Steel manufacturing scene-based quasi-issuing information prediction method and system, storage medium and program product

Abstract

This invention provides a method, system, storage medium, and program product for predicting on-time delivery information in a steel manufacturing scenario, relating to the field of intelligent logistics management technology. The method includes: responding to a query command for a target product, obtaining information on materials in use; inputting the information on materials in use into a manufacturing cycle prediction model to obtain the manufacturing time for the target product based on the materials in process, and obtaining initial on-time delivery information for the target product; obtaining the maximum daily on-time delivery quantity of the target product within a preset future time period through the on-time delivery prediction model; adjusting the initial on-time delivery information based on the maximum daily on-time delivery quantity within the preset future time period, generating a set of expected on-time delivery times and daily expected on-time delivery quantities, and obtaining the expected on-time delivery information for the target product. The embodiments of this invention exhibit strong cross-process adaptability and fast response speed in long-process scenarios.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics management technology, and in particular to a method, system, storage medium, and program product for predicting delivery information based on the steel manufacturing scenario. Background Technology

[0002] The steel manufacturing industry is a typical long-process industry, characterized by long product manufacturing cycles, numerous procedures, and complex production environments. Accurately predicting product delivery information, including delivery volume and delivery time, is crucial for developing reasonable production plans, meeting customer delivery deadlines, reducing work-in-process inventory, and improving operational efficiency.

[0003] Currently, traditional methods for predicting on-time delivery information rely on planners' personal experience for manual judgment and data entry, using only Excel spreadsheets or simple enterprise resource planning (ERP) systems as record-keeping aids. This approach suffers from low digitization and high subjectivity. Intelligent methods, on the other hand, generate accurate production plans and on-time delivery time predictions through advanced planning and scheduling systems. However, these systems typically require building and training highly specialized and complex independent stage models for different processes such as hot rolling and casting. These independent stage models are trained only on historical data for their specific process. When dynamic changes occur in the production environment, such as equipment failures, emergency orders, or raw material changes, their response speed is slow due to a lack of cross-process correlation knowledge.

[0004] Therefore, when the above-mentioned method of predicting and distributing information is applied to the long-process scenario of steel manufacturing, its adaptability and response speed are insufficient, which can easily lead to large deviations in the prediction, resulting in problems such as delivery delays or idle production resources. Summary of the Invention

[0005] To address the shortcomings of existing technologies, embodiments of the present invention provide a method, system, storage medium, and program product for predicting delivery information based on the steel manufacturing scenario. This solves the problems that existing methods for predicting delivery information are unable to meet the requirements of cross-process adaptive capabilities in the long-process scenario of steel manufacturing, and that the overall response speed is slow.

[0006] According to an embodiment of the present invention, a first aspect provides a method for predicting delivery information based on a steel manufacturing scenario, comprising: In response to a query command for a target product, information on materials to be used is obtained. The information on materials to be used includes work-in-process materials used to manufacture the target product and the manufacturing stage of the work-in-process materials. The target product is a product within a preset product category. The information on materials to be used is input into the manufacturing cycle prediction model to obtain the manufacturing time of the target product based on the work-in-process materials, and to obtain the initial ready-to-ship information of the target product. The manufacturing cycle prediction model is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity. The maximum daily shipment volume of the target product within a preset future time period is obtained through a pre-delivery prediction model. The initial issuance information is adjusted based on the maximum daily issuance volume within the preset future time period to generate a set of expected issuance times and a daily expected issuance volume, thereby obtaining the expected issuance information for the target product.

[0007] Optionally, the method for predicting delivery information based on the steel manufacturing scenario also includes: A basic prediction model is constructed based on the XGBoost algorithm; The manufacturing cycle prediction model is obtained by training the basic prediction model using the full-process manufacturing data. The full-process manufacturing data includes the manufacturing cycle and key influencing factors affecting the manufacturing cycle; When the manufacturing cycle prediction model identifies the information of the material to be used, it calculates the manufacturing time for obtaining the target product based on the work-in-process material according to the manufacturing stage of the work-in-process material.

[0008] Optionally, the manufacturing cycle prediction model includes a preparation cycle prediction model for the product before manufacturing begins, a process cycle prediction model for the product during manufacturing, and an evaluation cycle prediction model for the product after manufacturing is completed. When the manufacturing cycle prediction model identifies the information of the materials to be used, it calculates the manufacturing time for obtaining the target product based on the work-in-process material according to the manufacturing stage of the work-in-process material, including: The manufacturing time is calculated by calling one or more of the preparation cycle prediction model, the process cycle prediction model, and the evaluation cycle prediction model according to the manufacturing stage of the target product and the work-in-process material; wherein, when calling the process cycle prediction model, different branches and different nodes in the process cycle prediction model are also called according to the product category of the target product.

[0009] Optionally, the products within the preset product range include products involved in the sheet metal production route, products involved in the bar and wire production route, products involved in the profile production route, and products involved in the pipe production route. The product categories are divided according to the production route, the number of processing steps, and the process of each processing step, and are classified as primary materials and multiple materials in the process cycle prediction model; The process cycle prediction model includes a first process cycle prediction branch based on the initial processing flow, and an N+1th process cycle prediction branch based on N processing cycles, where N is a positive integer greater than or equal to 1. The step of calling different branches in the process cycle prediction model based on the product category of the target product includes: When the product category is a single material, the first process cycle prediction branch is invoked; When the product category is multi-material, at least the first process cycle prediction branch is called, and at most the first process cycle prediction branch to the N+1th process cycle prediction branch is called. When calling the second process cycle prediction branch to any one of the N+1th process cycle prediction branches, different nodes are called according to the product category of the target product itself.

[0010] Optionally, obtaining the maximum daily allowable shipment volume of the target product within a preset future time period through the allowable shipment prediction model includes: The maximum daily issuance volume within the preset future time period is output based on the target product and the issuance prediction sub-model. The on-time delivery prediction model includes multiple on-time delivery prediction sub-models. For each on-time delivery prediction sub-model, there is a unique product category, and the model stores the shipping date, optimal transportation route, and maximum daily on-time delivery volume corresponding to that product category. The product categories mentioned are classified as plates, bars and wires, profiles and pipes in the quasi-delivery prediction model.

[0011] Optionally, the method for predicting delivery information based on the steel manufacturing scenario also includes: Record the actual manufacturing time of the work-in-process material; If the error between the actual manufacturing time and the actual manufacturing time is greater than a preset error range, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model.

[0012] Optionally, the method for predicting delivery information based on the steel manufacturing scenario also includes: If the error between the last day of the expected delivery time set and the last day of the initial delivery time set is greater than a preset error number of days, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model and / or the delivery prediction sub-model that outputs outliers in the delivery prediction model.

[0013] The second aspect provides a prediction system for timely delivery information based on the steel manufacturing scenario, including: The material information acquisition module is used to acquire information on materials to be used. The information on materials to be used includes work-in-process materials used to manufacture the target product and the manufacturing stage of the work-in-process materials. The target product is a product within a preset product range. The cycle prediction module is used to input the information of the materials to be used into the manufacturing cycle prediction model, obtain the manufacturing time of the target product based on the work-in-process materials, and obtain the initial ready-to-ship information of the target product. The manufacturing cycle prediction model is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity. The pre-delivery prediction module is used to obtain the maximum daily pre-delivery volume of the target product within a preset time period through the pre-delivery prediction model; The issuance adjustment and output module is used to adjust the initial issuance information according to the maximum daily issuance volume within the preset future time period, generate a set of expected issuance times and a daily expected issuance volume, and obtain the expected issuance information of the target product.

[0014] A third aspect provides a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the aforementioned method for predicting quasi-issuance information based on a steel manufacturing scenario.

[0015] The fourth aspect provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for predicting quasi-issuance information based on a steel manufacturing scenario.

[0016] Compared to existing technologies, the embodiments of the present invention have the following beneficial effects: By working collaboratively with the manufacturing cycle prediction model and the on-time delivery prediction model, the expected on-time delivery information of the target product can be obtained quickly. This allows for the formulation of reasonable production plans based on the expected on-time delivery information, thereby meeting customer delivery deadlines, reducing work-in-process inventory, and improving enterprise operational efficiency. Furthermore, the manufacturing cycle prediction model used in the embodiments of the present invention is trained based on full-process manufacturing data, covering all manufacturing stages of products within a preset product category. For any product within the preset product category, such as the target product, it shares basic data from different processes. Therefore, the manufacturing cycle prediction model has strong cross-process adaptive capabilities in long-process scenarios and can supplement cross-process related knowledge to improve response speed when the production environment changes dynamically. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process flow for products within the preset product category of this invention. Figure 2This is a schematic diagram illustrating the implementation process of the method for predicting the issuance information based on a steel manufacturing scenario according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the composition structure of the steel manufacturing scenario-based information prediction system according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions in the embodiments of the present invention will be further described below with reference to the accompanying drawings and examples.

[0019] This invention proposes a method for predicting on-time delivery information based on a steel manufacturing scenario. It provides on-time delivery information predictions for products within a preset product category, including products involved in plate production routes, bar and wire production routes, profile production routes, and pipe production routes. In practical applications, the method can query the estimated on-time delivery information of any product, such as the target product, based on orders that have not yet been processed, to formulate a production plan. Alternatively, it can query the estimated on-time delivery information of the target product based on orders that are currently being processed, to determine if the orders are in a delayed state, and thus identify production delay nodes.

[0020] In this embodiment of the invention, the products within the preset product category include products involved in the sheet metal production route, products involved in the bar and wire rod production route, products involved in the profile production route, and products involved in the pipe production route. The aforementioned production routes also represent product categories in this embodiment of the invention. Different product categories have different production routes, and within the same production route, the number of processing steps and the use of the same or different processes for each processing step are also considered. Therefore, product categories in this embodiment of the invention are divided according to the production route, the number of processing steps, and the process used for each processing step. Taking products involved in the sheet metal production route as an example, such as... Figure 1As shown, based on different processing times and different processes for each processing step, five types of products can be obtained: galvanized coils / sheets, color-coated coils / sheets, silicon steel coils / sheets, commercial cold-rolled coils / sheets, and medium-thick plates. In other words, based on the production route of sheet metal and the different processing times and different processes for each processing step, the product categories include: galvanized products, color-coated products, silicon steel products, cold-rolled products, and medium-thick plates. Among them, galvanized coils / sheets belong to the galvanized product category, color-coated coils / sheets belong to the color-coated product category, silicon steel coils / sheets belong to the silicon steel product category, commercial cold-rolled coils / sheets belong to the cold-rolled product category, and medium-thick plates belong to the medium-thick plate product category. In specific steel manufacturing scenarios, after initial processing, such as steelmaking / continuous casting, the finished product from the initial processing—the cast billet—can be directly obtained. The shape of the cast billet can be defined during the initial processing, and medium-thick plates can be directly obtained from one shape. If the cast billet undergoes primary and secondary processing, such as hot rolling, the finished products from secondary processing can be obtained: silicon steel coils / plates and commercial cold-rolled coils / plates. Both silicon steel coils / plates and commercial cold-rolled coils / plates require primary processing, i.e., hot rolling, followed by secondary processing using different techniques to obtain silicon steel coils / plates and commercial cold-rolled coils / plates, respectively. If the cold-rolled coils / plates undergo tertiary processing, such as galvanizing, the finished product from tertiary processing can be obtained: galvanized coils / plates. If the galvanized coils / plates undergo quaternary processing, such as color coating, the finished product from quaternary processing can be obtained: color-coated coils / plates.

[0021] The different products obtained from the production routes of bars, wires, profiles, and tubes, as well as the resulting product categories, will not be elaborated here. They can be determined based on the actual production routes, processing times, and processing techniques.

[0022] Based on this, such as Figure 2 As shown, the method for predicting the issuance information based on the steel manufacturing scenario in this embodiment of the invention includes, but is not limited to, the following steps: S101. In response to the query command for the target product, obtain information on the materials to be used; S102. Input the information of the materials to be used into the manufacturing cycle prediction model to obtain the manufacturing time of the target product based on the work-in-process materials, and obtain the initial ready-to-ship information of the target product. S103. Obtain the maximum daily shipment volume of the target product within a preset time period using the shipment prediction model; S104. Adjust the initial issuance information according to the maximum daily issuance volume within the future preset time period, generate the expected issuance time set and the daily expected issuance volume, and obtain the expected issuance information of the target product.

[0023] Regarding step S101 above, in this embodiment of the invention, the information on materials to be used includes work-in-process materials for manufacturing the target product and the manufacturing stage of the work-in-process materials, wherein the target product is a product within a preset product category. (See also...) Figure 1 The target product can be a product involved in the sheet metal production route, such as galvanized coils / plates, color-coated coils / plates, silicon steel coils / plates, commercial cold-rolled coils / plates, and medium-thick plates, or it can be a product involved in the bar and wire rod production route, a product involved in the profile production route, or a product involved in the pipe production route.

[0024] It should be noted that there are multiple work-in-process materials for the target product. For example, assuming that the current manufacturing process includes multiple work-in-process materials for producing the target product, then there are multiple sets of material information to be used. For the i-th set of material information to be used, it includes the i-th set of work-in-process materials for producing the target product and the manufacturing stage of the i-th set of work-in-process materials.

[0025] For step S102 above, the manufacturing cycle prediction model provided in this embodiment of the invention is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity.

[0026] It is understandable that if there are multiple target products in production, then the initial release time set represents the time when one target product is obtained or several target products are obtained at the same time within a future period. If the range of the future period is set to 9:00 to 17:00 of the current day, then the number of target products obtained in this future period is the daily initial release quantity.

[0027] For example, for the kth part of work-in-process material, the kth target product M can be obtained based on the kth part of work-in-process material. k The manufacturing time is expressed as tM i The initial release information for the target product includes TM. k Assume tM k =1h, the response time to the query command for the target product is 12:00, then TM k =13:00. Based on this, the initial release time set for multiple target products can be represented as {13:00, 13:01, 13:02, 13:03, 13:04……17:00}, where one or more target products are obtained at each time point.

[0028] Regarding step S103 above, the delivery prediction model provided in this embodiment of the invention is used to calculate the maximum daily delivery volume based on the actual logistics situation. It can be understood that the latest date of the future preset time is later than the latest date of the initial delivery time set, and the earliest date of the future preset time is earlier than the earliest date of the initial delivery time set.

[0029] In one embodiment, obtaining the maximum daily permissible quantity of the target product within a preset future time period through a permissible issuance prediction model includes: The maximum daily issuance volume within the preset future time period is output based on the target product and the issuance prediction sub-model. The on-time delivery prediction model, as shown in Table 1, includes multiple on-time delivery prediction sub-models. Each sub-model corresponds to a unique product category and stores the corresponding shipping date, optimal transportation route, and maximum daily on-time delivery volume. It should be noted that the target product can be divided into different product categories and classified according to different standards in different models. In this on-time delivery prediction model, the product categories are classified as plates, bars and wires, profiles, and pipes. Furthermore, shipping dates can be discontinuous; for example, if only one of the next two days is a shipping day, the maximum on-time delivery volume for the other day is 0.

[0030] Table 1

[0031] For example, the galvanized coils / sheets, color-coated coils / sheets, silicon steel coils / sheets, commercial cold-rolled coils / sheets, and medium-thick plates mentioned above are products involved in the sheet metal production route, and therefore are all classified as sheet metal. In addition, for example, in practical applications, continuously rolled billets, square and round billets, high-strength wire rods, ordinary wire rods, and special-purpose wire rods are products involved in the wire rod production route, and are all classified as wire rods.

[0032] The embodiments of the present invention further illustrate the above table with practical application: If the initial set of approved shipment times for the target product covers the next 3 days, and the next 2 days are all shipment days for the target product, then the approved shipment prediction model outputs the maximum daily approved shipment volume of the target product for the next 1 day, 2-5 days according to the optimal transportation route and shipment day. For example, the XX route XX category is expected to be approved for shipment of 10,200 tons on the first day, 9,800 tons on the second day, and 0 tons on the third day.

[0033] For step S104 above, the initial approval information includes the daily initial approval quantity and the initial approval time set. Therefore, adjusting the initial approval information according to the daily maximum approval quantity within the future preset time is to adjust the daily initial approval quantity and the initial approval time set.

[0034] In step S102 above, the times in the initial set of approved launch times can be accurate to the hour or minute. However, the future preset times shown by the launch prediction model in step S103 are usually only accurate to the time interval, such as 9:00~12:00, 13:00~17:00, etc., and more often accurate to the date interval, such as the next 2 days, the next 3 days, etc. For example, the first day of the future is January 1st, and the second day of the future is January 2nd. Therefore, in step S104 above, the times in the expected set of approved launch times should be consistent with the launch prediction model, accurate to each time interval or date interval.

[0035] Taking a date range as an example, in practical applications, there may be situations where the initial allowable quantity for a certain day is greater than the maximum allowable quantity for that day. In this case, the maximum allowable quantity for the day needs to be used as the expected allowable quantity for that day. The remaining target products can be added to the initial allowable quantity for the next day. If the maximum allowable quantity for the next day is greater than the initial allowable quantity after adding, then the expected allowable quantity for the next day is the initial allowable quantity after adding. If the maximum allowable quantity for the next day is still less than the initial allowable quantity after adding, then the expected allowable quantity for the next day is still the maximum allowable quantity for the next day, until the maximum allowable quantity for a future day can send out all the remaining target products.

[0036] It is understandable that, since the latest date of the future preset time is later than the latest date of the initial approved release time set, and the earliest date of the future preset time is earlier than the earliest date of the initial approved release time set, there must exist a maximum release quantity on a future day that can release all the remaining target products.

[0037] Through steps S101 to S104 above, this embodiment of the invention rapidly obtains the expected delivery information of the target product through the collaborative work of the manufacturing cycle prediction model and the on-time delivery prediction model. This allows for the formulation of a reasonable production plan based on the expected delivery information to meet customer delivery deadlines, reduce work-in-process inventory, and improve enterprise operational efficiency. Furthermore, the manufacturing cycle prediction model used in this embodiment of the invention is trained based on full-process manufacturing data, covering all manufacturing stages of products within a preset product category. For any product within the preset product category, such as the target product, it shares basic data from different processes. Therefore, the manufacturing cycle prediction model has strong cross-process adaptive capabilities in long-process scenarios and can supplement cross-process related knowledge to improve response speed when the production environment dynamically changes.

[0038] This invention describes the training of a manufacturing cycle prediction model, including: A basic prediction model is constructed based on the XGBoost algorithm; The manufacturing cycle prediction model is obtained by training the basic prediction model using the full-process manufacturing data. The full-process manufacturing data includes the manufacturing cycle and key influencing factors affecting the manufacturing cycle.

[0039] It should be noted that the full-process manufacturing data includes the manufacturing cycle and the key influencing factors that affect the manufacturing cycle. This means that for any product within the preset product range, such as the target product, each manufacturing stage that the work-in-process material of the target product goes through corresponds to a manufacturing cycle, and there are also key influencing factors that affect the manufacturing cycle.

[0040] This invention also illustrates a detailed implementation of a manufacturing cycle prediction model obtained by constructing a basic prediction model based on the XGBoost algorithm and training the basic prediction model with full-process manufacturing data, including: A. Initialize multiple prediction models for multiple manufacturing stages. For a certain manufacturing stage, a prediction model is obtained by initializing the average manufacturing cycle value corresponding to that manufacturing stage.

[0041] B. For each prediction model, use all key influencing factors as input variables, all manufacturing cycles as output variables, and the target product as the independent variable, and iterate through the following steps: b1. Calculate the first and second gradients of the objective function for the current model; b2. Using information from the gradient and second derivative, recursively split the node with the goal of maximizing the gain. b3. Generate a structure tree based on the split node with the maximum gain; C. Add the new tree obtained in each iteration to the overall model and multiply it by the learning rate to prevent the step size from being too large; when stopping the iteration and outputting the final model, add the prediction results of all M trees to get Final Model = Tree1 + Tree2 + ... + Tree_M.

[0042] The gains shown in steps b2 and b3 above are calculated using the same objective function as in step b1. For example, the objective function includes a regularization term, expressed as: Obj = Loss + regularization; In step B above, the accuracy of the model is evaluated using functions such as mean absolute error (MAE), mean squared error (MSE), and coefficient of determination (R²). 2 To monitor the training process and determine when to stop iterations, for example, MAE, MSE, and R2 are represented as follows: ; ; ; Among them, y iy represents the true manufacturing cycle value corresponding to the training samples used to train a prediction model. i 'This is the manufacturing cycle prediction value output by the prediction model based on the same training samples. SST is the sum of squares of the deviations between the actual value and the mean of the dependent variable, reflecting the total variation of the data. SSR is the sum of squares of the deviations between the model prediction value and the mean, reflecting the variation explained by the model; SSE is the sum of squares of the residuals between the actual value and the prediction value, reflecting the variation not explained by the model.'

[0043] Step C above forms a manufacturing cycle prediction model covering all manufacturing stages of the target product, expressed as: ; Where j represents the manufacturing cycle corresponding to different manufacturing stages, α n This indicates the factors that influence each manufacturing cycle.

[0044] In addition, exemplary model configurations and tuning before training include: using gbtree as the ensemble method for individual learners in the XGBoost algorithm, selecting a maximum depth of 8 for individual learners, and setting the learning rate to 0.3 to form the hyperparameter combination of this model.

[0045] Based on this, for any product within the preset product category, such as the target product, the various manufacturing stages experienced by the work-in-process materials of the target product correspond to different branches or nodes on the same tree structure of the XGBoost model. Since these branches or nodes share the same training data, such as the manufacturing cycle corresponding to each manufacturing stage and the key influencing factors affecting the manufacturing cycle, and have automatically learned and internalized the inherent correlation and dependence between each manufacturing stage during the training stage, when the production environment changes dynamically, this manufacturing cycle prediction model can make rapid and coordinated predictions and responses to each manufacturing stage based on a unified global perspective, thereby effectively overcoming the response delay problem caused by the fragmentation of cross-process knowledge in traditional independent stage models.

[0046] Furthermore, embodiments of the present invention also illustrate methods for acquiring and generating end-to-end manufacturing data, including: Data Acquisition: Regularly collect and integrate data from the enterprise's manufacturing management system, PES systems of each production line, sales management system, and logistics management system, covering key execution nodes in the steel enterprise's production process. This includes start and end times for each stage, such as contract signing, external quality design, internal quality design, contract issuance, plan preparation, plan issuance, steelmaking production, continuous casting billet unloading, slow cooling and cleaning, primary material production, secondary material production, tertiary material production, quaternary material production, chemical testing sampling and delivery, quality control release, product packaging, and product release. Also included are product categories, specifications, and orders. Construct a fully traceable dataset related to the steel product manufacturing execution cycle, using order number, smelting number, billet number, and steel product number as primary keys.

[0047] Data Processing: Based on the key node data collected and integrated above, processing is performed. The accuracy, consistency, completeness, and timeliness of the data are verified. Effective governance measures are taken for data with quality issues, including filling missing values ​​and handling outliers, to ensure the data quality of manufacturing cycle-related data. For example, if the start or end time of production records is missing or exceeds the normal time range, data cleaning is required. When production records differ from material records, the source of the master data needs to be verified. If the standard deviation is abnormal, further verification of data accuracy is required, and deviations should be marked.

[0048] Statistics on various cycles in steel product manufacturing: Based on the high-quality dataset mentioned above, the cycles in the steel product manufacturing process are calculated, including the cycle from contract issuance to completion of material requisition planning, the cycle from completion of material requisition planning to start of converter smelting, the cycle from start of converter smelting to end of steelmaking / continuous casting billet output, the cycle from end of continuous casting billet output to billet loading, and for products requiring offline cleaning, additional statistics on the cycle from continuous casting billet output to cleaning completion, the cycle from billet loading to hot rolling completion, the cycle from hot rolling completion to comprehensive judgment completion, and for finished products, additional statistics on the cycle from rolling completion to finishing completion, the cycle from comprehensive judgment completion to packaging completion, and the cycle from packaging completion to ready-to-shipment completion, etc.

[0049] Impact Factor Statistics: Statistical analysis is performed based on the manufacturing cycle described above. If the statistical analysis results for any manufacturing cycle do not conform to a normal distribution, the impact factors affecting that manufacturing cycle are grouped and refined, and the importance of variables in the refined groupings is identified to obtain key impact factors. Specifically, the importance of variables in the refined groupings is identified by using Spearman's rank correlation coefficient for initial feature importance screening, and then ranking the features based on domain knowledge.

[0050] The above embodiments illustrate a method for constructing a manufacturing cycle prediction model. This invention further describes the detailed structure of the constructed manufacturing cycle prediction model. The manufacturing cycle prediction model covers all manufacturing stages, including the preparation stage before manufacturing begins, the process stage during manufacturing, and the evaluation stage after manufacturing is completed. Based on the XGBoost algorithm, each of the above manufacturing stages corresponds to multiple tree structures. Based on the manufacturing cycle prediction model, each of the above manufacturing stages corresponds to multiple subordinate models, and each subordinate model corresponds to multiple manufacturing cycles through branches and nodes. Therefore, the manufacturing cycle prediction model includes a preparation cycle prediction model for the product before manufacturing begins, a process cycle prediction model for the product during manufacturing, and an evaluation cycle prediction model for the product after manufacturing is completed. Here, the product refers to a product within a preset product category. Based on this, in step S102 above, when the manufacturing cycle prediction model identifies the information of the material to be used, it calculates the manufacturing time of obtaining the target product based on the manufacturing stage of the work-in-process material, including: calling one or more of the preparation cycle prediction model, the process cycle prediction model, and the evaluation cycle prediction model according to the manufacturing stage of the target product and the work-in-process material, and calculating the manufacturing time; wherein, when calling the process cycle prediction model, different branches and different nodes in the process cycle prediction model are also called according to the product category of the target product.

[0051] It should be noted that, in both the manufacturing cycle prediction model and the pre-delivery prediction model in the embodiments of the present invention, the product categories are divided according to the production route, the number of processing steps, and the process of each processing step. However, according to the above embodiments, in the pre-delivery prediction model, the product categories are classified as plates, bars and wires, profiles, and pipes, while here, in the manufacturing cycle prediction model, especially in the process cycle prediction model, the product categories are classified as primary materials and multiple materials. This is because, in this embodiment of the invention, the process cycle prediction model includes a first process cycle prediction branch based on the initial processing flow and an N+1th process cycle prediction branch based on N processing steps, where N is a positive integer greater than or equal to 1. Based on this, the step of calling different branches in the process cycle prediction model according to the product category of the target product includes: when the product category is a single-process material, calling the first process cycle prediction branch; when the product category is a multi-process material, calling at least the first process cycle prediction branch and at most the first process cycle prediction branch to the N+1th process cycle prediction branch; when calling the second process cycle prediction branch to any one of the N+1th process cycle prediction branches, calling different nodes according to the product category of the target product itself.

[0052] For example, for products within the preset product scope, namely products involved in the plate production route, products involved in the bar and wire production route, products involved in the profile production route, and products involved in the pipe production route, the same preparation cycle prediction model and evaluation cycle prediction model are used. The preparation cycle prediction model corresponds to one or more of the following: contract issuance to material requisition plan completion cycle, and material requisition plan completion to converter smelting start cycle. The evaluation cycle prediction model corresponds to one or more of the following: hot rolling completion to comprehensive judgment completion cycle, rolling completion to finishing completion cycle, comprehensive judgment completion to packaging completion cycle, and packaging completion to ready-to-ship completion cycle. The key influencing factors affecting the preparation cycle include one or more of the following: contract signing duration, external quality design duration, internal quality design duration, contract issuance duration, plan preparation duration, and plan issuance duration. The key factors affecting the evaluation cycle include one or more of the following: chemical inspection sampling and testing duration, quality blockade release duration, product packaging duration, product ready-to-ship stage duration, product category, product specifications, and product order type.

[0053] according to Figure 1 For the process cycle prediction model, this embodiment of the invention uses products involved in the plate production route as an example for illustration. The products involved in the plate production route can be classified into primary, secondary, tertiary, and quaternary materials, where secondary, tertiary, and quaternary materials are multiple materials. The process cycle prediction model includes a first process cycle prediction branch based on the initial processing flow, a second process cycle prediction branch based on primary processing, a third process cycle prediction branch based on secondary processing, a fourth process cycle prediction branch based on tertiary processing, and a fifth process cycle prediction branch based on quaternary processing. Specifically, the first process cycle prediction branch corresponds to the initial processing cycle from the start of converter smelting to the end of steelmaking / continuous casting billet unloading; the second process cycle prediction branch corresponds to the primary processing cycle from billet loading to the completion of hot rolling; the third process cycle prediction branch corresponds to the secondary processing cycle from the start of cold rolling after hot rolling to the completion of cold rolling; the fourth process cycle prediction branch corresponds to the tertiary processing cycle from the start of finishing after cold rolling to the completion of finishing; and the fourth process cycle prediction branch corresponds to the quaternary processing cycle from the start of coating after finishing. It should be noted that each process cycle corresponding to the above branches also includes process cycles for routine operations such as furnace loading and cleaning, such as the cycle from the end of continuous casting billet tapping to billet loading, and the cycle from continuous casting billet tapping to cleaning completion. These are not shown here, but the influencing factors affecting these process cycles are key influencing factors. Therefore, the key influencing factors affecting the process cycles include one or more of the following: steelmaking production time, continuous casting billet tapping time, slow cooling and cleaning time, secondary processing time, tertiary processing time, and quaternary processing time.

[0054] This embodiment of the invention continues to use products involved in the sheet metal production route as examples to illustrate how to call different branches in the process cycle prediction model according to the product category of the target product. For example, when the product category is primary material, such as medium and heavy plate, the first process cycle prediction branch is called. When the product category is secondary material, such as tertiary material, quaternary material, or multiple material, such as galvanized coil / plate, color-coated coil / plate, silicon steel coil / plate, or commercial cold-rolled coil / plate, at least the first and second process cycle prediction branches are called, and at most the first, second, third, fourth, and fifth process cycle prediction branches are called. When calling the second, third, fourth, and fifth process cycle prediction branches, different nodes are called according to the product category of the target product itself. Taking color-coated coil / plate as an example, the first to fifth process cycle prediction branches need to be called, and for each branch, different nodes need to be selected based on the target product. For example, the fifth process cycle prediction branch corresponds to the five processing cycles from the start of finishing to the completion of finishing after cold rolling. However, the fifth process cycle prediction branch includes not only the node corresponding to color coating treatment, but also the node corresponding to rolling treatment. Rolling treatment is used to obtain embossed plate coils / plates. Therefore, when calling the fifth process cycle prediction branch, it is also necessary to select the color coating treatment node according to the product category of the target product itself.

[0055] According to the above embodiments, the manufacturing cycle prediction model provided by the present invention has a basic feature system derived from actual production, strong interpretability, and deeply explores and utilizes the key influencing factors affecting the manufacturing cycle, which can meet the real-time or near-real-time prediction needs of the production site.

[0056] This invention also illustrates a method for applying the expected delivery information obtained in step S104 above. In one embodiment, the method further includes recording the actual manufacturing time of the work-in-process material. If the error between the actual manufacturing time and the actual manufacturing time is greater than a preset error range, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model.

[0057] It should be noted that the nodes that output outliers in the above manufacturing cycle prediction model can be any of the cycles in the preparation cycle prediction model, such as the cycle from contract issuance to completion of material requisition planning and the cycle from completion of material requisition planning to start of converter smelting; any of the cycles in the evaluation cycle prediction model, such as the cycle from completion of hot rolling to completion of comprehensive judgment, the cycle from completion of rolling to completion of finishing, the cycle from completion of comprehensive judgment to completion of packaging, and the cycle from completion of packaging to completion of shipment; or any of the branches in the process cycle prediction model, such as the first process cycle prediction branch, the second process cycle prediction branch, the third process cycle prediction branch, and the fourth process cycle prediction branch, as well as any node in any of these branches.

[0058] In another embodiment, the above method further includes: If the error between the last day of the expected delivery time set and the last day of the initial delivery time set is greater than a preset error number of days, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model and / or the delivery prediction sub-model that outputs outliers in the delivery prediction model.

[0059] It should be noted that, for the outlier prediction sub-model, according to Table 1 above, the production lag node can be either an outlier on the delivery day or the daily maximum on-time delivery volume.

[0060] Based on the above steps, this embodiment of the invention can determine whether the target product in an order is in a delayed state and identify the production delay node, providing accurate basic analysis data for materials that are delayed in delivery, quickly identifying dynamic changes in the production environment, such as the failure to prepare rolling plans in a timely manner or accidents occurring on-site equipment, improving the response speed to production anomalies, optimizing the scheduling efficiency of production resources, and thus ensuring the timely delivery of orders.

[0061] like Figure 3 As shown, this embodiment of the invention also provides a pre-delivery information prediction system 30 based on a steel manufacturing scenario, comprising: The material information acquisition module 31 is used to acquire the material information to be used, which includes the work-in-process material used to manufacture the target product and the manufacturing stage of the work-in-process material. The target product is a product within a preset product range. The cycle prediction module 32 is used to input the information of the materials to be used into the manufacturing cycle prediction model, obtain the manufacturing time of the target product based on the work-in-process materials, and obtain the initial ready-to-ship information of the target product. The manufacturing cycle prediction model is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity. The pre-delivery prediction module 33 is used to obtain the maximum daily pre-delivery quantity of the target product within a preset time period through the pre-delivery prediction model; The issuance adjustment and output module 34 is used to adjust the initial issuance information according to the maximum daily issuance volume within the future preset time, generate a set of expected issuance times and a daily expected issuance volume, and obtain the expected issuance information of the target product.

[0062] Therefore, the steel manufacturing scenario-based issuance information prediction system provided in this embodiment of the invention can execute the steel manufacturing scenario-based issuance information prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0063] This invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the quasi-issuance information prediction method based on a steel manufacturing scenario as described in the above embodiments.

[0064] This invention also provides a computer program product, which, when executed by a processor, is suitable for executing an initialization program with the following method steps: responding to a query instruction for a target product, obtaining information on materials to be used, the information including work-in-process materials for manufacturing the target product and the manufacturing stage of the work-in-process materials, the target product being a product within a preset product category; inputting the information on materials to be used into a manufacturing cycle prediction model to obtain the manufacturing time of the target product based on the work-in-process materials, and obtaining the initial release information of the target product; wherein, the manufacturing cycle prediction model is trained based on full-process manufacturing data, all products within the preset product category provide the full-process manufacturing data, and the initial release information includes a set of initial release times for the target product and a daily initial release quantity; obtaining the maximum daily release quantity of the target product within a preset future time period through the release prediction model; adjusting the initial release information according to the maximum daily release quantity within the preset future time period, generating a set of expected release times and a daily expected release quantity, and obtaining the expected release information of the target product.

[0065] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer programs and instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer programs and instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0069] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0070] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0071] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0072] It should be noted that the manufacturing data (including but not limited to the full-process manufacturing data, manufacturing cycle, influencing factors, key influencing factors, contract signing duration, etc. mentioned in the text), logistics data (including but not limited to the maximum allowable shipment volume, transportation routes, etc. mentioned in the text), and various data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in the embodiments of the present invention are all information and data authorized by the customer or fully authorized by all parties, and the collection, use and processing of related data must comply with the regulations and standards of the relevant regions.

[0073] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0074] The above are merely embodiments of the present invention and are not intended to limit the scope of the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of the present invention should be included within the scope of the claims of the embodiments of the present invention.

Claims

1. A method for predicting delivery information based on a steel manufacturing scenario, characterized in that, include: In response to a query command for a target product, information on materials to be used is obtained. The information on materials to be used includes work-in-process materials used to manufacture the target product and the manufacturing stage of the work-in-process materials. The target product is a product within a preset product category. The information on materials to be used is input into the manufacturing cycle prediction model to obtain the manufacturing time of the target product based on the work-in-process materials, and to obtain the initial ready-to-ship information of the target product. The manufacturing cycle prediction model is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity. The maximum daily shipment volume of the target product within a preset future time period is obtained through a pre-delivery prediction model. The initial issuance information is adjusted based on the maximum daily issuance volume within the preset future time period to generate a set of expected issuance times and a daily expected issuance volume, thereby obtaining the expected issuance information for the target product.

2. The method for predicting delivery information based on a steel manufacturing scenario according to claim 1, characterized in that, Also includes: A basic prediction model is constructed based on the XGBoost algorithm; The manufacturing cycle prediction model is obtained by training the basic prediction model using the full-process manufacturing data. The full-process manufacturing data includes the manufacturing cycle and key influencing factors affecting the manufacturing cycle; When the manufacturing cycle prediction model identifies the information of the material to be used, it calculates the manufacturing time for obtaining the target product based on the work-in-process material according to the manufacturing stage of the work-in-process material.

3. The method for predicting delivery information based on a steel manufacturing scenario according to claim 2, characterized in that, The manufacturing cycle prediction model includes a preparation cycle prediction model for the product before manufacturing begins, a process cycle prediction model for the product during manufacturing, and an evaluation cycle prediction model for the product after manufacturing is completed. When the manufacturing cycle prediction model identifies the information of the materials to be used, it calculates the manufacturing time for obtaining the target product based on the work-in-process material according to the manufacturing stage of the work-in-process material, including: The manufacturing time is calculated by calling one or more of the preparation cycle prediction model, the process cycle prediction model, and the evaluation cycle prediction model according to the manufacturing stage of the target product and the work-in-process material; wherein, when calling the process cycle prediction model, different branches and different nodes in the process cycle prediction model are also called according to the product category of the target product.

4. The method for predicting delivery information based on a steel manufacturing scenario according to claim 3, characterized in that, The products within the preset product scope include products involved in the sheet metal production route, products involved in the bar and wire production route, products involved in the profile production route, and products involved in the pipe production route. The product categories are divided according to the production route, the number of processing steps, and the process of each processing step, and are classified as primary materials and multiple materials in the process cycle prediction model; The process cycle prediction model includes a first process cycle prediction branch based on the initial processing flow, and an N+1th process cycle prediction branch based on N processing cycles, where N is a positive integer greater than or equal to 1. The step of calling different branches in the process cycle prediction model based on the product category of the target product includes: When the product category is a single material, the first process cycle prediction branch is invoked; When the product category is multi-material, at least the first process cycle prediction branch is called, and at most the first process cycle prediction branch to the N+1th process cycle prediction branch is called. When calling the second process cycle prediction branch to any one of the N+1th process cycle prediction branches, different nodes are called according to the product category of the target product itself.

5. The method for predicting quasi-delivery information based on a steel manufacturing scenario according to any one of claims 1 to 4, characterized in that, The step of obtaining the maximum daily allowable shipment volume of the target product within a preset future time period through the allowable shipment prediction model includes: The maximum daily issuance volume within the preset future time period is output based on the target product and the issuance prediction sub-model. The on-time delivery prediction model includes multiple on-time delivery prediction sub-models. For each on-time delivery prediction sub-model, there is a unique product category, and the model stores the shipping date, optimal transportation route, and maximum daily on-time delivery volume corresponding to that product category. The product categories mentioned are classified as plates, bars and wires, profiles and pipes in the quasi-delivery prediction model.

6. The method for predicting delivery information based on a steel manufacturing scenario according to claim 5, characterized in that, Also includes: Record the actual manufacturing time of the work-in-process material; If the error between the actual manufacturing time and the actual manufacturing time is greater than a preset error range, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model.

7. The method for predicting delivery information based on a steel manufacturing scenario according to claim 5, characterized in that, Also includes: If the error between the last day of the expected delivery time set and the last day of the initial delivery time set is greater than a preset error number of days, then the production lag node is determined based on the nodes that output outliers in the manufacturing cycle prediction model and / or the delivery prediction sub-model that outputs outliers in the delivery prediction model.

8. A prediction system for accurate delivery information based on a steel manufacturing scenario, characterized in that, include: The material information acquisition module is used to acquire information on materials to be used. The information on materials to be used includes work-in-process materials used to manufacture the target product and the manufacturing stage of the work-in-process materials. The target product is a product within a preset product range. The cycle prediction module is used to input the information of the materials to be used into the manufacturing cycle prediction model, obtain the manufacturing time of the target product based on the work-in-process materials, and obtain the initial ready-to-ship information of the target product. The manufacturing cycle prediction model is trained based on full-process manufacturing data. All products within the preset product category provide the full-process manufacturing data. The initial release information includes the initial release time set of the target product and the daily initial release quantity. The pre-delivery prediction module is used to obtain the maximum daily pre-delivery volume of the target product within a preset time period through the pre-delivery prediction model; The issuance adjustment and output module is used to adjust the initial issuance information according to the maximum daily issuance volume within the preset future time period, generate a set of expected issuance times and a daily expected issuance volume, and obtain the expected issuance information of the target product.

9. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, the processor is configured to perform the quasi-issuance information prediction method based on the steel manufacturing scenario as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the issuance information based on the steel manufacturing scenario as described in any one of claims 1 to 7.

Images