Ocean sample full life cycle digital management method

Abstract

The invention relates to the technical field of ocean sample management, and discloses an ocean sample full-life-cycle digital management method, which comprises the following steps of: firstly, collecting characteristic data such as sample types, source positions and chemical component information of ocean samples, and calculating a classification scheme and storage parameters according to the characteristic data; evaluating a management demand and generating a target management strategy by combining the real-time management data of the sample library; making and optimizing a preliminary management plan by using algorithms such as k-means clustering and a decision tree in combination with historical management data and sample use frequency data, and generating a personalized management scheme; and finally, implementing the scheme, monitoring a sample preservation rate index through a sensor, dynamically adjusting a temperature threshold value and a storage position of classified storage equipment, introducing a real-time feedback mechanism to iteratively correct and adjust parameters, and generating a full-life-cycle management scheme. According to the method, full-process digital management of the ocean samples is realized, and the management efficiency and the sample storage quality are improved.

Description

Technical Field

[0001] This invention relates to the field of marine sample management technology, specifically a digital management method for the entire lifecycle of marine samples. Background Technology

[0002] In marine scientific research and resource development, marine samples serve as crucial research objects and data carriers, and the scientific rigor and effectiveness of their management directly impact the accuracy and reliability of research findings. With the continuous advancement of marine exploration technology and the increasing frequency of marine development activities, the number of marine samples has exploded, and their types have become increasingly diverse, encompassing seawater, sediments, organisms, and other samples. This presents unprecedented challenges to marine sample management.

[0003] Traditional marine sample management methods largely rely on manual operation and simple information recording. In the sample collection stage, comprehensive and accurate collection of sample characteristic data is often lacking, making it difficult to accurately obtain key data such as sample type, origin, and chemical composition. This results in a lack of a reliable foundation for subsequent management. Regarding sample classification and storage, the lack of scientific calculation methods and data analysis tools makes it impossible to rationally determine classification schemes and storage parameters based on sample characteristic data. This leads to chaotic sample storage, making it difficult to meet the specific storage requirements of different samples, and consequently affecting the preservation quality of the samples.

[0004] In assessing the management needs of sample libraries, traditional methods fail to comprehensively consider factors such as classification schemes, storage parameters, and real-time management data, making it difficult to generate scientific and reasonable target management strategies, resulting in low efficiency in sample library management. Furthermore, the lack of effective analysis and utilization of historical management data and sample usage frequency data when formulating management plans hinders personalization and optimization of these plans, making it difficult to adapt to the management needs of different samples.

[0005] Furthermore, traditional methods lack real-time monitoring and dynamic adjustment mechanisms during management, making it impossible to dynamically adjust the temperature thresholds and storage locations of classification and storage equipment based on real-time data such as sample preservation rates. This makes it difficult to guarantee the good preservation status of samples throughout their entire lifecycle. As marine scientific research places increasingly higher demands on sample management, the shortcomings of traditional management methods in terms of data processing capabilities, management efficiency, and accuracy are becoming increasingly apparent. There is an urgent need for a digital and intelligent full lifecycle management method to solve these problems. Summary of the Invention

[0006] The purpose of this invention is to provide a digital management method for the entire life cycle of marine samples, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for digital management of marine samples throughout their entire lifecycle, the method comprising:

[0008] Collect characteristic data of marine samples, including sample type, source location, and chemical composition information;

[0009] Based on the aforementioned characteristic data, the classification scheme and storage parameters of the samples are calculated;

[0010] Based on the classification scheme and storage parameters, the management needs of the sample library are assessed to generate a target management strategy.

[0011] Based on the target management strategy, historical management data, and sample usage frequency data, a preliminary management plan is formulated, the preliminary management plan is optimized, and a personalized management solution is generated.

[0012] By implementing the personalized management scheme, a sample preservation rate index is obtained during the management process. Based on the sample preservation rate index, the temperature threshold and storage location of the classification and storage equipment are dynamically adjusted to generate a full life cycle management scheme.

[0013] Preferably, based on the target management strategy, historical management data, and sample usage frequency data, a preliminary management plan is formulated, the preliminary management plan is optimized, and a personalized management solution is generated, including:

[0014] The target management strategy, historical management data, and sample usage frequency data are grouped using the k-means clustering method to obtain the management task classification.

[0015] Based on the aforementioned management task classification, a decision tree algorithm is applied to determine the optimal management sequence and optimal management frequency in order to generate a preliminary management plan.

[0016] Real-time sample status data is acquired and analyzed to calculate sample status changes and anomaly distribution. Based on the sample status changes and anomaly distribution, a long short-term memory network is used to predict the long-term management trend of the management system to refine the preliminary management plan and obtain a refined management plan.

[0017] The simulated annealing algorithm is used to simulate the application effects of different management strategies in the refined management plan in order to search for the combination of optimal management time and optimal management frequency. Based on the combination of optimal management time and optimal management frequency, the gradient descent optimization algorithm is used to optimize the refined management plan and generate a personalized management scheme.

[0018] Preferably, real-time sample status data is acquired, and the data is analyzed to calculate sample status changes and anomaly distributions. Based on these changes and anomaly distributions, a Long Short-Term Memory (LSTM) network is used to predict the long-term management trend of the management system, thereby refining the initial management plan and obtaining a refined management plan, including:

[0019] Acquire real-time sample status data, and apply a recurrent neural network to identify target sample features in the real-time sample status data to obtain sample location information;

[0020] Based on the sample location information, the sample state changes and abnormal distribution are calculated. The self-organizing map network is used to classify the sample state changes and abnormal distribution to obtain the classification results. Based on the classification results, the abnormal distribution pattern is identified.

[0021] The sample state changes, the abnormal distribution, and the abnormal distribution pattern are analyzed using a long short-term memory network to obtain management analysis results. Based on the management analysis results, the long-term management trend of the management system is predicted, and a long-term management trend prediction result is generated.

[0022] Based on the long-term management trend forecast results, the management frequency and management time corresponding to each management stage are adjusted to obtain the adjusted management frequency and management time. Using the adjusted management frequency and management time, the preliminary management plan is refined to obtain the refined management plan.

[0023] Preferably, based on the characteristic data, the classification scheme and storage parameters of the sample are calculated, including:

[0024] Based on the aforementioned characteristic data, fused sample data is generated;

[0025] Feature extraction is performed on the fused sample data to identify sample type features and chemical composition features. Based on the sample type features and chemical composition features, a preliminary sample distribution map is generated. The preliminary sample distribution map is updated in real time using an ant colony optimization algorithm to obtain an updated sample distribution map.

[0026] Based on the updated sample distribution map, the sample classification scheme and storage parameters are calculated using the simulated annealing algorithm and the static path planning algorithm.

[0027] Preferably, the personalized management scheme is implemented to obtain the sample retention rate index during the management process. Based on the sample retention rate index, the temperature threshold and storage location of the classification and storage equipment are dynamically adjusted to generate a full lifecycle management scheme, including:

[0028] Implement the personalized management scheme and use multiple sensors to monitor the sample preservation rate index in real time during the implementation of the personalized management scheme;

[0029] Based on the sample retention rate index, a recurrent neural network is used to predict the retention rate change trend. According to the retention rate change trend and the management needs of the sample library, a proportional-integral-derivative control system is used to dynamically adjust the temperature threshold and storage location of the classification and storage equipment to obtain the adjusted management operation results.

[0030] Based on the adjusted management operation results, a full lifecycle management scheme is generated using a genetic algorithm.

[0031] Preferably, based on the classification scheme and storage parameters, the management needs of the sample library are assessed to generate a target management strategy, including:

[0032] Real-time management data is acquired from sensors in the sample library, including temperature and humidity data;

[0033] The classification scheme, the storage parameters, and the real-time management data are jointly analyzed using a decision tree model to generate management demand assessment results.

[0034] Based on the management needs assessment results and combined with historical management efficiency data, target management parameters are optimized through linear regression to generate target management strategies.

[0035] Preferably, a decision tree model is used to jointly analyze the classification scheme, the storage parameters, and the real-time management data to generate a management needs assessment result, including:

[0036] Feature extraction is performed on the real-time management data to identify areas of abnormal temperature distribution and humidity concentration;

[0037] Based on the classification scheme and the storage parameters, the dynamic management logic of the system is modeled using Bayesian networks to generate a management logic description.

[0038] The abnormal temperature distribution, the humidity concentration area, and the management logic description are input into the decision tree model, and the management demand assessment results are output.

[0039] Preferably, the step of grouping the target management strategy, historical management data, and sample usage frequency data using the k-means clustering method further includes:

[0040] Monte Carlo simulation is introduced to simulate extreme usage frequency scenarios, generating an enhanced dataset;

[0041] Based on the augmented dataset, a robust management model is trained to optimize the classification of the management tasks.

[0042] Preferably, the collection of the characteristic data specifically includes:

[0043] Sample temperature data is acquired using a temperature sensor, and the source location distribution is analyzed in conjunction with a GPS locator.

[0044] Based on the sample temperature data and the distribution of the source locations, a sample feature map is generated;

[0045] The spatial relationship between the target sample and the abnormal sample is identified using the sample feature map.

[0046] Preferably, the method further includes: after dynamically adjusting the temperature threshold and storage location of the classified storage equipment, introducing a real-time feedback mechanism to monitor the management effect after the adjustment;

[0047] Based on the deviation between the monitoring results and the target retention rate, the temperature threshold and storage location adjustment parameters are iteratively corrected.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] The digital management method for the entire lifecycle of marine samples provided by this invention comprehensively collects characteristic data of marine samples, including sample type, origin, and chemical composition information, providing an accurate and comprehensive data foundation for subsequent management. Based on this characteristic data, the classification scheme and storage parameters of the samples are calculated, enabling scientific classification and rational storage of samples, ensuring that samples with different characteristics can be preserved under suitable conditions, and effectively improving the preservation quality of the samples.

[0050] When assessing sample library management needs and generating target management strategies, a comprehensive consideration of factors such as classification schemes, storage parameters, and real-time management data allows for a more accurate grasp of the actual management requirements of the sample library, leading to more targeted and scientific management strategies and improving the overall management level of the sample library. In the process of developing preliminary management plans and optimizing them to generate personalized management solutions, various advanced algorithms are used to analyze and process target management strategies, historical management data, and sample usage frequency data. This enables the development of more reasonable management sequences, frequencies, and times based on the characteristics and usage needs of different samples, achieving personalized and precise management and improving management efficiency and effectiveness.

[0051] When implementing personalized management solutions, multiple sensors are used to monitor sample preservation rates in real time. Based on this, the temperature thresholds and storage locations of the categorized storage equipment are dynamically adjusted. This allows for timely responses to various changes during sample preservation, ensuring that the sample preservation rate remains at a high level. A real-time feedback mechanism is introduced to monitor the effectiveness of the adjusted management. The adjustment parameters are iteratively corrected based on the deviation between the monitoring results and the target preservation rate, further improving the accuracy and reliability of management and achieving dynamic and precise management of marine samples throughout their entire lifecycle.

[0052] This method deeply integrates digital technology with marine sample management. Through data collection, analysis, and utilization, it achieves intelligent and automated management processes, greatly reducing errors and workload from manual operations, improving management efficiency and quality, and providing stronger support for marine scientific research and resource development. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the digital management method for the entire lifecycle of marine samples as described in this invention.

[0054] Figure 2 Design drawings generated for personalized management solutions;

[0055] Figure 3 Design diagram for real-time sample status data analysis;

[0056] Figure 4 This is a design diagram for calculating sample classification schemes and storage parameters. Detailed Implementation

[0057] 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.

[0058] Please see Figures 1-4 The present invention relates to a digital management method for the entire lifecycle of marine samples, the specific implementation steps of which are as follows:

[0059] Based on characteristic data, the classification scheme and storage parameters of the samples are calculated. First, fused sample data is generated based on the characteristic data. Features are extracted to identify sample type characteristics and chemical composition characteristics, thereby generating a preliminary sample distribution map. The ant colony optimization algorithm is used to update the preliminary sample distribution map in real time to obtain an updated sample distribution map. Based on this, the simulated annealing algorithm and static path planning algorithm are used to calculate the sample classification scheme and storage parameters.

[0060] Based on the classification scheme and storage parameters, the management needs of the sample library are assessed to generate a target management strategy. Real-time management data, including temperature and humidity data, is acquired from sensors in the sample library. Feature extraction is performed on the real-time management data to identify abnormal temperature distributions and humidity clusters. Based on the classification scheme and storage parameters, a Bayesian network is used to model the dynamic management logic of the management system to generate a management logic description. The abnormal temperature distributions, humidity clusters, and management logic descriptions are input into a decision tree model to output the management needs assessment results. Based on these results and historical management efficiency data, the target management parameters are optimized through linear regression to generate a target management strategy.

[0061] Based on the target management strategy, historical management data, and sample usage frequency data, a preliminary management plan is developed, the preliminary management plan is optimized, and a personalized management solution is generated. Monte Carlo simulation is introduced to simulate extreme usage frequency scenarios and generate an enhanced dataset. Based on this, a robust management model is trained to optimize the management task classification obtained by grouping target management strategies, historical management data, and sample usage frequency data using the k-means clustering method. Based on the management task classification, a decision tree algorithm is applied to determine the optimal management sequence and optimal management frequency to generate a preliminary management plan. Real-time sample status data is acquired, and a recurrent neural network is applied to identify target sample features to obtain sample location information. Based on this, sample status changes and abnormal distributions are calculated, and a self-organizing map network is used to classify them and identify abnormal distribution patterns. A long short-term memory network is used to analyze sample status changes, abnormal distributions, and abnormal distribution patterns to obtain management analysis results. Based on this, the long-term management trend of the management system is predicted to generate long-term management trend prediction results. Based on this, the management frequency and management time corresponding to each management stage are adjusted to obtain the adjusted management frequency and management time. The preliminary management plan is refined to obtain a refined management plan. Simulated annealing algorithm is used to simulate the application effect of different management strategies in the refined management plan to search for the optimal combination of management time and optimal management frequency. Based on this, gradient descent optimization algorithm is used to optimize the refined management plan and generate a personalized management scheme.

[0062] A personalized management plan is implemented to obtain the sample preservation rate index during the management process. Based on the sample preservation rate index, the temperature threshold and storage location of the sorting and storage equipment are dynamically adjusted to generate a full life cycle management plan. The personalized management plan is implemented and multiple sensors are used to monitor the sample preservation rate index in real time. Based on this index, a recurrent neural network is used to predict the preservation rate change trend. According to the trend and the management needs of the sample library, a proportional-integral-derivative control system is used to dynamically adjust the temperature threshold and storage location of the sorting and storage equipment to obtain the adjusted management operation results. Based on this, a genetic algorithm is used to generate a full life cycle management plan. After dynamic adjustment, a real-time feedback mechanism is introduced to monitor the management effect of the adjustment. Based on the deviation between the monitoring results and the target preservation rate, the adjustment parameters of the temperature threshold and storage location are iteratively corrected.

[0063] Example 1:

[0064] When calculating the classification scheme and storage parameters for samples, the first step is to generate fused sample data based on the collected marine sample characteristic data. This characteristic data covers multiple dimensions, including sample type, origin, and chemical composition information. The fusion process standardizes these different types of data, integrating them under the same data structure to form a unified dataset containing all key information, laying the foundation for subsequent analysis.

[0065] After data fusion is completed, feature extraction is performed on the fused sample data. This step uses specific data processing algorithms and models to accurately identify sample type and chemical composition characteristics from the integrated data. For sample type characteristics, the specific type of each sample can be determined by parsing the sample type identifier field in the data and combining it with preset sample type classification criteria, such as distinguishing between plankton samples and seabed sediment samples. The extraction of chemical composition characteristics may utilize spectroscopic analysis techniques or chemical analysis methods to measure the various chemical components and their concentrations contained in the sample, transforming this information into quantifiable and analyzable data characteristics.

[0066] Based on the extracted sample type and chemical composition characteristics, a preliminary sample distribution map is generated. This map, grounded in spatial geographic information, marks samples with different characteristics on corresponding geographic coordinates according to their origin location information. Different types of samples or samples with different chemical compositions are distinguished by different icons or colors, thus visually demonstrating the spatial distribution of various samples and providing a visual reference for subsequent management and planning.

[0067] To ensure the sample distribution map reflects real-time dynamic changes in the samples, an ant colony optimization algorithm is used to update the initial map in real time. This algorithm simulates the pheromone transmission mechanism of ants foraging in a colony. Ants release pheromones while searching for a path, and subsequent ants choose paths based on pheromone concentration, gradually finding the optimal path. In this application, the locations of each sample and possible updated paths on the map are considered the ants' foraging environment. Through pheromone updates and diffusion, the algorithm optimizes and updates the sample distribution information on the map, promptly reflecting the addition of newly collected samples, changes in the location or state of existing samples, etc., ultimately resulting in an updated sample distribution map that more accurately reflects the current actual distribution status of the samples.

[0068] Based on the updated sample distribution map, simulated annealing and static path planning algorithms are used to calculate the sample classification scheme and storage parameters. The core idea of ​​simulated annealing is to simulate the gradual energy reduction characteristic of a physical system during annealing. During optimization, the algorithm accepts solutions worse than the current one with a certain probability, thus avoiding getting trapped in local optima and enabling the search for better classification and storage schemes within a larger solution space. In practice, a series of possible combinations of classification methods and storage locations are set as initial solutions. These solutions are then continuously adjusted, and new solutions are accepted or rejected according to the rules of simulated annealing, gradually optimizing the classification scheme and storage parameters.

[0069] Static path planning algorithms, given a known sample distribution and storage area layout, plan the optimal path from the sample collection or processing location to the storage location to determine the best storage location and order for each sample. This algorithm considers factors such as the spatial structure of the storage area, the storage requirements of different sample types, path length and convenience, and uses mathematical models and optimization methods to calculate the optimal storage path and location allocation scheme. For example, for samples requiring storage in specific temperature or humidity environments, the algorithm will prioritize allocating suitable storage locations and plan the shortest or most convenient transportation path to ensure safe and efficient sample storage.

[0070] By combining simulated annealing and static path planning algorithms, and comprehensively considering factors such as sample characteristics, storage environment requirements, and spatial distribution, a scientifically sound sample classification scheme and storage parameters are calculated. The classification scheme clearly defines the grouping method and classification criteria for different types of samples, while the storage parameters include the specific storage location of each sample, storage environment conditions (such as temperature and humidity), storage method (such as refrigeration and sealing), and storage order. This provides detailed guidance for subsequent sample management, ensuring that samples are properly preserved during storage and meeting the requirements of full lifecycle management.

[0071] Example 2:

[0072] When developing initial management plans and optimizing personalized management solutions, Monte Carlo simulations are needed to simulate extreme usage frequency scenarios. Monte Carlo simulation is a method that simulates uncertain events through random sampling. In this case, it specifically generates a large number of random samples that conform to extreme usage conditions based on sample usage frequency data. For example, it simulates scenarios where some samples are used frequently in a short period of time, or extreme cases where some samples are not used for a long period of time, thereby constructing an augmented dataset covering multiple extreme usage scenarios. Compared to the original data, this augmented dataset can more comprehensively reflect the distribution of sample usage frequency under different extreme conditions, providing richer input data for subsequent management model training.

[0073] Based on the generated augmented dataset, a robust management model is further trained to optimize the classification of management tasks. The training process of the robust management model involves learning and analyzing the target management strategies, historical management data, and sample usage frequency data in the augmented dataset using machine learning algorithms. The model automatically identifies patterns and features in the data, grouping tasks with similar management needs and usage patterns into the same category. For example, frequently used sample management tasks with similar management strategies are grouped into one category, while low-frequency used tasks with different management requirements are grouped into another. This optimizes the management task classification obtained by grouping the data using the k-means clustering method, making the classification results more robust and better able to adapt to various extreme situations that may arise in actual management processes.

[0074] After optimizing the classification of management tasks, a decision tree algorithm is applied based on the classification results to determine the optimal management order and frequency, thereby generating a preliminary management plan. The decision tree algorithm starts from the root node and uses the characteristics of the management tasks (such as usage frequency and management strategy complexity) as the decision basis to partition the data layer by layer, constructing a decision tree. During construction, the algorithm selects the features that most effectively partition the data as nodes until a preset stopping condition is met. This decision tree clearly determines the priority and management frequency of different management tasks. For example, for frequently used sample management tasks, the decision tree will provide a more frequent management frequency and a higher priority management order to ensure that these samples are managed and maintained in a timely manner, while for infrequently used tasks, a relatively lower management frequency is determined, thus generating a preliminary management plan framework.

[0075] To obtain sample real-time status data, a recurrent neural network (RNN) is applied to identify target sample features and determine sample location information. This real-time status data includes real-time monitoring data such as sample temperature, humidity, and positional changes. This data is collected in real-time by various sensors installed in the sample storage device and environment and transmitted to the system. The RNN has the ability to process sequential data, enabling it to learn and analyze real-time status data in time-series format. In application, the real-time status data is input into the RNN. Through its internal loop structure and memory units, the network models the temporal dependencies in the data, thereby identifying features related to the target sample, such as specific sample identification information and positional change trajectories, ultimately determining the sample's specific location within the storage area.

[0076] Based on the acquired sample location information, the changes in sample state and the distribution of anomalies are calculated. The calculation of sample state changes involves comparing sample location information and environmental parameters at different time points to analyze trends in sample state changes related to location movement and environmental adaptation. For example, comparing the current sample location with historical locations determines whether movement has occurred, and the frequency and range of such movement; analyzing changes in environmental parameters such as temperature and humidity determines whether the sample's environment is stable. The calculation of anomaly distribution involves setting a threshold range for normal states, considering state data exceeding this range as abnormal, and statistically analyzing the spatial and temporal distribution of these abnormal data, such as which regions and time periods are more prone to sample state anomalies.

[0077] Self-organizing map (SOM) networks are used to classify sample state changes and abnormal distributions, obtaining classification results and identifying abnormal distribution patterns. SOM networks are unsupervised learning neural networks that can map high-dimensional sample state data to a low-dimensional space while preserving the data's topological structure. In application, the calculated sample state change and abnormal distribution data are input into the SOM network, which automatically clusters and classifies the data, mapping data with similar state changes and abnormal distribution characteristics to nearby locations, thus obtaining classification results. Analysis of the classification results can identify different types of abnormal distribution patterns, such as persistent abnormal patterns in certain areas caused by environmental factors, or sudden abnormal patterns caused by changes in usage frequency over certain time periods.

[0078] Long Short-Term Memory (LSTM) networks are used to analyze sample state changes, anomaly distributions, and anomaly patterns to obtain management analysis results. Based on these results, long-term management trends of the management system are predicted, generating long-term management trend prediction results. As a special type of recurrent neural network, LSM networks can effectively handle the vanishing gradient problem in long-sequence data, making them suitable for analyzing long-term management data. Sample state change data, anomaly distribution data, and identified anomaly distribution pattern data are input into the LSM network. Through its unique memory unit structure, the network learns and models long-term dependencies in the data, thereby analyzing the operational status and potential problems of the management system at different stages, yielding management analysis results. Based on these results, the future long-term management trends of the management system are predicted, such as the changing trends in the usage frequency of different types of samples and the probability of possible anomalies over a future period, generating long-term management trend prediction results.

[0079] Based on long-term management trend forecasts, the management frequency and time for each management stage are adjusted to obtain the adjusted management frequency and time. Specifically, the management frequency and time set in the initial management plan are dynamically adjusted according to the changes in management needs at different stages as predicted. For example, if it is predicted that the usage frequency of a certain type of sample will increase significantly in the future, the management frequency of that type of sample will be increased accordingly, and the management interval will be shortened; if it is predicted that an environmental anomaly may occur in a certain area during a specific period, the management time for that area will be adjusted in advance, and the number of inspections will be increased. Using the adjusted management frequency and time, the initial management plan is refined to obtain the refined management plan, making the management plan more in line with future actual management needs.

[0080] Simulated annealing is used to simulate the application effects of different management strategies in the refined management plan, searching for the optimal combination of management time and frequency. Here, simulated annealing uses different combinations of management time and frequency as candidate solutions, simulating their application effects in real-world management scenarios to calculate the corresponding objective function values ​​(such as management efficiency and resource utilization). Starting from an initial solution, the algorithm generates new solutions according to certain rules and, based on the acceptance criteria of simulated annealing, accepts solutions worse than the current solution with a certain probability, thus conducting a broad search in the solution space and avoiding getting trapped in local optima. Through iterative searching, the optimal combination of management time and frequency is finally found. Based on this combination, gradient descent optimization is used to optimize the refined management plan, generating a personalized management scheme. Gradient descent optimization calculates the gradient of the objective function and continuously adjusts the parameters in the management plan along the gradient descent direction, gradually reducing the objective function value until the optimal personalized management scheme is obtained. This scheme can provide precise and efficient management strategies tailored to the characteristics and management needs of different samples.

[0081] Example 3:

[0082] When assessing the management needs of the sample library to generate a target management strategy, the first step is to acquire real-time management data from various sensors within the library. These sensors, including temperature and humidity sensors, are distributed across different areas of the library and continuously monitor environmental parameters in real time. Temperature sensors collect temperature data from each storage area in real time, while humidity sensors simultaneously acquire humidity data for their respective areas. This data is transmitted in real time to the management system's database via wired or wireless transmission, forming timestamped real-time management data records that provide the latest environmental status information for subsequent management needs assessments.

[0083] After acquiring real-time management data, feature extraction is required to identify abnormal temperature distributions and humidity clusters. During feature extraction, the system first sets normal threshold ranges for temperature and humidity, determined based on the storage requirements of different types of marine samples and the standard environmental parameters of the sample library. Then, the real-time temperature data is compared with these thresholds. When the temperature value in a certain area exceeds the normal range, it is determined to be a temperature anomaly, and the location and severity of the anomaly are recorded. Similarly, humidity data is compared with humidity thresholds to identify anomalies. Simultaneously, the spatial distribution characteristics of the humidity data are analyzed to identify areas where humidity values ​​are relatively concentrated and above the normal range—i.e., humidity clusters. For example, data analysis might reveal that the humidity in a corner of the sample library is consistently higher than in other areas, or that the temperature fluctuations of a certain storage rack frequently exceed the standard range. These are key pieces of information identified through feature extraction.

[0084] Based on the established classification scheme and storage parameters, a Bayesian network is used to model the dynamic management logic of the management system, generating a description of the management logic. The classification scheme clarifies the grouping methods and storage requirements for different types of marine samples, while the storage parameters specify the specific storage location, environmental conditions, and other information for each type of sample. A Bayesian network is a graphical model based on probabilistic reasoning. It uses nodes to represent various factors in the management system, such as sample type, storage location, and environmental parameters, and edges to represent the probabilistic dependencies between these factors. During the modeling process, the system determines the conditional probability distribution between nodes based on historical management data and expert experience, thereby constructing a Bayesian network model that reflects the dynamic management logic of the management system. For example, the model can describe the probability of changes in environmental temperature and humidity affecting the preservation status of a certain type of sample when stored in a specific location, as well as the probability distribution of the corresponding management measures that need to be adjusted. Ultimately, it generates a detailed description of the dynamic management logic of the management system, which reflects the interaction and influence between various management elements in the form of probabilistic relationships.

[0085] The identified temperature anomalies, humidity clusters, and management logic descriptions are input into a decision tree model. Through analysis and processing by the decision tree model, a management needs assessment result is output. The decision tree model is a hierarchical classification and regression model that constructs a decision tree by progressively dividing the input data. Each internal node represents a feature judgment, and each leaf node represents an assessment result. In application, the decision tree model first uses temperature anomalies and humidity clusters as initial partitioning features to analyze the impact of these environmental anomalies on different categories of samples. Then, combining the probabilistic relationships in the management logic description, it further determines the type and priority of management measures to be taken. For example, for areas with temperature anomalies and containing temperature-sensitive samples, the decision tree model will prioritize assessing the need for immediate environmental regulation in that area. For areas with humidity clusters but relatively high sample tolerance, it may assess a need for regular monitoring. Through this layered analysis and decision-making, the final output is a management needs assessment result for the current state of the sample library, which clarifies key information such as the management priority and type of management measures for each area and type of sample.

[0086] Based on the management needs assessment results and combined with historical management efficiency data, the system optimizes target management parameters through linear regression to generate target management strategies. Historical management efficiency data records the effects and efficiencies of different management measures taken under similar management needs in the past, including information such as the execution time of management measures, resource consumption, and improvement in sample preservation status. Linear regression is a statistical method used to establish linear relationships between variables; here, it is used to analyze the linear relationship between target management parameters (such as the operating parameters of environmental control equipment, the frequency of inspections by management personnel, etc.) and management efficiency. The system uses the various management needs from the management needs assessment results as input variables and the corresponding management efficiencies from the historical management efficiency data as output variables to construct a linear regression model. Through training and optimization of the model, the linear relationship expression between target management parameters and management efficiency is determined, thereby calculating the optimal target management parameter values ​​based on the current management needs assessment results. For example, based on the linear regression results, it is determined that in a certain temperature anomaly area, the cooling power of air conditioning equipment needs to be adjusted to a specific value, and the inspection frequency of management personnel needs to be increased to a corresponding number of times to achieve optimal management efficiency. Based on these optimized target management parameters, a specific target management strategy is generated. This strategy details the management measures, execution steps, timelines, and resource allocation for the current state of the sample library, providing clear guidance for subsequent sample management and ensuring that the management of the sample library can meet the needs of digital management of the entire life cycle of marine samples.

[0087] Example 4:

[0088] When implementing personalized management plans and generating full lifecycle management plans, the first step is to apply these plans to the actual management of marine samples. Taking a type of temperature-sensitive plankton sample as an example, the personalized management plan might specify the storage temperature range, inspection frequency, and specific environmental monitoring requirements for this type of sample. During implementation, various sensors are needed to monitor sample retention rates in real time. These sensors include temperature sensors, humidity sensors, pressure sensors, and image sensors for monitoring sample condition. For instance, a high-precision temperature sensor can be installed in the refrigerated cabinet storing plankton samples, collecting temperature data every 10 minutes. Simultaneously, an image sensor can periodically photograph the sample's appearance to promptly detect any deterioration or damage.

[0089] Based on the sample retention rate index obtained from real-time monitoring, a recurrent neural network (RNN) is used to predict the trend of retention rate changes. RNNs have the ability to process time-series data and can learn and analyze historical retention rate data. Taking planktonic samples as an example, retention rate data for this type of sample over the past week, month, or even longer, along with corresponding environmental parameters (such as temperature fluctuation range and humidity changes), are input into the RNN. The network models the time dependencies in the data through its internal memory units and loop structures, thereby predicting the retention rate trend of this type of sample over a future period. For example, the prediction results may show that if the current temperature control accuracy continues, the retention rate of this type of planktonic sample will remain at a high level for the next two weeks; however, if the temperature fluctuation range increases, the retention rate may show a downward trend.

[0090] Based on the retention rate trend and the management needs of the sample library, a proportional-integral-derivative (PID) control system is used to dynamically adjust the temperature threshold and storage location of the sorting and storage equipment. Taking planktonic samples as an example, if the recurrent neural network predicts a declining retention rate, and analysis reveals that this is due to temperature fluctuations in the refrigeration unit exceeding the tolerance range of this type of sample, the PID control system will automatically calculate the temperature threshold that needs adjustment based on the retention rate trend and preset management requirements (such as ensuring a retention rate of no less than 95%). For example, the original temperature threshold range may be adjusted from 2±0.5℃ to 2±0.3℃, and an adjustment command may be sent to the refrigeration unit's temperature control device to improve temperature control accuracy. Simultaneously, if poor temperature stability is detected in a certain storage area, the PID control system will also consider the sample's storage location information and suggest transferring that type of planktonic sample to a more temperature-stable area, thus obtaining the adjusted management operation results.

[0091] Based on the adjusted management operation results, a full life-cycle management plan is generated using a genetic algorithm. A genetic algorithm is an optimization algorithm that simulates the biological evolution process, finding the optimal solution through operations such as selection, crossover, and mutation. In generating the full life-cycle management plan, the adjusted management operation results (such as adjusted temperature thresholds and storage location adjustment suggestions) are first used as part of the initial population. This is combined with other possible management strategies and plans to construct a population containing multiple management scheme possibilities. Then, through the selection operation of the genetic algorithm, management schemes that perform well in the simulated environment are selected; through crossover, the advantages of different management schemes are combined; and through mutation, new management strategies and ideas are introduced, thereby continuously optimizing the population and ultimately generating a full life-cycle management plan for this type of marine sample. For example, the generated full life-cycle management plan may include detailed management measures for each stage from sample collection, transportation, storage to use and disposal, such as temperature control requirements during collection, shock protection measures during transportation, regular inspection plans during storage, and disposal procedures after use.

[0092] After dynamically adjusting the temperature thresholds and storage locations of the categorized storage equipment, a real-time feedback mechanism is needed to monitor the effectiveness of the adjustments. Taking planktonic samples as an example, after adjusting the temperature thresholds of the refrigerated cabinet and the sample storage location, the real-time feedback mechanism will continue to collect data on the sample preservation status and environmental parameters through various sensors, and compare and analyze this data with the data before the adjustment. For example, it monitors whether the temperature fluctuations after the adjustment are controlled within the new threshold range, and whether the sample preservation rate has improved or remained stable.

[0093] Based on the deviation between the monitoring results and the target preservation rate, the system iteratively adjusts the temperature threshold and storage location parameters. If monitoring reveals that the adjusted preservation rate still deviates from the target preservation rate (e.g., 95%), for example, the preservation rate only reaches 92%, the system analyzes the cause of the deviation. This could be due to insufficient adjustment of the temperature threshold or the storage location adjustment failing to fully address environmental stability issues. In this case, the system iteratively adjusts the temperature threshold and storage location parameters according to the magnitude and direction of the deviation. For example, it might further reduce the temperature threshold by 0.1℃, or reassess and select a more suitable storage area, and then implement adjustments again. The system monitors the new adjustment effects through a real-time feedback mechanism, repeating this iterative process until the deviation between the monitoring results and the target preservation rate meets the requirements, ensuring optimal preservation and management of marine samples throughout the entire management process.

[0094] Example 5:

[0095] When collecting characteristic data from marine samples, taking multiple samples collected during a deep-sea scientific expedition as an example, temperature sensors were used to acquire sample temperature data. This temperature sensor, integrated into the sampler, records the ambient temperature in real time when collecting seabed sediment and plankton samples. For instance, when collecting sediment at a depth of 500 meters, the sensor recorded a temperature of 4.2℃. Simultaneously, combined with GPS location analysis, the GPS precisely recorded the collection coordinates of the sediment sample as 28°15′N, 122°30′E. Based on the sample temperature data and source location distribution, a sample characteristic map is generated. This map uses geographical location as the horizontal axis and temperature as the vertical axis, marking different samples at corresponding coordinate points. For example, the sediment sample mentioned above is marked at (28°15′N, 122°30′E, 4.2℃). Plankton samples collected at different depths in the same area will also be marked with corresponding temperature and location information, thus visually presenting the temperature-location correlation characteristics of various samples. By using sample feature maps to identify the spatial relationship between target samples and anomalous samples, for example, if the temperature of planktonic samples in a certain area is found to be generally higher than that of other samples at the same depth in the map, it can be preliminarily determined that there may be anomalous heat sources such as hydrothermal vents in that area, and thus clarify the spatial differences between anomalous samples and surrounding normal samples.

[0096] When using k-means clustering to group target management strategies, historical management data, and sample usage frequency data, Monte Carlo simulation is introduced to simulate extreme usage frequency scenarios, taking the management data of a marine sample bank as an example. Assuming a certain type of coral sample has a historical usage frequency of twice per month, Monte Carlo simulation generates usage frequency scenarios for this type of sample under extreme conditions. These include an extreme high-frequency scenario (5 times per week during typhoon season when scientific research tasks increase) and an extreme low-frequency scenario (3 months without use during the winter off-season), thus generating an augmented dataset containing multiple extreme usage frequency scenarios. A robust management model is trained based on this augmented dataset to optimize management task classification. The model learns the management strategies and usage frequency characteristics under different extreme scenarios in the augmented dataset, classifying management tasks into high-frequency emergency management, low-frequency routine management, etc. For example, the simulated high-frequency coral sample management task is classified as a high-frequency emergency task, ensuring that management resources are prioritized during actual high-frequency use.

[0097] After dynamically adjusting the temperature thresholds and storage locations of the categorized storage equipment, a real-time feedback mechanism was introduced to monitor the management effectiveness, using a refrigerated cabinet storing shellfish samples as an example. Initially, the storage temperature for the shellfish samples was set at 8±1℃. However, fluctuations in the sample preservation status were observed during management, so the temperature threshold was adjusted to 7±0.5℃ and the samples were moved to the more temperature-stable middle section of the refrigerated cabinet. The real-time feedback mechanism continuously collects adjusted temperature data and sample appearance images through temperature sensors and sample status monitoring equipment installed inside the refrigerated cabinet. For example, the temperature is recorded hourly to ensure it is maintained between 7-7.5℃, and sample images are taken daily to observe for any abnormal changes on the shellfish shells.

[0098] The temperature threshold and storage location parameters were iteratively adjusted based on the deviation between the monitoring results and the target preservation rate. Assuming a target preservation rate of 98%, the initial monitoring showed a preservation rate of 95%, a 3% deviation. System analysis suggested the deviation might be due to insufficient temperature adjustment or slightly higher humidity at the new storage location despite stable temperature. Therefore, the temperature threshold was further adjusted to 6.5 ± 0.5℃, and a humidity control device was added to the new storage location. Real-time feedback was used for monitoring again. If the preservation rate increased to 96% but still did not reach the target, further analysis was conducted to determine if excessively low temperature was causing abnormal sample metabolism. The temperature was then fine-tuned to 7 ± 0.3℃, while simultaneously optimizing the humidity control parameters. This iterative adjustment process was repeated until the preservation rate approached or reached the target value, ensuring the accuracy of the adjusted parameters.

[0099] In practical implementation, the characteristic data acquisition stage achieves precise correlation between sample temperature and location information through the collaboration of temperature sensors and GPS positioning devices. The generation of sample feature maps makes the spatial location identification of abnormal samples more intuitive. K-means clustering combined with Monte Carlo simulation and robust model training enables management task classification to better cope with extreme usage scenarios and avoid management oversights due to frequency mutations. The dynamically adjusted real-time feedback and iterative correction mechanism, taking shellfish samples as an example, ensures a high degree of matching between management measures and sample preservation requirements through continuous optimization of temperature and location parameters.

[0100] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0101] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital management method for the entire lifecycle of marine samples, characterized in that, include: Collect characteristic data of marine samples, including sample type, source location, and chemical composition information; Based on the aforementioned characteristic data, the classification scheme and storage parameters of the samples are calculated; Based on the classification scheme and storage parameters, the management needs of the sample library are assessed to generate a target management strategy. Based on the target management strategy, historical management data, and sample usage frequency data, a preliminary management plan is formulated, the preliminary management plan is optimized, and a personalized management solution is generated. By implementing the personalized management scheme, a sample preservation rate index is obtained during the management process. Based on the sample preservation rate index, the temperature threshold and storage location of the classification and storage equipment are dynamically adjusted to generate a full life cycle management scheme.

2. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, Based on the aforementioned target management strategy, historical management data, and sample usage frequency data, a preliminary management plan is formulated, the preliminary management plan is optimized, and a personalized management solution is generated, including: The target management strategy, historical management data, and sample usage frequency data are grouped using the k-means clustering method to obtain the management task classification. Based on the aforementioned management task classification, a decision tree algorithm is applied to determine the optimal management sequence and optimal management frequency in order to generate a preliminary management plan. Real-time sample status data is acquired and analyzed to calculate sample status changes and anomaly distribution. Based on the sample status changes and anomaly distribution, a long short-term memory network is used to predict the long-term management trend of the management system to refine the preliminary management plan and obtain a refined management plan. The simulated annealing algorithm is used to simulate the application effects of different management strategies in the refined management plan in order to search for the combination of optimal management time and optimal management frequency. Based on the combination of optimal management time and optimal management frequency, the gradient descent optimization algorithm is used to optimize the refined management plan and generate a personalized management scheme.

3. The method for digital management of marine samples throughout their entire lifecycle according to claim 2, characterized in that, Real-time sample status data is acquired and analyzed to calculate sample status changes and anomaly distributions. Based on these changes and anomaly distributions, a Long Short-Term Memory (LSTM) network is used to predict the long-term management trend of the management system, thereby refining the initial management plan and obtaining a refined management plan, including: Acquire real-time sample status data, and apply a recurrent neural network to identify target sample features in the real-time sample status data to obtain sample location information; Based on the sample location information, the sample state changes and abnormal distribution are calculated. The self-organizing map network is used to classify the sample state changes and abnormal distribution to obtain the classification results. Based on the classification results, the abnormal distribution pattern is identified. The sample state changes, the abnormal distribution, and the abnormal distribution pattern are analyzed using a long short-term memory network to obtain management analysis results. Based on the management analysis results, the long-term management trend of the management system is predicted, and a long-term management trend prediction result is generated. Based on the long-term management trend forecast results, the management frequency and management time corresponding to each management stage are adjusted to obtain the adjusted management frequency and management time. Using the adjusted management frequency and management time, the preliminary management plan is refined to obtain the refined management plan.

4. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, Based on the aforementioned characteristic data, the classification scheme and storage parameters of the samples are calculated, including: Based on the aforementioned characteristic data, fused sample data is generated; Feature extraction is performed on the fused sample data to identify sample type features and chemical composition features. Based on the sample type features and chemical composition features, a preliminary sample distribution map is generated. The preliminary sample distribution map is updated in real time using an ant colony optimization algorithm to obtain an updated sample distribution map. Based on the updated sample distribution map, the sample classification scheme and storage parameters are calculated using the simulated annealing algorithm and the static path planning algorithm.

5. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, Implementing the personalized management scheme yields a sample retention rate index during the management process. Based on this index, the temperature threshold and storage location of the sorting and storage equipment are dynamically adjusted to generate a full lifecycle management scheme, including: Implement the personalized management scheme and use multiple sensors to monitor the sample preservation rate in real time during the implementation of the personalized management scheme; Based on the sample retention rate index, a recurrent neural network is used to predict the retention rate change trend. According to the retention rate change trend and the management needs of the sample library, a proportional-integral-derivative control system is used to dynamically adjust the temperature threshold and storage location of the classification and storage equipment to obtain the adjusted management operation results. Based on the adjusted management operation results, a full lifecycle management scheme is generated using a genetic algorithm.

6. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, Based on the classification scheme and storage parameters, the management needs of the sample library are assessed to generate a target management strategy, including: Real-time management data is acquired from sensors in the sample library, including temperature and humidity data; The classification scheme, the storage parameters, and the real-time management data are jointly analyzed using a decision tree model to generate management demand assessment results. Based on the management needs assessment results and combined with historical management efficiency data, target management parameters are optimized through linear regression to generate target management strategies.

7. The method for digital management of marine samples throughout their entire lifecycle according to claim 6, characterized in that, The classification scheme, storage parameters, and real-time management data are jointly analyzed using a decision tree model to generate management demand assessment results, including: Feature extraction is performed on the real-time management data to identify areas of abnormal temperature distribution and humidity concentration; Based on the classification scheme and the storage parameters, the dynamic management logic of the system is modeled using Bayesian networks to generate a management logic description. The abnormal temperature distribution, the humidity concentration area, and the management logic description are input into the decision tree model, and the management demand assessment results are output.

8. The method for digital management of marine samples throughout their entire lifecycle according to claim 2, characterized in that, The step of grouping the target management strategy, historical management data, and sample usage frequency data using the k-means clustering method further includes: Monte Carlo simulation is introduced to simulate extreme usage frequency scenarios, generating an enhanced dataset; Based on the augmented dataset, a robust management model is trained to optimize the classification of the management tasks.

9. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, The collection of the characteristic data specifically includes: Sample temperature data is acquired using a temperature sensor, and the source location distribution is analyzed in conjunction with a GPS locator. Based on the sample temperature data and the distribution of the source locations, a sample feature map is generated; The spatial relationship between the target sample and the abnormal sample is identified using the sample feature map.

10. The method for digital management of marine samples throughout their entire lifecycle according to claim 1, characterized in that, After dynamically adjusting the temperature threshold and storage location of the classified storage equipment, a real-time feedback mechanism is introduced to monitor the management effect after the adjustment. Based on the deviation between the monitoring results and the target retention rate, the temperature threshold and storage location adjustment parameters are iteratively corrected.

Images