Laboratory automatic process management and multi-source data fusion system based on Internet of Things

Abstract

The invention discloses a laboratory automatic process management and multi-source data fusion system based on the Internet of Things, and belongs to the technical field of laboratory management. The system comprises a sensing layer, an analysis layer, an intelligent management layer and an application layer. The sensing layer carries out global data acquisition and real-time analysis; the analysis layer integrates data, mines an association relationship and optimizes task matching; the intelligent management layer performs full-process automatic management and control on a laboratory, and comprises an equipment cooperation module, a dynamic detection module, a dynamic configuration module, a real-time simulation module, a preset trigger module and an intelligent auditing module; and the application layer performs multi-terminal collaborative interaction and intelligent assistant assistance processing. According to the system, intelligent collaboration of laboratory equipment, automatic detection of samples, dynamic optimization of equipment parameters and automatic early warning and processing of abnormal conditions are realized through technical means such as knowledge graph construction, a digital twinborn technology and multi-mode biological recognition, and the laboratory management efficiency and the data processing accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of laboratory management, and in particular to a laboratory automation process management and multi-source data fusion system based on the Internet of Things. Background Art

[0002] Laboratory process management refers to the systematic planning, organization, coordination and control of the entire process of laboratory activities from start to finish to ensure that experimental activities are carried out efficiently, orderly and safely and the expected goals are achieved. With the rapid development of modern scientific research, industrial testing and quality control, laboratories, as the core scenarios for data output and verification, need to manage the entire process of experimental data collection, storage, analysis, sharing and archiving.

[0003] For example, the "A laboratory full-process quality control management system based on paperless office" with publication number: CN119991028A includes an abnormality matching unit, an abnormality warning unit, an experiment monitoring unit, an experiment review unit and a performance evaluation unit, and also includes: an identity recognition unit, which is used to collect facial images of experimenters by setting up facial image collection equipment outside the laboratory to obtain facial images of experimenters, and extract features from the facial images through a trained feature extraction model to obtain a facial feature set.

[0004] In the existing technology, due to the complexity and intelligent development of scientific research experiments, the traditional laboratory management model faces the pain points of data silos, inefficient processes, and lagging equipment operation and maintenance. Testing business data, equipment operation data, personnel operation data and environmental monitoring data are separated from each other, forming "information islands", making it difficult to explore potential value through cross-dimensional correlation analysis, resulting in a lack of scientific basis for experimental condition optimization and task scheduling. Experimental processes mostly rely on manual design and manual equipment operation. There are delays in manual intervention in the generation of test work orders, equipment parameter configuration and multi-device collaborative operations, especially for high-risk sample identification and special processing processes. There is a lack of intelligent automatic triggering mechanism, which restricts experimental efficiency and safety. Summary of the Invention

[0005] The purpose of the present invention is to provide a laboratory automation process management and multi-source data fusion system based on the Internet of Things to solve the problem that the experimental process proposed in the above background technology relies on manual design and manual operation of equipment, has manual intervention delays, lacks an intelligent trigger mechanism, and leads to process inefficiency.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solutions: a laboratory automation process management and multi-source data fusion system based on the Internet of Things, comprising a perception layer, an analysis layer, an intelligent management layer, and an application layer;

[0007] The perception layer is used to collect global data and conduct real-time analysis and decision-making on the global data;

[0008] The analysis layer is used to integrate multi-source data, mine associations, optimize task matching, and monitor process anomalies in real time;

[0009] The intelligent management layer is used to automate the whole process of laboratory management;

[0010] The intelligent management layer includes equipment collaboration module, dynamic detection module, dynamic configuration module, real-time simulation module, preset trigger module and intelligent audit module; the equipment collaboration module is used to set up the experimental process in advance, integrate the workflow engine, perform task scheduling, resource allocation and process status tracking, automatically trigger equipment linkage based on the preset process, make multi-device parallel operation and conflict detection, and control equipment parameters in real time; the dynamic detection module is based on the knowledge graph, automatically parses the sample label information, matches the corresponding detection equipment, environmental conditions and operating steps, generates standardized detection work orders, and dynamically matches detection needs; the dynamic configuration module is used to integrate multiple devices, automatically configure detection conditions according to the work order parameters, and realize fully automatic detection; the real-time simulation module is used to build a digital twin of the detection equipment, mirror the equipment operation status in real time, predict the deviation of the detection result and optimize the equipment parameters; the preset trigger module is used to preset abnormal scenarios to trigger the automation plan, reducing the delay of manual intervention; the intelligent audit module uses a double audit mechanism, AI automatically verifies the data, and pushes the abnormal report to manual review;

[0011] The application layer is used for multi-terminal collaborative interaction and intelligent assistant-assisted processing.

[0012] Preferably, the perception layer includes an identity authentication module and an environment perception module;

[0013] The authentication module is used to integrate multimodal biosensors to authenticate the experimenter's authority. The multimodal biosensors include face, fingerprint and iris recognition.

[0014] The environmental perception module is used to deploy sensors and collect laboratory data in real time.

[0015] Preferably, the analysis layer includes a multi-data fusion module, an intelligent matching module and an abnormal response module;

[0016] The multi-data fusion module is used to integrate the collected data and implement data cleaning and association modeling through knowledge graphs;

[0017] The intelligent matching module integrates multi-dimensional data based on machine learning algorithms, mines associations, and performs task allocation;

[0018] The abnormal response module analyzes equipment operation data based on neural networks, builds equipment health models, predicts the remaining life of equipment, analyzes historical detection data, predicts the probability of sample defects, and warns of potential failures.

[0019] Preferably, in the multi-data fusion module, integrating multi-source data for association modeling includes the following:

[0020] S1. Data access: Determine the data source, access device sensors through the edge computing gateway, transmit device operating data and environmental parameters in real time, and clean and convert the data.

[0021] S2. Knowledge graph construction: extract indicator entities from the test standard document using regular expressions, parse sample labels using predefined templates, extract indicator attributes, define indicator rules, automatically generate indicator relationships between the equipment and maintenance records, and merge duplicate entities;

[0022] S3. Dynamic update: When the device value exceeds the threshold, the device status entity in the map is updated, and full data verification is performed regularly for comprehensive updates.

[0023] Preferably, in the dynamic detection module, after automatically parsing the sample label information, implicit requirements are mined, high-risk samples are automatically identified and risk assessment is performed, and a special processing flow is triggered to process high-risk samples.

[0024] Preferably, in the dynamic configuration module, dynamically adjusting calibration device parameters includes the following:

[0025] A1. Parse the work order parameters, extract the core parameters, identify the constraints, and match them with the parameter knowledge base;

[0026] A2. Wake up the device to perform status verification, perform zero-point calibration and linearity verification, dynamically adjust parameters for environmental compensation and sample characteristic adaptation, and predict the optimal parameter combination based on historical data;

[0027] A3. Conduct process arrangement and scheduling, generate equipment operation sequence diagrams, handle abnormality plans, monitor equipment data and operating status in real time, and perform fully automatic detection.

[0028] Preferably, in the real-time simulation module, the prediction of the detection result deviation specifically includes the following contents:

[0029] B1. Build a digital twin, using software to construct a 3D geometric model of the device, define material properties, set boundary conditions, and synchronize real-time data;

[0030] B2. Extract data features, train a prediction model, input the current device status and experimental parameters, predict the distribution range of the test results, and calculate the probability of deviation between the predicted value and the target value;

[0031] B3. Based on the global sensitivity analysis method, identify the key parameters that have the greatest impact on the results, perform multi-objective optimization based on the genetic algorithm, and generate a list of parameter adjustment plans.

[0032] Preferably, the application layer includes a multi-terminal collaboration module and an intelligent assistance module;

[0033] The multi-terminal collaboration module uses multi-terminal collaborative interaction to display real-time panoramic laboratory data, detection task progress, equipment utilization, environmental parameters and defect trends;

[0034] The intelligent assistance module is used to develop AI assistants. By inputting research objectives, the system automatically recommends experimental plans, reagent ratios, and equipment combinations, and predicts the success rate based on literature data.

[0035] Preferably, in the multi-terminal collaboration module, the multi-terminal collaborative interaction includes desktop interaction and mobile interaction. The desktop is used for process design, data monitoring and permission management, the mobile is used to receive real-time alerts and view experimental progress, and the AR interface is used for immersive operation guidance and remote equipment maintenance.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] In the present invention, multimodal biosensors and liveness detection technology are used to achieve accurate authentication of personnel permissions and improve laboratory safety. The environmental perception module monitors experimental conditions in real time to ensure data reliability. The analysis layer integrates multi-source data to construct a knowledge graph and optimize task allocation efficiency. At the same time, through the equipment health model and real-time stream calculation, fault response and predictive maintenance are achieved in seconds, reducing failure rate and operation and maintenance costs. The intelligent management layer relies on workflow engines and digital twin technology to achieve unmanned control of the entire process, improve the standardization and automation level of detection, intelligent identification of high-risk samples and dynamic parameter calibration to ensure detection accuracy. The application layer uses multi-terminal collaborative interaction and AI assistants to display panoramic data in real time, intelligently recommend experimental plans, support network-free data cache synchronization, improve operational convenience and scientific research efficiency, and realize unmanned, precise and intelligent management of the entire laboratory process. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of the laboratory automation process management and multi-source data fusion system based on the Internet of Things of the present invention;

[0039] Figure 2 This is a system block diagram of the laboratory automation process management and multi-source data fusion system based on the Internet of Things of the present invention;

[0040] In the figure: 1. Perception layer; 11. Identity authentication module; 12. Environmental perception module; 2. Analysis layer; 21. Multi-data fusion module; 22. Intelligent matching module; 23. Abnormal response module; 3. Intelligent management layer; 31. Device collaboration module; 32. Dynamic detection module; 33. Dynamic configuration module; 34. Real-time simulation module; 35. Preset trigger module; 36. Intelligent audit module; 4. Application layer; 41. Multi-terminal collaboration module; 42. Intelligent assistance module. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the implementation regulations described are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0042] Example 1, please refer to Figure 1 and Figure 2 This embodiment provides a laboratory automation process management and multi-source data fusion system based on the Internet of Things. The system includes a perception layer 1, an analysis layer 2, an intelligent management layer 3 and an application layer 4.

[0043] Perception layer 1 is used for global data collection, real-time analysis, and decision-making. It includes an authentication module 11 and an environmental perception module 12. The authentication module 11 integrates multimodal biometric sensors, including facial, fingerprint, and iris recognition, to authenticate laboratory personnel. The environmental perception module 12 deploys sensors to collect laboratory data in real time.

[0044] Identity verification module 11 utilizes triple biometric recognition technology to ensure laboratory safety and data confidentiality. Facial recognition utilizes a deep learning algorithm and infrared liveness detection to prevent photo spoofing, achieving 99.7% accuracy. Fingerprint recognition utilizes a capacitive sensor with a resolution of 500dpi, capable of detecting even tiny fingerprint features. Iris recognition utilizes near-infrared imaging technology, ensuring accurate recognition even through glasses. The system configures different authentication combinations for different security zones. Standard areas require only single-factor authentication, while core areas require three-factor authentication.

[0045] Environmental Perception Module 12 deploys a diverse sensor network, including temperature and humidity sensors, gas concentration sensors, light sensors, noise sensors, and electromagnetic field intensity sensors. The temperature and humidity sensors have an accuracy of ±0.1°C and ±1% RH, collecting data every 30 seconds. The gas concentration sensor can detect a variety of harmful gases, including formaldehyde, ammonia, and volatile organic compounds, with sensitivity reaching the ppb level. The light sensor measures laboratory illumination to ensure compliance with experimental requirements. The noise sensor monitors ambient noise to ensure that precision experiments are not disturbed. The electromagnetic field intensity sensor detects electromagnetic interference to prevent it from affecting precision instruments. All sensor data is transmitted to the edge computing gateway via Bluetooth Low Energy or ZigBee protocols, and after preliminary processing, it is uploaded to the cloud platform.

[0046] Analysis Layer 2 integrates global data, mines relationships, optimizes task matching, and monitors process anomalies in real time. Analysis Layer 2 includes a multi-data fusion module 21, an intelligent matching module 22, and an anomaly response module 23. The multi-data fusion module 21 integrates collected data and implements data cleansing and association modeling through a knowledge graph. The intelligent matching module 22 uses machine learning algorithms to fuse multi-dimensional data, mine relationships, and allocate tasks. The anomaly response module 23 uses neural networks to analyze equipment operating data, build equipment health models, predict remaining equipment life, and warn of potential failures.

[0047] The multi-data fusion module 21 integrates multi-source data for correlation modeling, including the following contents:

[0048] S1. Data access: Determine the data source, access device sensors through the edge computing gateway, transmit device operating data and environmental parameters in real time, and clean and convert the data.

[0049] S2. Knowledge graph construction: extract indicator entities from the test standard document using regular expressions, parse sample labels using predefined templates, extract indicator attributes, define indicator rules, automatically generate indicator relationships between the equipment and maintenance records, and merge duplicate entities;

[0050] S3. Dynamic update: When the device value exceeds the threshold, the device status entity in the graph is updated and full data verification is performed regularly.

[0051] During the data access step, the system first establishes a list of data sources, including laboratory equipment sensors, environmental monitoring equipment, sample information systems, and personnel management systems. The edge computing gateway uses an ARM Cortex-A72 architecture processor, 8GB of memory, and 128GB of storage space. It supports multiple communication protocols, including Modbus, OPC UA, MQTT, and HTTP. The gateway can process over 10,000 data points per second and preprocesses the raw data, including denoising, outlier detection, and data normalization. The data cleaning process uses a moving median filter algorithm to remove noise, a Z-score method to detect outliers, and min-max normalization to convert data of different dimensions to a unified interval. After conversion, the data is packaged in a unified JSON format, with timestamps, device IDs, and data type tags added, and then transmitted to the cloud platform via the MQTT protocol.

[0052] In the knowledge graph construction step, the system uses Python's re library to implement regular expression matching and extract key indicator entities from the test standard documents. The predefined templates contain a variety of sample label formats, which can identify sample identifiers of different manufacturers and different types. Indicator attribute extraction uses named entity recognition technology with an accuracy rate of over 95%. Indicator rule definition uses SWRL (Semantic Web Rule Language) expressions to support complex reasoning rules. The system automatically analyzes equipment maintenance records, extracts information such as maintenance time, maintenance content and maintenance personnel, and establishes associations with equipment entities. The knowledge graph is stored in the Neo4j graph database, contains more than 10,000 entity nodes and 50,000 relationship edges, and supports complex graph queries and reasoning.

[0053] During the dynamic update process, the system implements a multi-level threshold monitoring mechanism. When a device parameter reaches the warning threshold (typically 80% of the normal range), the system marks the device as "Caution"; when it reaches the danger threshold (typically 90% of the normal range), it is marked as "Warning"; and when it exceeds the extreme threshold, it is marked as "Danger" and an emergency plan is triggered. Graph updates utilize a combination of incremental updates and full verification. Incremental updates respond to status changes in real time, while full verification is performed every 24 hours to ensure data consistency.

[0054] The intelligent matching module 22 uses multiple machine learning algorithms to process multi-dimensional data. Principal component analysis (PCA) and t-SNE algorithms are first used to reduce feature dimensionality and extract key features from high-dimensional data. Then, ensemble learning algorithms such as random forest and XGBoost are used to build a predictive model with an accuracy of 92%. Association mining uses the Apriori and FP-Growth algorithms to discover hidden patterns and regularities in the data. Task allocation uses a reinforcement learning algorithm to automatically optimize allocation plans based on historical task completion, equipment status, and personnel expertise, improving resource utilization efficiency.

[0055] Abnormal response module 23 uses a deep neural network to analyze equipment operating data. The network structure includes a multi-layer LSTM (long short-term memory network) and an attention mechanism, which can capture the timing characteristics and abnormal patterns in equipment operating data. The equipment health model comprehensively considers multiple parameters such as equipment operating time, load conditions, temperature changes, vibration characteristics, etc. to generate a health score of 0-100. The remaining life prediction uses the Cox proportional risk model, combining historical equipment failure data and current operating status to predict the possible failure time of the equipment with an accuracy rate of 85%. The potential fault warning system can predict possible failures 7-14 days in advance, giving maintenance personnel ample time to prepare.

[0056] The intelligent management layer 3 is used to automate the entire process of the laboratory. The intelligent management layer 3 includes an equipment collaboration module 31, a dynamic detection module 32, a dynamic configuration module 33, a real-time simulation module 34, a preset trigger module 35, and an intelligent audit module 36. The equipment collaboration module 31 is used to set up the experimental process in advance, integrate the workflow engine, perform task scheduling, resource allocation, and process status tracking, and automatically trigger equipment linkage based on the preset process to achieve multi-device parallel operation and conflict detection; the dynamic detection module 32 automatically parses sample label information based on the knowledge graph, generates standardized inspection work orders, and dynamically matches inspection requirements; the dynamic configuration module 33 is used to integrate multiple devices, automatically configure inspection conditions according to work order parameters, and perform fully automatic inspection; the real-time simulation module 34 is used to build a digital twin of the inspection equipment, mirror the equipment operating status in real time, predict inspection result deviations, and optimize equipment parameters; the preset trigger module 35 is used to preset abnormal scenarios to trigger automated plans; the intelligent audit module 36 uses a double audit mechanism, AI automatically verifies data, and pushes abnormal reports to manual review.

[0057] The device collaboration module 31 uses the BPMN (Business Process Model and Notation) standard to define the experimental process and supports visual process design and editing. The workflow engine is developed based on the Activiti framework and supports complex conditional branches, parallel tasks and sub-process nesting. Task scheduling uses priority queues and resource competition mechanisms to ensure that high-priority tasks are executed first. Resource allocation uses a dynamic programming algorithm to automatically allocate the optimal resource combination based on the current resource status and task requirements. Process status tracking uses an event-driven architecture to monitor the execution status of each task in real time and push it to the front-end interface through the WebSocket protocol. Device linkage is automatically executed based on preset trigger conditions. For example, when the sample pretreatment is completed, the analytical instrument is automatically started for testing. The parallel operation of multiple devices uses a distributed lock mechanism to prevent resource conflicts. The conflict detection algorithm can identify potential device conflicts in advance and provide optimization suggestions.

[0058] Dynamic testing module 32 automatically parses sample label information based on knowledge graph technology. Label parsing supports multiple formats, including barcodes, QR codes, and RFID tags, with a recognition accuracy of 99.9%. Standardized testing work orders contain sample information, test items, test methods, and quality control requirements, in compliance with laboratory standard operating procedures. Dynamic testing requirement matching uses semantic analysis technology to understand the implicit requirements in the test request and automatically match the most appropriate test method.

[0059] In the dynamic detection module 32, after automatically parsing the sample label information, the implicit demand is mined, high-risk samples are automatically identified and risk assessment is performed, and a special handling process is triggered to handle high-risk samples. Implicit demand mining uses natural language processing technology to analyze the text description in the test application form and extract key information. High-risk sample identification is based on a preset risk factor library, including multi-dimensional assessments such as sample hazard, stability, and storage conditions. The risk assessment uses a multi-factor weighted scoring model to generate a risk score of 0-100. When the score exceeds 75 points, a special handling process is automatically triggered. The special handling process includes priority enhancement, dedicated personnel, full-process monitoring, and multiple verification of results to ensure that high-risk samples are properly handled.

[0060] Dynamic Configuration Module 33 integrates various laboratory testing equipment, including chromatographs, mass spectrometers, spectrometers, and electrochemical analyzers. This equipment integration utilizes the OPCUA protocol to enable device status monitoring and remote control. Testing conditions, including test methods, instrument parameters, and quality control requirements, are automatically configured based on work order parameters. This fully automated testing process, from sample pretreatment to data analysis, eliminates the need for human intervention.

[0061] In the dynamic configuration module 33, the dynamic adjustment of calibration equipment parameters includes the following:

[0062] A1. Parse the work order parameters, extract the core parameters, identify the constraints, and match them with the parameter knowledge base;

[0063] A2. Wake up the device to perform status verification, perform zero-point calibration and linearity verification, dynamically adjust parameters for environmental compensation and sample characteristic adaptation, and predict the optimal parameter combination based on historical data;

[0064] A3. Conduct process arrangement and scheduling, generate equipment operation sequence diagrams, handle abnormality plans, monitor equipment data and operating status in real time, and perform fully automatic detection.

[0065] During the work order parameter parsing step, the system uses natural language processing technology to analyze the work order text and extract core parameters, including test items, test methods, sample types, test ranges, and accuracy requirements. Constraint identification includes time limits, resource limitations, and special requirements. The parameter knowledge base contains over 10,000 parameter configuration records, covering common test scenarios and sample types. Parameter matching uses a combination of similarity calculation and rule-based reasoning to identify the most suitable parameter configuration solution.

[0066] During the device status verification step, the system wakes up the idle device via the network and performs a self-test to verify the device status. Zero-point calibration uses high-purity standard substances to ensure a stable measurement baseline. Linearity verification uses a multi-point standard curve to ensure good linearity within the measurement range. Environmental compensation dynamically adjusts the calibration coefficient based on environmental parameters such as current temperature, humidity, and air pressure. Sample property adaptation optimizes test parameters based on the sample's physicochemical properties, such as viscosity, concentration, and matrix effects. The optimal parameter combination prediction uses a Bayesian optimization algorithm to construct a parameter-performance mapping model based on historical test data to predict the optimal parameter combination under current conditions.

[0067] During the process orchestration and scheduling steps, the system generates a detailed operational sequence diagram based on the inspection task, including the startup time, operation sequence, and estimated completion time for each device. Exception response plans include solutions for various situations, including equipment failure, sample anomalies, and environmental disturbances. Real-time monitoring utilizes a distributed sensor network to monitor equipment operating data and environmental parameters. During the fully automated inspection process, the system continuously monitors inspection progress and quality control indicators to ensure the accuracy and reliability of test results.

[0068] The real-time simulation module 34 builds a digital twin of the inspection equipment, achieving real-time synchronization between the physical device and the virtual model. The digital twin utilizes a high-precision 3D model, combined with a physical simulation engine, to simulate the mechanical movement and physical processes of the equipment. Real-time mirroring technology uses sensor data to update the virtual model's status in real time, with a latency of less than 100 milliseconds.

[0069] In the real-time simulation module 34, the prediction of the detection result deviation specifically includes the following:

[0070] B1. Build a digital twin, using software to construct a 3D geometric model of the device, define material properties, set boundary conditions, and synchronize real-time data;

[0071] B2. Extract data features, train a prediction model, input the current device status and experimental parameters, predict the distribution range of the test results, and calculate the probability of deviation between the predicted value and the target value;

[0072] B3. Based on the global sensitivity analysis method, identify the key parameters that have the greatest impact on the results, perform multi-objective optimization based on the genetic algorithm, and generate a list of parameter adjustment plans.

[0073] During the digital twin construction step, the system uses Siemens NX software to construct a high-precision 3D geometric model of the device, achieving an accuracy of 0.01mm. Material properties, including physical parameters such as density, elastic modulus, and thermal expansion coefficient, are defined to ensure that simulation results align with actual device behavior. Boundary conditions include external factors such as ambient temperature, humidity, vibration, and electromagnetic fields. Real-time data synchronization utilizes the OPC UA protocol to acquire operational data from physical device sensors and update the virtual model status, with a synchronization frequency of 10Hz.

[0074] During the data feature extraction step, the system uses time-frequency domain analysis methods to extract features from the equipment operating data, including statistical, spectral, and wavelet features. The prediction model employs a deep learning architecture, combining CNN (convolutional neural network) and LSTM (long short-term memory network) to capture the spatial and temporal characteristics of the data. Model training uses historical inspection data, containing over 10,000 samples covering a wide range of operating conditions and equipment states. The distribution intervals for the prediction results are generated using the Monte Carlo method, which calculates the probability distribution of the results through multiple simulations. The Bayesian approach is used to calculate the probability of deviation, taking into account model uncertainty and parameter variability.

[0075] During the key parameter identification step, the system uses the Sobol method for global sensitivity analysis, calculating the first-order and total sensitivity indices for each parameter to the results. Parameter optimization employs the NSGA-II (Non-Dominated Sorting Genetic Algorithm II) for multi-objective optimization, simultaneously considering multiple objectives such as detection accuracy, detection speed, and resource consumption. The optimization process uses a population size of 100, 200 iterations, a crossover probability of 0.8, and a mutation probability of 0.1. A list of parameter adjustment options, containing multiple non-dominated solutions, is generated, sorted by overall performance, for operator selection.

[0076] The preset trigger module 35 pre-configures various abnormal scenarios, including equipment failure, sample anomalies, environmental disturbances, and operational errors. Each abnormal scenario corresponds to an automated response plan, including alarm triggering, emergency response, and recovery measures. The triggering conditions for these scenarios are defined using a rules engine that supports complex event processing and pattern matching. Automated response plan execution utilizes a state machine model to ensure reliable and consistent processing.

[0077] The Intelligent Audit Module 36 utilizes a dual audit mechanism, combining AI-powered verification and manual review. AI verification utilizes multiple algorithms, including anomaly detection, consistency checks, and trend analysis, achieving an accuracy rate exceeding 95%. Exception reports are automatically sent to relevant personnel, including anomaly descriptions, impact assessments, and action recommendations. Manual review utilizes a tiered review mechanism, assigning reviewers to different levels based on the severity of the anomaly.

[0078] Application layer 4 is used for multi-terminal collaborative interaction and intelligent assistant-assisted processing. Application layer 4 includes a multi-terminal collaborative module 41 and an intelligent assistance module 42. Multi-terminal collaborative module 41 uses multi-terminal collaborative interaction to display real-time panoramic laboratory data, detecting task progress, equipment utilization, environmental parameters, and defect trends. Intelligent assistance module 42 is used to develop AI assistants. After inputting research objectives, the system automatically recommends experimental plans, reagent ratios, and equipment combinations.

[0079] The multi-terminal collaboration module 41 supports multiple terminal devices such as PC, mobile and large-screen terminals. Real-time display uses WebSocket technology to ensure real-time data updates with a delay of less than 200 milliseconds. Panoramic data display includes information such as laboratory floor plans, equipment distribution, personnel locations and environmental parameters. Task progress tracking uses Gantt charts and progress bars for visual display, and supports drilling down to view detailed information. Equipment utilization statistics include daily utilization, weekly utilization and monthly utilization, and support multi-dimensional analysis and comparison. Environmental parameter monitoring includes real-time values ​​and historical trends of parameters such as temperature and humidity, gas concentration, noise and light. Defect trend analysis uses statistical charts to display the changing trends of indicators such as equipment failure rate, sample abnormality rate and detection deviation rate.

[0080] In the multi-terminal collaboration module 41, multi-terminal collaborative interaction includes desktop and mobile interactions. The desktop is used for process design, data monitoring, and permission management, while the mobile terminal is used to receive real-time alerts, view experimental progress, and use the AR interface for immersive operational guidance and remote equipment maintenance. The desktop terminal adopts a responsive design and supports multi-screen display and high-resolution output. The process design tool supports drag-and-drop operations and has a rich built-in process template and component library. The data monitoring interface adopts a dashboard design and supports custom layout and indicator selection. Permission management adopts the RBAC (Role-Based Access Control) model and supports fine-grained permission configuration. The mobile terminal adopts a hybrid application development framework and supports iOS and Android platforms. Real-time alerts use a push notification mechanism and support sound, vibration, and pop-up reminders. Experiment progress viewing supports list view and card view, providing progress percentage and estimated completion time. The AR interface uses ARKit and ARCore technologies to recognize the device through the camera and overlay virtual operation instructions. The remote maintenance function supports video calls, screen sharing, and remote control to help technicians solve problems remotely.

[0081] The AI ​​assistant developed by the Intelligent Assistance Module 42 uses natural language processing technology to support natural language interaction and intent recognition. Research objective input supports text and voice input, allowing the system to understand complex research requirements. Experimental plan recommendations are based on knowledge graphs and case-based reasoning, learning best practices from historical success stories. Reagent ratio recommendations consider experimental objectives, sample characteristics, and available resources to generate the optimal ratio. Equipment combination recommendations generate the most appropriate equipment combination based on equipment performance, availability, and compatibility.

[0082] Example 2, please refer to Figure 1 and Figure 2 This embodiment provides a laboratory automation process management and multi-source data fusion system based on the Internet of Things. The system includes a perception layer 1, an analysis layer 2, an intelligent management layer 3 and an application layer 4.

[0083] Perception layer 1 is used for global data collection, real-time analysis, and decision-making. It includes an authentication module 11 and an environmental perception module 12. The authentication module 11 integrates multimodal biometric sensors, including facial, fingerprint, and iris recognition, to authenticate laboratory personnel. The environmental perception module 12 deploys sensors to collect laboratory data in real time.

[0084] Identity verification module 11 utilizes triple biometric recognition technology to ensure laboratory safety and data confidentiality. Facial recognition utilizes a deep learning algorithm and infrared liveness detection to prevent photo spoofing, achieving 99.7% accuracy. Fingerprint recognition utilizes a capacitive sensor with a resolution of 500dpi, capable of detecting even tiny fingerprint features. Iris recognition utilizes near-infrared imaging technology, ensuring accurate recognition even through glasses. The system configures different authentication combinations for different security zones. Standard areas require only single-factor authentication, while core areas require three-factor authentication.

[0085] Environmental Perception Module 12 deploys a diverse sensor network, including temperature and humidity sensors, gas concentration sensors, light sensors, noise sensors, and electromagnetic field intensity sensors. The temperature and humidity sensors have an accuracy of ±0.1°C and ±1% RH, collecting data every 30 seconds. The gas concentration sensor can detect a variety of harmful gases, including formaldehyde, ammonia, and volatile organic compounds, with sensitivity reaching the ppb level. The light sensor measures laboratory illumination to ensure compliance with experimental requirements. The noise sensor monitors ambient noise to ensure that precision experiments are not disturbed. The electromagnetic field intensity sensor detects electromagnetic interference to prevent it from affecting precision instruments. All sensor data is transmitted to the edge computing gateway via Bluetooth Low Energy or ZigBee protocols, and after preliminary processing, it is uploaded to the cloud platform.

[0086] Analysis Layer 2 integrates global data, mines relationships, optimizes task matching, and monitors process anomalies in real time. Analysis Layer 2 includes a multi-data fusion module 21, an intelligent matching module 22, and an anomaly response module 23. The multi-data fusion module 21 integrates collected data and implements data cleansing and association modeling through a knowledge graph. The intelligent matching module 22 uses machine learning algorithms to fuse multi-dimensional data, mine relationships, and allocate tasks. The anomaly response module 23 uses neural networks to analyze equipment operating data, build equipment health models, predict remaining equipment life, and warn of potential failures.

[0087] The multi-data fusion module 21 integrates multi-source data for correlation modeling, including the following contents:

[0088] S1. Data access: Determine the data source, access device sensors through the edge computing gateway, transmit device operating data and environmental parameters in real time, and clean and convert the data.

[0089] S2. Knowledge graph construction: extract indicator entities from the test standard document using regular expressions, parse sample labels using predefined templates, extract indicator attributes, define indicator rules, automatically generate indicator relationships between the equipment and maintenance records, and merge duplicate entities;

[0090] S3. Dynamic update: When the device value exceeds the threshold, the device status entity in the graph is updated and full data verification is performed regularly.

[0091] During the data access step, the system first establishes a list of data sources, including laboratory equipment sensors, environmental monitoring equipment, sample information systems, and personnel management systems. The edge computing gateway uses an ARM Cortex-A72 architecture processor, 8GB of memory, and 128GB of storage space. It supports multiple communication protocols, including Modbus, OPC UA, MQTT, and HTTP. The gateway can process over 10,000 data points per second and preprocesses the raw data, including denoising, outlier detection, and data normalization. The data cleaning process uses a moving median filter algorithm to remove noise, a Z-score method to detect outliers, and min-max normalization to convert data of different dimensions to a unified interval. After conversion, the data is packaged in a unified JSON format, with timestamps, device IDs, and data type tags added, and then transmitted to the cloud platform via the MQTT protocol.

[0092] In the knowledge graph construction step, the system uses Python's re library to implement regular expression matching and extract key indicator entities from the test standard documents. The predefined templates contain a variety of sample label formats, which can identify sample identifiers of different manufacturers and different types. Indicator attribute extraction uses named entity recognition technology with an accuracy rate of over 95%. Indicator rule definition uses SWRL (Semantic Web Rule Language) expressions to support complex reasoning rules. The system automatically analyzes equipment maintenance records, extracts information such as maintenance time, maintenance content and maintenance personnel, and establishes associations with equipment entities. The knowledge graph is stored in the Neo4j graph database, contains more than 10,000 entity nodes and 50,000 relationship edges, and supports complex graph queries and reasoning.

[0093] During the dynamic update process, the system implements a multi-level threshold monitoring mechanism. When a device parameter reaches the warning threshold (typically 80% of the normal range), the system marks the device as "Caution"; when it reaches the danger threshold (typically 90% of the normal range), it is marked as "Warning"; and when it exceeds the extreme threshold, it is marked as "Danger" and an emergency plan is triggered. Graph updates utilize a combination of incremental updates and full verification. Incremental updates respond to status changes in real time, while full verification is performed every 24 hours to ensure data consistency.

[0094] The intelligent matching module 22 uses multiple machine learning algorithms to process multi-dimensional data. Principal component analysis (PCA) and t-SNE algorithms are first used to reduce feature dimensionality and extract key features from high-dimensional data. Then, ensemble learning algorithms such as random forest and XGBoost are used to build a predictive model with an accuracy of 92%. Association mining uses the Apriori and FP-Growth algorithms to discover hidden patterns and regularities in the data. Task allocation uses a reinforcement learning algorithm to automatically optimize allocation plans based on historical task completion, equipment status, and personnel expertise, improving resource utilization efficiency.

[0095] Abnormal response module 23 uses a deep neural network to analyze equipment operating data. The network structure includes a multi-layer LSTM (long short-term memory network) and an attention mechanism, which can capture the timing characteristics and abnormal patterns in equipment operating data. The equipment health model comprehensively considers multiple parameters such as equipment operating time, load conditions, temperature changes, vibration characteristics, etc. to generate a health score of 0-100. The remaining life prediction uses the Cox proportional risk model, combining historical equipment failure data and current operating status to predict the possible failure time of the equipment with an accuracy rate of 85%. The potential fault warning system can predict possible failures 7-14 days in advance, giving maintenance personnel ample time to prepare.

[0096] The intelligent management layer 3 is used to automate the entire process of the laboratory. The intelligent management layer 3 includes an equipment collaboration module 31, a dynamic detection module 32, a dynamic configuration module 33, a real-time simulation module 34, a preset trigger module 35, and an intelligent audit module 36. The equipment collaboration module 31 is used to set up the experimental process in advance, integrate the workflow engine, perform task scheduling, resource allocation, and process status tracking, and automatically trigger equipment linkage based on the preset process to achieve multi-device parallel operation and conflict detection; the dynamic detection module 32 automatically parses sample label information based on the knowledge graph, generates standardized inspection work orders, and dynamically matches inspection requirements; the dynamic configuration module 33 is used to integrate multiple devices, automatically configure inspection conditions according to work order parameters, and perform fully automatic inspection; the real-time simulation module 34 is used to build a digital twin of the inspection equipment, mirror the equipment operating status in real time, predict inspection result deviations, and optimize equipment parameters; the preset trigger module 35 is used to preset abnormal scenarios to trigger automated plans; the intelligent audit module 36 uses a double audit mechanism, AI automatically verifies data, and pushes abnormal reports to manual review.

[0097] The device collaboration module 31 uses the BPMN (Business Process Model and Notation) standard to define the experimental process and supports visual process design and editing. The workflow engine is developed based on the Activiti framework and supports complex conditional branches, parallel tasks and sub-process nesting. Task scheduling uses priority queues and resource competition mechanisms to ensure that high-priority tasks are executed first. Resource allocation uses a dynamic programming algorithm to automatically allocate the optimal resource combination based on the current resource status and task requirements. Process status tracking uses an event-driven architecture to monitor the execution status of each task in real time and push it to the front-end interface through the WebSocket protocol. Device linkage is automatically executed based on preset trigger conditions. For example, when the sample pretreatment is completed, the analytical instrument is automatically started for testing. The parallel operation of multiple devices uses a distributed lock mechanism to prevent resource conflicts. The conflict detection algorithm can identify potential device conflicts in advance and provide optimization suggestions.

[0098] Dynamic testing module 32 automatically parses sample label information based on knowledge graph technology. Label parsing supports multiple formats, including barcodes, QR codes, and RFID tags, with a recognition accuracy of 99.9%. Standardized testing work orders contain sample information, test items, test methods, and quality control requirements, in compliance with laboratory standard operating procedures. Dynamic testing requirement matching uses semantic analysis technology to understand the implicit requirements in the test request and automatically match the most appropriate test method.

[0099] In the dynamic detection module 32, after automatically parsing the sample label information, the implicit demand is mined, high-risk samples are automatically identified and risk assessment is performed, and a special handling process is triggered to handle high-risk samples. Implicit demand mining uses natural language processing technology to analyze the text description in the test application form and extract key information. High-risk sample identification is based on a preset risk factor library, including multi-dimensional assessments such as sample hazard, stability, and storage conditions. The risk assessment uses a multi-factor weighted scoring model to generate a risk score of 0-100. When the score exceeds 75 points, a special handling process is automatically triggered. The special handling process includes priority enhancement, dedicated personnel, full-process monitoring, and multiple verification of results to ensure that high-risk samples are properly handled.

[0100] Dynamic Configuration Module 33 integrates various laboratory testing equipment, including chromatographs, mass spectrometers, spectrometers, and electrochemical analyzers. This equipment integration utilizes the OPCUA protocol to enable device status monitoring and remote control. Testing conditions, including test methods, instrument parameters, and quality control requirements, are automatically configured based on work order parameters. This fully automated testing process, from sample pretreatment to data analysis, eliminates the need for human intervention.

[0101] In the dynamic configuration module 33, the dynamic adjustment of calibration equipment parameters includes the following:

[0102] A1. Parse the work order parameters, extract the core parameters, identify the constraints, and match them with the parameter knowledge base;

[0103] A2. Wake up the device to perform status verification, perform zero-point calibration and linearity verification, dynamically adjust parameters for environmental compensation and sample characteristic adaptation, and predict the optimal parameter combination based on historical data;

[0104] A3. Conduct process arrangement and scheduling, generate equipment operation sequence diagrams, handle abnormality plans, monitor equipment data and operating status in real time, and perform fully automatic detection.

[0105] During the work order parameter parsing step, the system uses natural language processing technology to analyze the work order text and extract core parameters, including test items, test methods, sample types, test ranges, and accuracy requirements. Constraint identification includes time limits, resource limitations, and special requirements. The parameter knowledge base contains over 10,000 parameter configuration records, covering common test scenarios and sample types. Parameter matching uses a combination of similarity calculation and rule-based reasoning to identify the most suitable parameter configuration solution.

[0106] During the device status verification step, the system wakes up the idle device via the network and performs a self-test to verify the device status. Zero-point calibration uses high-purity standard substances to ensure a stable measurement baseline. Linearity verification uses a multi-point standard curve to ensure good linearity within the measurement range. Environmental compensation dynamically adjusts the calibration coefficient based on environmental parameters such as current temperature, humidity, and air pressure. Sample property adaptation optimizes test parameters based on the sample's physicochemical properties, such as viscosity, concentration, and matrix effects. The optimal parameter combination prediction uses a Bayesian optimization algorithm to construct a parameter-performance mapping model based on historical test data to predict the optimal parameter combination under current conditions.

[0107] During the process orchestration and scheduling steps, the system generates a detailed operational sequence diagram based on the inspection task, including the startup time, operation sequence, and estimated completion time for each device. Exception response plans include solutions for various situations, including equipment failure, sample anomalies, and environmental disturbances. Real-time monitoring utilizes a distributed sensor network to monitor equipment operating data and environmental parameters. During the fully automated inspection process, the system continuously monitors inspection progress and quality control indicators to ensure the accuracy and reliability of test results.

[0108] The real-time simulation module 34 builds a digital twin of the inspection equipment, achieving real-time synchronization between the physical device and the virtual model. The digital twin utilizes a high-precision 3D model, combined with a physical simulation engine, to simulate the mechanical movement and physical processes of the equipment. Real-time mirroring technology uses sensor data to update the virtual model's status in real time, with a latency of less than 100 milliseconds.

[0109] In the real-time simulation module 34, the prediction of the detection result deviation specifically includes the following:

[0110] B1. Build a digital twin, using software to construct a 3D geometric model of the device, define material properties, set boundary conditions, and synchronize real-time data;

[0111] B2. Extract data features, train a prediction model, input the current device status and experimental parameters, predict the distribution range of the test results, and calculate the probability of deviation between the predicted value and the target value;

[0112] B3. Based on the global sensitivity analysis method, identify the key parameters that have the greatest impact on the results, perform multi-objective optimization based on the genetic algorithm, and generate a list of parameter adjustment plans.

[0113] During the digital twin construction step, the system uses Siemens NX software to construct a high-precision 3D geometric model of the device, achieving an accuracy of 0.01mm. Material properties, including physical parameters such as density, elastic modulus, and thermal expansion coefficient, are defined to ensure that simulation results align with actual device behavior. Boundary conditions include external factors such as ambient temperature, humidity, vibration, and electromagnetic fields. Real-time data synchronization utilizes the OPC UA protocol to acquire operational data from physical device sensors and update the virtual model status, with a synchronization frequency of 10Hz.

[0114] During the data feature extraction step, the system uses time-frequency domain analysis methods to extract features from the equipment operating data, including statistical, spectral, and wavelet features. The prediction model employs a deep learning architecture, combining CNN (convolutional neural network) and LSTM (long short-term memory network) to capture the spatial and temporal characteristics of the data. Model training uses historical inspection data, containing over 10,000 samples covering a wide range of operating conditions and equipment states. The distribution intervals for the prediction results are generated using the Monte Carlo method, which calculates the probability distribution of the results through multiple simulations. The Bayesian approach is used to calculate the probability of deviation, taking into account model uncertainty and parameter variability.

[0115] During the key parameter identification step, the system uses the Sobol method for global sensitivity analysis, calculating the first-order and total sensitivity indices for each parameter to the results. Parameter optimization employs the NSGA-II (Non-Dominated Sorting Genetic Algorithm II) for multi-objective optimization, simultaneously considering multiple objectives such as detection accuracy, detection speed, and resource consumption. The optimization process uses a population size of 100, 200 iterations, a crossover probability of 0.8, and a mutation probability of 0.1. A list of parameter adjustment options, containing multiple non-dominated solutions, is generated, sorted by overall performance, for operator selection.

[0116] The preset trigger module 35 pre-configures various abnormal scenarios, including equipment failure, sample anomalies, environmental disturbances, and operational errors. Each abnormal scenario corresponds to an automated response plan, including alarm triggering, emergency response, and recovery measures. The triggering conditions for these scenarios are defined using a rules engine that supports complex event processing and pattern matching. Automated response plan execution utilizes a state machine model to ensure reliable and consistent processing.

[0117] The Intelligent Audit Module 36 utilizes a dual audit mechanism, combining AI-powered verification and manual review. AI verification utilizes multiple algorithms, including anomaly detection, consistency checks, and trend analysis, achieving an accuracy rate exceeding 95%. Exception reports are automatically sent to relevant personnel, including anomaly descriptions, impact assessments, and action recommendations. Manual review utilizes a tiered review mechanism, assigning reviewers to different levels based on the severity of the anomaly.

[0118] Application layer 4 is used for multi-terminal collaborative interaction and intelligent assistant-assisted processing. Application layer 4 includes a multi-terminal collaborative module 41 and an intelligent assistance module 42. Multi-terminal collaborative module 41 uses multi-terminal collaborative interaction to display real-time panoramic laboratory data, detecting task progress, equipment utilization, environmental parameters, and defect trends. Intelligent assistance module 42 is used to develop AI assistants. After inputting research objectives, the system automatically recommends experimental plans, reagent ratios, and equipment combinations.

[0119] The multi-terminal collaboration module 41 supports multiple terminal devices such as PC, mobile and large-screen terminals. Real-time display uses WebSocket technology to ensure real-time data updates with a delay of less than 200 milliseconds. Panoramic data display includes information such as laboratory floor plans, equipment distribution, personnel locations and environmental parameters. Task progress tracking uses Gantt charts and progress bars for visual display, and supports drilling down to view detailed information. Equipment utilization statistics include daily utilization, weekly utilization and monthly utilization, and support multi-dimensional analysis and comparison. Environmental parameter monitoring includes real-time values ​​and historical trends of parameters such as temperature and humidity, gas concentration, noise and light. Defect trend analysis uses statistical charts to display the changing trends of indicators such as equipment failure rate, sample abnormality rate and detection deviation rate.

[0120] In the multi-terminal collaboration module 41, multi-terminal collaborative interaction includes desktop and mobile interactions. The desktop is used for process design, data monitoring, and permission management, while the mobile terminal is used to receive real-time alerts, view experimental progress, and use the AR interface for immersive operational guidance and remote equipment maintenance. The desktop terminal adopts a responsive design and supports multi-screen display and high-resolution output. The process design tool supports drag-and-drop operations and has a rich built-in process template and component library. The data monitoring interface adopts a dashboard design and supports custom layout and indicator selection. Permission management adopts the RBAC (Role-Based Access Control) model and supports fine-grained permission configuration. The mobile terminal adopts a hybrid application development framework and supports iOS and Android platforms. Real-time alerts use a push notification mechanism and support sound, vibration, and pop-up reminders. Experiment progress viewing supports list view and card view, providing progress percentage and estimated completion time. The AR interface uses ARKit and ARCore technologies to recognize the device through the camera and overlay virtual operation instructions. The remote maintenance function supports video calls, screen sharing, and remote control to help technicians solve problems remotely.

[0121] The AI ​​assistant developed by the Intelligent Assistance Module 42 uses natural language processing technology to support natural language interaction and intent recognition. Research objective input supports text and voice input, allowing the system to understand complex research requirements. Experimental plan recommendations are based on knowledge graphs and case-based reasoning, learning best practices from historical success stories. Reagent ratio recommendations consider experimental objectives, sample characteristics, and available resources to generate the optimal ratio. Equipment combination recommendations generate the most appropriate equipment combination based on equipment performance, availability, and compatibility.

[0122] It should be noted that both Example 1 and Example 2 are a type of laboratory automation process management and multi-source data fusion system based on the Internet of Things.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. 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 various embodiments of the present invention.

Claims

1. The laboratory automation process management and multi-source data fusion system based on the Internet of Things is characterized by: It includes the perception layer (1), analysis layer (2), intelligent management layer (3) and application layer (4); The perception layer (1) is used to collect global data and conduct real-time analysis and decision-making on the global data; The analysis layer (2) is used to integrate global data, mine associations, optimize task matching, and monitor process anomalies in real time; The intelligent management layer (3) is used to automate the whole process of laboratory management; The intelligent management layer (3) includes an equipment collaboration module (31), a dynamic detection module (32), a dynamic configuration module (33), a real-time simulation module (34), a preset trigger module (35) and an intelligent audit module (36); the equipment collaboration module (31) is used to set the experimental process in advance, integrate the workflow engine, perform task scheduling, resource allocation and process status tracking, automatically trigger equipment linkage based on the preset process, and make multi-equipment parallel operation and conflict detection; the dynamic detection module (32) automatically analyzes sample label information based on the knowledge graph, generates standardized detection work orders, and dynamically detects demand matching; the dynamic configuration module (33) is used to integrate multiple devices, automatically configure detection conditions according to work order parameters, and perform fully automatic detection; the real-time simulation module (34) is used to build a digital twin of the detection equipment, mirror the equipment operation status in real time, predict the deviation of the detection result and optimize the equipment parameters; the preset trigger module (35) is used to preset abnormal scenarios to trigger the automation plan; the intelligent audit module (36) uses a double audit mechanism, AI automatically verifies data, and pushes abnormal reports to manual review; The application layer (4) is used for multi-terminal collaborative interaction and intelligent assistant-assisted processing.

2. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: The perception layer (1) includes an identity authentication module (11) and an environment perception module (12); The identity authentication module (11) is used to integrate multimodal biosensors to authenticate the authority of the experimenter. The multimodal biosensors include face, fingerprint and iris recognition; The environmental perception module (12) is used to deploy sensors and collect laboratory data in real time.

3. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: The analysis layer (2) includes a multi-data fusion module (21), an intelligent matching module (22) and an abnormal response module (23); The multi-data fusion module (21) is used to integrate the collected data and realize data cleaning and association modeling through the knowledge graph; The intelligent matching module (22) integrates multi-dimensional data based on machine learning algorithms, mines associations, and performs task allocation; The abnormal response module (23) analyzes the equipment operation data based on the neural network, builds the equipment health model, predicts the remaining life of the equipment, and warns of potential failures.

4. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 3 is characterized by: In the multi-data fusion module (21), the integration of multi-source data for correlation modeling includes the following: S1. Data access: Determine the data source, access device sensors through the edge computing gateway, transmit device operating data and environmental parameters in real time, and clean and convert the data. S2. Knowledge graph construction: extract indicator entities from the test standard document using regular expressions, parse sample labels using predefined templates, extract indicator attributes, define indicator rules, automatically generate indicator relationships between the equipment and maintenance records, and merge duplicate entities; S3. Dynamic update: When the device value exceeds the threshold, the device status entity in the graph is updated and full data verification is performed regularly.

5. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: In the dynamic detection module (32), after automatically parsing the sample label information, the implicit demand is mined, high-risk samples are automatically identified and risk assessment is performed, and a special processing process is triggered to process the high-risk samples.

6. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: In the dynamic configuration module (33), the dynamic adjustment of calibration equipment parameters includes the following: A1. Parse the work order parameters, extract the core parameters, identify the constraints, and match them with the parameter knowledge base; A2. Wake up the device to perform status verification, perform zero-point calibration and linearity verification, dynamically adjust parameters for environmental compensation and sample characteristic adaptation, and predict the optimal parameter combination based on historical data; A3. Conduct process arrangement and scheduling, generate equipment operation sequence diagrams, handle abnormality plans, monitor equipment data and operating status in real time, and perform fully automatic detection.

7. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: In the real-time simulation module (34), the prediction of the detection result deviation specifically includes the following contents: B1. Build a digital twin, using software to construct a 3D geometric model of the device, define material properties, set boundary conditions, and synchronize real-time data; B2. Extract data features, train a prediction model, input the current device status and experimental parameters, predict the distribution range of the test results, and calculate the probability of deviation between the predicted value and the target value; B3. Based on the global sensitivity analysis method, identify the key parameters that have the greatest impact on the results, perform multi-objective optimization based on the genetic algorithm, and generate a list of parameter adjustment plans.

8. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 1 is characterized by: The application layer (4) includes a multi-terminal collaboration module (41) and an intelligent assistance module (42); The multi-terminal collaborative module (41) uses multi-terminal collaborative interaction to display the laboratory's panoramic data in real time, detecting task progress, equipment utilization, environmental parameters and defect trends; The intelligent assistance module (42) is used to develop an AI assistant. After inputting the research objectives, the system automatically recommends experimental plans, reagent ratios, and equipment combinations.

9. The laboratory automation process management and multi-source data fusion system based on the Internet of Things according to claim 8 is characterized by: In the multi-terminal collaboration module (41), multi-terminal collaborative interaction includes desktop interaction and mobile interaction. The desktop is used for process design, data monitoring and permission management, and the mobile is used to receive real-time alerts, view experimental progress, and use the AR interface for immersive operation guidance and remote equipment maintenance.

Images