Predictive scheduling optimization method for intelligent logistics system

Abstract

The invention provides an intelligent logistics system predictive scheduling optimization method and system, and the method specifically comprises the steps: collecting the operation state data of equipment in an intelligent logistics system in real time, carrying out the preprocessing of the operation state data, and dividing the operation state data into a training set and a test set; time domain analysis is carried out on the preprocessed operation state data, key fault features are extracted from the operation state data, and health state evaluation is carried out on the equipment; a predictive scheduling model is established, scheduling strategy switching is carried out based on the obtained health state evaluation result, and multi-objective optimization is carried out on a production logistics scheduling scheme; a three-dimensional modeling technology is adopted to carry out visual management, and virtual simulation is combined to carry out scheduling optimization verification. The method provided by the invention can effectively deal with the problems of equipment failure and the like in a dynamic logistics environment, optimize the logistics scheduling performance, reduce the distribution delay, improve the overall efficiency and resource utilization rate of a logistics system, and realize high-intelligence and high-efficiency production logistics scheduling.

Description

Technical Field

[0001] The present invention belongs to the technical field of dynamic scheduling in industrial manufacturing, and particularly relates to a predictive scheduling optimization method for an intelligent logistics system. Background Art

[0002] With the increase in the complexity and uncertainty of logistics systems, traditional logistics scheduling methods are no longer sufficient to meet the logistics demands in a dynamic environment. Especially when dealing with emergencies such as order fluctuations, traffic congestion, and resource allocation conflicts, they often show a slow response, resulting in a decline in logistics efficiency and an increase in costs. Existing logistics scheduling solutions usually lack the ability to predict the key states and potential problems in the logistics process, making it difficult to achieve efficient utilization of resources and scheduling optimization.

[0003] In the patent with publication number CN118886808A, deep learning is combined with multi-objective optimization to improve prediction accuracy and scheduling efficiency. However, the model has a high complexity, requires a large amount of computing resources, and is difficult to deploy and maintain in practice. In the patent with publication number CN117610817A, the predictability of production plans is improved, and it can adapt to inventory changes. However, there are problems such as adaptability to complex and changeable environments and the accuracy of prediction models.

[0004] The main problems in the predictive scheduling of intelligent logistics systems are as follows: 1) Existing scheduling methods are usually based on static planning or historical data, lacking real-time monitoring and prediction of the states of key nodes in the logistics chain, and it is difficult to identify risks that may cause delays or resource waste in advance. This passive response method is prone to problems such as transportation delays and order backlogs. 2) Existing logistics scheduling systems are slow to respond when facing complex situations such as sudden increases in orders, adjustments to delivery routes, and changes in traffic conditions in a dynamic environment, relying on manual intervention for adjustment, resulting in low scheduling efficiency and lacking real-time and intelligent support. 3) Most existing scheduling systems are based on fixed distribution and maintenance strategies and cannot flexibly optimize resource allocation according to the actual logistics operation status. Fixed strategies may lead to resource redundancy or insufficiency, increasing operating costs and affecting the overall stability and response ability of the logistics system. Therefore, there is an urgent need for a method that can predict the key states of logistics and achieve dynamic scheduling optimization to improve the flexibility and efficiency of the system. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a predictive scheduling optimization method for an intelligent logistics system, which can effectively address problems such as equipment failures in a dynamic logistics environment, optimize logistics scheduling performance, reduce delivery delays, and improve the overall efficiency and resource utilization rate of the logistics system.

[0006] To achieve the above technical objectives, the present invention provides the following technical solutions:

[0007] A predictive scheduling optimization method for an intelligent logistics system, specifically including the following steps:

[0008] S1. Real-time collect the operation status data of the equipment in the intelligent logistics system, preprocess the operation status data, and divide it into a training set and a test set;

[0009] S2. Conduct time-domain analysis on the preprocessed operation status data, extract key fault features from it, and evaluate the health status of the equipment;

[0010] S3. Establish a predictive scheduling model, switch the scheduling strategy based on the health status evaluation result obtained in step S2, and perform multi-objective optimization on the production logistics scheduling plan;

[0011] S4. Adopt three-dimensional modeling technology for visual management, and combine virtual simulation to verify the scheduling optimization.

[0012] Furthermore, step S1 specifically includes:

[0013] S11. Utilize industrial Internet of Things technology to real-time collect the operation status data of the processing equipment in the intelligent logistics system through sensors. The collected data includes the temperature, current, and load of the processing equipment, and transmit the data to the edge computing unit for real-time processing through wireless communication. The processed data is stored in the cloud database in a time-series format;

[0014] S12. Conduct data cleaning, normalization, and feature engineering on the collected operation status data to remove outliers and noise, ensuring the effectiveness of the data; perform data augmentation to improve the generalization ability of the model; and divide the data set into a training set and a test set.

[0015] Furthermore, step S2 specifically includes:

[0016] S21. Extract the mean and variance of the temperature, current, and load signals as key fault features through statistical analysis methods in the time domain, and normalize the key fault features to avoid the influence of different dimensions; pay special attention to the equipment with a high historical failure rate during the extraction process;

[0017] S22. Construct an equipment health status coefficient based on the key fault features, and divide the equipment health status into normal, mildly abnormal, and severely abnormal;

[0018] S23. Select the random forest model as the equipment health status evaluation model; first perform dimensionality reduction on the key fault features through the principal component analysis method before model training, then use the training set to train the model, and evaluate the generalization ability of the model through K-fold cross-validation during the model training process, and optimize the hyperparameters through the Bayesian optimization method; use the test set to evaluate the model performance to obtain the optimal equipment health status evaluation model.

[0019] More specifically, the device health status coefficient and the device health status classification in step S22 are specifically as follows:

[0020] Device health status coefficient H:

[0021]

[0022] Among them, are the mean values of the normalized temperature, current, and load respectively; are the variances of the normalized temperature, current, and load respectively; w T 、w I 、w L are the influence weights of temperature, current, and load on the device health status respectively, and different influence weights are set according to the importance of different physical quantities to the health status;

[0023] The calculation formulas for the normalized temperature mean value and variance are:

[0024]

[0025] Among them, μ T ,max, μ T ,min represent the upper and lower bounds of the normal value of the temperature mean respectively, σ 2 T ,max, σ 2 T ,min represent the upper and lower bounds of the normal value of the temperature variance respectively; the upper and lower bounds of the normal value are given by historical fault data;

[0026] The mean values and variances of the normalized current and load adopt the same calculation method;

[0027] The device health status classification is specifically as follows:

[0028] Set different abnormal thresholds. When the calculated device health status coefficient H exceeds the threshold, it is judged as an abnormal state, including:

[0029] Normal state, at this time H < h1;

[0030] Mild abnormality, h1 ≤ H < h2, and max(H T , H I , H L ) ≥ h1;

[0031] Severe abnormality, H ≥ h2, and max(H T , H I , H L ) ≥ h1;

[0032] Among them, respectively represent the influence coefficients of temperature, current, and load on the health state of the device.

[0033] Furthermore, step S3 specifically includes:

[0034] S31. Construct a flexible job shop model FJSP, record all processing equipment and its operating status, all to-be-executed logistics tasks, and the optimal driving routes of AGVs to reflect the operating status of the production logistics system in real time; the FJSP formula is expressed as:

[0035] M = (E, T, P);

[0036] where E is the set of equipment, T is the set of tasks, and P is the set of logistics paths;

[0037] S32. Based on the constructed FJSP model, design the objective function for multi-objective optimized production logistics scheduling, generate an initial scheduling plan through the BRKGA algorithm, and allocate processing equipment, logistics tasks, and optimal driving routes; and check whether the initial scheduling plan meets the constraints of the FJSP model. If it meets the requirements, the plan is considered feasible, and the initial scheduling plan is executed;

[0038] S33. During the execution of the initial scheduling plan, evaluate the health state of the equipment through real-time detected equipment operating status data, and perform pre-reactive scheduling according to the evaluation results. Specifically:

[0039] When the health state is evaluated as normal within a certain time period, maintain the production logistics operation plan according to the initial scheduling plan;

[0040] When the health state is evaluated as slightly abnormal within a certain time period, perform rolling scheduling, and add the production logistics operation tasks within this time period as new tasks to the end of the initial scheduling plan;

[0041] When the health state is evaluated as severely abnormal within a certain time period, perform complete rescheduling. Similarly, through the BPKGA algorithm, based on the objective function of multi-objective optimized production logistics scheduling, completely rearrange all tasks that have not yet started, and reallocate tasks to other available equipment to ensure that the production logistics process is not interrupted;

[0042] S34. The plan after complete rescheduling is also checked for feasibility of constraints and real-time detection is maintained until all production logistics tasks are completed.

[0043] More specifically, the objective function for multi-objective optimized production logistics scheduling designed in step S32 is specifically:

[0044] minF(x) = α1C max +α2T idle +α3Ecost ;

[0045] Among them, T idle is the equipment idle time; E cost is the total energy consumption cost of the scheduling scheme; C max = max(C k ) is the makespan, and C k represents the completion time of any workpiece k; α1, α2, α3 are the weight coefficients for optimizing the scheduling.

[0046] The formula for the equipment idle time T idle is expressed as:

[0047]

[0048] Among them, M is the number of devices, N i is the number of tasks assigned to the i-th device, and are respectively the start time of the (j + 1)-th task and the end time of the j-th task of the i-th device;

[0049] The calculation process of the total energy consumption cost E cost is as follows: First, for each device, calculate its energy consumption cost E i as:

[0050] E i = P i × T i × C e ;

[0051] Add up the energy consumption costs of all devices to obtain the total energy consumption cost E cost , and the formula is expressed as:

[0052]

[0053] Among them, C e is the unit price of energy, M is the number of devices, P i is the energy consumption rate of the i-th device, and T i is the running time of the i-th device;

[0054] The completion time C k of a workpiece is the end time of the last process for producing this workpiece;

[0055] α1, α2, α3 are dynamically adjusted according to changes in the production environment; during the production peak period, increase the weight of α1 and prioritize minimizing the makespan; during equipment maintenance, increase the weight of α2 and prioritize minimizing the equipment idle time; when energy prices fluctuate, increase the weight of α3 and prioritize minimizing the energy consumption cost.

[0056] Further, step S4 is as follows:

[0057] Based on CAD modeling software, establish a digital twin model of the logistics equipment and the production line to achieve production scheduling simulation in a virtual environment; combine Internet of Things technology to monitor the operating status of the equipment in real time, and perform dynamic task adjustment through AI prediction and optimization algorithms; the virtual simulation environment is used to evaluate the execution effects of different scheduling schemes, optimize the production scheduling strategy, improve equipment utilization and production efficiency, and reduce production interruptions caused by equipment failures.

[0058] Based on the above technical solutions, the present invention has at least the following beneficial effects:

[0059] 1. The present invention extracts various time-domain key fault features from the equipment operating status data, and combines machine learning algorithms to accurately evaluate the equipment health status, accurately predict the faults of the equipment in high-risk links, and obtain accurate fault time periods; thereby reducing equipment downtime and improving the efficiency of equipment maintenance;

[0060] 2. Based on the equipment health status evaluation results, the present invention uses hybrid rescheduling to dynamically adjust the production logistics tasks, uses rolling scheduling and complete rescheduling in the two scenarios of mild equipment abnormality and severe equipment abnormality respectively, and assigns tasks to non-faulty equipment to ensure that the production process is not interrupted while ensuring that the rescheduling strategy does not occupy too many resources; in addition, the genetic algorithm is combined to optimize the production scheduling of the workshop from multiple target angles, improving the overall efficiency and equipment utilization rate of the production logistics system and reducing energy consumption;

[0061] 3. By establishing a three-dimensional model and a virtual environment, combining with the actual process flow, test and verify the fault prediction and scheduling optimization algorithms; evaluate the actual effects of the methods according to the data of the real production environment, ensuring the feasibility and applicability of the methods proposed by the present invention. Description of the Drawings

[0062] Figure 1 is the overall flowchart of a predictive scheduling optimization method for an intelligent logistics system proposed by the present invention;

[0063] Figure 2 is the schematic diagram of the operation process of the predictive scheduling model proposed by the present invention. Detailed Embodiments

[0064] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present 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 only used to explain the present invention and are not used to limit the present invention.

[0065] Although the steps in the present invention are arranged with reference numerals, they are not used to limit the order of the steps. Unless the order of the steps is clearly stated or the execution of a certain step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0066] As Figure 1 shown, a predictive scheduling optimization method for an intelligent logistics system proposed by the present invention is shown, which specifically includes the following steps:

[0067] S1. Real-time collect the operation status data of the devices in the intelligent logistics system, preprocess the operation status data, and divide it into a training set and a test set;

[0068] As a preferred embodiment, step S1 specifically includes:

[0069] S11. Utilize industrial Internet of Things technology to real-time collect the operation status data of the processing devices in the intelligent logistics system through sensors. The collected data includes the temperature, current, and load of the processing devices, and transmit it to the edge computing unit through wireless communication for real-time processing. The processed data is stored in the cloud database in a time series format; the so-called storage in a time series format means that, for example, a single piece of data stored in the database contains information such as a timestamp, device number, sensor type, and value, so as to achieve efficient query and support subsequent data mining and analysis;

[0070] S12. Perform data cleaning, normalization, and feature engineering on the collected operation status data to remove outliers and noise and ensure the effectiveness of the data; perform data augmentation to improve the generalization ability of the model; and divide the data set into a training set and a test set. In this embodiment, the ratio of the training set to the test set is 8:2.

[0071] S2. Perform time-domain analysis on the preprocessed operation status data, extract key fault features from it, and evaluate the health status of the devices;

[0072] As a preferred embodiment, step S2 specifically includes:

[0073] S21. Extract the mean and variance of the temperature, current, and load signals from the time domain through statistical analysis methods as key fault features, and normalize the key fault features to avoid the influence of different dimensions; especially pay attention to the devices with high historical failure rates during the extraction process;

[0074] Generally, when a device fails, its operating data will show obvious anomalies. Therefore, this application selects to analyze the abnormal temperature peak, load fluctuation, and current fluctuation of the device during the failure period; then extracts key failure features from these failure data, and focuses on devices with high historical failure rates (such as punching, grinding, etc.) during the feature extraction process; however, the extracted key failure features are subsequently used to evaluate the health status of the processing device; that is, the device status is divided into normal, slightly abnormal, and severely abnormal; in this application, the health assessment of the device is carried out from a certain time period rather than a single moment. On the one hand, this can avoid the influence of occasionally occurring abnormal values on the device health assessment, and on the other hand, it can grasp the correlation and dependence of the failure signal over a long period of time, so as to predict the state pattern of a specific device within a certain time period, thereby providing decision support for predictive scheduling.

[0075] In this embodiment, the means and methods of temperature, current, and load signals within a time period of length X are calculated as follows:

[0076] Temperature signal:

[0077]

[0078] Among them, T x is the temperature signal value of the device at the current x moment; μ T is the mean of the temperature signal, and σ 2 T is the variance of the temperature signal.

[0079] Current signal:

[0080]

[0081] Among them, I x is the current signal value of the device at the current x moment, μ I is the mean of the current signal, and σ 2 I is the variance of the current signal.

[0082] Load signal:

[0083]

[0084] Among them, L x is the load signal value of the device at the current x moment, μ L is the mean of the load signal, and σ 2 L is the variance of the load signal; further, due to the different dimensions of different features, it is necessary to normalize the mean and variance so that they can be compared on the same scale; the normalized mean and variance are respectively:

[0085]

[0086] Among them, are respectively the means of the normalized temperature, current, and load; are respectively the variances of the normalized temperature, current, and load; μ T ,max, μ T ,min, μ I ,max, μ I ,min, μ L ,max, μ L ,min respectively represent the upper and lower bounds of the normal values of the temperature, current, and load means; σ 2 T ,max, σ 2 T ,min, σ 2 I ,max, σ 2 I ,min, σ 2 L ,max, σ 2 L ,min respectively represent the upper and lower bounds of the normal values of the temperature variance; the upper and lower bounds of the normal values are given by historical fault data

[0087] S22. Construct the equipment health state coefficient based on the key fault characteristics, and divide the equipment health state into normal, slightly abnormal, and severely abnormal;

[0088] More specifically, the equipment health state coefficient and the equipment health state division in step S22 are specifically as follows:

[0089] Equipment health state coefficient H:

[0090]

[0091] Among them, w T , w I , w L are respectively the influence weights of temperature, current, and load on the equipment health state. Different influence weights are set according to the importance of different physical quantities for the health state; for different equipment, different weights can be assigned to different physical quantities according to their characteristics. For example, in this embodiment, it is considered that the temperature signal can best reflect the equipment health state, followed by the current. Therefore, w T , w I , w L are assigned values of 0.6, 0.3, and 0.1 in sequence;

[0092] The equipment health state division is specifically as follows:

[0093] Set different anomaly thresholds. When the calculated device health status coefficient H exceeds the threshold, it is determined to be in an abnormal state, including:

[0094] Normal state, where H < h1;

[0095] Mild anomaly, h1 ≤ H < h2, and max(H T ,H I ,H L ) ≥ h1;

[0096] Severe anomaly, H ≥ h2, and max(H T ,H I ,H L ) ≥ h1;

[0097] Among them, respectively represent the influence coefficients of temperature, current, and load on the device health status; in this embodiment, h1 = 0.6 and h2 = 0.8 are taken;

[0098] It should be noted here that this application analyzes whether the device is faulty from abnormal operation data. Therefore, the mean and variance within a certain time period are normalized with the mean range and variance range when the device is normal, and the constructed health status coefficient can better reflect the deviation between abnormal and normal. Analyzing the three physical quantities and setting different weights make the result of health status evaluation more accurate and reliable; classifying the fault anomaly conditions into mild and severe cases also makes the rescheduling strategy selected according to the health status more reasonable in the follow-up;

[0099] In addition, considering that there may be a situation where all three physical quantities are within the normal range, but the sum is judged to be in a mild anomaly or even a severe anomaly, this application also sets max(H T ,H I ,H L ) ≥ h1 to ensure that at least one of the temperature, current, and load deviates from the normal range before determining that the device is in a mild anomaly or severe anomaly, so as to avoid misjudgment.

[0100] S23. Select the random forest model as the device health status evaluation model; before model training, first reduce the dimensionality of the key fault features through the principal component analysis method, then use the training set to train the model, and evaluate the generalization ability of the model through K-fold cross-validation during the model training process, and optimize the hyperparameters through the Bayesian optimization method; use the test set to evaluate the model performance to obtain the optimal device health status evaluation model.

[0101] In this embodiment, K-fold cross-validation ensures the stability of the model on different data sets, and the Bayesian optimization method adjusts the hyperparameters of the random forest model, including the number of trees, the maximum depth, and the minimum sample split, so as to obtain an optimal configuration model.

[0102] S3. Establish a predictive scheduling model, switch the scheduling strategy based on the health status evaluation result obtained in step S2, and perform multi-objective optimization on the production logistics scheduling plan;

[0103] As a preferred embodiment, as Figure 2 shown, step S3 specifically includes:

[0104] S31. Construct a flexible job shop model FJSP, record all processing equipment and its operating status, all pending logistics tasks, and the optimal driving route of the AGV, so as to reflect the operating status of the production logistics system in real time; the flexible job shop model includes an equipment layer, a task layer, and a logistics path layer; the equipment layer contains all processing equipment and its operating status, the task layer records all pending logistics tasks, and the logistics path layer defines the optimal driving route of the AGV; the FJSP formula is expressed as:

[0105] M = (E, T, P);

[0106] where E is the equipment set, T is the task set, and P is the logistics path set;

[0107] S32. Based on the constructed FJSP model, design the objective function for multi-objective optimization of production logistics scheduling, generate an initial scheduling plan through the BRKGA algorithm (the BRKGA algorithm generates multiple scheduling plans through operations such as initializing the population, crossover, and mutation, and evaluates the fitness of each plan; and based on the designed objective function, optimize through the algorithm to find the optimal scheduling plan as the initial scheduling plan), allocate processing equipment, logistics tasks, and the optimal driving route; and check whether the initial scheduling plan meets the constraints of the FJSP model. If it meets, the plan is considered feasible, and the initial scheduling plan is executed; the constraints of the FJSP model include:

[0108] Job processing sequence constraint: Each job must be processed on each equipment in the technological route sequence;

[0109] Equipment capacity constraint: Each equipment can only process one job at the same time;

[0110] Non-interruptible constraint: Once a job starts processing, it cannot be interrupted;

[0111] Equipment failure constraint: When a device fails, it needs to be shut down for maintenance and can continue processing only after the maintenance is completed;

[0112] Such constraint conditions ensure that the initial scheduling plan is feasible. As for how to achieve optimal scheduling, this application considers optimization objectives from multiple perspectives. As shown below, the objective function of the multi-objective optimized production logistics scheduling designed in this application is specifically:

[0113] minF(x)=α1C max +α2T idle +α3E cost ;

[0114] Among them, T idle is the equipment idle time; E cost is the total energy consumption cost of the scheduling plan; C max =max(C k ) is the makespan, and C k represents the completion time of any workpiece k; α1, α2, and α3 are the weight coefficients for optimizing the scheduling. Through such an objective function, it is ensured that each generated scheduling plan can: shorten the production cycle, ensure that all tasks are completed as soon as possible; improve equipment utilization rate and reduce equipment idle time; reduce energy consumption costs and achieve green production.

[0115] The following describes how the equipment idle time, total energy consumption cost, makespan, and weight coefficients for optimizing the scheduling are calculated and determined:

[0116] The equipment idle time T idle refers to the time when the equipment is not used during the production process, that is, the time when the equipment is in an idle state. According to the scheduling plan, first determine the tasks assigned to each equipment and their start time and end time. For each equipment, calculate the time interval between its adjacent tasks, that is, the equipment idle time. The formula is expressed as:

[0117]

[0118] Among them, M is the number of equipment, and N i is the number of tasks assigned to the i-th equipment, and are the start time and end time of the (j + 1)-th task of the i-th equipment respectively;

[0119] The total energy consumption cost E cost refers to the energy cost consumed by the equipment in the entire production logistics system during the production process, which is usually related to the running time and energy consumption rate of the equipment. For each equipment, calculate its energy consumption cost E i as:

[0120] E i =P i ×T i ×C e ;

[0121] Add up the energy consumption costs of all devices to obtain the total energy consumption cost E cost , which is expressed by the formula:

[0122]

[0123] where C e is the unit price of energy, M is the number of devices, P i is the energy consumption rate of the i-th device, and T i is the operating time of the i-th device;

[0124] The completion time C k of a workpiece is the time when all operations of workpiece k are completed, that is, the end time of the last operation of this workpiece; to calculate the completion time of a workpiece, first determine the operation sequence of the workpiece, then according to the scheduling plan, determine the start time and end time of each operation on the device, and finally obtain the completion time of the workpiece; for example, assume that workpiece k has 3 operations, and its operation sequence and processing time are as follows: Operation 1: The processing time is 2 hours, the start time is 0 hour, and the end time is 2 hours. Operation 2: The processing time is 3 hours, the start time is 2 hours, and the end time is 5 hours. Operation 3: The processing time is 1 hour, the start time is 5 hours, and the end time is 6 hours. Then the completion time of this workpiece is:

[0125] C k = max(2, 5, 6) = 6 hours;

[0126] α1, α2, α3 are dynamically adjusted according to changes in the production environment; during the production peak period, increase the weight of α1 and give priority to minimizing the makespan; during equipment maintenance, increase the weight of α2 and give priority to minimizing equipment idle time; when energy prices fluctuate, increase the weight of α3 and give priority to minimizing energy consumption costs.

[0127] S33. During the execution of the initial scheduling plan, by real-time detecting the device operation status data and evaluating the device health status, perform pre-reactive scheduling according to the evaluation results (that is, select a rescheduling strategy according to the severity of the device health status), specifically:

[0128] When the health status is evaluated as normal within a certain time period, keep producing the logistics operation plan according to the initial scheduling plan;

[0129] When the health status is evaluated as slightly abnormal within a certain time period, perform rolling scheduling and add the production logistics operation tasks within this time period as new tasks to the end of the initial scheduling plan;

[0130] When the health status is evaluated as severely abnormal within a certain period, a complete rescheduling is performed. Similarly, through the BPKGA algorithm, based on the objective function of multi-objective optimized production logistics scheduling, all unstarted tasks are completely rearranged, and tasks are reallocated to other available devices to ensure the uninterruption of the production logistics process.

[0131] S34. The solution after complete rescheduling is also subjected to constraint checking to determine feasibility and real-time detection is maintained (i.e., the perturbation detection as shown in Figure 2 When perturbations and anomalies such as temperature, current, and load signals are detected, it is determined again whether local correction is required, i.e., rolling scheduling and complete rescheduling) until all production logistics tasks are completed.

[0132] S4. Three-dimensional modeling technology is adopted for visual management, and virtual simulation is combined for scheduling optimization verification.

[0133] As a preferred implementation manner, step S4 specifically includes:

[0134] Based on CAD modeling software, a digital twin model of logistics equipment and production lines is established to realize production scheduling simulation in a virtual environment; combined with Internet of Things technology, the operation status of equipment is monitored in real time, and task dynamic adjustment is performed through AI prediction and optimization algorithms; the virtual simulation environment is used to evaluate the execution effects of different scheduling schemes, optimize production scheduling strategies, improve equipment utilization rate and production efficiency, and reduce production interruptions caused by equipment failures.

[0135] In an intelligent logistics system, real-time monitoring of equipment status and scheduling optimization are the keys to improving production efficiency; to further verify the effectiveness of the method proposed in this application, in this embodiment, a three-dimensional digital twin model of production lines and logistics equipment is established using CAD modeling software to accurately reflect the equipment, workstations, and work processes in the production environment, and the method proposed in this application is verified. Combined with Internet of Things technology, sensors are deployed to monitor the operation status of equipment in real time, obtain the working data and fault information of the equipment, and transmit these data to the control system to synchronize the virtual environment with the actual production situation. Different scheduling schemes are simulated through virtual simulation technology, their impacts on production processes, equipment utilization rate, and workstation load are tested, and AI prediction and optimization algorithms are used for dynamic adjustment of tasks, thereby optimizing production scheduling in real time.

[0136] At the same time, this method uses IoT devices to collect the operation status of equipment, including parameters such as temperature, current, and load, and evaluates the health status of the equipment through big data analysis and AI algorithms. The predicted equipment failure time period through virtual simulation can be compared with and simultaneously adopted in the evaluation results during the actual production process, making two preparations for advance planning of maintenance tasks and avoiding the impact of sudden failures on production plans.

[0137] This application ensures the maximization of production efficiency, the improvement of equipment utilization rate, and the reduction of production interruptions caused by equipment failures through the simultaneous implementation of predictive scheduling for the intelligent logistics system in combination with virtual simulation and actual production, ultimately achieving intelligent and efficient logistics scheduling.

[0138] In summary, the present invention provides a method for optimizing the predictive scheduling of an intelligent logistics system, which can effectively address problems such as equipment failures in a dynamic logistics environment, optimize the logistics scheduling performance, reduce delivery delays, and improve the overall efficiency and resource utilization rate of the logistics system.

[0139] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this application can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0140] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the inventive concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A predictive scheduling optimization method for an intelligent logistics system, characterized in that It includes the following steps: S1. Collect the operation status data of the devices in the intelligent logistics system in real time, preprocess the operation status data, and divide it into a training set and a test set; S2. Conduct time-domain analysis on the preprocessed operation status data, extract key fault features from it, and evaluate the health status of the devices; S3. Establish a predictive scheduling model, switch the scheduling strategy based on the health status evaluation result obtained in step S2, and perform multi-objective optimization on the production logistics scheduling plan; S4. Adopt three-dimensional modeling technology for visual management, and combine virtual simulation to verify the scheduling optimization.

2. The predictive scheduling optimization method for an intelligent logistics system according to claim 1, wherein Step S1 specifically includes the following steps: S11. Utilize industrial Internet of Things technology to collect the operation status data of the processing devices in the intelligent logistics system in real time through sensors. The collected data includes the temperature, current, and load of the processing devices, and transmits the data to the edge computing unit for real-time processing through wireless communication. The processed data is stored in the cloud database in a time-series format; S12. Clean, normalize, and perform feature engineering on the collected operation status data to remove outliers and noise and ensure the effectiveness of the data; perform data augmentation to improve the generalization ability of the model; and divide the data set into a training set and a test set.

3. The predictive scheduling optimization method for an intelligent logistics system according to claim 1, wherein Step S2 specifically includes: S21. Extract the mean and variance of the temperature, current, and load signals from the time domain through statistical analysis methods as key fault features, and normalize the key fault features to avoid the influence of different dimensions; pay special attention to the devices with high historical failure rates during the extraction process; S22. Construct a device health status coefficient based on the key fault features, and divide the device health status into normal, mildly abnormal, and severely abnormal; S23. Select the random forest model as the device health status evaluation model; perform dimensionality reduction on the key fault features through principal component analysis before model training, then use the training set to train the model, and evaluate the generalization ability of the model through K-fold cross-validation during the model training process, and perform hyperparameter optimization through the Bayesian optimization method; use the test set to evaluate the model performance to obtain the optimal device health status evaluation model.

4. An intelligent logistics system predictive scheduling optimization method according to claim 3, characterized in that, The device health status coefficient and the device health status division in step S22 are specifically as follows: Device health status coefficient H: wherein, are the mean values of the normalized temperature, current, and load respectively; are the variances of the normalized temperature, current, and load respectively; w T 、w I 、w L are the influence weights of temperature, current, and load on the health state of the device respectively, and different influence weights are set according to the importance of different physical quantities for the health state; The calculation formulas for the normalized mean and variance of temperature are: Among them, μ T ,max and μ T ,min respectively represent the upper and lower bounds of the normal value of the average temperature, and σ 2 T ,max and σ 2 T ,min respectively represent the upper and lower bounds of the normal value of the temperature variance; the upper and lower bounds of the normal value are given by historical fault data; The normalized mean and variance of current and load adopt the same calculation method; The device health status division is specifically as follows: Set different anomaly thresholds. When the calculated device health status coefficient H exceeds the threshold, it is judged as an abnormal state, including: Normal state, where H < h1; Mild abnormality, h1 ≤ H < h2, and max(H T , H I , H L ) ≥ h1; Severe abnormality, H≥h2, and max(H T ,H I ,H L )≥h1; Among them, respectively represent the influence coefficients of temperature, current, and load on the health state of the device.

5. A predictive scheduling optimization method for an intelligent logistics system according to claim 1, characterized in that Step S3 specifically includes: S31. Construct a flexible job shop model FJSP, record all processing devices and their operation status, all pending logistics tasks, and the optimal driving routes of AGVs to reflect the operation status of the production logistics system in real time; The FJSP formula is expressed as: M = (E, T, P); where E is the set of devices, T is the set of tasks, and P is the set of logistics paths; S32. Based on the constructed FJSP model, design the objective function for multi-objective optimization of production logistics scheduling. Generate an initial scheduling plan through the BRKGA algorithm, allocate processing equipment, logistics tasks, and the optimal driving route; and check whether the initial scheduling plan meets the constraints of the FJSP model. If it meets the requirements, the plan is considered feasible, and the initial scheduling plan is executed. S33. During the execution of the initial scheduling plan, evaluate the health status of the equipment based on the real-time detected equipment operation status data, and perform pre-reactive scheduling according to the evaluation results. Specifically: When the health status is evaluated as normal within a certain time period, maintain the production logistics operation plan according to the initial scheduling plan. When the health status is evaluated as slightly abnormal within a certain time period, perform rolling scheduling, and add the production logistics operation tasks within this time period as new tasks to the end of the initial scheduling plan. When the health status is evaluated as severely abnormal within a certain time period, perform complete rescheduling. Similarly, through the BPKGA algorithm, based on the objective function of multi-objective optimization of production logistics scheduling, completely rearrange all tasks that have not yet started, and reallocate tasks to other available equipment to ensure that the production logistics process is not interrupted. S34. The plan after complete rescheduling is also checked for constraints to determine its feasibility, and real-time detection is maintained until all production logistics tasks are completed.

6. The predictive scheduling optimization method for an intelligent logistics system according to claim 5, characterized in that, The specific objective function for multi-objective optimization of production logistics scheduling designed in step S32 is: minF(x) = α1C max + α2T idle + α3E cost ; Among them, T idle is the device idle time; E cost is the total energy consumption cost of the scheduling scheme; C max = max(C k ) is the makespan, and C k represents the completion time of any workpiece k; α1, α2, α3 are the weight coefficients for optimizing the scheduling; Device idle time T idle is expressed by the formula: Among them, M is the number of devices, and N i is the number of tasks assigned to the i-th device, and are respectively the start time of the (j + 1)-th task and the end time of the j-th task of the i-th device; Total energy consumption cost E cost The calculation process is as follows: for each device, calculate its energy consumption cost E i which is E i = P i × T i × C e ; Add up the energy consumption costs of all devices to obtain the total energy consumption cost E cost , which is expressed by the formula: Among them, C e is the unit price of energy, M is the number of devices, and P i is the energy consumption rate of the i-th device, and T i is the operating time of the i-th device; Completion time C of the workpiece k It is the end time of the last process for producing the workpiece; α1, α2, and α3 are dynamically adjusted according to changes in the production environment; during the production peak period, increase the weight of α1 to prioritize minimizing the makespan; during equipment maintenance, increase the weight of α2 to prioritize minimizing equipment idle time; when energy prices fluctuate, increase the weight of α3 to prioritize minimizing energy consumption costs.

7. An optimization method for predictive scheduling of an intelligent logistics system according to claim 1, characterized in that, Step S4 is specifically: Based on CAD modeling software, establish a digital twin model of logistics equipment and production lines to achieve production scheduling simulation in a virtual environment; combined with Internet of Things technology, real-time monitor the equipment operation status, and perform dynamic task adjustment through AI prediction and optimization algorithms; the virtual simulation environment is used to evaluate the execution effects of different scheduling plans, optimize production scheduling strategies, improve equipment utilization and production efficiency, and reduce production interruptions caused by equipment failures.

Images