Automated reagent storage method, device and system

By dividing areas and setting temperature monitoring nodes in the reagent storage device, building a temperature correlation network, identifying and labeling abnormal reagents, the problem of reagent performance degradation caused by uneven temperature in the reagent storage device is solved, and the efficiency and safety of reagent management are improved sex.

CN118365142BActive Publication Date: 2024-11-26SHENZHEN YIZHILIAN JEWELRY TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410782883.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-18
Publication Date
2024-11-26
Estimated Expiration
2044-06-18

AI Technical Summary

Technical Problem

Existing reagent storage devices cause uneven temperatures under frequent operation, affecting the biological activity and chemical properties of the reagents, making it difficult for users to identify potential abnormal reagents, resulting in performance degradation or failure.

Method used

By dividing independent areas within the reagent storage device and setting up multiple temperature monitoring nodes, a temperature correlation network is built, cluster analysis and time series anomaly detection, potential abnormal areas are identified, initial attribute data are collected, manifold learning and sparse encoding are performed Processing, combining distributed hashing and decision tree algorithms, identifying and labeling abnormalities kits.

Benefits of technology

It realizes accurate monitoring and abnormal identification of internal temperature changes of reagent storage equipment, improves reagent management efficiency and safety, and ensures that reagent quality meets standards.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118365142B_ABST
    Figure CN118365142B_ABST
Patent Text Reader

Abstract

The present invention provides an automated reagent storage method, device and system, which includes dividing each area of ​​a reagent storage device into multiple nodes; analyzing and processing the temperature data of each node to obtain a temperature correlation network and a temperature risk map; identifying potential anomalies between areas through a temperature correlation network and updating the temperature risk map of each area, performing time series anomaly detection on the temperature risk map of each area to obtain the environmental risk level of each area; obtaining an initial attribute data set marked as a potential abnormal state and performing manifold learning and sparse coding processing, comparing the detection data of a potential abnormal test kit with a predetermined reference standard, retrieving matching features through a distributed hash algorithm, and comparing the detection feature data through a decision tree algorithm to obtain a test kit that does not meet the standard. The method can timely discover reagents that have failed or degraded in performance in the equipment and visually mark them.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of reagent storage, and in particular to an automated reagent storage method, device and system. Background Art

[0002] Existing reagent storage methods mainly rely on reagent storage equipment that can provide a low-temperature environment. In actual operation, the reagent storage equipment needs to be opened and closed frequently to take out or put in reagents. Each time the device door is opened, the outside heat will enter the device, which will cause the internal temperature of the device to rise in a short period of time and cause uneven temperature distribution inside the device.

[0003] In addition, some reagent storage devices have large storage spaces, so even without the influence of external heat, the temperature at different locations inside is uneven. This is because the distribution of cold air flow is affected by the internal structure of the equipment. For example, the location of the refrigeration elements and the layout of the fans may cause some areas to have a lower cooling effect than other areas. At the same time, the circulation and distribution of cold air may be limited by the physical arrangement of the equipment, the reagent storage method, and the overall design of the device, making it impossible to reach all storage areas evenly.

[0004] Reagents, especially chemical and biological reagents, are extremely sensitive to environmental conditions, especially temperature. Once they are subjected to inappropriate temperature fluctuations, their biological activity or chemical properties may be affected. Under the influence of the above-mentioned phenomena, existing reagent storage devices often have some reagents with degraded performance or even failed, but users cannot know whether there are problems with the reagents they obtain from the device. In view of this, technical personnel in this field are urgently needed to make improvements. Summary of the invention

[0005] The main purpose of the present invention is to solve the technical problem that during the storage process of the reagent in the reagent storage device, the performance of the reagent may be degraded or fail, while the user cannot know whether there is a problem with the obtained reagent.

[0006] A first aspect of the present invention provides an automated reagent storage method, the automated reagent storage method comprising:

[0007] Dividing the internal space of the reagent storage device into multiple independent areas, and dividing each area into multiple nodes;

[0008] The temperature data of each node is collected, and the correlation analysis is performed on the temperature data of each node to obtain a temperature correlation network for identifying the temperature correlation between nodes. The temperature data of each node is clustered to generate a temperature risk map for each area.

[0009] Analyzing the temperature correlation network to identify potential anomalies between regions and updating the temperature risk map of each region, performing time series anomaly detection on the temperature risk map of each region to obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, marking the test kit of the corresponding region as a potential abnormal state;

[0010] Collecting initial attribute data of the reagent kit marked as being in a potential abnormal state, wherein the initial attribute data includes optical data of the reagent kit and temperature data of the reagent kit, to obtain an initial attribute data set;

[0011] Manifold learning and sparse coding are performed on the initial attribute data set to determine whether there are potential abnormal test kits, the detection data of the potential abnormal test kits are compared with the predetermined reference standards, matching features are retrieved through a distributed hash algorithm, and the detection feature data are compared through a decision tree algorithm to obtain test kits that do not meet the standards and perform visual marking processing on them.

[0012] Preferably, the collecting of temperature data of each node and performing correlation analysis on the temperature data of each node to obtain a temperature correlation network for identifying temperature correlations between nodes includes:

[0013] The temperature data of each node is collected according to the preset sampling frequency, and denoised to obtain a continuous temperature data set of each node;

[0014] The temperature data sets of each node are compared pairwise using a statistical correlation algorithm to obtain a temperature correlation matrix;

[0015] Using each node in the temperature correlation matrix as a node in a temperature correlation network;

[0016] Setting the correlation strength threshold for screening node pairs with high correlation;

[0017] Connecting the node pairs whose association strength reaches or exceeds the association strength threshold, generating edges between these node pairs, and forming an edge set representing the temperature association between the nodes;

[0018] The nodes and edge sets are processed comprehensively to obtain the temperature correlation network representing the temperature associations between nodes.

[0019] Preferably, the association strength threshold is calculated by the following formula:

[0020] ;

[0021] Where T represents the calculated association strength threshold, is the average mutual information value between all node pairs, k is the adjustment parameter, is the standard deviation of the mutual information values ​​between all node pairs, is the weight parameter, is the sum of entropy values ​​of all nodes, and N is the total number of nodes.

[0022] Preferably, clustering the temperature data of each node to generate a temperature risk map of each area includes:

[0023] Performing data preprocessing on the temperature data set of each node, wherein the data preprocessing includes removing outliers, filling missing values ​​and normalizing to obtain a continuous temperature data set on the same scale;

[0024] Perform feature extraction on the continuous temperature data set at the same scale to obtain the feature vector of each node;

[0025] A clustering algorithm is used to cluster the feature vectors of each node, identify node groups with similar temperature distribution characteristics, and assign cluster labels to obtain a cluster label set for temperature risk classification.

[0026] Different risk levels are set according to the temperature characteristics of the cluster label set, and each label is assigned a high, medium, or low risk label to obtain the temperature risk level of each cluster label.

[0027] According to the clustering label and its corresponding risk level, a color or shape mark is assigned to each node and mapped to the spatial layout diagram of the reagent storage device to obtain a temperature risk map representing the temperature risk level distribution of each node.

[0028] Preferably, the kit for performing manifold learning and sparse coding processing on the initial attribute data set to determine whether there is a potential abnormality includes:

[0029] Performing dimensionality reduction processing on the initial attribute data set using a manifold learning algorithm to convert the multi-dimensional optical and temperature data into a low-dimensional manifold feature vector;

[0030] Performing sparse coding on the manifold feature vector to obtain a sparse feature vector set;

[0031] Performing statistical attribute analysis on the sparse feature vector set to obtain an abnormality score for each test kit;

[0032] The status of each test kit is determined based on a pre-set scoring threshold, and a list of potentially abnormal test kits and their identification information is generated.

[0033] Preferably, the sparse feature vector set is subjected to statistical attribute analysis to obtain an abnormality score for each kit, which is calculated by the following formula:

[0034] ;

[0035] in, is the final abnormality score of the test kit, M is the number of statistical attributes in the sparse feature vector, is the weight of the ith statistical attribute, is the value of the i-th statistical attribute.

[0036] A second aspect of the present invention provides an automated reagent storage device, the automated reagent storage device comprising:

[0037] A region division module is used to divide the internal space of the reagent storage device into multiple independent regions, and each region is divided into multiple nodes;

[0038] The temperature monitoring module is used to collect the temperature data of each node, perform correlation analysis on the temperature data of each node, obtain a temperature correlation network for identifying the temperature correlation between nodes, perform clustering on the temperature data of each node, and generate a temperature risk map for each area;

[0039] An anomaly detection module is used to analyze the temperature correlation network to identify potential anomalies between regions and update the temperature risk map of each region, perform time series anomaly detection on the temperature risk map of each region, and obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, the test kit of the corresponding region is marked as a potential abnormal state;

[0040] An attribute collection module is used to collect initial attribute data of the reagent box marked as a potential abnormal state, wherein the initial attribute data includes optical data of the reagent box and temperature data of the reagent box, to obtain an initial attribute data set;

[0041] The risk management module is used to perform manifold learning and sparse coding processing on the initial attribute data set to determine whether there are potential abnormal test kits, compare the detection data of the potential abnormal test kits with predetermined reference standards, retrieve matching features through a distributed hash algorithm, and compare the detection feature data through a decision tree algorithm to obtain test kits that do not meet the standards and perform visual marking processing on them.

[0042] The third aspect of the present invention provides an automated reagent storage system, including a reagent storage device, a reagent in-and-out storage mechanism, and a control device;

[0043] The reagent storage device comprises a body, wherein a plurality of reagent kit storage positions are formed in the body;

[0044] The reagent in-and-out storage mechanism comprises a conveying member and a transporting member provided on the body, wherein the conveying member is used to transport the reagent kit between the inside of the body and the outside of the body, and the transporting member is used to transport the reagent kit between the reagent kit storage position and the conveying member;

[0045] The control device includes a processor and an automated reagent storage device, the processor can call instructions in the automated reagent storage device, and the automated reagent storage device includes a region division module, a temperature monitoring module, an abnormality detection module, an attribute acquisition module and a risk management module;

[0046] The area division module is used to divide the internal space of the reagent storage device into multiple independent areas, and divide each area into multiple nodes;

[0047] The temperature monitoring module is used to collect the temperature data of each node, perform correlation analysis on the temperature data of each node, obtain a temperature correlation network for identifying the temperature correlation between nodes, perform clustering on the temperature data of each node, and generate a temperature risk map for each area;

[0048] The anomaly detection module is used to analyze the temperature correlation network to identify potential anomalies between regions and update the temperature risk map of each region, perform time series anomaly detection on the temperature risk map of each region, and obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, the test kit of the corresponding region is marked as a potential abnormal state;

[0049] The attribute collection module is used to collect initial attribute data of the reagent box marked as a potential abnormal state, wherein the initial attribute data includes optical data of the reagent box and temperature data of the reagent box, to obtain an initial attribute data set;

[0050] The risk management module is used to perform manifold learning and sparse coding processing on the initial attribute data set, determine whether there are potential abnormal test kits, compare the detection data of the potential abnormal test kits with predetermined reference standards, retrieve matching features through a distributed hash algorithm, and compare the detection feature data through a decision tree algorithm to obtain test kits that do not meet the standards and perform visual marking processing on them.

[0051] The automated reagent storage method of the present invention has the following advantages:

[0052] By dividing the reagent storage device into independent areas and further subdividing each area into multiple monitoring nodes, a fine-grained temperature monitoring network is constructed. This structure allows real-time collection and detailed correlation analysis of temperature data at each node, thus forming a comprehensive temperature correlation network. This network can not only accurately track the temperature changes that occur during frequent use of storage devices, but also effectively identify high-risk areas where temperature fluctuations occur due to frequent operations.

[0053] After using a clustering algorithm to analyze the temperature data of each node, the system can generate detailed temperature risk maps, which are then analyzed using time series anomaly detection technology to assess the environmental risk level of each area. Once the environmental risk level of an area exceeds a preset threshold, the area is marked as a potential abnormal state.

[0054] Next, optical and temperature data are collected for the test kits in the areas marked as potential abnormal conditions, and the data are processed using manifold learning and sparse coding techniques to optimize the data dimensionality reduction and encoding process. By making comprehensive use of these processed data and the preset scoring threshold, the system can effectively identify and judge the abnormal state of each test kit, thereby generating an accurate list of potential abnormal test kits and identification information. Through visual marking, users can easily identify possible risk items in storage devices, that is, test kits whose performance may be affected by temperature fluctuations. These marked risk items enable users to conduct appropriate inspections and evaluations before use through obvious visual prompts, ensuring that only reagents that meet quality standards are used for medical and scientific research. This process significantly improves the management efficiency and safety of test kits, and provides strong support for experimental and medical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of an embodiment of an automated reagent storage method in an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of an embodiment of an automated reagent storage device in an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the structure of an embodiment of the automated reagent storage system of the present invention.

[0059] Description of Figure Numbers:

[0060] 1. Reagent storage device; 11. Machine body; 12. Reagent reagent storage position; 2. Reagent in and out of the warehouse mechanism; 21. Conveying parts; 22. Transport parts; 3. Control equipment.

[0061] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0064] To facilitate understanding of this embodiment, firstly, an automated reagent storage method disclosed in an embodiment of the present invention is described in detail. Figure 1 As shown, the method comprises the following steps:

[0065] S100, dividing the internal space of the reagent storage device into multiple independent areas, and dividing each area into multiple nodes;

[0066] Independent areas are multiple spaces divided inside the reagent storage device. Each area is relatively independent and has its own temperature monitoring mechanism. Usually the size and number of these independent areas are set according to the specific storage requirements of the reagents.

[0067] Nodes are multiple monitoring points or collection points in an independent area. They usually represent the installation location of temperature sensors and are used to collect temperature data in independent areas in real time. By setting up multiple nodes in each independent area, the temperature changes in the entire area can be more comprehensively monitored, abnormal temperature fluctuations can be captured, and local temperature problems can be accurately identified, providing reliable basic data for subsequent data analysis and temperature management.

[0068] For example, the internal space of the reagent storage device can be divided into three areas: upper, middle, and lower to ensure the uniformity of temperature monitoring. Each area is further divided into three nodes: front, middle, and back, representing the installation positions of temperature sensors. In this way, there are three nodes in each area to collect real-time data through temperature sensors.

[0069] S200, collecting temperature data of each node, performing correlation analysis on the temperature data of each node, obtaining a temperature correlation network for identifying temperature correlations between nodes, performing clustering processing on the temperature data of each node, and generating a temperature risk map for each area;

[0070] The specific steps include:

[0071] S201, collecting temperature data of each node according to a preset sampling frequency, and performing denoising processing to obtain a continuous temperature data set of each node;

[0072] In order to ensure that the temperature data of each node can accurately reflect the environmental conditions of the independent area, it is necessary to regularly collect and record the temperature data of each node. By setting the preset sampling frequency, real-time temperature data can be obtained within a fixed time interval. This can not only maintain the integrity of the data, but also detect temperature fluctuations or anomalies in a timely manner.

[0073] During the collection process, temperature data may be affected by factors such as sensor accuracy and environmental interference (such as opening and closing the door of the device), resulting in noise or anomalies in the data. Therefore, after collecting the temperature data, it is necessary to perform denoising processing on it to eliminate the interference of the noise data. Denoising methods include moving average, filtering or median filtering, which can effectively smooth the data curve, retain the main trend of temperature changes, and ensure that the final result is a stable and continuous temperature data set.

[0074] S202, using a statistical correlation algorithm to perform pairwise comparisons on the temperature data sets of each node to obtain a temperature correlation matrix;

[0075] The temperature data sets of each node are compared pairwise, and the correlation between each node is calculated. Statistical correlation algorithms, such as the Pearson correlation coefficient or the Spearman correlation coefficient, are used to quantify the temperature correlation between each pair of nodes. Finally, the correlation results between all nodes are organized into a matrix structure to generate a temperature correlation matrix. Each element of the matrix represents the degree of correlation between two nodes. This matrix will be used to identify the temperature correlation relationship within the region.

[0076] Take the nodes in an independent area as an example. Assume that there are four nodes A, B, C, and D in the independent area. The correlation between the nodes is calculated by the statistical correlation algorithm to generate the following matrix:

[0077] .

[0078] Each value in the matrix represents the correlation between two nodes. For example, the correlation coefficient between A and B is 0.8, indicating that the temperature data between the two nodes is strongly correlated.

[0079] S203, using each node in the temperature correlation matrix as a node in a temperature correlation network;

[0080] After generating the temperature correlation matrix, each node needs to be mapped to the temperature correlation network as an independent node. The temperature correlation network is a graph structure, in which each node represents a temperature monitoring point in the reagent storage device. Through this graph structure, the temperature correlation relationship between the monitoring points in the reagent storage device can be intuitively reflected.

[0081] Based on the temperature correlation matrix generated in step S202, A, B, C and D will be respectively used as nodes in the temperature correlation network.

[0082] S204, setting a correlation strength threshold for screening node pairs with high correlation;

[0083] In order to identify node pairs with high correlation in the temperature correlation network, it is necessary to set a correlation strength threshold. Node pairs with correlation coefficients higher than the threshold will be identified as node pairs with strong temperature correlation, and thus will be further analyzed and processed in the temperature correlation network. Assuming that the correlation strength threshold is set to 0.7, then in the temperature correlation matrix, the correlation coefficients of A and B, A and D, and C and D all exceed 0.7, and will be regarded as high correlation node pairs.

[0084] S205, connecting the node pairs whose association strength reaches or exceeds the association strength threshold, generating edges between these node pairs, and forming an edge set representing the temperature association between the nodes;

[0085] For the highly correlated node pairs screened out in step S204, edges will be created between these node pairs in the temperature correlation network to build a complete temperature correlation relationship. The formation of the temperature correlation edge set will help identify which nodes have a high correlation in temperature changes, facilitating further temperature risk analysis.

[0086] In the above example, since the correlations between A and B, A and D, and C and D exceed the threshold, these node pairs will be connected into edges in the temperature correlation network to form associated edge sets of AB, AD, and CD.

[0087] In the foregoing, the association strength threshold is calculated by the following formula:

[0088] .

[0089] Among them, T represents the calculated correlation strength threshold, which is used to determine whether the temperature correlation between the node pairs is significant enough. When the correlation strength between the node pairs exceeds this threshold, it indicates that the correlation between the two nodes is high. It is the average value of the mutual information between all node pairs, reflecting the overall level of correlation distribution in the entire node data set and indicating the central trend of data distribution. k is an adjustment parameter, and the k value is generally between 1 and 2. is the standard deviation of the mutual information values ​​between all node pairs, is the weight parameter, is the sum of the entropy values ​​of all nodes, and N is the total number of nodes, indicating how many nodes are contained in the entire data set.

[0090] Specifically:

[0091] .

[0092] in, is the mutual information value between node i and node j, and n is the total number of node pairs, that is, the number of all non-diagonal elements in the calculation matrix.

[0093] .

[0094] in, is the mutual information value between node i and node j, is the average mutual information value between all node pairs, and n is the total number of node pairs.

[0095] .

[0096] is the temperature data value in node i probability.

[0097] Specifically, using the mutual information matrix of nodes A, B, C, and D in the previous example, calculate the mean and standard deviation:

[0098] .

[0099] Calculate the average :

[0100] .

[0101] Calculate the average :

[0102] .

[0103] Assume that the entropy values ​​of nodes A, B, C, and D are:

[0104] .

[0105] .

[0106] .

[0107] .

[0108] .

[0109] Setting adjustment parameters and .

[0110] but .

[0111] According to the calculated threshold of 0.945, in the above mutual information matrix, the mutual information value of any node pair does not exceed 0.945. Therefore, no node pair has a high correlation in this example.

[0112] S206 , comprehensively process the nodes and edge sets to obtain a temperature correlation network representing the temperature correlation between nodes.

[0113] In the generated temperature correlation matrix, based on the threshold set previously, node pairs that meet the conditions are screened out as edges, and together with related nodes, they form a complete temperature correlation network. Each node represents a monitoring point, and each edge represents the association relationship between two nodes. Through this network, the temperature correlation pattern of the entire system can be identified, and it can be clarified which nodes are associated in temperature data.

[0114] Based on the previously calculated threshold of 0.945, in the temperature correlation network, no mutual information value of any node pair exceeds the threshold, so these nodes do not form an associated edge set. It should be understood that in practical applications, the threshold can be adjusted as needed to find suitable node pairs to construct a temperature correlation network.

[0115] S207, performing data preprocessing on the temperature data set of each node, wherein the data preprocessing includes removing outliers, filling missing values, and normalizing to obtain a continuous temperature data set on the same scale;

[0116] In the preprocessing of temperature data, first of all, it is necessary to remove outliers that are obviously deviated from the reasonable range to prevent these data from misleading the analysis results. Secondly, for missing values, appropriate filling methods need to be adopted, such as using the average or median of neighboring data to fill the gaps. Finally, through normalization processing, the temperature data of different nodes are uniformly scaled to the same scale range to facilitate subsequent clustering and analysis.

[0117] Assume that the temperature data set of node A is: [21.2, 21.4, NaN, 21.7, 21.9], where NaN represents missing values. In data preprocessing, if the temperature value exceeds the normal range (such as 10-40°C), it is considered an outlier and removed, and the mean of the adjacent data is used to fill the missing value, resulting in [21.2, 21.4, 21.55, 21.7, 21.9]. The data is scaled to [0, 1] for normalization.

[0118] S208, extracting features from the continuous temperature data set on the same scale to obtain a feature vector of each node;

[0119] In the preprocessed temperature data set, key feature information needs to be extracted for clustering. Feature extraction can include key statistical information such as the mean, standard deviation, kurtosis, skewness, maximum value, and minimum value of the data.

[0120] S209, using a clustering algorithm to perform clustering processing on the feature vectors of each node, identifying a node group with similar temperature distribution characteristics, and assigning clustering labels to obtain a clustering label set for temperature risk classification;

[0121] The feature vectors of all nodes are aggregated into a matrix, and then clustering algorithms (such as K-means, hierarchical clustering) are used to cluster nodes with similar temperature distribution characteristics. Each cluster group represents a typical temperature distribution pattern. Each node is assigned a cluster label to facilitate node management and temperature risk classification.

[0122] Assume that the eigenvector matrix of all nodes is as follows:

[0123] .

[0124] A clustering algorithm is used to divide these nodes into three categories, and each node will be assigned a cluster label (such as 1, 2, 3).

[0125] S210, setting different risk levels according to the temperature characteristics of the cluster label set, assigning high, medium, and low risk labels to each label, and obtaining the temperature risk level of each cluster label;

[0126] Each cluster label group represents a temperature distribution feature, and different risk levels are set according to its average temperature, temperature change and other information. For example, the label group with high temperature and large fluctuation is assigned a high risk level, and the label group with stable temperature and within the normal range is assigned a low risk level.

[0127] As in the above example, cluster label 1 represents a node group with a higher average temperature and is assigned a high-risk label; cluster label 2 represents a node group with a relatively stable temperature and is assigned a low-risk label.

[0128] S211. According to the clustering labels and their corresponding risk levels, a color or shape mark is assigned to each node, and mapped to the spatial layout diagram of the reagent storage device to obtain a temperature risk map representing the temperature risk level distribution of each node.

[0129] Assuming that the high-risk node group is assigned red and the low-risk node group is assigned green, the labels of each node are mapped to the layout diagram of the reagent storage device. The high-risk nodes are displayed as red dots and the low-risk nodes are displayed as green dots on the diagram, thereby forming a temperature risk map.

[0130] S300, analyzing the temperature correlation network to identify potential anomalies between regions and updating the temperature risk map of each region, performing time series anomaly detection on the temperature risk map of each region to obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, marking the test kit in the corresponding region as a potential abnormal state;

[0131] It is not difficult to understand that the temperature correlation network is a graph structure. The nodes represent various temperature monitoring points, and the edges represent the temperature correlation relationship between the nodes. Through graph analysis methods, such as community detection or cluster analysis, different areas in the temperature correlation network can be identified. For each area, its overall temperature change trend is calculated, and areas that are significantly different from historical trends are found. When analyzing the temperature change trend of each area, if the correlation between a certain area and other areas is significantly reduced or shows abnormal fluctuations, the area can be marked as a potential abnormal area. Further check the temperature monitoring points in these potential abnormal areas to identify specific abnormal nodes. According to the identified abnormal area, adjust its risk level to high risk, update the temperature risk map of the area, and maintain its existing risk level for low-risk or normal areas to monitor the overall environmental conditions in real time.

[0132] The predetermined threshold is a numerical upper or lower limit used to determine the regional temperature risk level, and the threshold is usually determined based on historical data analysis or experimental experience.

[0133] Assume that there are two regions A and B in the temperature correlation network. The temperature change trend of region A remains stable compared with historical data, and the correlation between nodes is consistent with the predetermined threshold, so it is rated as a normal risk region. The temperature fluctuation of region B is large, and the correlation with other regions is significantly reduced. Through graph analysis, it is found that the correlation between region B and other regions is lower than the predetermined threshold of 0.8, and the node temperature fluctuates abnormally. Therefore, region B is marked as a potential abnormal area, and the temperature risk map is updated to a high-risk state.

[0134] From the area that has been identified as abnormal, obtain a list of all test kits stored therein, and record the basic information of these test kits, including the kit ID, storage location, and specific attributes. These test kits will be given abnormal labels or marks, which are usually recorded in the management system, or physical identification is directly added to the test kits for subsequent processing.

[0135] S400, collecting initial attribute data of the reagent kit marked as being in a potential abnormal state, wherein the initial attribute data includes optical data of the reagent kit and temperature data of the reagent kit, to obtain an initial attribute data set;

[0136] Once an area is marked as a potential abnormal state, it is necessary to further collect initial attribute data for the reagent kits in the area, including optical data and temperature data. Optical data can reflect the color, transparency and other characteristics of the reagents, while temperature data directly reflects the ambient temperature of the reagents. By collecting this data, a complete initial attribute data set is formed, providing a basis for further analysis and abnormality detection.

[0137] S500, performing manifold learning and sparse coding processing on the initial attribute data set to determine whether there is a potential abnormal test kit, comparing the test data of the potential abnormal test kit with a predetermined reference standard, retrieving matching features through a distributed hash algorithm, and comparing the test feature data through a decision tree algorithm to obtain test kits that do not meet the standards, and visually marking them.

[0138] The specific steps include:

[0139] S501, performing dimensionality reduction processing on the initial attribute data set using a manifold learning algorithm to convert the multi-dimensional optical and temperature data into a low-dimensional manifold feature vector;

[0140] Specifically, before processing, the initial property data set is standardized to ensure that the optical and temperature data are comparable on the same scale. Data standardization usually uses normalization or Z-score standardization to scale all feature values ​​to the same range or mean.

[0141] The multidimensional distance between each pair of kits was calculated, characterized by optical and temperature data, and the commonly used distance calculation methods included Euclidean distance and cosine similarity.

[0142] Local Linear Embedding (LLE) or t-SNE is selected as the manifold learning algorithm to map the high-dimensional optical and temperature data into a low-dimensional space using a distance matrix. The low-dimensional feature representation preserves the local relationships and data structure in the high-dimensional space.

[0143] For ease of understanding, the following example assumes that the initial property data set includes the following temperature and optical data:

[0144] Temperature data: [21.5, 21.8, 22.0, 22.3, 22.5];

[0145] Optical data: [0.8, 0.85, 0.9, 0.75, 0.7] (representing properties such as color or transparency).

[0146] By using the t-SNE algorithm, these data are reduced to a two-dimensional manifold feature space and the following feature vectors are obtained:

[0147] .

[0148] S502, sparsely encode the manifold feature vector to obtain a sparse feature vector set;

[0149] In the implementation process, a dictionary containing sparse basis vectors is first created. These basis vectors are used to represent the manifold feature vectors. Using sparse representation methods such as LASSO or OMP (Orthogonal Matching Pursuit), each manifold feature vector is represented as a linear combination of the dictionary basis vectors. During the encoding process, the number of basis vectors used in the combination is minimized to obtain the sparse feature vector. Through the above encoding, a sparse representation of each manifold feature vector is finally obtained.

[0150] S503, performing statistical attribute analysis on the sparse feature vector set to obtain an abnormality score for each test kit;

[0151] Specifically, the sparse feature vector set is analyzed, and the statistical properties of each kit are calculated using the components of each feature vector, including the mean, variance, skewness and other indicators. These statistical properties will be used as feature functions The input is used to construct the abnormality score of each kit. It is calculated by the following formula:

[0152] ;

[0153] in, is the final abnormality score of the test kit, M is the number of statistical attributes in the sparse feature vector, is the weight of the ith statistical attribute, is the value of the i-th statistical attribute.

[0154] Suppose we have a sparse feature vector of the test kit [0.7, 0.9, 0.2, 0.0]. Based on this sparse feature vector, calculate the mean ,variance , skewness , assuming the weights are and .

[0155] Calculating anomaly score .

[0156] By comparing the anomaly scores With the pre-set scoring threshold, it can be determined whether the kit is in a potential abnormal state and marked or included in the abnormal list.

[0157] S504: Determine the status of each test kit based on a preset scoring threshold, and generate a list of potential abnormal test kits and their identification information.

[0158] The pre-set scoring threshold is a numerical standard used to determine whether the abnormal score of the test kit indicates an abnormal state. This threshold can be a fixed value or a result of dynamic calculation based on a statistical model. The determination of its specific value is usually based on historical data statistics, business needs (the strictness of the threshold is adjusted according to the purpose of the test kit and the sensitivity to tolerance for misclassification. For example, for high-value or critical reagents, a lower threshold may be required to conservatively handle potential anomalies) and risk assessment.

[0159] After that, the test data of the potentially abnormal test kit is compared with the predetermined reference standard. The reference standard may include optical properties, temperature data and other characteristic values, which usually represent the typical data range and pattern that a healthy test kit should have. Comparing the test data with the standard can preliminarily screen out those test kits that may not meet the standard. However, the simple comparison operation is inefficient when the amount of data is large, so a fast retrieval algorithm is needed - the distributed hashing algorithm.

[0160] The distributed hash algorithm maps feature data into hash values ​​and stores them in a distributed hash table to achieve efficient matching of a large amount of detection data. Specifically, the detection data can be encoded through the hash function h to obtain the hash value of each detection data. , and quickly find the corresponding position in the hash table for matching:

[0161] .

[0162] in, To detect the hash value of data X, h is the hash function, is the value of each feature in the detection data.

[0163] The decision tree algorithm is a commonly used machine learning model that can screen out kits that do not meet the standards step by step by dividing nodes with features. When building a decision tree model, it is necessary to use training data to determine the division criteria for each node (such as information gain or Gini coefficient). In the specific process of comparing test data, each node will check whether the feature value of the test data meets the standard, and finally screen out the kits that do not meet the standards step by step.

[0164] The judgment formula for feature partitioning nodes is as follows:

[0165] .

[0166] in, is the output of the decision node, is the eigenvalue used in the current node, and t is the threshold for eigenvalue division.

[0167] Suppose we find a potentially abnormal test kit in the test data: [21.5, 0.9, 0.2] (temperature, optics, and other attributes). Use the hash function to encode the test data and get the hash value , locates the matching position in the distributed hash table, and finds that the position indicates that there may be an anomaly.

[0168] The first-layer node of the decision tree determines whether the temperature attribute is less than or equal to the predetermined threshold value 22.0. The result is True, and the next layer of nodes is entered. The second-layer node determines whether the optical attribute is less than or equal to 0.8. The result is False, indicating that the optical attribute is abnormal. Through the judgment of this node, the test kit is marked as a risky product.

[0169] In this way, the combination of the distributed hashing algorithm and the decision tree algorithm can quickly and efficiently screen out test kits that do not meet the standards and visually mark them (visual marking can be achieved in a variety of ways, including but not limited to using color coding, flashing lights, reflective labels, LED indicators or other luminous elements) for easy user identification.

[0170] The above describes the automated reagent storage method in the embodiment of the present invention. The following describes the automated reagent storage device in the embodiment of the present invention. Figure 2 , an embodiment of the automated reagent storage device in the embodiment of the present invention includes:

[0171] The area division module 201 is used to divide the internal space of the reagent storage device into multiple independent areas, and divide each area into multiple nodes;

[0172] The temperature monitoring module 202 is used to collect temperature data of each node, perform correlation analysis on the temperature data of each node, obtain a temperature correlation network for identifying temperature correlations between nodes, perform clustering processing on the temperature data of each node, and generate a temperature risk map for each area;

[0173] The anomaly detection module 203 is used to analyze the temperature correlation network to identify potential anomalies between regions and update the temperature risk map of each region, perform time series anomaly detection on the temperature risk map of each region, and obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, the test kit of the corresponding region is marked as a potential abnormal state;

[0174] The attribute collection module 204 is used to collect initial attribute data of the reagent kit marked as a potential abnormal state, wherein the initial attribute data includes optical data of the reagent kit and temperature data of the reagent kit, to obtain an initial attribute data set;

[0175] The risk management module 205 is used to perform manifold learning and sparse coding processing on the initial attribute data set to determine whether there are potential abnormal test kits, compare the detection data of the potential abnormal test kits with a predetermined reference standard, retrieve matching features through a distributed hash algorithm, and compare the detection feature data through a decision tree algorithm to obtain test kits that do not meet the standards and perform visual marking processing on them.

[0176] The above describes in detail the automated reagent storage device in the embodiment of the present invention from the perspective of modular functional entities, and the following describes in detail the automated reagent storage system in the embodiment of the present invention from the perspective of entity structure.

[0177] Figure 3 1 is a schematic diagram of the structure of an automated reagent storage system device provided by an embodiment of the present invention, wherein the automated reagent storage system comprises a reagent storage device 1, a reagent in-and-out storage mechanism 2, and a control device 3;

[0178] The reagent storage device 1 comprises a body 11 , in which a plurality of reagent kit storage positions 12 are formed; specifically, the reagent kit storage positions 12 are used for storing the reagent kits, ensuring that the reagent kits can be taken or stored when needed.

[0179] The reagent in-and-out storage mechanism 2 includes a conveying member 21 and a transporting member 22 provided on the body 11, wherein the conveying member 21 is used to transport the reagent kit between the inside of the body 11 and the outside of the body 11, and the transporting member 22 is used to transport the reagent kit between the reagent kit storage position 12 and the conveying member 21;

[0180] In specific implementation, in some embodiments, the conveyor 21 includes a conveyor belt, which can be an annular or linear conveyor belt, responsible for transporting the reagent kit between the inside and outside of the body 11. In other embodiments, the conveyor 21 includes a slide rail system, which realizes the linear movement of the reagent kit through a slide rail installed on the body 11 and a pulley installed on the slide rail.

[0181] The transporting member 22 may include a manipulator equipped with a multi-degree-of-freedom manipulator arm that can accurately grasp and place the reagent kit, and the end of the manipulator may be equipped with a dedicated grasping tool, such as a vacuum suction cup or a clamp. The transporting member 22 may also include a lifting platform, which is used to vertically transport the reagent kit to achieve the movement of the reagent kit between storage locations at different heights. In addition, the transporting member 22 may also include a mobile robot or a turntable system, etc., which is not specifically limited here.

[0182] The control device 3 includes a processor (central processing unit, CPU) and an automated reagent storage device, and the processor can call instructions in the automated reagent storage device.

[0183] The control device 3 may be a computer or an industrial controller capable of running dedicated software. In the process of the processor calling instructions, the steps of the above-mentioned automated reagent storage method may be implemented.

[0184] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automated reagent storage method, characterized in that: include: Dividing the internal space of the reagent storage device into multiple independent areas, and dividing each area into multiple nodes; The temperature data of each node is collected, and the correlation analysis is performed on the temperature data of each node to obtain a temperature correlation network for identifying the temperature correlation between nodes. The temperature data of each node is clustered to generate a temperature risk map for each area. Analyzing the temperature correlation network to identify potential anomalies between regions and updating the temperature risk map of each region, performing time series anomaly detection on the temperature risk map of each region to obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, marking the test kit of the corresponding region as a potential abnormal state; Collecting initial attribute data of the reagent kit marked as being in a potential abnormal state, wherein the initial attribute data includes optical data of the reagent kit and temperature data of the reagent kit, to obtain an initial attribute data set; The initial attribute data set is subjected to manifold learning and sparse coding processing to determine whether there is a potential abnormal test kit, and the detection data of the potential abnormal test kit is compared with a predetermined reference standard, wherein the comparison process includes retrieving matching features through a distributed hash algorithm, and comparing the detection feature data through a decision tree algorithm to obtain a test kit that does not meet the standard and perform visual marking processing on it.

2. The automated reagent storage method according to claim 1, wherein: The temperature data of each node is collected, and correlation analysis is performed on the temperature data of each node to obtain a temperature correlation network for identifying temperature correlations between nodes, including: The temperature data of each node is collected according to the preset sampling frequency, and denoised to obtain a continuous temperature data set of each node; The temperature data sets of each node are compared pairwise using a statistical correlation algorithm to obtain a temperature correlation matrix; Using each node in the temperature correlation matrix as a node in a temperature correlation network; Setting the correlation strength threshold for screening node pairs with high correlation; Connecting the node pairs whose association strength reaches or exceeds the association strength threshold, generating edges between these node pairs, and forming an edge set representing the temperature association between the nodes; The nodes and edge sets are processed comprehensively to obtain the temperature correlation network representing the temperature associations between nodes.

3. The automated reagent storage method according to claim 2, wherein: The association strength threshold is calculated by the following formula: ; in, represents the calculated association strength threshold, is the average mutual information value between all node pairs, To adjust the parameters, is the standard deviation of the mutual information values ​​between all node pairs, is the weight parameter, is the sum of entropy values ​​of all nodes, is the total number of nodes.

4. The automated reagent storage method according to claim 2, wherein: The temperature data of each node is clustered to generate a temperature risk map of each area, including: Performing data preprocessing on the temperature data set of each node, wherein the data preprocessing includes removing outliers, filling missing values ​​and normalizing to obtain a continuous temperature data set on the same scale; Perform feature extraction on the continuous temperature data set at the same scale to obtain the feature vector of each node; A clustering algorithm is used to cluster the feature vectors of each node, identify node groups with similar temperature distribution characteristics, and assign cluster labels to obtain a cluster label set for temperature risk classification. Different risk levels are set according to the temperature characteristics of the cluster label set, and each label is assigned a high, medium, or low risk label to obtain the temperature risk level of each cluster label. According to the clustering label and its corresponding risk level, a color or shape mark is assigned to each node and mapped to the spatial layout diagram of the reagent storage device to obtain a temperature risk map representing the temperature risk level distribution of each node.

5. The automated reagent storage method according to claim 1, wherein: The kit for performing manifold learning and sparse coding processing on the initial attribute data set to determine whether there is a potential abnormality includes: Performing dimensionality reduction processing on the initial attribute data set using a manifold learning algorithm to convert the multi-dimensional optical and temperature data into a low-dimensional manifold feature vector; Performing sparse coding on the manifold feature vector to obtain a sparse feature vector set; Performing statistical attribute analysis on the sparse feature vector set to obtain an abnormality score for each test kit; The status of each test kit is determined based on a pre-set scoring threshold, and a list of potentially abnormal test kits and their identification information is generated.

6. The automated reagent storage method according to claim 5, characterized in that: The statistical attribute analysis is performed on the sparse feature vector set to obtain the abnormality score of each kit, which is calculated by the following formula: ; in, The final abnormality score of the kit, is the number of statistical attributes in the sparse feature vector, For the The weight of a statistical attribute, For the The value of a statistical attribute.

7. An automated reagent storage device, characterized in that: The automated reagent storage device comprises: A region division module is used to divide the internal space of the reagent storage device into multiple independent regions, and each region is divided into multiple nodes; The temperature monitoring module is used to collect the temperature data of each node, perform correlation analysis on the temperature data of each node, obtain a temperature correlation network for identifying the temperature correlation between nodes, perform clustering on the temperature data of each node, and generate a temperature risk map for each area; An anomaly detection module is used to analyze the temperature correlation network to identify potential anomalies between regions and update the temperature risk map of each region, perform time series anomaly detection on the temperature risk map of each region, and obtain the environmental risk level of each region, wherein if the environmental risk level exceeds a predetermined threshold, the test kit of the corresponding region is marked as a potential abnormal state; An attribute collection module is used to collect initial attribute data of the reagent box marked as a potential abnormal state, wherein the initial attribute data includes optical data of the reagent box and temperature data of the reagent box, to obtain an initial attribute data set; The risk management module is used to perform manifold learning and sparse coding processing on the initial attribute data set to determine whether there are potential abnormal test kits, and compare the detection data of the potential abnormal test kits with a predetermined reference standard, wherein the comparison process includes retrieving matching features through a distributed hash algorithm, and comparing the detection feature data through a decision tree algorithm to obtain test kits that do not meet the standards and perform visual marking processing on them.

8. An automated reagent storage system, characterized in that: Including reagent storage equipment, reagent in and out of the warehouse mechanism and control equipment; The reagent storage device comprises a body, wherein a plurality of reagent kit storage positions are formed in the body; The reagent in-and-out storage mechanism comprises a conveying member and a transporting member provided on the body, wherein the conveying member is used to transport the reagent kit between the inside of the body and the outside of the body, and the transporting member is used to transport the reagent kit between the reagent kit storage position and the conveying member; The control device includes a processor and an automated reagent storage device as described in claim 7, and the processor can call instructions in the automated reagent storage device.