An auxiliary system for artemisinin extraction

Abstract

This invention proposes an auxiliary system for artemisinin extraction, comprising a processor and a memory. The processor acquires dynamic data, static data, and parameter data affecting extraction efficiency and purity at each node during the artemisinin extraction process, generating multidimensional feature data. The dynamic data includes real-time dynamic data of each node at different time points during the artemisinin extraction process, and the static data includes the geometry, capacity, weight, and volume of each node. The nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence. Node discrimination coefficients are initialized, and the node discrimination coefficients are weights of the parameters affecting extraction efficiency and purity. The node discrimination coefficients are trained and updated to obtain a discrimination threshold. Abnormal nodes are identified based on the comparison between the discrimination threshold and the node discrimination coefficients, and the real-time dynamic data, static data, and corresponding parameter data affecting extraction efficiency and purity of the abnormal nodes are used as the first type of data.

Description

Technical Field

[0001] This invention belongs to the field of artemisinin extraction, and specifically relates to an auxiliary system for artemisinin extraction. Background Technology

[0002] Currently, artemisinin is an important antimalarial drug, and the efficiency and quality of its extraction process directly affect its efficacy and production cost. To improve the extraction efficiency and purity of artemisinin, auxiliary systems play a crucial role in the extraction process.

[0003] Currently available artemisinin extraction auxiliary systems suffer from significant shortcomings in several aspects, limiting their effectiveness and reliability in practical applications. Firstly, existing auxiliary systems often fail to meet the high standards of modern pharmaceutical industry in terms of extraction efficiency. Many systems rely on traditional extraction methods, such as solvent extraction, which are often inefficient and time-consuming when processing large quantities of artemisia annua raw materials, leading to extended production cycles and increased production costs.

[0004] Traditional extraction methods are easily affected by environmental factors, such as temperature and humidity variations, further reducing the stability and consistency of the extraction. Secondly, current auxiliary systems are insufficient in terms of extraction purity. Artemisinin extraction is often accompanied by the generation of other impurities, and existing systems are not technologically advanced enough in separating and purifying artemisinin, resulting in low purity of the final product and affecting the drug's efficacy and safety. Furthermore, many systems fail to effectively control impurity formation during processing, leading to complex and costly subsequent purification steps.

[0005] Many auxiliary systems are complex to operate and maintain, lacking user-friendly interfaces and intuitive operating instructions. Operators may require lengthy training to master the system, increasing labor and training costs. Furthermore, system malfunctions lead to cumbersome maintenance and troubleshooting processes, potentially causing production line downtime and increased production losses. In addition, existing auxiliary systems are inefficient in terms of energy. Many systems consume high energy during operation, especially under high loads, resulting in increased production costs. Moreover, inadequate energy management can lead to energy waste, further exacerbating energy consumption problems. In today's increasingly environmentally conscious world, developing more energy-efficient and effective auxiliary systems is of paramount importance.

[0006] Current auxiliary systems generally have low levels of intelligence and automation, lack data monitoring and feedback mechanisms, and cannot monitor the extraction process in real time. This makes it difficult for the system to provide effective fault warnings and adjustments during operation, reducing production flexibility and response speed. Therefore, to solve these problems, there is an urgent need for a new type of auxiliary system for artemisinin extraction that can achieve high efficiency, high purity, a user-friendly interface, and intelligent monitoring and control, thereby improving the overall efficiency and product quality of artemisinin extraction and meeting the high standards of modern pharmaceutical industry production processes. Summary of the Invention

[0007] This invention proposes an auxiliary system for artemisinin extraction, which solves the problems of efficiency and purity monitoring in the traditional extraction process. Through comprehensive analysis of real-time dynamic data and static data, it intelligently identifies abnormal nodes and optimizes extraction parameters, thereby improving the extraction efficiency and product purity of artemisinin.

[0008] The technical solution of the present invention is implemented as follows: an auxiliary system for artemisinin extraction, comprising a processor and a memory, wherein the processor acquires dynamic data, static data, and parameter data affecting extraction efficiency and purity at each node during the artemisinin extraction process, and generates multidimensional feature data;

[0009] The dynamic data includes real-time dynamic data of each node at different time points during the artemisinin extraction process, and the static data includes the geometry, capacity, weight, and volume of each node; the nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence; the node discrimination coefficients are initialized, and the node discrimination coefficients are the weights of the parameters affecting extraction efficiency and purity; the node discrimination coefficients are trained and updated to obtain the discrimination threshold.

[0010] Abnormal nodes are identified based on the comparison between the discrimination threshold and the node discrimination coefficient. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the abnormal nodes are taken as the first type of data. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the remaining nodes are taken as the second type of data. Through association rule analysis, frequent itemsets containing each parameter data are selected to generate association rules.

[0011] Based on the association rules, the parameter data affecting the extraction efficiency and purity are determined, adjustment commands are generated, and the adjustment commands are sent to the corresponding nodes to adjust the parameter data and the process parameters during the artemisinin extraction process.

[0012] Traditional artemisinin extraction systems often rely on manual monitoring and experience-based judgment, lacking real-time data acquisition at each stage of the extraction process. This leads to fluctuations in extraction efficiency and instability in product purity. Our system, however, uses a processor to acquire dynamic and static data at each stage of the artemisinin extraction process, enabling real-time monitoring of each step. This real-time data acquisition capability allows the system to respond promptly to changes during extraction, ensuring the stability and consistency of the extraction process.

[0013] Existing technologies for parameter optimization typically employ single empirical formulas or fixed process parameters, lacking flexibility and adaptability. Artemisinin extraction is influenced by various factors, including temperature, pressure, and time, and traditional methods struggle to effectively adjust these parameters to adapt to different extraction conditions. This system, however, intelligently identifies key parameters affecting extraction efficiency and purity by generating multidimensional feature data and training and updating node discrimination coefficients. This data-driven parameter optimization approach enables the system to adaptively adjust under varying extraction conditions, improving the flexibility and targeting of the extraction process.

[0014] Existing technologies have limited capabilities in anomaly detection and handling. Traditional systems often rely on manual judgment to identify problem nodes, which may lead to delays and misjudgments. This system, however, automatically identifies anomalous nodes by initializing node discrimination coefficients and comparing them with a discrimination threshold. This intelligent anomaly detection mechanism enables the system to respond quickly to potential problems, reducing losses caused by equipment failures or abnormal parameters.

[0015] Existing artemisinin extraction systems often lack depth in data analysis and decision support. Our system, however, utilizes association rule analysis to identify frequent itemsets affecting extraction efficiency and purity, thereby generating association rules. This data-driven analysis approach enables the system to extract valuable information from historical data, providing a scientific basis for subsequent process optimization. Traditional methods often rely on experience and intuition, lacking systematic analysis and verification, resulting in insufficient scientific rigor and accuracy in decision-making.

[0016] Existing technologies often lack automation and intelligence in adjusting process parameters, resulting in cumbersome operations. This system, however, generates adjustment commands and sends them to the corresponding nodes, achieving automated parameter adjustment. This automated design not only improves operational efficiency and reduces the need for manual intervention but also ensures precise control of the extraction process, further enhancing the extraction efficiency and product purity of artemisinin.

[0017] In a preferred embodiment, the process of collecting dynamic data, static data, and parameter data affecting extraction efficiency and purity at each node during artemisinin extraction to generate multidimensional feature data includes: normalizing the dynamic and static data; arranging the dynamic and static data in chronological order to obtain an original data sequence; segmenting the original data sequence to obtain multiple data subsequences; constructing a first type of data matrix with the data subsequences as rows and a first feature vector as columns; constructing a second feature vector from the parameter data affecting extraction efficiency and purity corresponding to each node, and arranging the second feature vector in chronological order to obtain a second type of data matrix; wherein, the first feature vector includes the normalized dynamic and static data of each node; and the second feature vector includes the normalized parameter data affecting extraction efficiency and purity of each node.

[0018] In a preferred embodiment, the nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence. Then, a third type of data matrix is ​​constructed according to the arrangement order of the node sequence, with each node as a row and the third feature vector as a column. The second feature vector includes parameter data of each node that affect the extraction efficiency and purity. The discrimination threshold is multiplied by the parameter data of each node in the second type of data matrix that affect the extraction efficiency and purity to obtain the third type of data matrix.

[0019] In a preferred embodiment, the node discrimination coefficients are trained and updated to obtain a discrimination threshold, which includes: acquiring historical training data, which includes historical process data and label data; wherein, the label data consists of labels marked with the corresponding parameter data of each node; constructing a classification data matrix based on the historical process data; and training a neural network model based on the classification data matrix to obtain a judgment threshold; wherein, the network model includes a feature extraction layer, an activation layer, and a classification layer connected in sequence.

[0020] In a preferred embodiment, the process of determining the abnormal node includes: taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the abnormal node as the first type of data; otherwise, taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the remaining nodes as the second type of data, and using association rule analysis to filter out frequent itemsets containing each parameter data to generate association rules.

[0021] By adopting the above technical solution, the beneficial effects of this invention are as follows: Through real-time data acquisition and analysis, the system can monitor various parameters during the extraction process in a timely manner, ensuring the stability of the extraction process. The acquisition of real-time dynamic data enables operators to quickly identify potential problems and adjust parameters promptly, avoiding losses caused by delays. This timely monitoring not only improves production efficiency but also reduces downtime caused by problems, ensuring the continuity and efficiency of the production process.

[0022] The system's intelligent parameter optimization capabilities enhance the flexibility and adaptability of the extraction process. By analyzing historical and real-time data, the system can automatically identify key factors affecting extraction efficiency and purity, and make corresponding adjustments based on actual conditions. This intelligent decision support helps manufacturers better adapt to market demands, optimize production processes, and improve product quality.

[0023] The introduction of anomaly detection and handling mechanisms significantly improves the system's security and reliability. By automatically identifying abnormal nodes, the system can implement corresponding measures in a timely manner, reducing the risks caused by equipment failures or parameter anomalies. This rapid response capability enables enterprises to improve their production stability, ensure product consistency and reliability, and thus enhance their market competitiveness.

[0024] The application of the data analytics module provides enterprises with a more scientific basis for decision-making. Through association rule analysis, the system can uncover potential optimization paths, helping manufacturing enterprises make more informed decisions in future production. This data-driven management approach improves the operational efficiency and management level of enterprises, providing strong support for continuous improvement.

[0025] The system's automated design reduces reliance on manual operation, lowers labor costs, and improves production efficiency. Operators can focus more on production monitoring and data analysis, enhancing the intelligence level of the production process. This automated operating model not only improves the company's production capacity but also lays the foundation for future intelligent manufacturing transformation. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below 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.

[0027] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example:

[0030] like Figure 1 As shown, an auxiliary system for artemisinin extraction includes a processor and a memory. The processor acquires dynamic data, static data, and parameter data affecting extraction efficiency and purity at each node during the artemisinin extraction process, and generates multidimensional feature data.

[0031] The dynamic data includes real-time dynamic data of each node at different time points during the artemisinin extraction process, and the static data includes the geometry, capacity, weight, and volume of each node; the nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence; the node discrimination coefficients are initialized, and the node discrimination coefficients are the weights of the parameters affecting extraction efficiency and purity; the node discrimination coefficients are trained and updated to obtain the discrimination threshold.

[0032] Abnormal nodes are identified based on the comparison between the discrimination threshold and the node discrimination coefficient. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the abnormal nodes are taken as the first type of data. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the remaining nodes are taken as the second type of data. Through association rule analysis, frequent itemsets containing each parameter data are selected to generate association rules.

[0033] Based on the association rules, the parameter data affecting the extraction efficiency and purity are determined, adjustment commands are generated, and the adjustment commands are sent to the corresponding nodes to adjust the parameter data and the process parameters during the artemisinin extraction process.

[0034] The core of the entire workflow lies in the acquisition of dynamic and static data, node identification, anomaly detection, and the generation and execution of adjustment commands. The system begins with the data acquisition phase, where the processor collects dynamic and static data from each node in the artemisinin extraction process. Dynamic data includes real-time data from each node at different time points, such as temperature, pressure, and flow rate; this data is crucial for assessing the real-time status of the extraction process. Static data involves parameters such as the geometry, capacity, weight, and volume of each node; this information helps in understanding the physical characteristics of the equipment and its performance during the extraction process. By integrating this data, the system generates multidimensional feature data, laying the foundation for subsequent analysis.

[0035] Next, the system arranges the nodes according to the order of the artemisinin extraction process, forming a node sequence. This serialization provides a logical order for subsequent node discrimination and data analysis, enabling the system to more clearly identify the role of each node in the entire extraction process. Initializing the node discrimination coefficients is another crucial step; these coefficients represent the influence of each parameter on extraction efficiency and purity. By training and updating these coefficients, the system can derive the discrimination threshold for each parameter, facilitating effective comparisons in subsequent processing.

[0036] During anomaly detection, the system compares a discrimination threshold with the node's discrimination coefficient. If a node's discrimination coefficient exceeds the set threshold, the system marks it as an anomaly. At this point, the anomaly node's real-time dynamic data, static data, and corresponding parameter data affecting extraction efficiency and purity are categorized into the first type of data for in-depth analysis. The data for the remaining nodes are labeled as the second type of data. By analyzing these two types of data, the system can identify key parameters that may affect the extraction process.

[0037] During the data analysis phase, the system utilizes association rule analysis technology to filter out frequent itemsets containing data for each parameter and generate association rules. These rules reveal the potential relationships between different parameters, thereby helping the system understand which factors significantly affect the extraction efficiency and purity of artemisinin.

[0038] Based on the generated association rules, the system can identify specific parameter data affecting extraction efficiency and purity, and generate adjustment commands accordingly. These adjustment commands will be sent to the corresponding nodes to adjust the relevant parameters to optimize the artemisinin extraction process. This process may involve real-time adjustment of process parameters such as temperature, pressure, and flow rate to ensure that the extraction process is always in optimal condition, maximizing the yield and purity of artemisinin.

[0039] The process of collecting dynamic and static data, as well as parameter data affecting extraction efficiency and purity at each node during artemisinin extraction, to generate multidimensional feature data includes: normalizing the dynamic and static data; arranging the dynamic and static data in chronological order to obtain the original data sequence; segmenting the original data sequence to obtain multiple data subsequences; constructing a first-class data matrix with the data subsequences as rows and the first feature vector as columns; constructing a second-class feature vector from the parameter data affecting extraction efficiency and purity corresponding to each node, and arranging the second-class feature vectors in chronological order to obtain a second-class data matrix; wherein, the first feature vector includes the normalized dynamic and static data of each node; and the second feature vector includes the normalized parameter data affecting extraction efficiency and purity of each node.

[0040] The steps involved in collecting and processing dynamic and static data, as well as parameter data affecting extraction efficiency and purity, at each stage of artemisinin extraction to generate multidimensional feature data demonstrate a significant difference from existing technologies. Traditional extraction processes often rely on single, static data collection methods, lacking comprehensive analysis and processing of dynamic changes. This approach, however, normalizes both dynamic and static data and arranges them chronologically to construct an original data sequence. This method not only improves data comparability but also ensures that the impact of time on extraction efficiency and purity is fully considered during analysis. Furthermore, by segmenting the original data sequence to generate multiple data subsequences and constructing first and second type data matrices, this systematic feature extraction method makes data analysis more comprehensive and detailed, providing a solid data foundation for subsequent efficiency and purity optimization.

[0041] The nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence. Then, a third type of data matrix is ​​constructed according to the arrangement order of the node sequence, with each node as a row and the third feature vector as a column. The second feature vector includes the parameter data of each node that affect the extraction efficiency and purity. The discrimination threshold is multiplied by the parameter data of each node in the second type of data matrix that affect the extraction efficiency and purity to obtain the third type of data matrix.

[0042] Arranging the nodes according to the order of the artemisinin extraction process, the step of constructing the third type of data matrix demonstrates the further innovation of this scheme. Existing technologies often lack in-depth analysis of the interactions between nodes, while this scheme, by combining node sequences with feature vectors, can systematically analyze the impact of each node on extraction efficiency and purity. By multiplying the discrimination threshold by the parameter data in the second type of data matrix, the third type of data matrix is ​​obtained. This method not only enhances the operability of the data but also provides a quantitative basis for subsequent judgments and decisions. Compared with traditional methods, this scheme has greater depth in the dynamic analysis of relationships between nodes and can more scientifically guide the optimization of the extraction process.

[0043] The node discrimination coefficients are trained and updated to obtain the discrimination threshold, which includes: acquiring historical training data, which includes historical process data and label data; wherein, the label data consists of labels marked with the corresponding parameter data of each node; constructing a classification data matrix based on the historical process data; and training a neural network model based on the classification data matrix to obtain the judgment threshold; wherein, the network model includes a feature extraction layer, an activation layer, and a classification layer connected in sequence.

[0044] Regarding the training and updating of node discrimination coefficients, this scheme acquires historical training data, including historical process data and label data, constructs a classification data matrix, and trains it using a neural network model to derive the judgment threshold. The main difference between this process and existing technologies is that traditional methods often rely on experience or simple statistical analysis, lacking a systematic learning and adjustment mechanism. This scheme, however, introduces a neural network model and utilizes deep learning methods to mine potential relationships in the data, thereby improving the accuracy and adaptability of the discrimination threshold. The stepwise connection design of the feature extraction layer, activation layer, and classification layer provides stronger expressive power for learning complex data relationships, significantly improving the level of intelligence.

[0045] The process of determining abnormal nodes includes: taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the abnormal node as the first type of data; otherwise, taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the remaining nodes as the second type of data, and using association rule analysis to filter out frequent itemsets containing each parameter data to generate association rules.

[0046] In the process of identifying anomalous nodes, this solution demonstrates its difference from existing technologies by systematically analyzing the real-time dynamic data, static data, and corresponding parameter data affecting extraction efficiency and purity of the nodes. Existing technologies often lack real-time monitoring and analysis of anomalous states, while this solution uses association rule analysis to filter frequent itemsets and generate association rules, thereby more effectively identifying anomalous nodes. By defining the data discriminant coefficients of nodes and their differences, and combining preset thresholds with the distance between the node discriminant coefficient vectors, a quantitative method for anomaly detection is provided. This method not only improves the accuracy of anomaly detection but also searches for potential anomaly patterns in multi-dimensional data, offering a more comprehensive and in-depth approach compared to traditional single-indicator analysis methods.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An auxiliary system for artemisinin extraction, characterized in that, Includes a processor and a memory. The processor acquires dynamic data, static data, and parameter data affecting extraction efficiency and purity at each node during the artemisinin extraction process, and generates multidimensional feature data. The dynamic data includes real-time dynamic data of each node at different time points during the artemisinin extraction process, and the static data includes the geometry, capacity, weight and volume of each node. The nodes are arranged in the order of the artemisinin extraction process to obtain the node sequence; Initialize the node discrimination coefficients, which are the weights of parameters affecting extraction efficiency and purity; The node discrimination coefficients are trained and updated to obtain the discrimination threshold; Abnormal nodes are identified based on the comparison between the discrimination threshold and the node discrimination coefficient. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the abnormal nodes are taken as the first type of data. The real-time dynamic data, static data, and corresponding parameter data affecting the extraction efficiency and purity of the remaining nodes are taken as the second type of data. Through association rule analysis, frequent itemsets containing each parameter data are selected to generate association rules. Based on the association rules, the parameter data affecting the extraction efficiency and purity are determined, adjustment commands are generated, and the adjustment commands are sent to the corresponding nodes to adjust the parameter data and the process parameters in the artemisinin extraction process.

2. The auxiliary system for artemisinin extraction as described in claim 1, characterized in that: The process of collecting dynamic and static data, as well as parameter data affecting extraction efficiency and purity at each node during artemisinin extraction, to generate multidimensional feature data includes: normalizing the dynamic and static data; arranging the dynamic and static data in chronological order to obtain the original data sequence; segmenting the original data sequence to obtain multiple data subsequences; constructing a first-class data matrix with the data subsequences as rows and the first feature vector as columns; constructing a second-class feature vector from the parameter data affecting extraction efficiency and purity corresponding to each node, and arranging the second-class feature vectors in chronological order to obtain a second-class data matrix; wherein, the first feature vector includes the normalized dynamic and static data of each node; and the second feature vector includes the normalized parameter data affecting extraction efficiency and purity of each node.

3. The auxiliary system for artemisinin extraction as described in claim 2, characterized in that: The nodes are arranged according to the order of the artemisinin extraction process to obtain a node sequence. Then, a third type of data matrix is ​​constructed according to the arrangement order of the node sequence, with each node as a row and the third feature vector as a column. The second feature vector includes the parameter data of each node that affect the extraction efficiency and purity. The discrimination threshold is multiplied by the parameter data of each node in the second type of data matrix that affect the extraction efficiency and purity to obtain the third type of data matrix.

4. The auxiliary system for artemisinin extraction as described in claim 1, characterized in that: The node discrimination coefficients are trained and updated to obtain the discrimination threshold, which includes: acquiring historical training data, which includes historical process data and label data; wherein, the label data consists of labels marked with the corresponding parameter data of each node; constructing a classification data matrix based on the historical process data; and training a neural network model based on the classification data matrix to obtain the judgment threshold; wherein, the network model includes a feature extraction layer, an activation layer, and a classification layer connected in sequence.

5. The auxiliary system for artemisinin extraction as described in claim 1, characterized in that: The process of determining abnormal nodes includes: taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the abnormal node as the first type of data; otherwise, taking the real-time dynamic data, static data, and corresponding parameter data that affect the extraction efficiency and purity of the remaining nodes as the second type of data, and using association rule analysis to filter out frequent itemsets containing each parameter data to generate association rules.

Images