fusion evaluation: based on the time requirement, material type and regional level of the distribution task, a priority scoring model is constructed, and the scoring result is introduced into the scheduling optimization, so that the optimal path scheduling scheme tends to timely complete the high priority task.

[0073] The path state prediction in S1 includes:

[0074] S11, road section traffic data integration: collecting historical operation data, current traffic state data and vehicle scheduling plan of each distribution path section, and combining with the plant road network structure, the traffic data is time-aligned and format-unified, and a standardized traffic state data set is constructed;

[0075] S12, congestion level prediction: based on the constructed traffic state data set, a time series prediction model is used to predict the congestion level of each road section in the future time period, and the traffic condition of the predicted path is output.

[0076] The road section traffic data integration in S11 includes:

[0077] S111, path segment division and coding: divide the internal distribution path of the factory into wherein, is the total number of path segments, is the first path segment, is the last path segment, a unique code is assigned to each path segment , and a mapping relationship with the geographical location information of the factory is established , forming a path segment coding table ;

[0078] S112, data collection: for each path segment and time step , the traffic data of the path segment is obtained from the factory data system, including:

[0079] historical operation data: including the vehicle passing time of each path segment in the past time window ;

[0080] current traffic state data: including current speed and current density ;

[0081] vehicle scheduling plan: including the expected vehicle passing time under the assigned task ;

[0082] S113, time alignment: all traffic data are resampled to a uniform time step ;

[0083] S114, normalization processing: the speed, density, passing time and passing time are uniformly normalized as:

[0084] ;

[0085] wherein, is the normalized traffic data, is the original traffic data of the path segment at time step , , are the global minimum and maximum values of the traffic data respectively;

[0086] S115, standardized data set construction: the traffic data after time alignment and normalization processing is organized into a standardized traffic state data set , expressed as:

[0087] .

[0088] The congestion level prediction in S12 includes:

[0089] S121, Road segment history sequence construction: Based on the standardized traffic state dataset, extract path segment state sequence within a continuous time window, constitute an input tensor , denoted as:

[0090] ;

[0091] where, is the feature vector at time point , including speed, density, passing duration, passing time, is the time window length, is the feature vector at time point ;

[0092] S122, congestion state convolution prediction: use the time convolution network (TCN) constructed by the causal convolution to carry out multi-layer convolution calculation on the input tensor, predict the passing state of the future time step, denoted as:

[0093] ;

[0094] ;

[0095]

[0096] ;

[0097] where, , , are the output feature sequences of the th layer, the th layer, and the th layer, respectively, is the ReLU nonlinear activation function, , , are the weight matrices of the th layer, the th layer, and the output layer, respectively, , , are the bias vectors of the th layer, the th layer, and the output layer, respectively, is the predicted passing state of the path segment at the future time points;

[0098] S123, congestion level division: compare the predicted value with the set threshold set Comparing, output the predicted congestion level , including level 0 (free), level 1 (light congestion), level 2 (moderate congestion), level 3 (serious congestion), represented as:

[0099] ;

[0100] Wherein, is the path segment The predicted congestion level at the future time point , , , is the preset congestion level threshold is the path segment The predicted traffic state at the future time point ;

[0101] S124, path traffic state output: all path segments in the future time congestion level set , represented as:

[0102] .

[0103] The equipment idle degree evaluation in S2 includes:

[0104] S21, loading and unloading equipment state data acquisition: acquire the state data of loading and unloading equipment in each distribution target area, including the current task state of the equipment, the expected completion time and the queuing queue information;

[0105] S22, use time window atlas construction: based on the acquired loading and unloading equipment state data, combined with the task scheduling plan, construct the use time window atlas of each equipment within a certain time range (1 hour in the future) in the future;

[0106] S23, equipment idle degree calculation: according to the time window atlas, the total available duration of the equipment in the predicted time period is calculated, and combined with the task average processing time, the number of waiting tasks and other parameters, the predicted idle degree of the equipment is quantified.

[0107] The loading and unloading equipment state data acquisition in S21 includes:

[0108] S211, loading and unloading equipment identification and target area association establishment: according to the target area number of distribution task, calling the factory area GIS system and equipment management system interface, acquiring the unique equipment number of all loading and unloading equipment in the area And the current bound position information, establish the target area-equipment mapping table , represented as:

[0109] ;

[0110] wherein, is the number of the th delivery target area, is the set of devices within the area;

[0111] S212, current task state collection: through the device execution system interface (MES or scheduling control platform), query each device whether it is currently in idle, working, maintenance or failure;

[0112] S213, predicted completion time acquisition: for the device currently in working state , read its scheduling task information, get its working job number and predicted end time , if it is a multi-task queue, only record the predicted completion time of the current task;

[0113] S214, queue information collection: query each device in the scheduling system The task queue that has been scheduled but has not been executed, count the number of tasks , and record the predicted working time of each task , form the task queue information set , denoted as:

[0114] .

[0115] The use of time window atlas in S22 includes:

[0116] S221, time axis division and window initialization: evenly divide the prediction time range of 1 hour in the future Into Fixed length time window, denoted as:

[0117] ;

[0118] wherein, is the current time (the starting point of atlas construction), is the number of time windows (5 minutes for each window), is the length of a single time window;

[0119] S222, mapping of working tasks to time windows: for each device , map all tasks in its current task and queue to the window number interval According to the start time and end time of the task, represent the occupied section of the task in the window graph, denoted as:

[0120] ;

[0121] ;

[0122] S223, Usage time window graph generation: define the device 's usage time window graph as , where each element represents the device's state in the th time window, i.e. , the final graph represents the device's occupancy and idle situation in each time period within the next 1 hour.

[0123] The device idle degree calculation in S23 includes:

[0124] S231, Idle window statistics: based on the device usage time window graph , count the total number of idle windows in the future prediction period , represented as:

[0125] ;

[0126] where, is whether the device is occupied (1 for occupied, 0 for idle) in the th time window, is the total number of time windows;

[0127] S232, Serviceable task number estimation: calculate the average task processing time according to the historical job duration of the device in the past scheduling records , then the theoretical number of tasks that the device can complete in the prediction period is , represented as:

[0128] ;

[0129] where, is the length of a single time window;

[0130] S233, Prediction idle degree quantification: define the device's current waiting task number as , then the device's idle degree index in the prediction period is defined as , represented as:

[0131] ;

[0132] where, is the device's ​The predicted idle degree, the value closer to 1 indicates the higher idle degree, and closer to 0 indicates busy or congestion.

[0133] The distribution task path scheduling optimization in S3 includes:

[0134] S31, three-dimensional scheduling model construction: constructing a three-tuple combination in the scheduling domain , wherein, is the first distribution task, is the first feasible path, is the first loading and unloading device, in the three-dimensional scheduling model, each set of three-tuple combinations corresponds to a set of scheduling decision variables, including path predicted travel time , device predicted waiting time , and task allocation Boolean variable ;

[0135] S32, multi-objective weighted scheduling optimization: constructing a target function to minimize the total travel time of the vehicle and the waiting time of the loading and unloading device, represented as:

[0136] ;

[0137] wherein, , are time weight coefficients, , , are the number of tasks, the number of paths, and the number of devices, respectively;

[0138] S33, constraint condition setting and optimal solution output: the three-dimensional scheduling model satisfies the device availability constraint and the path smoothness constraint, the optimal solution of the target function is solved and the optimal combination scheme of each task is output, represented as:

[0139] Device availability constraint: ;

[0140] wherein, is the departure time of the task , and is the idle time window set of the device ;

[0141] Path smoothness constraint: ;

[0142] wherein, is the maximum path travel time tolerated by the scheduling;

[0143] Optimal combination scheme: ;

[0144] wherein, is the optimal solution combination, representing the selected path and equipment combination scheme, such that the corresponding , , respectively represent the optimal path and optimal loading and unloading equipment number index allocated for the distribution task after the scheduling optimization solution.

[0145] The task priority fusion evaluation in S4 includes:

[0146] S41, task attribute quantification: obtain the attribute indicators of each distribution task, including task timeliness requirement (latest arrival time), material type level and target area level, and respectively map them into standardized score vectors, including task timeliness score , task material type score , and task area level score , represented as:

[0147] ;

[0148] wherein, is the timeliness score of the task , is the timeliness score function, is the deadline of the task , is the planned scheduling start time of the task , is the current system time;

[0149] ;

[0150] wherein, is the material score function, is the material type identifier required for the distribution of the task ;

[0151] ;

[0152] wherein, is the area level score function, is the target area level of the task (positive integer, the higher the area level, the more important it is), is the maximum area level value in the factory area, used for normalization (set to 5);

[0153] S42, priority score calculation: based on the task timeliness score , task material type score , task area level score , build a comprehensive priority score function to calculate the comprehensive priority score of the task , expressed as:

[0154] ;

[0155] wherein, , , respectively, the importance weights of task time limit, material level and area level;

[0156] S43, scheduling model fusion optimization: introduce priority weighting coefficient in path scheduling, take task priority score as penalty item to adjust scheduling weight, make scheduling model tend to complete high priority task first, expressed as:

[0157] .

[0158] The present application encompasses any alternatives, modifications, equivalent methods and solutions made to the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0159] The above is only the preferred embodiment of the present application, it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can also be made, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method for optimizing material distribution in a factory area, characterized in that: The following steps are involved: S1, route status prediction: Obtain traffic data for each section of the delivery route, and predict the congestion level of each section in the future time period based on the time series prediction model to generate predicted route congestion information; S2, Equipment Idleness Assessment: Collect the status data of loading and unloading equipment in each delivery target area, including the current task status, estimated completion time, and queue information, build an equipment usage time window map, and calculate the predicted idleness of each loading and unloading equipment in the specified future time period; S3, delivery task route scheduling optimization: Based on predicted route congestion information and predicted idleness, a three-dimensional task-route-equipment scheduling model is constructed. The objective function is used to perform a weighted summation of vehicle travel time and waiting time. Under the conditions of satisfying the equipment availability window and route smoothness constraints, the delivery route and arrival time are dynamically optimized to output the optimal route scheduling solution. S4, Task Priority Fusion Assessment: A priority scoring model is constructed based on the time requirements, material types, and regional levels of the delivery tasks, and the scoring results are introduced into the scheduling optimization, so that the optimal path scheduling plan tends to favor the timely completion of high-priority tasks.

2. A factory material distribution optimization method according to claim 1, characterized in that: The path state prediction in S1 includes: S11, road segment traffic data integration: Collect historical operation data, current traffic status data, and vehicle dispatch plans for each delivery route segment. Combined with the factory road network structure, the traffic data is time-aligned and formatted to construct a standardized traffic status dataset. S12, congestion level prediction: Based on the constructed traffic status dataset, the time series prediction model is used to predict the congestion level of each road section in the future time period, and the traffic conditions of the predicted path are output.

3. A factory material distribution optimization method according to claim 2, characterized in that: The integration of road traffic data in S11 includes: S111, road segment division and coding: Divide the internal distribution routes of the factory into ,in, is the total number of path segments, Respectively path segments, for each path segment Assign a unique code , and establish geographical location information of the factory The mapping relationship constitutes the path segment coding table ; S112, data collection: for each path segment and time steps , obtain the traffic data of the road section from the factory data system, including: Historical operation data: including the vehicle passing time of each route segment in the past several time windows ; Current traffic status data: including current vehicle speed and current density ; Vehicle dispatch plan: including the estimated vehicle travel time under the assigned tasks ; S113, Time Alignment: All traffic data are resampled to a uniform time step ; S114, normalization processing: performing minimum-maximum normalization processing on vehicle speed, density, passing time, and travel time; S115, Standardized Dataset Construction: Organize the time-aligned and normalized traffic data into a standardized traffic status dataset .

4. A factory material distribution optimization method according to claim 3, characterized in that: The congestion level prediction in S12 includes: S121, road segment history sequence construction: extracting path segments based on standardized traffic status dataset The state sequence in the continuous time window constitutes the input tensor ; S122, Congestion Status Convolution Prediction: Using a temporal convolutional network built using causal convolution to perform multi-layer convolution calculations on the input tensor to predict the future The transit status of the time step; S123, congestion level classification: the predicted value With the set threshold Compare and output the predicted congestion level , including level 0, level 1, level 2, level 3; S124, path traffic status output: the congestion levels of all path segments in the future are combined into a set .

5. A factory material distribution optimization method according to claim 1, characterized in that: The device idleness evaluation in S2 includes: S21, collection of loading and unloading equipment status data: collects status data of loading and unloading equipment in each delivery target area, including the equipment's current task status, estimated completion time, and queue information; S22, using time window map construction: Based on the collected loading and unloading equipment status data and combined with the task scheduling plan, a usage time window map of each equipment within a certain time range in the future is constructed; S23, device idleness calculation: Based on the time window graph, the total available time of the device in the predicted time period is counted, and the predicted idleness of the device is quantified by combining parameters such as the average task processing time and the number of waiting tasks.

6. A factory material distribution optimization method according to claim 5, characterized in that: The collection of loading and unloading equipment status data in S21 includes: S211, the association between the loading and unloading equipment identification and the target area is established: according to the target area number of the delivery task, the plant GIS system and the equipment management system interface are called to obtain the unique equipment number of all loading and unloading equipment in the area and its currently bound location information, and establish a target area-device mapping table ; S212, current task status collection: query each device through the device execution system interface Whether it is currently idle, in operation, under maintenance, or in failure; S213, Estimated completion time acquisition: For devices currently in the working state , read its scheduled task information, obtain the job number it is executing and the estimated end time ,If it is a multi-task queue, only the estimated completion time of the current task is recorded; S214, Queue information collection: query the task queue of each device that has been scheduled but not yet executed in the scheduling system, and count the number of tasks , and record the estimated duration of each task , forming a task queue information set .

7. A factory material distribution optimization method according to claim 6, characterized in that: The construction of the time window graph in S22 includes: S221, time axis division and window initialization: the prediction time range for the next hour Divide evenly into A fixed-length time window; S222, mapping of job tasks to time windows: for each device , its current task and all tasks in the queue are sorted according to the task start time and end time Mapping to window number range , indicating the task Occupancy segment in the window diagram; S223, Using Time Window Graph Generation: Defining Devices The usage time window map is , where each element Indicates that the device is The state within a time window, that is, , the final map Representation device The occupancy and idleness of each time period in the next hour.

8. A factory material distribution optimization method according to claim 7, characterized in that: The device idleness calculation in S23 includes: S231, Idle Window Statistics: Based on the Device Usage Time Window Graph , count the total number of idle windows in the future prediction time period ; S232, Estimation of the number of serviceable tasks: Calculate the average task processing time based on the historical operation duration of the equipment in the previous scheduling records , then the number of tasks that the equipment can theoretically complete within the forecast period is ; S233, predict idle quantification: define the number of tasks currently waiting for the device as , then the device The idleness index in the forecast period is defined as .

9. A factory material distribution optimization method according to claim 1, characterized in that: The delivery task path scheduling optimization in S3 includes: S31, 3D Scheduling Model Construction: Constructing Triple Combinations in the Scheduling Domain ,in, For the A delivery task, For the feasible paths, For the loading and unloading equipment, in the three-dimensional scheduling model, each triple combination Corresponding to a set of scheduling decision variables, including the estimated travel time of the path , device predicted waiting time , and task assignment Boolean variables ; S32, multi-objective weighted scheduling optimization: with the goal of minimizing the total vehicle travel time and the waiting time of loading and unloading equipment, construct the objective function, which is expressed as: ； in, 、 is the time weight coefficient, 、 、 They are the number of tasks, the number of paths, and the number of devices respectively; S33, constraint setting and optimal solution output: The three-dimensional scheduling model satisfies the equipment availability constraints and path smoothness constraints, and solves the optimal solution of the objective function and outputs each task. The optimal combination solution is expressed as: Equipment availability constraints: ; in, For the task The departure time, For devices The set of idle time windows; Path patency constraint: ; in, The maximum path travel time tolerated by scheduling; The best combination: ; in, is the optimal solution combination, which represents the selected path and device combination solution, so that the corresponding , 、 They represent the distribution tasks after the scheduling optimization solution. The number index of the assigned optimal path and optimal loading and unloading equipment.

10. A factory material distribution optimization method according to claim 9, characterized in that: The task priority fusion evaluation in S4 includes: S41, Task attribute quantification: Obtain attribute indicators for each delivery task, including task timeliness requirements, material type level and target area level, and map them into standardized scoring vectors, including task timeliness score , Mission Material Type Rating , Mission Area Rating ; S42, priority score calculation: based on task timeliness score , Mission Material Type Rating , Mission Area Rating , build a comprehensive priority scoring function to calculate the comprehensive priority score of the task ; S43, Scheduling model fusion optimization: Introduce priority weighting coefficient in path scheduling, and use task priority score as a penalty item to adjust scheduling weight.

Images