Feed production center automatic control method and system based on cloud platform
By collecting and analyzing high-dimensional spectral data in real time on a cloud platform, and combining deep learning and feature matching, the problem of difficulty in identifying the inherent differences between batches of feed raw materials in existing technologies has been solved, and the stability of product quality and dynamic optimization of the production process have been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-05
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies rely on macroscopic physicochemical indicators, which are insufficient to accurately characterize the inherent differences between batches of feed raw materials, resulting in poor product quality consistency and difficulty in effective control during the production process.
By collecting and analyzing high-dimensional spectral data in real time on a cloud platform, and combining deep learning and feature matching, process optimization parameters and fault warning information are generated. Knowledge graphs and transfer learning paths are constructed to achieve accurate identification and dynamic optimization of the intrinsic properties of raw materials.
It enables precise control over the production process, improves the stability of product quality and the flexibility of production, and ensures rapid adaptability to new raw materials and efficient collaboration between equipment.
Smart Images
Figure CN121657602A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of industrial automation, Internet of Things and artificial intelligence, and in particular to an automated control method and system for a feed production center based on a cloud platform. Background Technology
[0002] The feed industry is the material foundation of modern animal husbandry and aquaculture. Its technological level and product quality directly affect breeding efficiency, animal health, and food safety. With the continuous improvement of the scale and precision of farming, the market has placed higher demands on the nutritional balance, batch-to-batch quality stability, and economic efficiency of feed products. Therefore, how to improve the intelligence level of feed production processes through advanced automation control technology and achieve precise control of product quality has become a key aspect of industry development and a focus of technological research and development. Against this backdrop, production enterprises are generally seeking to utilize automation systems to optimize production processes, aiming to improve production efficiency and reduce operating costs while ensuring product quality.
[0003] In related technologies, technicians typically sample and test incoming raw materials, obtaining key physicochemical indicators such as moisture, crude protein, and crude ash content through laboratory analysis. Based on these test results and the production plan, a set of the closest production process parameters is selected or matched from a pre-established expert database or formula library. This scheme is then programmed into the central control system to guide the operation of various equipment on the production line (such as crushers, mixers, and granulators). During production, sensors deployed on the field equipment continuously monitor operating parameters such as temperature, pressure, and motor current, and closed-loop control logic ensures that these parameters fluctuate within preset ranges to maintain the apparent stability of the production process.
[0004] However, feed ingredients are mostly agricultural products such as corn and soybean meal, whose inherent chemical composition and physical structure vary significantly due to factors such as origin, variety, harvesting season, and storage conditions. The macroscopic physicochemical indicators obtained from sampling tests, which rely on related technologies, are insufficient to comprehensively and precisely characterize these inherent differences between batches of raw materials. For example, two batches of soybean meal with the same crude protein content may have completely different amino acid compositions and protein denaturation levels. Therefore, when the central control system calls upon a set of standardized process parameters based on these macroscopic indicators, these parameters are actually based on an "average" or "typical" assumption about the inherent condition of the raw materials. Once the inherent characteristics of the raw materials actually used in production deviate from this assumption, the preset process parameters become disconnected from the actual processing requirements of the raw materials. Even if the operating parameters at the field equipment level (such as temperature and pressure) are strictly controlled within the set values, it is difficult to avoid unpredictable fluctuations in core quality indicators such as digestibility, palatability, and particle uniformity of the final product, resulting in poor product quality consistency and difficulty in effective control. Summary of the Invention
[0005] This application provides an automated control method and system for a feed production center based on a cloud platform, which addresses the problem that the production process parameters set by relevant technologies based on monitoring results are out of sync with the actual batch conditions of raw materials during feed production, resulting in poor consistency of final product quality and difficulty in effective control.
[0006] Firstly, this application provides an automated control method for a feed production center based on a cloud platform, applied to an automated control system for a feed production center, including a central control system, a field equipment layer, and a cloud platform. The method includes: The process parameters and equipment operating status data of each stage of feed production are acquired in real time through the field equipment layer, which includes sensors deployed on multiple target production equipment. The process parameters and the equipment operating status data are simultaneously uploaded to the central control system and the cloud platform; The analysis model mounted on the cloud platform is used to analyze the process parameters and the equipment operating status data to generate process optimization parameters and fault early warning information. The central control system provides a visual display of the process parameters, equipment operating status data, process optimization parameters, and fault warning information. In response to the user's adjustment instructions, the process optimization parameters are sent to the target production equipment through the central control system for parameter control.
[0007] By adopting the above technical solution, the system uploads process parameters and equipment status data collected on-site to the cloud platform in real time. Utilizing its built-in analytical model, the platform performs in-depth analysis to generate forward-looking process optimization parameters and fault warning information. This information is then visualized in the central control system, confirmed by the user, and distributed to the production equipment. This method overcomes the limitations of traditional control systems that rely solely on preset fixed parameters, achieving dynamic optimization and intelligent control of the production process. This improves production flexibility and control precision, thereby ensuring the stability of the final product quality.
[0008] In some embodiments, before the step of analyzing the process parameters and the equipment operating status data through the analysis model mounted on the cloud platform to generate process optimization parameters and fault early warning information, the method further includes: Acquire high-dimensional spectral data of the raw material to be produced, wherein the high-dimensional spectral data is a high-dimensional data matrix describing the intrinsic chemical composition and physical structure of the raw material; The high-dimensional spectral data is matched with existing raw material data clusters in the historical database to generate a matching result.
[0009] By adopting the above technical solution, the system can accurately characterize the intrinsic differences of each batch of raw materials from the microscopic level of chemical composition and physical structure by introducing the acquisition and analysis of high-dimensional spectral data of raw materials. It can also match these characteristics with historical data, thus solving the core problem of the disconnect between production parameters and actual needs caused by batch differences of raw materials. This makes the matching of process parameters no longer dependent on the rough estimation of macroscopic indicators, and improves the pertinence and accuracy of process matching.
[0010] In some embodiments, after the step of performing feature matching between the high-dimensional spectral data and existing raw material data clusters in a historical database to generate a matching determination result, the method further includes: When the matching determination result is a mismatch, a target perturbation strategy is matched from the parameter strategy library based on the high-dimensional spectral data; A multi-dimensional perturbation parameter set containing multiple combinations of process parameters is generated based on the target perturbation strategy; The multi-dimensional disturbance parameter set is used to drive the production equipment for testing and operation, and process data and quality data corresponding to each combination of process parameters are collected. An analysis model is constructed based on the high-dimensional spectral data, the process data, and the quality data and then uploaded to the cloud platform.
[0011] By adopting the above technical solution, when faced with entirely new raw materials that do not match historical data, the system can proactively trigger an "exploration-learning" process that includes test runs and data collection. By applying a set of perturbation parameters to drive production and collecting feedback data, the system can build a dedicated, high-precision analysis model for the new raw material based on measured data. This enables the dynamic expansion and self-improvement of the production knowledge base, ensuring the system's rapid adaptability to new materials and the effectiveness of production control.
[0012] In some embodiments, after the step of constructing an analysis model based on the high-dimensional spectral data, the process data, and the quality data and uploading it to a cloud platform, the method further includes: Based on the analysis model, optimization calculations are performed to generate process optimization parameters, which are then sent to the central control system. The high-dimensional spectral data, the analysis model, and the process optimization parameters are correlated to generate a new raw material data cluster; The newly added raw material data clusters are stored in the historical database.
[0013] By adopting the above technical solution, after generating the optimal process parameters for the new raw material through optimization calculation, the system further associates and packages the spectral data of the raw material, the newly established analysis model, and the optimized parameters to form a complete "raw material-model-parameter" knowledge unit, namely the newly added raw material data cluster, and stores it in the historical database, thus constructing a closed-loop knowledge accumulation and iteration mechanism. This allows the experience of each new raw material to be accumulated and reused, continuously enhancing the system's knowledge reserves and decision-making efficiency.
[0014] In some embodiments, prior to the step of storing the newly added raw material data cluster into the historical database, the method further includes: The high-dimensional spectral data is correlated and mapped with the origin information, seasonal factors, storage time and environmental parameters of the raw materials to obtain a raw material feature knowledge graph; The analysis model is annotated in multiple dimensions based on the raw material feature knowledge graph to generate an enhanced production model that includes information on the entire life cycle of raw materials. The correlation between the enhanced production model and existing models in the historical database is calculated using a similarity algorithm to establish a transfer learning path between the models. The enhanced production model, the raw material feature knowledge graph, and the transfer learning path are stored in the knowledge base of the cloud platform.
[0015] By adopting the above technical solution, the system constructs a raw material feature knowledge graph, deeply associating the spectral fingerprint of raw materials with their entire lifecycle information, such as origin and season, generating an enhanced production model with richer information dimensions. Furthermore, it utilizes transfer learning to establish association paths between models, improving the model's generalization ability and prediction accuracy. Through knowledge association and transfer, the system can achieve rapid cold start on new tasks.
[0016] In some embodiments, after the step of performing feature matching between the high-dimensional spectral data and existing raw material data clusters in a historical database to generate a matching determination result, the method further includes: When the matching determination result is a partial match, the difference vector between the high-dimensional spectral data and the matched raw material data cluster is calculated; Retrieve the historical production case with the highest matching degree to the difference vector from the historical database, and construct a difference compensation model based on the historical production case; Based on the difference compensation model, the production parameters corresponding to the matched raw material data clusters are corrected to generate adaptive process parameters; The adaptive process parameters are then sent to the central control system.
[0017] By adopting the above technical solution, the system can quantify the "difference vector" between raw materials that partially match historical data and existing data clusters. It can also intelligently retrieve "compensation" experience from historical cases to address similar differences, constructing a difference compensation model to finely adjust existing production parameters. This "fine-tuning" mechanism avoids the huge overhead of rebuilding models for similar raw materials, achieving accurate and efficient reuse and local optimization of existing production knowledge.
[0018] In some embodiments, after the step of visually displaying the process parameters, equipment operating status data, process optimization parameters, and fault warning information through the central control system, the method further includes: A multi-device collaborative state matrix is constructed based on real-time collected equipment operation status data. The multi-device collaborative state matrix represents the operational correlation and temporal dependency between different production equipment. The multi-device collaborative state matrix is dynamically analyzed using deep learning algorithms to identify implicit coupling patterns and potential bottleneck nodes among production equipment. When an abnormal coupling pattern is detected between production equipment, a fault propagation prediction model is constructed based on historical fault propagation paths, and the probability and time window of fault propagation between different production equipment are predicted based on the fault propagation prediction model. Based on the propagation probability and the time window, a decoupling control strategy is generated between production equipment, and the operating parameters of the production equipment are pre-adjusted by the central control system according to the decoupling control strategy.
[0019] By adopting the above technical solutions, the system elevates the perspective from individual devices to the collaborative level of the entire production system by constructing and analyzing a multi-device collaborative state matrix. It can leverage deep learning algorithms to gain insights into the complex implicit coupling relationships and potential bottlenecks between devices. When an anomaly is detected, the system not only identifies the problem but also predicts its propagation path and time based on a fault propagation model. This allows for the generation of decoupling control strategies for pre-adjustment, ensuring the continuous, stable operation and collaborative efficiency of the entire production line.
[0020] Secondly, this application provides an automated control system for a feed production center, the system comprising: one or more processors and a memory; The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the system can implement the cloud-based automated control method for a feed production center provided in the above embodiments, which will not be described in detail here.
[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on an automated control system for a feed production center, enable the system to implement the cloud-based automated control method for a feed production center provided in the above embodiments, which will not be elaborated further here.
[0022] Fourthly, this application provides a computer program product that, when running on an automated control system for a feed production center, enables the system to implement a cloud-based automated control method for a feed production center provided in the above embodiments, which will not be elaborated further here.
[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Abandoning the traditional method of relying on macroscopic physicochemical indicators, it instead utilizes high-dimensional spectral data to establish a unique "chemical fingerprint" for each batch of raw materials, enabling the system to accurately identify subtle differences between raw materials. Faced with new raw materials, it can intelligently select adaptive strategies: for partially matching raw materials, it fine-tunes parameters by constructing a difference compensation model; for completely unknown raw materials, it can proactively trigger testing processes to build a dedicated production model from scratch, fundamentally solving the problem of quality inconsistencies caused by raw material fluctuations and achieving precise control of the production process.
[0024] 2. Not only can it create analytical models for new raw materials, but it also packages the spectral data of the new raw materials, optimization parameters, and the newly created models into a complete "knowledge cluster" and stores it in the database, forming a closed-loop learning and memory mechanism. Furthermore, by constructing a knowledge graph, it deeply links the entire lifecycle information of raw materials with the production model and establishes transfer learning paths between models. This means the system no longer solves problems in isolation, but transforms each production experience into reusable and transferable structured knowledge, allowing the experience of the entire production system to continuously grow over time.
[0025] 3. By constructing a multi-device collaborative state matrix and using deep learning for analysis, the system can gain insight into the hidden coupling relationships and potential bottlenecks between production equipment. When an anomaly is detected, it not only issues an alarm but also predicts the propagation path and time window of the fault between devices based on historical data. Furthermore, it proactively generates decoupling control strategies to intervene, ensuring the stability and operational efficiency of the entire production line. Attached Figure Description
[0026] Figure 1 This is a flowchart illustrating an automated control method for a feed production center based on a cloud platform, as described in an embodiment of this application. Figure 2 This is another schematic diagram of a cloud-based automated control method for a feed production center, as described in this application embodiment. Figure 3 This is a schematic diagram of the physical device structure of an automated control system for a feed production center, as described in this application. Detailed Implementation
[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.
[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.
[0029] The following describes the process of the method provided in this implementation. Please refer to [link / reference]. Figure 1This is a flowchart illustrating an automated control method for a cloud-based feed production center, as described in an embodiment of this application.
[0030] S101. Real-time acquisition of process parameters and equipment operating status data at each stage of feed production through the field equipment layer.
[0031] This step is performed after feed production begins and continues throughout all production stages, including raw material processing, mixing, pelleting, and packaging. It is suitable for large-scale feed production centers where real-time data needs to be continuously acquired during the production process while the system is operational.
[0032] Among them, the field equipment layer refers to the collection of equipment deployed at the feed production site, including multiple target production equipment and supporting sensors; process parameters refer to the key technical parameters that affect product quality and production efficiency during the feed production process, used to represent the core control indicators in the production process; equipment operation status data refers to various data generated during the operation of the target production equipment, used to represent the working status of the equipment; target production equipment refers to the equipment on the feed production line that undertakes specific production tasks, such as crushers, mixers, pellet mills, etc.; sensors refer to the detection devices deployed on the target production equipment for collecting process parameters and equipment operation status data.
[0033] Specifically, after the system starts, the sensors at the field equipment level will continuously collect key information at each stage of feed production according to the preset collection frequency. In the raw material crushing stage, the sensors will collect data such as crushed particle size, crushing motor operating current, and temperature inside the crushing chamber; in the mixing stage, they will collect data such as mixing time, mixer speed, and pressure inside the mixing chamber; in the pelleting stage, they will collect data such as pelleting temperature, pelleting pressure, and pellet moisture content.
[0034] S102. Upload process parameters and equipment operating status data to the central control system and cloud platform simultaneously.
[0035] The central control system refers to the central control system of the feed production center, which is used for centralized monitoring of the production process, data display, and instruction issuance; the cloud platform refers to a remote cloud platform with data storage, analysis, and processing capabilities, which is used to carry analytical models and historical data storage.
[0036] Specifically, data collected and preliminarily processed by the field equipment is simultaneously sent to the central control system and the cloud platform via a pre-defined data transmission protocol. Encryption algorithms are used during data transmission to ensure security and prevent tampering or leakage. Data uploaded to the central control system is primarily used for local real-time monitoring and rapid response, allowing field operators to stay informed about production status. Data uploaded to the cloud platform is used for in-depth analysis, historical data storage, and model training, providing data support for generating process optimization parameters and issuing fault warnings. Asynchronous data upload ensures that the data acquisition efficiency of the field equipment is not affected, guaranteeing the continuity of the production process.
[0037] S103. The analysis model on the cloud platform analyzes the process parameters and equipment operating status data to generate process optimization parameters and fault warning information.
[0038] Among them, the analysis model refers to the data analysis model built on artificial intelligence algorithms and deployed on the cloud platform, which is used to perform in-depth processing of process parameters and equipment operating status data; the process optimization parameters refer to the optimized process parameters obtained after calculation by the analysis model, which can improve production efficiency and product quality; the fault early warning information refers to the prompt information generated by the analysis model after discovering potential equipment faults or production anomalies through data monitoring, which is used to inform relevant personnel in advance for handling.
[0039] Specifically, after receiving the data, the cloud platform inputs it into the built-in analysis model. The analysis model first analyzes the process parameters, compares them with the historical optimal range, and calculates optimized process parameters that can make product quality more stable and production efficiency higher, taking into account the characteristics of the current production stage. At the same time, the analysis model performs trend analysis and anomaly detection on the equipment operating status data. By comparing it with historical data during normal equipment operation, it identifies abnormal features in the current data, such as a sudden increase in equipment operating current or a continuous rise in temperature. Then, it determines the possible type of equipment failure, the probability of failure, and the expected time of occurrence, and generates corresponding fault warning information.
[0040] In some embodiments, after receiving the data, the cloud platform performs data preprocessing, including data standardization, missing value imputation, and outlier removal, to ensure data quality. The preprocessed data is then input into a process optimization sub-model and a fault warning sub-model. The process optimization sub-model uses machine learning algorithms, trained based on historical high-quality production data, and generates process optimization parameters by calculating the deviation between the current data and the historical best data. The fault warning sub-model uses deep learning algorithms to predict potential faults by identifying abnormal patterns in equipment operating status data. The outputs of the two sub-models are integrated, and a result credibility score and generation timestamp are added to form the final process optimization parameters and fault warning information.
[0041] In other embodiments, the system can also partition the received process parameters and equipment operating status data according to the production stage, such as raw material processing area, mixing area, granulation area, etc.; for the data of different partitions, the corresponding special analysis model is called for parallel analysis. For example, the raw material processing area model focuses on analyzing parameters related to particle size, while the granulation area model focuses on analyzing parameters related to granulation temperature and pressure; the analysis results of each special model are collected and integrated through a weighted fusion algorithm to eliminate the differences between the results of the models, generate unified process optimization parameters and fault warning information, and mark the key influencing factors of each partition.
[0042] It is understandable that other methods can be used to achieve data analysis, such as combining expert rule bases to correct the model analysis results, which is not limited here.
[0043] S104. The process parameters, equipment operating status data, process optimization parameters, and fault warning information are visualized through the central control system.
[0044] Visualization refers to presenting various data and information in an intuitive way, such as charts, text, and indicator lights, so that operators can quickly view and understand them.
[0045] Specifically, after receiving relevant data and information, the central control system categorizes and visualizes it according to a preset display logic and interface layout. Real-time process parameters and equipment operating status data are displayed as dynamic dashboards and trend graphs, with normal range thresholds marked. When data exceeds the threshold, a red font or flashing indicator is displayed as a warning. Process optimization parameters are displayed in tables or prominent text boxes, with differences from current process parameters clearly indicated. Fault warning information is presented simultaneously via pop-ups, red indicator lights, and voice prompts, clearly indicating the equipment involved, fault type, warning level, and suggested handling measures. Operators can quickly and comprehensively grasp production dynamics and potential problems through the central control system's visual interface.
[0046] Optionally, the system can classify the received process parameters, equipment operating status data, process optimization parameters, and fault warning information through the central control system into real-time monitoring, optimization suggestions, and early warning prompts. A visual interface layout is designed according to the classification results, with an equipment status navigation bar on the left, a core data display area (including dynamic dashboards and trend charts) in the middle, and an optimization parameter and early warning information display area on the right. The display content of various data types is updated in real time. Process parameters and equipment operating status data update trend curves at a frequency of once per second. Process optimization parameters are indicated by a flashing notification when updated, and fault warning information is accompanied by an audio alarm. A display log is also recorded.
[0047] Optionally, the system can also adopt a multi-dimensional display method, with both a comprehensive data overview interface for the entire production line and a detailed data interface for individual equipment; it supports operators to switch between different display interfaces, adjust the time range of trend charts, and view historical data through mouse clicks or touch screen operation; when clicking on fault warning information, operators can drill down to view detailed data on the operating status of the equipment corresponding to the fault, historical fault handling records, and relevant process parameters, helping operators to analyze problems in depth.
[0048] It is understandable that other methods can be used to achieve visualization, such as supporting custom interface layouts and display methods to meet the usage habits of different operators; no restrictions are imposed here.
[0049] S105. In response to the user's adjustment instructions, process optimization parameters are sent to the target production equipment through the central control system for parameter control.
[0050] Among them, the user's adjustment instruction refers to the operator issuing an instruction to the system to perform parameter adjustment after confirming that the process optimization parameters are appropriate, based on the information displayed by the central control system; parameter control refers to the target production equipment adjusting its own operating status according to the received process optimization parameters to ensure that the production process meets the optimization requirements.
[0051] Specifically, after viewing the process optimization parameters through the central control system's visual interface, operators, if they deem the parameters reasonable, can issue an adjustment command by clicking the "Confirm and Send" button on the interface or through a dedicated operating terminal. Upon receiving the command, the central control system converts the process optimization parameters to a format compatible with the target production equipment's parameter receiving format and encrypts the command. Subsequently, the central control system sends the process optimization parameters to the corresponding target production equipment via the industrial control bus. After receiving the parameters, the equipment automatically adjusts its operating parameters, such as motor speed, heating temperature, and feed rate, to ensure the production process follows the optimized parameters. Once the parameter adjustments are complete, the target production equipment sends confirmation information to the central control system, which then records the adjustment results in its system log.
[0052] In other embodiments, the system can also construct a multi-device collaborative state matrix based on real-time collected equipment operating status data, thereby characterizing the operational correlation and temporal dependency between different production devices; dynamically analyze the multi-device collaborative state matrix using deep learning algorithms to identify implicit coupling patterns and potential bottleneck nodes between production devices; when an abnormal coupling pattern is detected between production devices, a fault propagation prediction model is constructed based on historical fault propagation paths, and the propagation probability and time window of the fault between different production devices are predicted according to the fault propagation prediction model; a decoupling control strategy between production devices is generated according to the propagation probability and the time window, and the operating parameters of the production devices are pre-adjusted by the central control system according to the decoupling control strategy.
[0053] In the above embodiments, the system uploads process parameters and equipment status data collected on-site to a cloud platform in real time. Using its built-in analytical model, the platform performs in-depth analysis to generate forward-looking process optimization parameters and fault warning information. This information is then visualized in the central control system, confirmed by the user, and distributed to the production equipment. This method overcomes the limitations of traditional control systems that rely solely on preset fixed parameters, achieving dynamic optimization and intelligent control of the production process. This improves production flexibility and control accuracy, thereby ensuring the stability of the final product quality.
[0054] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating an automated control method for a cloud-based feed production center, as described in this application.
[0055] S201. Obtain high-dimensional spectral data of the raw material to be produced.
[0056] Specifically, the system is equipped with a spectral detection device in the raw material storage area or production inlet. When the raw materials to be produced arrive at the detection area in a preset batch (e.g., 500 kg per batch), the detection device automatically starts the spectral scanning process. During the scanning process, the device emits spectral signals within a specific wavelength range, penetrates the raw material sample, and receives the reflected and transmitted spectral response signals. These signals are then converted into a high-dimensional data matrix through signal conversion. This data matrix contains the content distribution of chemical components such as proteins, amino acids, and cellulose in the raw materials, as well as physical structure information such as particle structure and density, which can comprehensively and accurately reflect the intrinsic characteristics of each batch of raw materials.
[0057] Optionally, the system can randomly select multiple sub-samples (e.g., 5 samples, 1 kg each) from the batch of raw materials to be produced, remove impurities, mix them uniformly, and prepare standard test samples. The samples are placed in the detection chamber of the near-infrared spectroscopy detection equipment, and the equipment performs full-spectrum scanning according to a preset wavelength range (e.g., 780nm-2526nm) to collect spectral response signals. The built-in signal processing module of the equipment performs preprocessing such as noise reduction and baseline correction on the original spectral signals, converts the processed signals into a high-dimensional spectral data matrix, and adds metadata such as sample number and detection time.
[0058] S202. Perform feature matching between the high-dimensional spectral data and the existing raw material data clusters in the historical database to generate a matching judgment result.
[0059] Among them, the historical database refers to the database that stores data related to various raw materials in past production, including existing raw material data clusters, production cases and other information; the existing raw material data clusters refer to the high-dimensional spectral data sets that have been stored in the historical database and correspond to specific raw material types, with each data cluster associated with a set of mature production process parameters; the matching judgment result refers to the conclusion generated after feature matching, including three types: non-match, partial match and complete match.
[0060] Specifically, the system inputs the high-dimensional spectral data obtained in step S201 into the feature matching module of the cloud platform. The module first extracts the core feature vector of the high-dimensional spectral data, including key indicators such as the position of the feature peak, the peak intensity, and the slope of the feature curve. Then, it calls a similarity algorithm (such as the cosine similarity algorithm) to compare the feature vector with the feature vectors of existing raw material data clusters in the historical database one by one, and calculates the similarity value. According to the preset similarity threshold (such as a complete match threshold ≥95%, a partial match threshold 60%-94%, and a non-match threshold <60%), the corresponding matching judgment result is generated. If the similarity reaches the complete match threshold, it is judged as a complete match; if it is within the partial match threshold range, it is judged as a partial match; if it is below the non-match threshold, it is judged as a non-match.
[0061] S203. When the matching result is a mismatch, the target perturbation strategy is matched from the parameter strategy library based on the high-dimensional spectral data.
[0062] Among them, the parameter strategy library refers to a database that stores various perturbation strategies, and the perturbation strategies correspond to the characteristics of raw materials; the target perturbation strategy refers to the process parameter exploration strategy matched from the parameter strategy library that is applicable to the current mismatched raw materials, and is used to guide the generation of a multi-dimensional perturbation parameter set.
[0063] Specifically, when the matching result is a mismatch, the system inputs the high-dimensional spectral data of the raw material to be produced into the matching engine of the parameter strategy library. The engine analyzes the core characteristics of the raw material reflected in the high-dimensional spectral data, such as key indicators like hardness, moisture content, and nutrient concentration, and then retrieves the corresponding perturbation strategy from the strategy library based on these indicators. The parameter strategy library stores perturbation rules for different raw material characteristics. For example, for high-moisture raw materials, the strategy focuses on adjusting the perturbation range of drying-related process parameters; for high-hardness raw materials, it focuses on adjusting the perturbation gradient of crushing and granulation parameters. The matching engine finds the perturbation strategy that best fits the current raw material characteristics through a feature association algorithm, and determines it as the target perturbation strategy. This strategy clarifies the perturbation dimension, range, and step size of subsequent process parameters.
[0064] Optionally, the system can perform characteristic analysis on high-dimensional spectral data to extract key characteristic indicators such as raw material moisture content, particle hardness, and protein concentration, and quantify them into specific values. Based on these values, search conditions are constructed, and perturbation strategies containing the same or similar characteristic indicator ranges are searched in the parameter strategy library to initially screen out multiple candidate strategies. The fit between each candidate strategy and the current raw material characteristics is calculated, and the strategies are sorted from high to low fit. The strategy with the highest fit is selected as the target perturbation strategy, and the screening process is logged.
[0065] S204. Generate a multi-dimensional perturbation parameter set containing multiple combinations of process parameters based on the target perturbation strategy.
[0066] Among them, the multi-dimensional perturbation parameter set refers to a set of multiple different process parameter combinations generated based on the target perturbation strategy, with each combination corresponding to a test scheme; the process parameter combination refers to a set of parameters composed of multiple process dimensions, such as the combination of parameters such as particle size, mixing time, and granulation temperature.
[0067] Specifically, after receiving the target perturbation strategy, the system analyzes the clearly defined perturbation dimensions (such as the three core production stages of crushing, mixing, and granulation), the parameter ranges for each dimension (such as crushing particle size of 20-80 mesh), and the adjustment step size (such as every 10 mesh as a step size). Then, following the principles of full factorial design, the system combines the parameters of each dimension to generate multiple sets of process parameter combinations. For example, if the perturbation dimensions include crushing particle size (20 mesh, 40 mesh, 60 mesh), mixing time (5 minutes, 10 minutes), and granulation temperature (80℃, 90℃), then 3×2×2=12 sets of process parameter combinations are generated. Each combination is labeled with a unique test number, associated with the corresponding parameter value, ensuring accurate traceability during the test. Simultaneously, the parameter combinations exclude values exceeding the equipment's operating capabilities, ensuring the safety of the test operation.
[0068] S205. Apply a multi-dimensional disturbance parameter set to drive the production equipment for testing and operation, and collect process data and quality data corresponding to each set of process parameter combinations.
[0069] Among them, test operation refers to the small-scale, short-cycle production operation of production equipment according to the process parameter combination of multi-dimensional disturbance parameter set; process data refers to the process execution data and equipment operation data generated during the test operation, such as actual mixing time, real-time granulation pressure, etc.; quality data refers to the quality index data of the samples produced by the test operation, such as particle uniformity, digestibility, palatability related indicators, etc.
[0070] Specifically, the system sequentially distributes each set of process parameters to the corresponding production equipment according to the test number order of the multi-dimensional disturbance parameter set. After receiving the parameters, the equipment starts a small-scale test run (e.g., producing 50kg of sample per set of parameters). During the run, sensors at the field equipment level collect process data in real time, including the actual executed process parameters (e.g., actual particle size, mixing time) and equipment operating status data (e.g., motor current, equipment temperature). After the test run is completed, the system automatically collects the quality data of the produced sample. Online quality testing equipment is used to detect physical indicators such as particle uniformity, moisture content, and hardness, while laboratory testing equipment is used to detect nutritional indicators such as crude protein content and amino acid digestibility. The process data and quality data corresponding to each set of process parameters are stored in association with the test number, forming a complete data chain of "parameter-process-quality," ensuring data traceability and relevance.
[0071] S206. An analysis model is built based on high-dimensional spectral data, process data, and quality data and then uploaded to the cloud platform.
[0072] Specifically, the system first performs data preprocessing on the high-dimensional spectral data acquired by S201 and the process and quality data collected by S205. This includes data cleaning (removing outliers and filling missing values), data standardization (mapping data from different dimensions to the same numerical range), and feature fusion (integrating the key features of the three types of data into a unified feature matrix). Then, using the high-dimensional spectral data and process data as input features and quality data (such as particle uniformity and digestibility) as output labels, a suitable algorithm (such as gradient boosting trees or neural networks) is selected for model training. During training, model parameters (such as learning rate and tree depth) are adjusted through cross-validation to ensure that the model's prediction accuracy meets preset requirements (e.g., quality indicator prediction error ≤ 5%). After model training is complete, performance testing is performed. Once the model is confirmed to meet the requirements, it is encapsulated as a callable service interface and deployed to the model repository on the cloud platform, enabling it to receive real-time production data and output analysis results.
[0073] Optionally, the system can extract core features by performing dimensionality reduction processing on high-dimensional spectral data (such as principal component analysis), perform outlier detection (such as the 3σ principle) and normalization on process data and quality data, align the features of the three types of data according to the sample dimensions, and form a training dataset; use the gradient boosting tree algorithm to build a regression model, taking "high-dimensional spectral features + process parameters" as input and "quality indicators" as output, use the training dataset to train the model, and optimize the model hyperparameters through grid search; convert the trained model to ONNX format, upload it to the model deployment module of the cloud platform, generate an API call interface, verify the interface response speed (≤1 second / time) and prediction accuracy by inputting test data, and officially activate the model after confirming that there are no errors.
[0074] S207. The analysis model on the cloud platform analyzes the process parameters and equipment operating status data to generate process optimization parameters and fault warning information.
[0075] Specifically, the cloud platform receives process parameters and equipment operating status data uploaded by the field equipment layer in real time through data interfaces and inputs them into the built-in analysis model. The model optimizes the process parameters: combining the "parameter-quality" correlation when building the model in S206, it calculates the predicted value of product quality under the current process parameters, and adjusts the process parameters through iterative optimization (such as adjusting the granulation temperature from 85℃ to 88℃) to achieve the optimal predicted quality value, generating optimized process parameters. At the same time, the model performs anomaly detection on the equipment operating status data: comparing it with historical data thresholds during normal equipment operation (such as the normal range of motor current 10A-15A), if the current data exceeds the threshold or shows an abnormal trend (such as a continuous rise in temperature), it is judged as a potential fault. The risk level is divided according to the severity of the fault (such as Level 1 warning: immediate shutdown is required; Level 2 warning: inspection is required within 1 hour), generating fault warning information including fault location, possible causes, and handling suggestions.
[0076] Optionally, the system can convert the real-time received process parameters and equipment operating status data through a cloud platform to ensure consistency with the input requirements of the analysis model, and add metadata such as data acquisition time and equipment number. The process optimization module inside the model uses the current process parameters as initial values, combines the raw material characteristics corresponding to the high-dimensional spectral data, and uses a genetic algorithm to find the optimal parameter combination and output the process optimization parameters. The fault early warning module monitors the changing trend of equipment operating status data through a sliding window. If the data exceeds the normal range for 5 consecutive acquisition cycles, a fault early warning is triggered and an early warning message is generated. The process optimization parameters and fault early warning message are integrated in a preset format (such as JSON), and the result generation time and credibility score (such as 92% credibility of optimization parameters) are added before being sent to the central control system.
[0077] S208. Link the high-dimensional spectral data, analysis model, and process optimization parameters to generate a new raw material data cluster.
[0078] The newly added raw material data cluster refers to a structured data unit formed by associating and packaging the high-dimensional spectral data of the raw materials to be produced, the analytical model constructed for them, and the corresponding process optimization parameters, which is used to supplement the historical database.
[0079] Specifically, the system establishes association mapping rules: the high-dimensional spectral data acquired in S201 is used as the "feature identifier" of the raw material; the metadata of the analysis model constructed in S206 (such as model structure, training dataset information, and prediction accuracy) is used as the "model association information"; and the process optimization parameters generated in S207 (including specific parameter values for each production stage and applicable scenario descriptions) are used as the "parameter scheme." Subsequently, according to the preset data cluster format, the three are integrated into a complete data unit, and management information such as the data cluster number, generation time, and corresponding raw material type (e.g., "soybean meal-202405 batch") is added. Simultaneously, key information in the data cluster is verified to ensure the accurate correspondence between the high-dimensional spectral data and the process optimization parameters, and the completeness of the analysis model metadata, avoiding association errors. After successful verification, a new raw material data cluster is formed.
[0080] S209. The high-dimensional spectral data is correlated and mapped with the origin information, seasonal factors, storage time and environmental parameters of the raw materials to obtain the raw material characteristic knowledge graph.
[0081] Among them, origin information refers to the production area information of the raw materials to be produced (such as "Heilongjiang corn" and "Shandong soybean meal"), including related attributes such as regional climate and soil conditions; seasonal factors refer to the climate characteristics of the raw material harvest season (such as rainy season, dry season, and temperature range); storage time refers to the storage time of the raw materials from harvest to production; environmental parameters refer to the environmental conditions during the storage process of the raw materials (such as temperature, humidity, and ventilation); and the raw material characteristic knowledge graph is a visualized knowledge network with the raw material characteristics as the core and related to various influencing factors, used to intuitively show the relationship between raw material characteristics and external factors.
[0082] Specifically, the system collects information on the origin of raw materials, seasonal factors, storage duration, and environmental parameters: Origin information is extracted from raw material purchase orders and includes the origin name, regional soil type, and average annual precipitation; seasonal factors are determined based on the raw material harvest time and are associated with climate data such as average temperature and number of rainy days in the corresponding season; storage duration is calculated by comparing the raw material's entry time with the current time; environmental parameters are obtained from the raw material storage management system and include the average daily temperature, humidity, and ventilation frequency during storage. Subsequently, using the core features of high-dimensional spectral data (such as protein content characteristics and particle structure characteristics) as core nodes, and using origin, season, storage duration, and environmental parameters as associated nodes, a mapping relationship is established between nodes: for example, "high protein content characteristics" may be associated with "Northeast origin," "autumn harvest," "storage duration < 3 months," and "storage humidity < 60%." These relationships are visualized using knowledge graph building tools (such as Neo4j) to form a raw material feature knowledge graph. Each node in the graph contains detailed attribute information, and each relationship edge is labeled with its association strength (e.g., "Heilongjiang origin - high protein content" association strength 0.85).
[0083] S210. Based on the raw material feature knowledge graph, the analysis model is annotated in multiple dimensions to generate an enhanced production model containing information on the entire life cycle of raw materials.
[0084] Among them, multi-dimensional labeling refers to adding descriptive tags and related information to the analysis model from multiple dimensions (origin, season, storage, environment) throughout the entire life cycle of raw materials; raw material life cycle information refers to the key information of raw materials throughout the entire process from planting, harvesting, storage to input into production; enhanced production model refers to a model that, based on the original analysis model, links raw material life cycle information to have more comprehensive data support and higher generalization ability.
[0085] Specifically, the system extracts multi-dimensional information related to the analysis model from the raw material characteristic knowledge graph: categorized by "origin-season-storage-environment," it extracts key factors affecting raw material characteristics under each dimension (such as soil type for origin, precipitation for season, storage duration, and humidity for environment). Subsequently, multi-dimensional annotations are added to the analysis model: tags such as "applicable origin range," "optimal harvest season," and "recommended storage conditions" are added to the model metadata. The correlation between factors in the knowledge graph and the model input features (e.g., "for every 10% increase in storage humidity, the drying parameters in the model need to be increased by 5%) are embedded as annotations in the model document. Next, the annotation information is integrated with the original analysis model: during model inference, if the raw material background information associated with the input real-time production data (e.g., the current raw material storage humidity is high) matches the annotation information, the model automatically adjusts the parameter prediction logic (e.g., prioritizing the optimization of drying process parameters). Finally, an enhanced production model containing information about the entire life cycle of the raw materials and possessing scenario-adaptive capabilities is generated.
[0086] S211. Calculate the correlation between the enhanced production model and existing models in the historical database using a similarity algorithm, and establish transfer learning paths between models.
[0087] Specifically, the system first extracts metadata from the historical database of all existing production models, including key information such as the model's input feature dimensions (e.g., whether it contains high-dimensional spectral data, equipment operating parameters), output indicator types (e.g., quality prediction indicators, fault warning types), core algorithm framework (e.g., gradient boosting trees, neural networks), and applicable raw material characteristic range (e.g., corn, soybean meal). Then, a cosine similarity algorithm is selected to convert the metadata of the enhanced production model and the historical model into feature vectors, respectively. The cosine similarity value (i.e., correlation) between the two is calculated, and a correlation threshold is set (e.g., ≥0.7 is considered high correlation). Based on the calculation results, historical models with a correlation higher than the threshold are selected. The transferable knowledge components of these historical models (e.g., optimization logic for process parameters of similar raw materials, equipment fault warning rules) are analyzed. A knowledge transfer link from the historical model to the enhanced production model is constructed, clarifying the transferred knowledge types, transmission methods, and fusion rules, forming a complete transfer learning path.
[0088] S212. When the matching result is a partial match, calculate the difference vector between the high-dimensional spectral data and the matched raw material data cluster.
[0089] Specifically, the system identifies the existing raw material data cluster matched in S202 and extracts the core feature vectors (such as spectral peak intensity, peak position, characteristic curve slope, and other key indicators) from the high-dimensional spectral data of this data cluster. Subsequently, features of the same dimension are extracted from the high-dimensional spectral data of the raw material to be produced, resulting in the feature vector to be analyzed. The two feature vectors are subtracted according to their corresponding dimensions to obtain the difference value for each dimension. These difference values are arranged in the original feature dimension order to form a difference vector. Each element in the difference vector represents the degree of difference in the corresponding feature dimension. For example, "peak intensity difference value -0.2" indicates that the intensity of the spectral peak of the raw material to be produced is 0.2 units lower than that of the historical raw material; positive and negative numbers indicate the direction of the difference (e.g., a positive number indicates that the feature value of the raw material to be produced is higher). Through the difference vector, the system can intuitively identify the core difference dimensions between the raw material to be produced and the historical raw material, avoiding blindly adjusting process parameters.
[0090] S213. Retrieve the historical production cases with the highest matching degree with the difference vector from the historical database, and construct a difference compensation model based on the historical production cases.
[0091] Specifically, the system extracts all historical production cases containing raw material difference vector records from the historical database's production case library (e.g., a case record states "raw material protein characteristic peak intensity is 0.1 lower, corresponding to granulation temperature adjustment +3℃"). Then, the Euclidean distance algorithm is used to calculate the distance between the difference vector of the raw material to be produced and the difference vectors of each historical case (the smaller the distance, the higher the matching degree), and the historical production case with the smallest distance value is selected as the optimal reference case. Next, the core information of this optimal case is analyzed: including the raw material difference dimension, the corresponding process parameter adjustment direction (e.g., increasing / decreasing a parameter), the adjustment magnitude (e.g., ±5%), and the quality improvement effect after adjustment (e.g., particle uniformity improved by 3%). Based on this information, a difference compensation model is constructed: taking the values of each dimension of the difference vector as input and the process parameter adjustment amount as output, a mapping relationship between the two is established through linear regression or decision tree algorithms to ensure that the model can output corresponding parameter adjustment suggestions based on the specific difference value of the raw material to be produced.
[0092] S214. Based on the difference compensation model, the production parameters corresponding to the matched raw material data clusters are corrected, adaptive process parameters are generated, and sent to the central control system.
[0093] Specifically, the system extracts the baseline production parameters associated with the raw material data cluster matched in S202 from the historical database (e.g., 40 mesh particle size, 8 minutes mixing time, 85℃ granulation temperature). Then, the difference vector calculated in S212 is input into the difference compensation model, which outputs the specific adjustment amount for each process parameter (e.g., "particle size +5 mesh, granulation temperature +3℃, mixing time unchanged"). The system corrects the baseline production parameters parameter by parameter according to the adjustment amount, obtaining preliminary adaptive process parameters. After correction, the parameters are validated for rationality: checking whether the corrected parameters are within the equipment's operating limits (e.g., granulation temperature not exceeding 120℃) and whether they conform to the basic logic of raw material processing (e.g., high-moisture raw materials require increased drying time). After successful validation, the adaptive process parameters are encapsulated in a preset format (e.g., JSON), adding metadata such as parameter generation time and corresponding difference vector ID, and sent to the central control system via an encrypted data link. The central control system receives and stores the parameters in the parameter database and marks them as "currently available parameters" in the visualization interface.
[0094] The automated control system for the feed production center in this embodiment of the invention is applied to electronic equipment. Figure 3 A schematic diagram of the architecture of an electronic device suitable for implementing embodiments of the present invention is shown.
[0095] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.
[0096] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in the embodiments of the present invention.
[0097] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.
[0098] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and may communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.
[0099] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.
[0100] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A cloud-based automated control method for a feed production center, applied to an automated control system for a feed production center, comprising a central control system, a field equipment layer, and a cloud platform, characterized in that, The method includes: The process parameters and equipment operating status data of each stage of feed production are acquired in real time through the field equipment layer, which includes sensors deployed on multiple target production equipment. The process parameters and the equipment operating status data are simultaneously uploaded to the central control system and the cloud platform; The analysis model mounted on the cloud platform analyzes the process parameters and the equipment operating status data to generate process optimization parameters and fault early warning information. The central control system provides a visual display of the process parameters, equipment operating status data, process optimization parameters, and fault warning information. In response to the user's adjustment instructions, the process optimization parameters are sent to the target production equipment through the central control system for parameter control.
2. The method according to claim 1, characterized in that, Before the step of analyzing the process parameters and equipment operating status data through the analysis model mounted on the cloud platform to generate process optimization parameters and fault early warning information, the method further includes: Acquire high-dimensional spectral data of the raw material to be produced, wherein the high-dimensional spectral data is a high-dimensional data matrix describing the intrinsic chemical composition and physical structure of the raw material; The high-dimensional spectral data is matched with existing raw material data clusters in the historical database to generate a matching result.
3. The method according to claim 2, characterized in that, After the step of performing feature matching between the high-dimensional spectral data and existing raw material data clusters in the historical database to generate a matching determination result, the method further includes: When the matching determination result is a mismatch, a target perturbation strategy is matched from the parameter strategy library based on the high-dimensional spectral data; A multi-dimensional perturbation parameter set containing multiple combinations of process parameters is generated based on the target perturbation strategy; The multi-dimensional disturbance parameter set is used to drive the production equipment for testing and operation, and process data and quality data corresponding to each combination of process parameters are collected. An analysis model is constructed based on the high-dimensional spectral data, the process data, and the quality data and then uploaded to the cloud platform.
4. The method according to claim 3, characterized in that, After the step of constructing an analysis model based on the high-dimensional spectral data, the process data, and the quality data and uploading it to the cloud platform, the method further includes: Based on the analysis model, optimization calculations are performed to generate process optimization parameters, which are then sent to the central control system. The high-dimensional spectral data, the analysis model, and the process optimization parameters are correlated to generate a new raw material data cluster; The newly added raw material data clusters are stored in the historical database.
5. The method according to claim 4, characterized in that, Before the step of storing the newly added raw material data cluster into the historical database, the method further includes: The high-dimensional spectral data is correlated and mapped with the origin information, seasonal factors, storage time and environmental parameters of the raw materials to obtain a raw material feature knowledge graph; The analysis model is annotated in multiple dimensions based on the raw material feature knowledge graph to generate an enhanced production model that includes information on the entire life cycle of raw materials. The correlation between the enhanced production model and existing models in the historical database is calculated using a similarity algorithm to establish a transfer learning path between the models. The enhanced production model, the raw material feature knowledge graph, and the transfer learning path are stored in the knowledge base of the cloud platform.
6. The method according to claim 2, characterized in that, After the step of performing feature matching between the high-dimensional spectral data and existing raw material data clusters in the historical database to generate a matching determination result, the method further includes: When the matching determination result is a partial match, the difference vector between the high-dimensional spectral data and the matched raw material data cluster is calculated; Retrieve the historical production case with the highest matching degree to the difference vector from the historical database, and construct a difference compensation model based on the historical production case; Based on the difference compensation model, the production parameters corresponding to the matched raw material data clusters are corrected to generate adaptive process parameters; The adaptive process parameters are then sent to the central control system.
7. The method according to claim 1, characterized in that, After the step of visually displaying the process parameters, equipment operating status data, process optimization parameters, and fault warning information through the central control system, the method further includes: A multi-device collaborative state matrix is constructed based on real-time collected equipment operation status data. The multi-device collaborative state matrix represents the operational correlation and temporal dependency between different production equipment. The multi-device collaborative state matrix is dynamically analyzed using deep learning algorithms to identify implicit coupling patterns and potential bottleneck nodes among production equipment. When an abnormal coupling pattern is detected between production equipment, a fault propagation prediction model is constructed based on historical fault propagation paths, and the probability and time window of fault propagation between different production equipment are predicted based on the fault propagation prediction model. Based on the propagation probability and the time window, a decoupling control strategy is generated between production equipment, and the operating parameters of the production equipment are pre-adjusted by the central control system according to the decoupling control strategy.
8. An automated control system for a feed production center, characterized in that, The system includes: one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.
9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the automated control system of the feed production center, the system performs the method as described in any one of claims 1-7.
10. A computer program product, characterized in that, When the computer program product is run on the automated control system of the feed production center, the system performs the method as described in any one of claims 1-7.