Material delivery dynamic residual buffer time judgment method and system, terminal and medium

Abstract

The invention relates to the field of supply chain management, and particularly discloses a material delivery dynamic residual buffer time judgment method and system, a terminal and a medium, and the method comprises the steps: obtaining the initial plan buffer time of an order; calculating the accumulated consumed buffer time based on the deviation between the actual progress and the planned progress of the order in each link of the supply chain; based on the real-time internal and external risk data, obtaining a dynamic risk adjustment amount through a rule engine or a pre-trained risk adjustment amount prediction model; and calculating the current dynamic residual buffer time according to the initial plan buffer time, the accumulated consumed buffer time and the dynamic risk adjustment amount. The authenticity and timeliness of time judgment are improved, and the reliability and efficiency of material delivery are improved.

Description

Technical Field

[0001] This invention relates to the field of supply chain management, specifically to a method, system, terminal, and medium for determining the dynamic remaining buffer time for material delivery. Background Technology

[0002] In existing supply chain management systems, the monitoring of material delivery time often adopts a static time management approach. This means the system tracks progress based on preset fixed time nodes and assesses delivery risk by simply comparing the planned and actual dates. However, linear calculations based solely on calendar time fail to reflect the dynamic consumption of overall buffer time due to the actual execution efficiency of each stage. Furthermore, data from different stages are not interconnected, and progress deviations cannot be uniformly quantified into the overall buffer time system, hindering a comprehensive risk assessment and affecting the accuracy of buffer time assessments. Moreover, the system can only react passively after risks occur, lacking foresight regarding potential risks and the ability to adjust buffer time expectations in advance. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a method, system, terminal, and medium for determining the dynamic remaining buffer time for material delivery, thereby improving the accuracy and timeliness of time determination and enhancing the reliability and efficiency of material delivery.

[0004] In a first aspect, the technical solution of the present invention provides a method for determining the dynamic remaining buffer time for material delivery, comprising the following steps: Initial planned buffer time for order acquisition; Calculate the cumulative buffer time consumed based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain. Based on real-time internal and external risk data, dynamic risk adjustment is obtained through a rule engine or a pre-trained risk adjustment prediction model. The current dynamic remaining buffer time is calculated based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

[0005] Secondly, the technical solution of the present invention provides a dynamic remaining buffer time determination system for material delivery, comprising: The initial plan buffer time acquisition module is used to obtain the initial plan buffer time of an order. The consumed buffer time calculation module is used to calculate the cumulative consumed buffer time based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain. The dynamic risk adjustment generation module is used to obtain dynamic risk adjustment based on real-time internal and external risk data, through a rule engine or a pre-trained risk adjustment prediction model. The dynamic remaining buffer time calculation module is used to calculate the current dynamic remaining buffer time based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

[0006] Thirdly, the technical solution of the present invention provides a terminal, comprising: The memory is used to store the program for determining the dynamic remaining buffer time for material delivery. The processor is configured to implement the steps of the material delivery dynamic remaining buffer time determination method as described above when executing the material delivery dynamic remaining buffer time determination program.

[0007] Fourthly, the present invention provides a computer-readable storage medium storing a dynamic remaining buffer time determination program for material delivery, wherein the dynamic remaining buffer time determination program for material delivery is executed by a processor to implement the steps of the dynamic remaining buffer time determination method for material delivery as described in any of the above claims.

[0008] As can be seen from the above technical solutions, this application has the following advantages: By combining order target characteristic data to determine the initial planned buffer time, it breaks the limitations of the traditional fixed threshold, realizes personalized adaptation of buffer time, and avoids resource waste caused by redundant buffering; Based on the deviation between the actual progress and the planned progress of each link in the supply chain, the cumulative consumed buffer time is calculated, and at the same time, real-time internal and external risk data is integrated to quantify the dynamic risk adjustment amount, so that the remaining buffer time can reflect the process progress and potential risk impact in real time, overcome the problem of the disconnect between traditional static calculation and actual scenario, and improve the authenticity and timeliness of time judgment; Through the rule engine or pre-trained risk adjustment amount prediction model, real-time internal and external risk data is analyzed and quantified, realizing the early identification and impact prediction of risk events, so that the buffer time can reserve risk response space in advance, changing the traditional passive risk response mode, and effectively reducing the probability of delivery delay; By integrating the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount, the intuitive and accurate dynamic remaining buffer time is output, providing a quantitative basis for decision-making such as progress monitoring, resource allocation, and risk intervention in supply chain management, optimizing management processes, and improving the reliability and efficiency of material delivery. Attached Figure Description

[0009] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1This is a schematic flowchart of a method for determining the dynamic remaining buffer time for material delivery, provided in an embodiment of the present invention.

[0011] Figure 2 This is a schematic block diagram of a dynamic remaining buffer time determination system for material delivery provided in an embodiment of the present invention.

[0012] Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0015] Figure 1 This is a flowchart illustrating a method for determining the dynamic remaining buffer time for material delivery, provided in an embodiment of the present invention. Figure 1 The executing entity can be a dynamic remaining buffer time determination system for material delivery. The dynamic remaining buffer time determination method for material delivery provided in this embodiment of the invention is executed by a computer device; correspondingly, the dynamic remaining buffer time determination system for material delivery runs on the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0016] like Figure 1 As shown, the method includes the following steps.

[0017] S1, Obtain the initial planned buffer time for the order.

[0018] S2 calculates the cumulative buffer time consumed based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain.

[0019] S3, based on real-time internal and external risk data, obtains dynamic risk adjustment amounts through a rule engine or a pre-trained risk adjustment amount prediction model.

[0020] S4. Calculate the current dynamic remaining buffer time based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

[0021] As a refinement and extension of the specific implementation of the above embodiments, in order to fully explain the specific implementation process of this embodiment, the following will provide possible embodiments to describe the specific implementation of the above steps in a non-limiting manner.

[0022] In some alternative implementations, step S1 obtains the initial planned buffer time for the order, specifically including the following steps.

[0023] S101, obtain target characteristic data of the order, including material complexity, transportation distance, and supplier historical performance indicators.

[0024] Material complexity is a score quantified by predefined rules based on the material's weight, volume, whether it is a hazardous material, and whether it requires constant temperature transportation. Supplier historical performance indicators are a comprehensive performance score based on the supplier's historical on-time delivery rate, average quality defect rate, and response and communication timeliness over a preset period.

[0025] (1) Quantification of material complexity Material complexity needs to be comprehensively considered in terms of material physical properties and special transportation requirements. It is quantified through predefined multi-dimensional scoring rules and finally outputs a complexity score of 0-100.

[0026] Extract basic attribute information of the current order materials from the enterprise ERP system and material management database, including material unit weight, total volume, whether it is a dangerous good, and whether it requires temperature-controlled transportation. An example of the dimensional scoring rules is shown in Table 1 below.

[0027] Table 1: Example of scoring rules for material complexity dimension

[0028] The total material complexity score is obtained by summing the scores of the above four dimensions. The formula is: Material Complexity Score = Weight Score + Volume Score + Hazardous Goods Attribute Score + Constant Temperature Requirement Score.

[0029] (2) Determination of transportation distance The transportation distance is the actual transportation path length of the ordered materials from the supplier's production base to the target delivery location. The system calls a third-party map service API, inputs the supplier's address and delivery address, selects the optimal route for the truck, and obtains the planned path length returned by the API.

[0030] (3) Calculation of supplier historical performance indicators Historical data from the target supplier was extracted from the Supply Chain Management (SCM) system, quality inspection system, and procurement communication record database, including historical on-time delivery rate, average defect rate, and response communication timeliness. Specifically, the historical on-time delivery rate was the number of orders delivered on time by the supplier over the past 12 months divided by the total number of delivered orders; the average defect rate was the cumulative number of defects in materials delivered by the supplier over the past 12 months divided by the cumulative number of deliveries; and the response communication timeliness was the average response time of the supplier to the buyer's communication requests over the past 12 months. An example of standardized scoring for these metrics is shown in Table 2 below.

[0031] Table 2: Examples of Standardized Scoring for Indicators

[0032] The standardized scores of the three dimensions are summed to obtain the total score of the supplier's historical performance indicators, namely, the supplier's historical performance indicators = on-time delivery rate score + quality defect rate score + response and communication timeliness score.

[0033] S102, the target feature data is input into a pre-trained initial buffer time prediction model to obtain the initial planned buffer time for the order; the initial buffer time prediction model is a neural network model.

[0034] The initial buffer time prediction model adopts a three-layer fully connected neural network model. After training with historical order data, it outputs a personalized initial plan buffer time by inputting target feature data. The specific process includes four stages: model building, data preprocessing, model training, and model application.

[0035] (1) Model building The initial buffer time prediction model in this embodiment includes the following structure.

[0036] Input layer: There are 3 neurons, which correspond to the three target features: "material complexity score", "transportation distance (km)" and "supplier historical performance index score". Hidden layers: Two hidden layers are set. The first hidden layer has 16 neurons and uses the ReLU activation function; the second hidden layer has 8 neurons and uses the ReLU activation function. Output layer: It has 1 neuron, and the output value is the initial planned buffer time. The activation function is the Linear function. Optimizer and Loss Function: The Adam optimizer is used (learning rate set to 0.001), and the mean squared error (MSE) loss function is used to minimize the deviation between the predicted buffer time and the actual buffer time.

[0037] (2) Training data preprocessing Valid order data from the past three years are selected from the company's historical order database. The data must include complete target feature data and actual buffer time data. More than 1,000 samples are collected. If the sample size is insufficient, it can be expanded by data augmentation or by supplementing historical data.

[0038] Each sample must include material complexity score, transportation distance, and supplier historical performance indicators, calculated according to the quantification rules in step S101. The actual buffer time for historical orders is defined as the difference between the actual total delivery time and the planned total delivery time.

[0039] The samples were cleaned and standardized.

[0040] (3) Model training and validation The preprocessed historical data was divided into training set, validation set, and test set in a ratio of 7:2:1.

[0041] The training epochs are set to 100, the batch size to 32, and the model is trained using the training set. At the same time, the model loss value is monitored through the validation set. If the loss value on the validation set does not decrease for 10 consecutive epochs, the early stopping mechanism is triggered to avoid model overfitting.

[0042] The model performance is evaluated using a test set, with key metrics including mean squared error (MSE) and mean absolute error (MAE). MSE requires a minimum of 2, meaning the average squared deviation between the predicted and actual buffer times is less than 2 days. MAE requires a minimum of 1, meaning the average absolute deviation between the predicted and actual buffer times is less than 1 day.

[0043] If the model performance does not meet the target, the network structure needs to be adjusted, the learning rate optimized, or additional training data added, and the model retrained until it meets the target.

[0044] (4) Model application The target feature data of the current order obtained by quantization in step S101 is standardized according to the standardization rules of the training phase. The standardized feature data is then input into the trained neural network model, and the model output layer outputs the initial planned buffer time for the current order.

[0045] In some optional implementations, step S2 calculates the cumulative consumed buffer time based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain, specifically including the following steps.

[0046] S201: For each predefined step in the supply chain process, obtain the planned time and actual time for that step.

[0047] S202, calculate the time consumption deviation for each step, whereby the time consumption deviation is the difference between the actual time consumed and the planned time consumed for that step.

[0048] S203, sum up the time deviations of all steps to obtain the total consumed buffer time.

[0049] Specifically, the system predefines a complete sequence of key supply chain links for each order. For example, for a power transformer, the sequence of links can be defined as: "raw material procurement, core component production, main assembly, in-plant testing, factory shipment, in-transit transportation, and on-site installation." The system collects the planned start and finish times, as well as the actual start and finish times, for each link in real time or at scheduled intervals through an integrated data interface.

[0050] For each completed step, the system calculates its time deviation, i.e., time deviation = actual time spent - planned time spent. When the time deviation is greater than 0, it indicates that the step was delayed, consuming buffer time; when the time deviation is less than or equal to 0, it indicates that the step was completed on time or ahead of schedule, without consuming buffer time. In a preferred embodiment of the invention, only positive deviations (i.e., delays) are included in the consumption to reflect the cumulative nature of risk. Then, the positive deviations of all completed steps are summed to obtain the cumulative buffer time consumed at the current moment.

[0051] In this embodiment, step S3 obtains the dynamic risk adjustment amount based on real-time internal and external risk data through a rule engine or a pre-trained risk adjustment amount prediction model, including the calculation methods of the dynamic risk adjustment amount by both the rule engine and the pre-trained risk adjustment amount prediction model.

[0052] Method 1: Obtain dynamic risk adjustment amount through the rules engine.

[0053] S301, construct a risk knowledge base, in which multiple risk event types are predefined, and corresponding data sources, feature extraction rules, and event recognition rules are configured for each risk event type.

[0054] Risk event types can include "road closures," "severe weather," "port congestion," and "production delays," among others. A structured risk knowledge base is maintained, predefined with the following content: Data source configuration: Specify which data sources to monitor. For example, for the "road interruption" event, the configured data sources are "XXX real-time traffic API" and "mainstream map service provider traffic event interface". Feature extraction rules: Define how to extract key features from the raw data, such as extracting fields like event_type, location, start_time, and expected_duration from the JSON data returned by the API; Event identification rules: Define the logical conditions that constitute a valid risk event. For example, a valid "road interruption" event must simultaneously satisfy: (event_type=="bridge repair" OR event_type=="major accident") AND (location is located within the buffer area of ​​the order's planned transportation route).

[0055] S302 obtains real-time heterogeneous data from internal and external data sources through a data interface.

[0056] Through pre-configured interfaces, real-time data can be pulled or received concurrently from various internal and external data sources. The format of this raw data may include JSON, XML, database row records, or text.

[0057] S303 performs feature extraction and standardization on real-time heterogeneous data to generate structured risk feature vectors.

[0058] Based on the feature extraction rules in the risk knowledge base, the collected raw data is parsed and cleaned.

[0059] Extract "warning type", "affected area", "wind force level" and "estimated start time" from meteorological API data.

[0060] Extract "planned working hours for the day", "actual working hours completed for the day", and "number of delayed orders" from the supplier's MES system data.

[0061] Extract "current average speed", "whether it has deviated from the planned route", and "duration of being stationary" from the logistics GPS data.

[0062] The extracted features are converted into a unified, standardized structured risk feature vector.

[0063] S304 matches the structured risk feature vector with the event identification rules to identify risk event instances.

[0064] The rule engine matches the structured risk feature vector generated in the previous step with all event identification rules in the risk knowledge base. The matching process is a logical judgment process. For example, the system matches the extracted "bridge maintenance" feature vector with the identification rule for the "road interruption" event. Since the feature satisfies all the conditions of the rule, the event is successfully identified.

[0065] When a match is successful, a specific risk event instance is created. This instance contains all relevant information, including the event type, identification time, and key features.

[0066] S305. For each risk event instance, based on its risk event type, query the predefined delay impact mapping relationship to obtain the expected delay time corresponding to the risk event.

[0067] S306 sums up the estimated delay times corresponding to all identified risk event instances to obtain the dynamic risk adjustment amount.

[0068] For each identified risk event instance, the system queries the predefined delay impact mapping relationship in the knowledge base according to its event type. For example, "bridge maintenance" is found to be +3 days after querying the mapping table; "typhoon" is calculated by the formula: wind force level * 0.5 to get +2.5 days, thereby calculating the corresponding expected delay time for each instance.

[0069] Finally, the estimated delay times corresponding to all active risk event instances are summed up to synthesize the dynamic risk adjustment amount.

[0070] Method 2: Obtain the dynamic risk adjustment amount through a pre-trained risk adjustment amount prediction model.

[0071] The risk adjustment prediction model is a deep learning-based temporal multimodal fusion network, whose architecture includes an input layer, a feature embedding and encoding layer, a multimodal feature fusion layer, and an output layer.

[0072] Input layer: Responsible for receiving and standardizing multi-source heterogeneous real-time and historical data. This layer includes multiple sub-input channels, each corresponding to a different type of data source.

[0073] Feature embedding and encoding layer: Located after the input layer, including: Temporal feature encoder: For temporal data, long short-term memory networks or temporal convolutional networks are used to extract temporal dependency features; Spatial Feature Encoder: For geospatial data, graph neural networks or spatial convolutional networks are used to extract spatial correlation features; Unstructured feature embedder: For text-based data, a natural language processing model is used to extract semantic features and output them as structured feature vectors; Structured feature processing module: For numerical and categorical structured data, embedding layers and fully connected layers are used for processing.

[0074] Multimodal feature fusion layer: The high-dimensional feature vectors output by the different encoders are concatenated and dimensionality reduced. Then, through an attention mechanism network, the contribution weight of different feature modalities to the final delay risk prediction is dynamically evaluated and weighted to form a unified and information-rich joint feature representation.

[0075] Output layer: Using one or more fully connected layers as prediction heads, the fused joint feature representation is mapped to the final prediction result. In this embodiment, the output layer is designed as a regression output or a staged predictive output. The regression output directly outputs a continuous value, namely the dynamic risk adjustment amount. The staged predictive output can output the expected delay time of each subsequent link in the supply chain, and then sum them to obtain the total dynamic risk adjustment amount.

[0076] During training, the parameters of each layer of the network are first initialized using a Xavier normal distribution to ensure gradient stability in the early stages of training. The training set is input into the model in batches, and the predicted values ​​are calculated using forward propagation. The deviation between the predicted values ​​and the true labels is calculated using a loss function. The gradient is calculated using the backpropagation algorithm (BP algorithm), and the network parameters of each layer are updated using the AdamW optimizer. After each training epoch, the model loss value is calculated using the validation set, and the model parameters corresponding to the optimal loss are recorded.

[0077] A cosine annealing learning rate scheduling strategy is adopted, in which the learning rate is periodically adjusted with each training epoch to avoid getting stuck in local optima in the later stages of training. Dropout layers are added to fully connected layers to randomly deactivate some neurons, reducing redundant dependencies between features; layer normalization is applied to LSTM and GCN layers to accelerate model convergence and improve generalization ability; a regularization term is introduced, and L2 regularization is added to the loss function to constrain the network parameter size.

[0078] The model performance was evaluated using multi-dimensional indicators, including MSE < 4 (mean squared deviation between predicted and actual delay times is less than 4 hours), MAE < 1.5 (mean absolute deviation between predicted and actual delay times is less than 1.5 hours), and the coefficient of determination R0. 2 A value >0.85 indicates that the model can explain more than 85% of the risk delay variation; the maximum absolute error MaxAE is less than 6 hours, avoiding excessive prediction bias in extreme risk scenarios.

[0079] The five-fold cross-validation method is adopted, which divides the training set and the validation set into five non-overlapping subsets. Four subsets are used as the training set and one subset is used as the validation set. The training is repeated five times, and the average of the five validation results is calculated as the final performance index to ensure that the model performance is stable and reliable.

[0080] Based on a pre-trained risk adjustment prediction model, the dynamic risk adjustment is obtained through the following steps.

[0081] S311 inputs real-time internal and external risk data into the input layer of the risk adjustment prediction model to obtain standardized data to be processed, which includes time series data, spatial geographic data, text data and structured data.

[0082] The model acquires real-time and historical data from both internal and external data sources. Internal data sources include the enterprise's ERP (Enterprise Resource Planning), MES (Manufacturing Execution System), TMS (Transportation Management System), and WMS (Warehouse Management System). The provided data includes: real-time order progress status, supplier production reports, complete lifecycle data of historical orders (for training and reference), and logistics GPS track point sequences. External data sources are public data obtained through API interfaces, including: real-time traffic conditions, weather forecasts and warnings, port / airport congestion indices, and news event streams related to supply chain nodes.

[0083] The input layer performs operations such as standardization, noise reduction, timestamp alignment, and missing value processing on the collected raw data to form a multimodal input tensor.

[0084] S312, the input features of the data to be processed are embedded in the encoding layer, and features are extracted from time series data, spatial geographic data, text data and structured data respectively to obtain the corresponding time series features, spatial features, text features and structured features.

[0085] The preprocessed data is fed into the corresponding modules of the feature embedding and encoding layers. Real-time and historical logistics GPS trajectories and production reporting sequences are fed into the temporal feature encoder; the current location of the goods and key node information on the planned route are fed into the spatial feature encoder; the acquired news text is fed into the unstructured feature embedder; and the current congestion index, supplier performance score, etc., are fed into the structured feature processing module. Each encoder works in parallel, extracting high-level abstract feature vectors from different dimensions.

[0086] S313 uses a multimodal feature fusion layer to weightedly fuse temporal features, spatial features, textual features, and structured features to generate a joint feature representation.

[0087] All feature vectors are fed into a multimodal feature fusion layer. The attention mechanism in this layer dynamically calculates the importance weight of each feature vector. For example, for orders in transit, the model may assign higher weights to "real-time traffic" and "weather" features; while for orders still in production, it will focus more on "supplier production reports" and "quality inspection" data.

[0088] S314, based on joint feature representation, obtains the dynamic risk adjustment amount through output layer prediction.

[0089] The fused joint feature representation is fed into the output layer, which performs a nonlinear transformation on the joint feature representation, ultimately outputting a scalar value. This value is the model's prediction of the total expected delay time in the future due to current and foreseeable risks, which is also known as the dynamic risk adjustment.

[0090] In some optional implementations, step S4 sums up the time deviations of all stages to obtain the cumulative consumed buffer time. The time deviation of each stage obtained in the previous steps represents the degree of deviation between the actual execution of each stage and the original plan. The algebraic summation of the time deviations of all completed stages reflects the transmission and cumulative effect of supply chain risk. Specifically, the system monitors the status updates of each supply chain stage. Whenever a stage is marked as "completed," the system automatically triggers the deviation calculation for that stage and immediately updates the cumulative value.

[0091] In some optional implementations, the method also identifies and quantifies positive factors that may shorten delivery time in the future as dynamic opportunity gains, and the calculation function for dynamic remaining buffer time further includes the dynamic opportunity gains as a positive adjustment factor. Specifically, it includes the following steps.

[0092] a) Identify predefined positive opportunity events by monitoring the status data of the supply chain execution system and external data sources.

[0093] Maintain a database of opportunity events, which predefines various types of positive opportunity events and their identification rules.

[0094] Opportunity events include production schedule ahead of schedule and logistics route optimization.

[0095] Through data interfaces, monitor the following data sources to trigger identification. Specifically, obtain real-time work reporting data from the supplier's MES, production order status from the ERP, and planned and executed routes from the TMS. Obtain alternative route recommendation interfaces from logistics service providers.

[0096] The monitored real-time data is matched with rules in the opportunity event library to identify specific positive opportunity event instances. The current completion percentage and current system time of orders in the MES are obtained. By querying the planned progress curve of the order (which defines the expected completion percentage at any given time), if the current completion percentage > the expected percentage corresponding to the planned progress curve, a production schedule ahead-of-time event is determined to have been triggered.

[0097] By monitoring the TMS interface to see if there are confirmed alternative routes that are expected to take less time than the original planned route, if such route changes are detected, it is determined that a logistics route optimization event has been triggered.

[0098] b) For each identified positive opportunity event, the expected time savings are obtained based on predefined rules.

[0099] For each identified opportunity event instance, the expected time savings are calculated using a predefined quantification model. Regarding "production ahead of schedule," a regression prediction model is used to predict the final production completion time based on the current percentage ahead and remaining workload. The predicted completion time is compared to the planned completion time for that stage, and the expected time savings are the difference between the two.

[0100] Regarding "logistics route optimization", the estimated transportation time of the new route and the estimated transportation time of the original route are obtained directly from TMS, and the expected time saving is calculated as the difference between the two.

[0101] c) Sum the expected time savings corresponding to all identified positive opportunity events to obtain the dynamic opportunity gain.

[0102] Accordingly, when calculating the current dynamic remaining buffer time, the dynamic opportunity gain is also used as a positive adjustment factor. Specifically, the formula for calculating the dynamic remaining buffer time that integrates the opportunity gain is as follows: Dynamic Remaining Buffer Time = Initial Planned Buffer Time - Cumulative Consumed Buffer Time - Dynamic Risk Adjustment + Dynamic Opportunity Gain Among them, the dynamic opportunity gain, as a positive adjustment factor in the formula, together with the dynamic risk adjustment, as a negative adjustment factor, constitutes a two-way dynamic correction to the baseline remaining time.

[0103] The above text describes in detail an embodiment of a method for determining the dynamic remaining buffer time for material delivery. Based on the method for determining the dynamic remaining buffer time for material delivery described in the above embodiment, this invention also provides a system for determining the dynamic remaining buffer time for material delivery corresponding to the method.

[0104] Figure 2 This is a schematic block diagram of a dynamic remaining buffer time determination system for material delivery provided in an embodiment of the present invention. In this embodiment, the dynamic remaining buffer time determination system 200 for material delivery can be divided into multiple functional modules according to the functions it performs, such as... Figure 2 As shown. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and which are stored in memory.

[0105] The initial plan buffer time acquisition module 210 is used to acquire the initial plan buffer time of an order.

[0106] The consumed buffer time calculation module 220 is used to calculate the cumulative consumed buffer time based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain.

[0107] The dynamic risk adjustment generation module 230 is used to obtain dynamic risk adjustment based on real-time internal and external risk data through a rule engine or a pre-trained risk adjustment prediction model.

[0108] The dynamic remaining buffer time calculation module 240 is used to calculate the current dynamic remaining buffer time based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

[0109] In some optional implementations, system 200 further includes a dynamic opportunity gain calculation module 250, used to identify the existence of predefined positive opportunity events by monitoring the status data of the supply chain execution system and external data sources; for each identified positive opportunity event, the expected time saving is obtained based on predefined rules; and the expected time saving corresponding to all identified positive opportunity events is accumulated to obtain the dynamic opportunity gain. Positive opportunity events include supplier production progress exceeding the planned schedule and the use of a better-than-planned transportation mode or route in the logistics process. The expected time saving for the event "supplier production progress exceeding the planned schedule" is calculated as follows: based on real-time work reporting data from the production management system, the final production completion time is predicted and compared with the planned completion time; the difference between the two is the expected time saving. The expected time saving for the event "the use of a better-than-planned transportation mode or route in the logistics process" is calculated as follows: the difference between the estimated transportation time of the original planned transportation route and the estimated transportation time of the newly used transportation route is calculated; the difference between the two is the expected time saving.

[0110] Correspondingly, when the dynamic remaining buffer time calculation module 240 calculates the current dynamic remaining buffer time, it also calculates the dynamic opportunity gain as a positive adjustment factor.

[0111] The material delivery dynamic remaining buffer time determination system of this embodiment is used to implement the aforementioned material delivery dynamic remaining buffer time determination method. Therefore, the specific implementation of this system can be found in the embodiment section of the material delivery dynamic remaining buffer time determination method above. So, its specific implementation can be referred to the description of the corresponding embodiment, and will not be described in detail here.

[0112] Furthermore, since the material delivery dynamic remaining buffer time judgment system in this embodiment is used to implement the aforementioned material delivery dynamic remaining buffer time judgment method, its function corresponds to the function of the above method, and will not be repeated here.

[0113] Figure 3This is a schematic diagram of a terminal 300 provided in an embodiment of the present invention, including: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the process steps of the above-mentioned dynamic remaining buffer time determination method for material delivery when implementing the material delivery dynamic remaining buffer time determination program stored in the memory 320.

[0114] This invention also provides a computer storage medium, which may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. The computer storage medium stores a dynamic remaining buffer time determination program for material delivery. When the processor executes the dynamic remaining buffer time determination program for material delivery, it implements the process steps of the aforementioned dynamic remaining buffer time determination method for material delivery.

[0115] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining the dynamic remaining buffer time for material delivery, characterized in that, Includes the following steps: Initial planned buffer time for order acquisition; Calculate the cumulative buffer time consumed based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain. Based on real-time internal and external risk data, dynamic risk adjustment is obtained through a rule engine or a pre-trained risk adjustment prediction model. The current dynamic remaining buffer time is calculated based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

2. The method for determining the dynamic remaining buffer time for material delivery according to claim 1, characterized in that, The initial planned buffer time for obtaining orders includes: Obtain target characteristic data for orders, including material complexity, transportation distance, and supplier historical performance indicators; The target feature data is input into a pre-trained initial buffer time prediction model to obtain the initial planned buffer time for the order; the initial buffer time prediction model is a neural network model. Material complexity is a score quantified by predefined rules based on the material's weight, volume, whether it is a hazardous material, and whether it requires constant temperature transportation. Supplier historical performance indicators are a comprehensive performance score based on the supplier's historical on-time delivery rate, average quality defect rate, and response and communication timeliness over a preset period.

3. The method for determining the dynamic remaining buffer time for material delivery according to claim 1, characterized in that, Based on the deviation between the actual and planned progress of orders at each stage of the supply chain, the cumulative consumed buffer time is calculated, specifically including: For each predefined step in the supply chain process, obtain the planned time and actual time spent on that step. Calculate the time consumption deviation for each step, whereby the time consumption deviation is the difference between the actual time consumed and the planned time consumed for that step. The total buffer time consumed is obtained by summing up the time deviations of all steps.

4. The method for determining the dynamic remaining buffer time for material delivery according to claim 1, characterized in that, Based on real-time internal and external risk data, a dynamic risk adjustment amount is obtained through a rules engine, specifically including: Build a risk knowledge base, which predefines multiple risk event types and configures corresponding data sources, feature extraction rules, and event recognition rules for each risk event type; Real-time heterogeneous data is obtained from internal and external data sources through data interfaces; Feature extraction and standardization are performed on real-time heterogeneous data to generate structured risk feature vectors; The structured risk feature vector is matched with the event identification rules to identify risk event instances; For each risk event instance, based on its risk event type, query the predefined delay impact mapping relationship to obtain the expected delay time corresponding to the risk event; The estimated delay times corresponding to all identified risk event instances are summed to obtain the dynamic risk adjustment amount.

5. The method for determining the dynamic remaining buffer time for material delivery according to claim 4, characterized in that, Based on real-time internal and external risk data, dynamic risk adjustment is obtained through a pre-trained risk adjustment prediction model, specifically including: Real-time internal and external risk data are input into the input layer of the risk adjustment prediction model to obtain standardized data to be processed, which includes time series data, spatial geographic data, text data and structured data. The input features of the data to be processed are embedded in the encoding layer, and features are extracted from time series data, spatial geographic data, text data and structured data respectively to obtain the corresponding time series features, spatial features, text features and structured features; A multimodal feature fusion layer is used to weight and fuse temporal features, spatial features, text features, and structured features to generate a joint feature representation. Based on joint feature representation, the dynamic risk adjustment amount is obtained by predicting through the output layer.

6. The method for determining the dynamic remaining buffer time for material delivery according to claim 1, characterized in that, The method also includes: By monitoring the status data of the supply chain execution system and external data sources, we can identify whether there are predefined positive opportunity events. For each identified positive opportunity event, the expected time savings are obtained based on predefined rules; The expected time savings corresponding to all identified positive opportunity events are summed up to obtain the dynamic opportunity gain. Among these, positive opportunities include suppliers producing ahead of schedule and the use of better-than-planned transportation modes or routes in the logistics process. The expected time savings for the event "supplier production progress is ahead of schedule" are calculated as follows: based on the real-time work reporting data of the production management system, the final production completion time is predicted and compared with the planned completion time. The difference between the two is the expected time savings. The expected time savings for the event "a better transportation mode or route than originally planned was used in the logistics process" are calculated as follows: calculate the difference between the estimated transportation time of the original planned transportation route and the estimated transportation time of the newly used transportation route. The difference between the two is the expected time savings.

7. The method for determining the dynamic remaining buffer time for material delivery according to claim 6, characterized in that, When calculating the current dynamic remaining buffer time, the dynamic opportunity gain is also used as a positive adjustment factor.

8. A system for determining the dynamic remaining buffer time for material delivery, characterized in that, include: The initial plan buffer time acquisition module is used to obtain the initial plan buffer time of an order. The consumed buffer time calculation module is used to calculate the cumulative consumed buffer time based on the deviation between the actual progress and the planned progress of the order at each stage of the supply chain. The dynamic risk adjustment generation module is used to obtain dynamic risk adjustment based on real-time internal and external risk data, through a rule engine or a pre-trained risk adjustment prediction model. The dynamic remaining buffer time calculation module is used to calculate the current dynamic remaining buffer time based on the initial planned buffer time, the cumulative consumed buffer time, and the dynamic risk adjustment amount.

9. A terminal, characterized in that, include: The memory is used to store the program for determining the dynamic remaining buffer time for material delivery. The processor is configured to implement the steps of the material delivery dynamic remaining buffer time determination method as described in any one of claims 1 to 7 when executing the material delivery dynamic remaining buffer time determination program.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores a dynamic remaining buffer time determination program for material delivery, which, when executed by a processor, implements the steps of the dynamic remaining buffer time determination method for material delivery as described in any one of claims 1 to 7.

Images