Aluminum ingot component grouping system and method based on sliding window and intelligent combination optimization

Abstract

The invention discloses an aluminum ingot component grouping system and method based on a sliding window and intelligent combination optimization, and belongs to the technical field of aluminum smelting and casting. According to the method, weighted calculation is carried out on the weights of the multi-element chemical components (such as Al, Si, Fe, Cu, Ga, Zn and Mg) of the molten aluminum, and three decimal numbers are reserved by adopting a banker rounding method, so that the precision of the element mean value is remarkably improved; meanwhile, by presetting upper and lower limit standards of 70 aluminum and 85 aluminum, a mechanism of'automatic rejection below the lowest standard and automatic excitation above the high standard 'is established; the optimal combination is automatically screened after multiple sets of evaluation are carried out in the bag arranging process, it is ensured that the quality fluctuation of each bag of molten aluminum is reduced to the lower level, and therefore the follow-up aluminum blending difficulty is greatly reduced, meanwhile, the production line efficiency is improved, the system supports automatic sliding window bag arranging, window combinations can be automatically constructed from overall data, and evaluation decisions can be completed; by means of a rapid algorithm implementation mode, evaluation of each group of windows can be completed in several milliseconds.

Description

Technical Field

[0001] This invention relates to the field of aluminum smelting and casting technology, specifically to an aluminum ingot composition grouping system and method based on sliding window and intelligent combination optimization. Background Technology

[0002] In high-temperature and high-pressure industrial production environments such as metallurgy and casting, the smelting and casting process of molten aluminum is crucial to product quality and production efficiency. Traditional methods for combining molten aluminum baths mainly rely on operator experience or simple linear proportioning algorithms to determine the proportions of each element. In practical applications, when the composition of the molten aluminum or production requirements change, adjustments can often only be made manually or according to fixed rules, making it difficult to adapt to complex and changing on-site conditions in a timely manner.

[0003] Currently, these traditional empirical or linear proportioning methods are usually based on historical averages or static settings of a single parameter, such as pre-estimating the target content of certain key elements (such as Si, Fe, Al, etc.) or simply splicing together the tanks within a certain range. However, with the gradual increase in aluminum liquid production and the emergence of multi-variety and segmented demands, the original "point-to-point" or "fixed threshold" methods are increasingly unable to meet the requirements of high precision and high efficiency in the context of multiple elements coexisting and rapid switching of large batches of tanks.

[0004] In current aluminum smelting production processes, the elemental composition (e.g., silicon Si, iron Fe, copper Cu, magnesium Mg, gallium Ga, zinc Zn, aluminum Al, etc.) of molten aluminum fluctuates across multiple slots. To ensure the quality of aluminum ingots meets standards, companies typically use a "package arrangement" based on slot data, dividing several slots into multiple "packages" and judging their composition according to average elemental values. The two main practices currently employed by companies are as follows:

[0005] ① Subcontracting method based on human experience rules

[0006] Operators rely on experience to judge the composition of each batch, manually combining multiple batches to form a "package," and then averaging them to determine if they meet the standards. While this method offers some flexibility, it is heavily influenced by subjective factors, making it difficult to consistently guarantee quality, and it is also inefficient.

[0007] ② Combination of static templates or heuristic algorithms

[0008] In recent years, some companies have introduced simple rule models, such as fixed slot spacing, single-element priority sorting, or greedy algorithms, for combination. These methods can improve efficiency to some extent, but they cannot simultaneously address multi-objective optimization involving multi-element constraints, sliding continuity control, and high-standard incentives, resulting in limited combination quality. Furthermore, they cannot handle intelligent replenishment or rearrangement strategies in cases of insufficient tail slots.

[0009] The specific technical and operational issues are as follows:

[0010] ① When it is impossible to satisfy the constraints of multiple elements at the same time, especially when it involves the lower limit of aluminum content and the upper limit of other impurity content, it is necessary to introduce weighted average calculation and precision control.

[0011] ② Ignoring continuity constraints or span constraints can easily lead to combinations with strong jumps, which is not conducive to the stability of production line processes.

[0012] ③ The combination method is not intelligent, lacks a global scoring mechanism, cannot automatically select the optimal combination scheme, and results in resource waste.

[0013] ④ Incomplete tail-end slot combinations have poor processing capacity, ultimately leading to "wasted remaining slots" or "excessive elements within the package." Real-time monitoring and analysis are difficult to achieve.

[0014] Specifically:

[0015] Limitations of the combination method: Traditional methods lack a comprehensive assessment of multi-element composition and its dynamic changes, usually focusing only on a single major element (such as aluminum content) while ignoring temperature, impurity control, and the interactions between different baths. Because it is impossible to systematically and in real-time optimize the combination of multi-element requirements, it often leads to product quality fluctuations or resource waste, making it difficult to achieve a better balance between production efficiency and finished product consistency.

[0016] Lack of ability to balance multi-dimensional standards and objectives: Different smelting or casting requirements have different preferences and restrictions on the composition of molten aluminum, such as needing to meet high purity standards of below 70% aluminum or close to 99.85% aluminum. Traditional proportioning methods often can only meet a certain objective or restriction, while not taking into account other constraints, making it difficult to achieve a comprehensive balance among multiple elements and objectives, thus leading to unsatisfactory product quality or economic benefits.

[0017] Real-time computing and the challenge of large-volume smelting: As the daily aluminum molten metal production and the number of smelting cycles increase in smelting enterprises, traditional linear proportioning or manual experience-based methods are often inefficient when processing large amounts of smelting data. Rapid changes in external demand or internal smelting conditions require frequent and time-consuming manual adjustments, which are insufficient to meet the real-time or near-real-time intelligent decision-making requirements of industrial sites, and also cannot adapt to the rapid response demands of continuous production.

[0018] Difficulty in achieving flexible integration and expansion: In the context of accelerated industrial informatization, aluminum molten casting often needs to be interconnected with various detection systems, execution systems, or data centers. If existing methods rely too heavily on manual operation or fixed scripts, it is difficult to achieve flexible interface integration with modern industrial control platforms or information systems, thus preventing the full realization of automation and digitalization potential.

[0019] Therefore, a new solution is needed to address the above problems. Summary of the Invention

[0020] The purpose of this invention is to provide an aluminum ingot composition grouping system and method based on sliding window and intelligent combination optimization, so as to solve the technical problems mentioned in the background art.

[0021] To achieve the above objectives, the present invention provides the following technical solution: an aluminum ingot composition grouping system based on sliding window and intelligent combination optimization, comprising at least a request input module, a data preprocessing module, a sliding window combination module, a composition mean calculation and banker rounding module, a multidimensional fitness evaluation module, and an optimal combination output module;

[0022] The request input module is used to receive structured data with various information and to interact with external systems through the Flask interface. The various information includes at least the aluminum liquid tank number, the composition of various elements, and the planned weight.

[0023] The data preprocessing module is used to filter core fields (including slot number, chemical composition and planned weight) from the data received by the request input module, convert the slot number into an integer and sort it, and generate structured data for subsequent algorithm processing after removing invalid fields.

[0024] The sliding window combination module is used to continuously slide and combine slot data according to a preset window length and step size. The window size is 2×slots. When there are fewer than 2×slots remaining slots, it supports backtracking the previous part of the slots to form the most optimized candidate combination.

[0025] The component mean calculation and banker's rounding module is used to calculate the weighted average of each element in each candidate combination. The weighted average calculation uses the planned weight corresponding to the slot as the weight and uses banker's rounding (IEEE 754 standard) to retain three decimal places to ensure the accuracy of component values ​​and audit consistency.

[0026] The multidimensional fitness evaluation module is used to score each candidate combination in multiple dimensions. The scoring dimensions include at least the constraints of aluminum below 70, the target standard of aluminum 99.85, the span limit of the number of tanks in the combination, and the continuity requirements. Corresponding penalties or bonuses are given when the key conditions are not met or are met.

[0027] The optimal combination output module is used to select the combination with the highest fitness score in each window after evaluating all candidate combinations, and output them in ascending order of slot number, ultimately generating the optimal grouping result that matches the aluminum ingot composition requirements.

[0028] The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization, used in the aforementioned aluminum ingot composition grouping system based on sliding window and intelligent combination optimization, includes at least the following steps:

[0029] S1: Data preprocessing, which includes at least extracting key fields and converting them into a structured data table. The key fields to be extracted include slot number, composition, and weight.

[0030] S2: Sliding window combination, which combines the slot list by sliding continuously. Each group contains 2×slots slots. If there are fewer than 2×slots remaining, backtracking is performed to form the optimal candidate combination.

[0031] S3: Weighted average calculation, calculate the weighted average of the element content for each combination, where the weighting coefficient is the planned weight of each tank;

[0032] S4: Banker's rounding, using the banker's rounding method (IEEE 754 standard) to retain three decimal places to ensure consistent component precision;

[0033] S5: Fitness score, evaluate the fitness of all combinations, and the scoring dimensions include at least aluminum content standards, impurity limits, continuity requirements and span control;

[0034] S6: Optimal combination filtering. Select the highest-scoring combination from each sliding window and output it. The slots in the combination are arranged in ascending order.

[0035] Furthermore, the sliding window combination method in S2 supports window overlap, with a window step size of slots and a window size of 2×slots, to ensure sufficient combination possibilities and avoid blind spots caused by too many or too few slots.

[0036] Furthermore, the banker rounding process retains three decimal places for the average of each element according to the ROUND_HALF_EVEN strategy to ensure minimal deviation of the overall mean of the result and improve calculation consistency.

[0037] Furthermore, the fitness score comprehensively judges whether the combination meets the constraints and gives bonus points or deduction points as penalties, forming a unified scoring standard. For combinations that do not meet the high-purity aluminum standard (such as 99.85 aluminum), a bonus of 200 points is given; for combinations that violate the standard of less than 70 aluminum, a deduction of no less than 400 points is given; and additional penalty items are added for cases where the span of the inner tank exceeds the set threshold.

[0038] Furthermore, in all groups output by S6, the slot order is arranged in ascending order and returned to the calling system in nested array or JSON format.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] 1. Significantly improves compositional stability. This invention calculates the aluminum liquid using a weighted average of multiple chemical components (such as Al, Si, Fe, Cu, Ga, Zn, and Mg) and employs banker's rounding to three decimal places, significantly improving the accuracy of the elemental mean. Simultaneously, by presetting upper and lower limits for 70% and 85% aluminum, a mechanism is established that automatically rejects items below the minimum standard and automatically incentivizes items above the maximum standard. During the batching process, multiple evaluations are conducted to automatically select the optimal combination, ensuring that the quality fluctuation of each batch of aluminum liquid is reduced to a lower level, thereby significantly reducing the difficulty of subsequent aluminum blending.

[0041] 2. Improved production line efficiency: The system supports automated sliding window packing, which can automatically construct window combinations from the overall data and complete evaluation decisions. With the help of fast algorithm implementation, the evaluation of each window group can be completed in just a few milliseconds, meeting the real-time calculation and result return requirements of large-scale slot data. At the same time, by using window overlap sliding processing and intelligent tail completion, all slots can be fully utilized without omission, greatly improving the overall efficiency of the production line.

[0042] 3. Intelligent operation, free from reliance on human experience: This invention transforms the traditional aluminum blending knowledge that relies on operator experience into a quantifiable model, fitness function, and rule set. In the system, users only need to upload the original test data, and the algorithm can automatically complete the combination judgment and generate the packing results based on the chemical composition of each batch. In this way, enterprises can obtain accurate and stable packing solutions without repeated manual trials.

[0043] 4. Improve control accuracy: The system sets a maximum slot span limit to avoid instability caused by excessive slot spacing. At the same time, it applies additional penalties to non-continuous combinations, thereby guiding the system to prioritize the output of continuous slot combinations. In the tail replenishment stage, a reverse sliding window logic is also introduced to effectively alleviate or utilize the remaining slot resources, thereby comprehensively improving the accuracy and controllability of aluminum liquid grouping.

[0044] 5. Independent modules and flexible deployment: This invention achieves decoupling of functional modules and unified interface management under the Python+Flask framework. All core algorithm modules can be deployed independently on local or cloud servers. In addition, the logging and exception handling functions are complete, providing feasible guarantees for subsequent auditing, tracing, playback and algorithm optimization, and also making it easier for enterprises to integrate the system with existing industrial control platforms or data centers.

[0045] 6. Green and environmentally friendly, and resource-saving: By optimizing the combination strategy, the situation of "inferior aluminum liquid mixed with good material" is effectively avoided, reducing the waste of resources caused by substandard or unqualified products; at the same time, the probability of remelting or waste disposal caused by aluminum mixing failure is reduced, thereby improving the utilization efficiency of aluminum liquid resources from the source and better aligning with the concept of green production and sustainable development. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a schematic diagram of the process provided by the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0049] Example 1:

[0050] The aluminum ingot composition grouping system based on sliding window and intelligent combination optimization includes at least a request input module, a data preprocessing module, a sliding window combination module, a composition mean calculation and banker rounding module, a multidimensional fitness evaluation module, and an optimal combination output module.

[0051] The request input module is used to receive structured data with various information and to interact with external systems through the Flask interface. The various information includes at least the aluminum liquid tank number, the composition of various elements, and the planned weight.

[0052] The data preprocessing module is used to filter core fields (including slot number, chemical composition and planned weight) from the data received from the request input module, convert the slot number to an integer and sort it, and generate structured data for subsequent algorithm processing after removing invalid fields.

[0053] The sliding window combination module is used to continuously slide and combine slot data according to a preset window length and step size. The window size is 2×slots. When there are fewer than 2×slots remaining slots, it supports backtracking the previous part of the slots to form the most optimized candidate combination.

[0054] The component mean calculation and banker's rounding module is used to calculate the weighted average of each element in each candidate combination. The weighted average calculation uses the planned weight corresponding to the slot as the weight and uses banker's rounding (IEEE 754 standard) to retain three decimal places to ensure the accuracy of component values ​​and audit consistency.

[0055] The multidimensional fitness assessment module is used to score each candidate combination in multiple dimensions. The scoring dimensions include at least the constraints of aluminum below 70, the target standard of aluminum 99.85, the span limit of the combination's internal tanks, and the continuity requirements. Corresponding penalties or bonuses are given when the key conditions are not met or are met.

[0056] The optimal combination output module is used to select the combination with the highest fitness score in each window after evaluating all candidate combinations, and output them in ascending order of slot number, ultimately generating the optimal grouping result that meets the requirements of aluminum ingot composition.

[0057] Example 2:

[0058] The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization, used in the aluminum ingot composition grouping system based on sliding window and intelligent combination optimization mentioned in Embodiment 1 above, includes at least the following steps:

[0059] S1: Data preprocessing, which includes at least extracting key fields and converting them into structured data tables. Key fields to be extracted include slot number, composition, and weight.

[0060] S2: Sliding window combination, which combines the slot list by sliding continuously. Each group contains 2×slots slots. If there are fewer than 2×slots remaining, backtracking is performed to form the optimal candidate combination.

[0061] S3: Weighted average calculation, calculate the weighted average of the element content for each combination, and the weighting coefficient is the planned weight of each tank;

[0062] S4: Banker's rounding, using the banker's rounding method (IEEE 754 standard) to retain three decimal places to ensure consistent component precision;

[0063] S5: Fitness score, evaluate the fitness of all combinations, and the scoring dimensions include at least aluminum content standards, impurity limits, continuity requirements and span control;

[0064] S6: Optimal combination filtering. Select the highest-scoring combination from each sliding window and output it. The slots in the combination are arranged in ascending order.

[0065] The sliding window combination method in S2 supports window overlap, with a window step size of slots and a window size of 2×slots, to ensure sufficient combination possibilities and avoid blind spots caused by too many or too few slots.

[0066] Banker rounding retains the average of each element to three decimal places using the ROUND_HALF_EVEN strategy to ensure minimal deviation of the overall mean and improve computational consistency.

[0067] The fitness score comprehensively judges whether the combination meets the constraints and gives bonus points or deduction points as penalties, forming a unified scoring standard. For combinations that do not meet the high-purity aluminum standard (such as 99.85 aluminum), a bonus of 200 points is given; for combinations that violate the standard of less than 70 aluminum, a deduction of no less than 400 points is given; and additional penalty items are added for the case where the span of the inner tank exceeds the set threshold.

[0068] In all groups output by S6, the slot order is arranged in ascending order and returned to the calling system in nested array or JSON format.

[0069] Further supplement to Example 2:

[0070] Please see Figure 1 The first module receives aluminum molten metal tank data submitted by external systems. The request format is JSON, containing key fields such as each tank's number, elemental composition (e.g., Si, Fe, Al), and planned weight. The system uses the Flask framework to build a unified service interface, ensuring that the requested data can be parsed using standard methods.

[0071] After the API module receives the data, it enters the data preprocessing module. During this stage, the system will:

[0072] 1. Filter out the core fields used for modeling calculations (slot number, various chemical components, planned weight);

[0073] 2. Convert slot numbers to integers and sort them;

[0074] 3. Remove invalid fields;

[0075] 4. Finally, a structured pandas.DataFrame is generated for subsequent algorithm modules to use.

[0076] This module uses a sliding window approach, selecting six consecutive slots (window size = 2 × number of slots per packet) as a candidate grouping set each time. All possible combination grouping strategies are executed within the window, such as splitting the six slots into 3+3 groups_a and group_b.

[0077] If there are fewer than 6 remaining slots, the system will automatically fill the remaining slots (backtracking by 1 bag) to construct a window that does not overlap but is optimized as much as possible.

[0078] Within each candidate combination, the component mean will be calculated using a weighted average method:

[0079]

[0080] Where (x) i ) represents the element value, (w i ) represents the planned weight corresponding to the slot.

[0081] To ensure accuracy and audit consistency, all results will be rounded to three decimal places using the banker's rounding method (IEEE 754 standard) before comparison and judgment.

[0082] For each combination, a multi-dimensional evaluation logic is executed, including the following scoring dimensions:

[0083] 1. Does it meet the 70 aluminum baseline standard? (Points will be deducted if it does not meet the standard)

[0084] 2. Does it meet the target standard for 85 aluminum? (Bonus points if it does;)

[0085] 3. Does the span within the combination exceed the limit? (If it exceeds the limit, it will be judged as invalid)

[0086] 4. Whether the combination of slots is continuous (the better the continuity, the better);

[0087] 5. The rationality of the slot combination's position in the window (preferring to prioritize the earlier ones);

[0088] Finally, a fitness score is output as the basis for judging the quality of the combination.

[0089] The system selects the combination with the highest fitness (Group A + Group B) from the current sliding window and appends it to the final output for use by downstream applications (such as casting execution systems).

[0090] Repeat this process until all slots have been combined and judged. The system will then output all combination results and slot numbers in JSON format.

[0091] In summary, the present invention has the following technical advantages:

[0092] 1. A component evaluation mechanism based on component weighted average and banker rounding is provided to solve the problems of floating-point precision error and boundary misjudgment.

[0093] 2. A method for optimizing the sorting of packages under a sliding window is provided, which preserves the continuity of the sliding window and the adjustable step size mechanism during the combination of slots, thereby improving the diversity of combinations;

[0094] 3. A multi-objective fitness function construction method supporting incentives and penalties is proposed to flexibly evaluate high-standard incentives (such as reaching the 85 aluminum standard) and cross-standard penalties (such as falling below the 70 aluminum standard);

[0095] 4. Provide an intelligent tail pack replenishment mechanism, which merges the original slots forward and recombines them when the last group of slots is insufficient for a complete combination, thereby minimizing waste.

[0096] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An aluminum ingot composition grouping system based on sliding window and intelligent combination optimization, characterized in that: It includes at least a request input module, a data preprocessing module, a sliding window combination module, a component mean calculation and banker rounding module, a multidimensional fitness evaluation module, and an optimal combination output module; The request input module is used to receive structured data with various information and to interact with external systems through the Flask interface. The various information includes at least the aluminum liquid tank number, the composition of various elements, and the planned weight. The data preprocessing module is used to filter core fields from the data received by the request input module, convert slot numbers into integers and sort them, remove invalid fields and generate structured data for subsequent algorithm processing; The sliding window combination module is used to continuously slide and combine slot data according to a preset window length and step size. The window size is 2×slots. When there are fewer than 2×slots remaining slots, it supports backtracking the previous part of the slots to form an optimized candidate combination that is close to the optimal one. The component mean calculation and banker's rounding module is used to calculate the weighted average of each element in each candidate combination. The weighted average calculation uses the planned weight corresponding to the slot as the weight and uses banker's rounding to retain three decimal places to ensure the accuracy of component values ​​and audit consistency. The multidimensional fitness evaluation module is used to score each candidate combination in multiple dimensions and to give corresponding penalties or bonuses when the key conditions are not met or are met. The optimal combination output module is used to select the combination with the highest fitness score in each window after evaluating all candidate combinations, and output them in ascending order of slot number, ultimately generating the optimal grouping result that matches the aluminum ingot composition requirements.

2. A method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization, used in the aluminum ingot composition grouping system based on sliding window and intelligent combination optimization as described in claim 1, characterized in that, At least the following steps are included: S1: Data preprocessing, which includes at least extracting key fields and converting them into a structured data table. The key fields to be extracted include slot number, composition, and weight. S2: Sliding window combination, which combines the slot list by sliding continuously. Each group contains 2×slots slots. If there are fewer than 2×slots remaining, backtracking is performed to form the optimal candidate combination. S3: Weighted average calculation, calculate the weighted average of the element content for each combination, where the weighting coefficient is the planned weight of each tank; S4: Banker's rounding, which uses banker's rounding to retain three decimal places to ensure consistent component precision; S5: Fitness score, which evaluates the fitness of all combinations. The scoring dimensions include at least aluminum content standards, impurity limits, continuity requirements, and span control. S6: Optimal combination filtering. Select the highest-scoring combination from each sliding window and output it. The slots in the combination are arranged in ascending order.

3. The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization according to claim 2, characterized in that: The sliding window combination method in S2 supports window overlap, with a window step size of slots and a window size of 2×slots, to ensure sufficient combination possibilities and avoid blind spots caused by too many or too few slots.

4. The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization according to claim 2, characterized in that: The banker rounding process retains three decimal places for the average of each element according to the ROUND_HALF_EVEN strategy, in order to ensure the minimum deviation of the overall mean of the result and improve the consistency of calculation.

5. The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization according to claim 2, characterized in that: The fitness score comprehensively judges whether the combination meets the constraints and gives bonus points as an incentive or deduction points as a penalty, forming a unified scoring standard, and adds an additional penalty item for the case where the slot span in the combination exceeds the set threshold.

6. The method for grouping aluminum ingot composition based on sliding window and intelligent combination optimization according to claim 2, characterized in that: In all groups output by S6, the slot order is arranged in ascending order and returned to the calling system in nested array or JSON format.

Images