Management method and device of intelligent laboratory based on artificial intelligence

By establishing data type certificates and association relationships in smart laboratories, and using intelligent control and scheduling modules to optimize resource allocation, the problem of data chaos is solved, and efficient management of laboratory resources and experimental security is achieved.

CN118982216AInactive Publication Date: 2024-11-19NANCHANG XIEDA TECH DEV CO LTD

Patent Information

Application Number
CN202411468349.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-21
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

There is a risk of data confusion in smart laboratories, which affects the normal work of the laboratory. How to efficiently utilize laboratory resources has become a difficult point.

Method used

By interconnecting various implementation units in the laboratory with the data processing module, generating data type certificates, establishing associations, using intelligent control modules to predict the health status of the equipment and monitoring the experimental environment, intelligent scheduling module optimizes resource allocation, and intelligent security management module detects abnormal behavior , realize data classification storage and real-time processing.

Benefits of technology

It realizes comprehensive monitoring of laboratory environment and equipment and optimizes resource allocation, ensuring the smooth progress and security of the experimental process, and optimizing laboratory resource management and equipment scheduling through continuous machine learning and data accumulation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982216A_ABST
    Figure CN118982216A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of smart laboratories, in particular to a smart laboratory management method and device based on artificial intelligence, and the method achieves the comprehensive monitoring of the environment and equipment of a laboratory through the real-time collection of the environment data, equipment states and personnel operation behaviors of the laboratory, and the system achieves the real-time monitoring of the environment and equipment of the laboratory according to the incidence relation lt; data type certificates and data classification gt; according to the method, the type of received data is efficiently identified, analysis and prediction are carried out through an AI algorithm, so that smooth proceeding of the experiment process is ensured, then the system intelligently schedules resources and optimizes allocation of experiment equipment according to experiment requirements and equipment availability, the system monitors the experiment equipment and personnel operation in real time in the experiment process, and the experiment efficiency is improved. Feedback is provided, abnormal conditions are handled in time, and experiment safety and effectiveness are ensured. Through continuous machine learning and data accumulation, the system can continuously optimize laboratory resource management, equipment scheduling and experiment scheduling, and the beneficial technical effect of overall efficient management of the laboratory is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of smart laboratories, and in particular to a management method and device for smart laboratories based on artificial intelligence. Background Art

[0002] Smart laboratories refer to scientific research sites that use artificial intelligence technology to carry out comprehensive intelligent transformation of laboratories. Their advantages are that they can realize the automation of experimental processes, the optimal allocation of experimental resources, and the intelligent monitoring and early warning of experimental safety. They are of great significance for improving the scientific research level of universities and cultivating innovative talents. A search of the prior art revealed that a data management method for a smart laboratory management platform is disclosed in a Chinese invention patent with a publication number of CN115471216B. This method uses a neural network model based on deep learning as a feature extractor to extract the dynamic features of the experimental operation of the learner's experimental operation monitoring video, and combines a classifier to classify the experimental operation dynamics to achieve monitoring and intelligent analysis of the learner's experimental operation, thereby avoiding laboratory accidents caused by incorrect operations. However, with the increasing frequency and complexity of scientific research activities in universities, how to efficiently use smart laboratories has become a major research hotspot and difficulty. Smart laboratories have a wide range of data sources, involving the storage and transmission of a large amount of experimental data, and there is a risk of data confusion. These risks have a significant impact on the normal operation of smart laboratories. Summary of the invention

[0003] The purpose of the present invention is to address the problems existing in the background technology and to propose a management method and device for a smart laboratory based on artificial intelligence.

[0004] The technical solution of the present invention is a management method of a smart laboratory based on artificial intelligence, comprising the following specific implementation steps: S1. Connect each implementation unit in the laboratory with the data processing module. The data processing module generates a data type certificate for each implementation unit and establishes an association relationship <identity identification, data type certificate> and <data type certificate, data classification>; S2. In the data collection and management stage, each implementation unit in the data collection module attaches its own identity identification and data type certificate to the collected data, and transmits the collected data to the data processing module; S3, the data processing module determines the data type of the received data according to the association relationship <identity identifier, data type certificate>, and then identifies the data classification according to the association relationship <data type certificate, data classification>, and then performs preliminary processing on the received data, and classifies and stores the data after preliminary processing according to the data classification; S4, the intelligent control module uses the equipment health prediction model to predict the health status of the equipment, and monitors the real-time experimental environment of the laboratory. If the environmental data exceeds the set range, the intelligent control module automatically adjusts the environmental conditions and sends a warning message to the experimenter; S5, the intelligent scheduling module builds a resource scheduling model, analyzes the current use of experimental equipment, and predicts resource requirements based on historical data to determine the availability of experimental equipment and materials; S6. The intelligent safety management module automatically analyzes the content of harmful gases in the experimental environment. When an abnormality is found, it immediately issues an alarm and notifies the laboratory personnel. On the other hand, it uses a behavioral detection model to detect violations by laboratory personnel, and issues real-time reminders and warnings when abnormal behavior is found.

[0005] Preferably, the classification standard for data classification is: classification according to data source and purpose, and the collected data is classified into: environmental data, equipment status data and personnel behavior image data.

[0006] Preferably, the data type certificate is generated as follows: S3-1. Implementation unit i Generate a random number r i , calculate the certificate application parameter P~CA i =r i ×G, then implement unit i {Identity ID i , certificate application parameters P~CA i}Send to data processing module; Where G is the base point of order q on the elliptic curve E, and E is defined in the finite field F p Elliptic curves on ; S3-2, the data processing module selects a random number R i ∈F p , calculate the first certificate generation parameter P~Ⅰ~GA i =R i ×G, and calculate the second certificate generation parameter P~Ⅱ~GA i =P~Ⅱ~GA i +P~CA i ; S3-3. The data processing module generates a data type certificate cert i =H(P~Ⅱ~GA i ||ID i )×R i +G×R i ; Among them, H represents the hash function, and || represents the string concatenation symbol.

[0007] Preferably, the equipment health prediction model introduces the attention mechanism into the LSTM neural network, performs in-depth analysis on equipment operation data, predicts the health status of the equipment, and outputs the possibility of equipment failure.

[0008] Preferably, the specific implementation steps of S5 are as follows: S5-1, the experimenter inputs the experiment requirements through the intelligent terminal. After receiving the experiment requirements, the intelligent scheduling module first analyzes the experiment content to determine whether the equipment required for the experiment meets the current laboratory resource status. Then, it uses historical data, equipment status, and experimental material inventory information to evaluate whether the experiment has the required resources. If the current laboratory resources are sufficient, the intelligent scheduling module confirms the experiment plan and continues with subsequent allocation. If the resources are insufficient, the system will prompt the experimenter to adjust the experiment time or the order of equipment use; S5-2, the intelligent scheduling module calls the equipment data in the equipment database, analyzes the current usage of experimental equipment through the resource scheduling model, and predicts resource requirements based on historical data to determine the availability of experimental equipment and materials; S5-3. The intelligent scheduling module automatically allocates experimental equipment according to the experimental requirements and the status of the experimental equipment. If resources are tight, the intelligent scheduling module will adjust the experimental time and equipment allocation order.

[0009] Preferably, the scheduling process of the resource scheduling model is as follows: S6-1, the state of laboratory resources and the local state of each agent constitute the global state space, the global state S t At time step t, it contains the following information: S t ={(s1 t , s2 t , ..., s n t )}; In the formula, s i t represents the local state of the ith agent at time t, including the agent's current needs and experimental progress; Among them, the agent represents the experiment in the laboratory; S6-2. Action a1 taken by each agent t It includes the following types: select a specific device to perform the experiment, adjust the start time of the experiment, and delay the execution of a certain experiment; For each time step t, the intelligent scheduling module records the action combination of all agents: A t =(a1 t , a2 t , ..., a n t ); In the formula, a i t represents the action of the ith agent at time t; S6-3. Based on global resource utilization and experiment completion indicators, all agents share a joint reward R t , if multiple experiments can use the equipment efficiently, the overall reward increases, when high-priority experiments are completed on time, the system receives higher rewards, and if lower-priority experiments block resource use, the system receives negative rewards: ; In the formula, R i t represents the local reward of each agent, α i is the weight parameter that determines the contribution of each experiment to the overall reward; S6-4. Resource scheduling model learns the joint strategy of all agents π={π1, π2, ..., π n}, each agent has an independent strategy π i , which is used to decide the actions to be taken in different states. During the training process, all agents share a global environment model and train the strategies of all agents at the same time, using a centralized value function Q(S t , A t ), indicating that in the global state S t Next, execute action combination A t The obtained reward, the update formula of the value function is: ; In the formula, β is the learning rate and γ is the discount factor; S6-5. Each agent learns deterministic actions through the policy network, and evaluates the joint actions of all agents through a central evaluator network. Its goal is to optimize the strategy of each agent so that all agents can obtain the best return in the global reward, that is, to obtain the final experimental distribution: Local Q-network: Each agent has an independent Q-network Q i (s i , a i ), used to learn its local action value; Hybrid network: The resource scheduling model uses a hybrid network to combine the local Q values ​​of each agent into a global Q value Q total : Q total =f(Q1, Q2, ..., Q n ); Where f is a nonlinear function modeled by a neural network to ensure the global Q value Q totalMeet the optimization needs of joint strategies.

[0010] Preferably, the detection process of the behavior detection model is as follows: first, call the real-time personnel behavior image data in the behavior database, and input the personnel behavior image data into the fast detection channel and the slow detection channel respectively when inputting. The fast detection channel takes one frame of image every 2 frames, and the slow detection channel takes one frame of image every 16 frames. The fast detection channel samples 32 frames of images, and the slow detection channel samples 4 frames of images. Then, they are respectively input into the transformer module after 3D convolution, and then the multi-head self-attention mechanism is performed in the transformer module, and then the fusion is performed in 2 times, and finally the detection result is output.

[0011] The technical solution of the present invention is: a management device for a smart laboratory based on artificial intelligence, which is applicable to the management method for a smart laboratory based on artificial intelligence proposed above, comprising: The data acquisition module uses the Internet of Things technology to comprehensively monitor the experimental equipment, environmental parameters and personnel activities in the laboratory; The data processing module pre-processes the collected laboratory environment, equipment and personnel data, and stores the preliminarily processed data in categories according to source and purpose: environmental data is stored in the environment database, equipment status data is stored in the equipment database, and personnel behavior image data is stored in the behavior database; The intelligent control module uses artificial intelligence algorithms to conduct in-depth data analysis, and automatically adjusts laboratory temperature and humidity parameters based on the analysis results. It also automatically controls the operation of experimental equipment based on abnormal equipment signals detected in advance to prevent unexpected failures. Intelligent scheduling module, which uses AI algorithms to intelligently allocate resources within the laboratory, analyze the appointment needs of laboratory personnel and the use of existing resources, and dynamically allocate and optimize resources; The intelligent safety management module uses AI algorithms to analyze and judge the behavior of laboratory personnel and identify potential risks in personnel operation behavior.

[0012] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects: The present invention designs a management method and device for a smart laboratory based on artificial intelligence. The method realizes comprehensive monitoring of the laboratory environment and equipment by real-time collection of laboratory environmental data, equipment status and personnel operation behavior, and using technologies such as Internet of Things sensors. After that, the system uses the association relationship <data type certificate, data classification> established in the initialization phase to efficiently identify the received data type, and analyzes and predicts it through AI algorithms, including equipment health status prediction, environmental condition regulation and experimental progress prediction, so as to ensure the smooth progress of the experimental process. In terms of resource allocation, the system intelligently schedules resources according to experimental requirements and equipment availability, and optimizes the allocation of experimental equipment. During the experiment, the system monitors the experimental equipment and personnel operations in real time, provides feedback and handles abnormal situations in a timely manner, to ensure the safety and effectiveness of the experiment. Through continuous machine learning and data accumulation, the system can continuously optimize laboratory resource management, equipment scheduling and experimental scheduling, and achieve the beneficial technical effect of efficient management of the overall laboratory. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is an architectural diagram of the management device of the smart laboratory based on artificial intelligence proposed by the present invention; Figure 2 A network structure diagram of a device health prediction model proposed by the present invention; Figure 3 This is a network structure diagram of a behavior detection model proposed in the present invention. DETAILED DESCRIPTION

[0014] Embodiment 1, as Figure 1 As shown, the present invention proposes an artificial intelligence-based smart laboratory management device, which includes a data acquisition module, a data processing module, an intelligent scheduling module, an intelligent control module and an intelligent safety management module.

[0015] The data acquisition module uses the Internet of Things (IoT) technology to deploy sensors, cameras and other implementation units in the laboratory to comprehensively monitor the experimental equipment, environmental parameters and personnel activities in the laboratory; The data processing module pre-processes the collected laboratory environment, equipment and personnel data, and stores the preliminarily processed data in categories according to source and purpose: environmental data is stored in the environment database, equipment status data is stored in the equipment database, and personnel behavior image data is stored in the behavior database; The intelligent control module uses artificial intelligence (AI) algorithms to conduct in-depth data analysis, and automatically adjusts laboratory temperature and humidity parameters based on the analysis results. It also automatically controls the operation of experimental equipment based on abnormal equipment signals detected in advance to prevent unexpected failures. The intelligent scheduling module uses AI algorithms to intelligently allocate resources within the laboratory, analyze the appointment needs of laboratory personnel and the use of existing resources, and dynamically allocate and optimize resources; The intelligent safety management module uses AI algorithms to analyze and judge the behavior of laboratory personnel and identify potential risks in personnel operation behavior.

[0016] Embodiment 2, a management method of a smart laboratory based on artificial intelligence proposed by the present invention, which is applicable to a management device of a smart laboratory based on artificial intelligence proposed in Embodiment 1, specifically comprises the following implementation steps: Initialization phase: In order to ensure that the data processing module can efficiently identify the data type transmitted by the implementation unit, this embodiment designs a method for rapid data identification, and the specific process is as follows: (1) Install temperature and humidity sensors, air pressure sensors, light sensors, vibration sensors, and cameras at key locations in the laboratory, install chemical sensors to monitor the concentration of volatile gases during the experiment, and connect all sensors to the data processing module through the Internet of Things network; (2) Interconnect the main experimental equipment with the data processing module and use the equipment's built-in status monitoring system to collect the equipment's operating data; (3) Define parameters a, b, p, q, E, and G; Where E is defined in the finite field F p The elliptic curve on which the point (x, y) satisfies y 2 ≡(x 3 +ax+b)mod p,G=(x G ,y G ) is the base point of order q on the curve; (4) Any implementation unit i (various sensors, cameras, and state monitoring systems of experimental equipment) Generate a random number r i , calculate the certificate application parameter P~CA i =r i ×G, then implement unit i {Identity ID i , certificate application parameters P~CA i}Send to data processing module; (5) The data processing module selects a random number R i ∈F p , calculate the first certificate generation parameter P~Ⅰ~GA i =R i ×G, and calculate the second certificate generation parameter P~Ⅱ~GA i =P~Ⅱ~GA i+P~CA i ; (6) The data processing module generates a data type certificate cert i =H(P~Ⅱ~GA i ||ID i )×R i +G×R i ; Among them, H represents the hash function, || represents the string concatenation symbol; (7) The data processing module establishes an identity ID i , data type certificate cert i The relationship between: <Identity ID i , data type certificate cert i > and set the data type to cert i Send to the corresponding implementation unit i ; (8) Data processing is classified according to data source and purpose, and a data type certificate is established i and data classification type j The relationship between: <Data type certificate cert i , data classification type j > Among them, the data processing module sets three types of data categories: environmental data, equipment status data, and personnel behavior image data; Data collection and management stage: S1. Each implementation unit in the data acquisition module i The collected data i Attach your own identity and data type certificate, with cert i #ID i #data i The data collected by itself is transmitted to the data processing module in the packet format; S2, data processing module according to the association relationship < identity ID i , data type certificate cert i > and associated relationship < data type certificate cert i , data type type j >, determine that the received packet format is cert i #ID i #data i The data type of the data is then processed initially, including but not limited to denoising, formatting, and data cleaning operations to ensure data accuracy and consistency. The data processing module then processes the initially processed data according to the data classification typej Classified storage is performed, with environmental data stored in the environmental database, equipment status data stored in the equipment database, and personnel behavior image data stored in the behavior database; S3, the intelligent control module uses AI algorithms to analyze the health status of equipment, environmental conditions and experimental progress, and predict equipment failures or experimental abnormalities in advance. The specific contents include: (1) Call the equipment status data in the equipment database, build an equipment health prediction model, use LSTM (long short-term memory network) to conduct in-depth analysis of equipment operation data, predict the health status of the equipment, and the intelligent control module outputs the possibility of equipment failure and issues an early warning to help experimenters perform equipment maintenance in a timely manner; like Figure 2 As shown in the figure, the prediction process of the equipment health prediction model is as follows: The device health prediction model consists of five layers: input layer, network layer, attention layer, regression layer and output layer; Input layer: receives the features of samples. Each sample contains two parts: features and regression values. Network layer: The network layer is used to model the data. It includes two independent LSTM neuron layers, which process the sequence data output by the input layer respectively. The feature vectors output by the two LSTM layers are concatenated through the concatenation function (Concat) as the output of the network layer [h0, h1, ..., h n ]; The calculation is as follows: k n =LSTM([x n ,y n ]); m n =LSTM([x n ,y n ]); h n =Concat(k n , m n ); In the formula, k n represents the output of the first LSTM layer, m n represents the output of the second LSTM layer, h n Indicates that k n and m n The concatenated vector, x n Represents the characteristics of the nth sample, y n Represents the regression value of the nth sample; Attention layer: The attention layer is used to assign different weight parameters, i.e., attention values, according to the importance of the input information. The output of the network layer [h0, h1, ..., h n] is used as the input of the attention layer. First, the similarity s between each feature in the input vector and the regression value quality inspection is calculated through the attention function. Then, the softmax function is used to normalize the attention to obtain the probability distribution of the attention. Finally, the attention distribution value is matrix-multiplied with the output of the network layer to obtain the attention value. The calculation methods of the attention function, attention distribution value and attention value are as follows: Score(h n ,y n )=V T tanh[Wh n +Uy n ]; a n =softmax(Score(h n ,y n )); t n =h n ×a n ; Where V, W, and U are learnable parameters in the neural network training process, and a n is the attention distribution value, t n is the attention value; Regression layer: It contains one Flatten layer and four fully connected layers (Dense layers), which are used to map the previous input results into predicted values. To prevent overfitting, a Dropout layer is added after each fully connected layer. Output layer: outputs the predicted value corresponding to the input sequence; When the predicted value is ≥0.8, the operating status of the experimental equipment is consistent with the historical data mode, there is no abnormal signal, and the intelligent control module does not issue an alarm message; When 0.4<predicted value<0.8, the operating data of the experimental equipment begins to fluctuate, which may be slight overheating, abnormal vibration amplitude, etc., but it will not immediately affect the experimental process. The intelligent control module prompts the experimenter to monitor the equipment status; When the predicted value is ≤0.4, the operating data of the experimental equipment shows significant abnormal signals, such as drastic changes in temperature or vibration. The intelligent control module issues an alarm and recommends that the equipment be stopped immediately for inspection. (2) Call the environmental data in the environmental database to monitor the laboratory test environment in real time to ensure that it meets the test conditions. If the environmental data exceeds the set range, the intelligent control module automatically adjusts the environmental conditions and sends a warning message to the experimenter; S4. Based on experimental requirements and equipment availability, the intelligent scheduling module uses AI algorithms to optimize resource allocation to ensure the effective use of experimental equipment, materials and space. The specific implementation steps are as follows: S4-1, the experimenter inputs the experiment requirements (experimental objectives, equipment requirements, required materials, and experiment duration) through the smart terminal (tablet, mobile phone, computer). After receiving the experiment requirements, the intelligent scheduling module first analyzes the experiment content to determine whether the equipment required for the experiment meets the current laboratory resource status. The analysis includes equipment type, material type, and laboratory space requirements. Then, it uses historical data, equipment status, experimental material inventory and other information to evaluate whether the experiment has the required resources. If the current laboratory resources are sufficient, the intelligent scheduling module confirms the experiment plan and continues with subsequent allocation. If the resources are insufficient, the system will prompt the experimenter to adjust the experiment time or the order of equipment use; S4-2, the intelligent scheduling module calls the equipment data in the equipment database, analyzes the current use of experimental equipment through the resource scheduling model, and predicts resource requirements based on historical data to determine the availability of experimental equipment and materials. The structure of the resource scheduling model is as follows: (1) State representation and modeling: The state of laboratory resources and the local state of each agent (experiment) constitute the global state space. The global state S t At time step t, it contains the following information: S t ={(s1 t , s2 t , ..., s n t )}; In the formula, s i t Represents the local state of the ith agent at time t, including the agent's current requirements (equipment, time) and experimental progress; (2) Action definition and modeling: The action a1 taken by each agent (experiment) t It includes the following types: select a specific device to perform the experiment, adjust the start time of the experiment, and delay the execution of a certain experiment; For each time step t, the intelligent scheduling module records the action combination of all agents: A t =(a1 t , a2 t , ..., a n t ); In the formula, a i t represents the action of the ith agent at time t; (3) Reward design: Based on global resource utilization and experiment completion indicators, all agents share a joint reward R t, if multiple experiments can use the device efficiently (such as reducing device idle time), the overall reward increases. When high-priority experiments are completed on time, the system receives higher rewards. If lower-priority experiments block resource usage, the system receives negative rewards: ; In the formula, R i t represents the local reward of each agent, α i is the weight parameter that determines the contribution of each experiment to the overall reward; (4) Joint strategy and learning: The resource scheduling model learns the joint strategy of all agents π = {π1, π2, ..., π n}, each agent has an independent strategy π i , which is used to decide the actions to be taken in different states. During the training process, all agents share a global environment model and train the strategies of all agents at the same time, using a centralized value function Q(S t , A t ), indicating that in the global state S t Next, execute action combination A t The obtained reward, the update formula of the value function is: ; In the formula, β is the learning rate and γ is the discount factor; (5) Each agent learns deterministic actions through the policy network, and evaluates the joint actions of all agents through a central critic network. Its goal is to optimize the strategy of each agent so that all agents obtain the best return in the global reward, that is, to obtain the final experimental allocation, including: Local Q-network: Each agent has an independent Q-network Q i (s i , a i ), used to learn its local action value; Hybrid network: The resource scheduling model uses a hybrid network to combine the local Q values ​​of each agent into a global Q value Q total : Q total =f(Q1, Q2, ..., Q n ); Where f is a nonlinear function modeled by a neural network to ensure the global Q value Q total Meet the optimization needs of joint strategies; S4-3, the intelligent scheduling module automatically allocates experimental equipment according to experimental requirements and the status of experimental equipment. If resources are scarce, the intelligent scheduling module will adjust the experimental time and equipment allocation order; S5. During the experiment, the intelligent safety management module monitors the laboratory environment and personnel operation behavior in real time, provides timely feedback and handles exceptions to ensure the safety and smooth progress of the experiment. The specific implementation process is as follows: S5-1. The intelligent safety management module calls the environmental data in the environmental database to automatically analyze the content of harmful gases in the environment. When an abnormality is found, it immediately issues an alarm and notifies the laboratory personnel. S5-2, the intelligent security management module calls the real-time personnel behavior image data in the behavior database to build a behavior detection model, such as Figure 3 As shown in the figure, first, the personnel behavior image data is input into two channels, namely the fast detection channel and the slow detection channel. The fast detection channel samples the input image in the time domain according to the rule of taking one frame every 2 frames, and the slow detection channel samples one frame every 16 frames. The fast detection channel samples 32 frames of images, and the slow detection channel samples 4 frames of images. Then, they are respectively input into the transformer module after 3D convolution (3D conv). Then, the multi-head self-attention mechanism is performed in the transformer module, and then the fusion is performed in 2 times. Finally, the detection result is output; When violations are detected, the intelligent security management module provides real-time alerts and warnings.

[0017] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A management method for a smart laboratory based on artificial intelligence, characterized in that: The specific implementation steps include the following: S1. Connect each implementation unit in the laboratory with the data processing module. The data processing module generates a data type certificate for each implementation unit and establishes an association relationship <identity identification, data type certificate> and <data type certificate, data classification>; S2. In the data collection and management stage, each implementation unit in the data collection module attaches its own identity identification and data type certificate to the collected data, and transmits the collected data to the data processing module; S3, the data processing module determines the data type of the received data according to the association relationship <identity identifier, data type certificate>, and then identifies the data classification according to the association relationship <data type certificate, data classification>, and then performs preliminary processing on the received data, and classifies and stores the data after preliminary processing according to the data classification; S4, the intelligent control module uses the equipment health prediction model to predict the health status of the equipment, and monitors the real-time experimental environment of the laboratory. If the environmental data exceeds the set range, the intelligent control module automatically adjusts the environmental conditions and sends a warning message to the experimenter; S5, the intelligent scheduling module builds a resource scheduling model, analyzes the current use of experimental equipment, and predicts resource requirements based on historical data to determine the availability of experimental equipment and materials; S6. The intelligent safety management module automatically analyzes the content of harmful gases in the experimental environment. When an abnormality is found, it immediately issues an alarm and notifies the laboratory personnel. On the other hand, it uses a behavioral detection model to detect violations by laboratory personnel, and issues real-time reminders and warnings when abnormal behavior is found.

2. According to the management method of a smart laboratory based on artificial intelligence according to claim 1, it is characterized in that: The classification standard for data classification is: classification based on data source and purpose, and classification of collected data into: environmental data, equipment status data and personnel behavior image data.

3. According to the management method of a smart laboratory based on artificial intelligence according to claim 1, it is characterized in that: The data type certificate is generated as follows: S3-1. Implementation unit i Generate a random number r i , calculate the certificate application parameter P~CA i =r i ×G, then implement unit i {Identity ID i , certificate application parameters P~CA i }Send to data processing module; Where G is the base point of order q on the elliptic curve E, and E is defined in the finite field F p Elliptic curves on ; S3-2, the data processing module selects a random number R i ∈F p , calculate the first certificate generation parameter P~Ⅰ~GA i =R i ×G, and calculate the second certificate generation parameter P~Ⅱ~GA i =P~Ⅱ~GA i +P~CA i ; S3-3. The data processing module generates a data type certificate cert i =H(P~Ⅱ~GA i ||ID i )×R i +G×R i ; Among them, H represents the hash function, and || represents the string concatenation symbol.

4. The management method of a smart laboratory based on artificial intelligence according to claim 1, characterized in that: The equipment health prediction model introduces the attention mechanism into the LSTM neural network, conducts in-depth analysis of equipment operation data, predicts the health status of the equipment, and outputs the possibility of equipment failure.

5. According to the management method of a smart laboratory based on artificial intelligence in claim 1, it is characterized in that: The specific implementation steps of S5 are as follows: S5-1, the experimenter inputs the experiment requirements through the intelligent terminal. After receiving the experiment requirements, the intelligent scheduling module first analyzes the experiment content to determine whether the equipment required for the experiment meets the current laboratory resource status. Then, it uses historical data, equipment status, and experimental material inventory information to evaluate whether the experiment has the required resources. If the current laboratory resources are sufficient, the intelligent scheduling module confirms the experiment plan and continues with subsequent allocation. If the resources are insufficient, the system will prompt the experimenter to adjust the experiment time or the order of equipment use; S5-2, the intelligent scheduling module calls the equipment data in the equipment database, analyzes the current usage of experimental equipment through the resource scheduling model, and predicts resource requirements based on historical data to determine the availability of experimental equipment and materials; S5-3. The intelligent scheduling module automatically allocates experimental equipment according to the experimental requirements and the status of the experimental equipment. If resources are tight, the intelligent scheduling module will adjust the experimental time and equipment allocation order.

6. According to claim 5, a management method of a smart laboratory based on artificial intelligence is characterized in that: The scheduling process of the resource scheduling model is as follows: S6-1, the state of laboratory resources and the local state of each agent constitute the global state space, the global state S t At time step t, it contains the following information: S t ={(s1 t ,s2 t ,...,s n t )}; In the formula, s i t represents the local state of the ith agent at time t, including the agent's current needs and experimental progress; Among them, the agent represents the experiment in the laboratory; S6-2. Action a1 taken by each agent t It includes the following types: select a specific device to perform the experiment, adjust the start time of the experiment, and delay the execution of a certain experiment; For each time step t, the intelligent scheduling module records the action combination of all agents: <h2 style=";text-align:left;direction:ltr">A<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> =(a1<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> a2<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> ,...,a<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> ); In the formula, a i t represents the action of the ith agent at time t; S6-3. Based on global resource utilization and experiment completion indicators, all agents share a joint reward R t , if multiple experiments can use the equipment efficiently, the overall reward increases, when high-priority experiments are completed on time, the system receives higher rewards, and if lower-priority experiments block resource use, the system receives negative rewards: ; In the formula, R i t represents the local reward of each agent, α i is the weight parameter that determines the contribution of each experiment to the overall reward; S6-4. Resource scheduling model learns the joint strategy of all agents π={π1, π2, ..., π n }, each agent has an independent strategy π i , which is used to decide the actions to be taken in different states. During the training process, all agents share a global environment model and train the strategies of all agents at the same time, using a centralized value function Q(S t , A t ), indicating that in the global state S t Next, execute action combination A t The obtained reward, the update formula of the value function is: ; In the formula, β is the learning rate and γ is the discount factor; S6-5. Each agent learns deterministic actions through the policy network, and evaluates the joint actions of all agents through a central evaluator network. The goal is to optimize the strategy of each agent so that all agents can obtain the best return in the global reward, that is, to obtain the final experimental allocation, including: Local Q-network: Each agent has an independent Q-network Q i (s i , a i ), used to learn its local action value; Hybrid network: The resource scheduling model uses a hybrid network to combine the local Q values ​​of each agent into a global Q value Q total : Q total =f(Q1,Q2,...,Q n ); Where f is a nonlinear function modeled by a neural network to ensure the global Q value Q total Meet the optimization needs of joint strategies.

7. According to claim 1, a management method of a smart laboratory based on artificial intelligence is characterized in that: The detection process of the behavior detection model is as follows: First, call the real-time personnel behavior image data in the behavior database, and input the personnel behavior image data into the fast detection channel and the slow detection channel respectively. The fast detection channel takes one frame of the input image every 2 frames, and the slow detection channel takes one frame of the input image every 16 frames. The fast detection channel samples 32 frames of images, and the slow detection channel samples 4 frames of images. Then they are 3D convolved and input into the transformer module respectively. Then, the multi-head self-attention mechanism is performed in the transformer module, and then fusion is performed in 2 times. Finally, the detection result is output.

8. A management device for a smart laboratory based on artificial intelligence, which is applicable to the management method for a smart laboratory based on artificial intelligence according to any one of claims 1 to 7, characterized in that: include: The data acquisition module uses the Internet of Things technology to comprehensively monitor the experimental equipment, environmental parameters and personnel activities in the laboratory; The data processing module pre-processes the collected laboratory environment, equipment and personnel data, and stores the preliminarily processed data in categories according to source and purpose: environmental data is stored in the environment database, equipment status data is stored in the equipment database, and personnel behavior image data is stored in the behavior database; The intelligent control module uses artificial intelligence algorithms to conduct in-depth data analysis, and automatically adjusts laboratory parameters such as temperature and humidity based on the analysis results. It also automatically controls the operation of experimental equipment based on abnormal equipment signals detected in advance to prevent unexpected failures. Intelligent scheduling module, which uses AI algorithms to intelligently allocate resources within the laboratory, analyze the appointment needs of laboratory personnel and the use of existing resources, and dynamically allocate and optimize resources; The intelligent safety management module uses AI algorithms to analyze and judge the behavior of laboratory personnel and identify potential risks in personnel operation behavior.

Citation Information

Patent Citations

  • DAA (Data Access Arrangement) authentication method and system based on elliptical curve divergence logarithm intractability assumption

    CN102638345A

  • Network resource scheduling method and system based on deep reinforcement learning

    CN113254197A

  • Taking identification method and device based on double-channel cross attention mechanism

    CN113936339A

  • Human interaction behavior detection method based on self-attention mechanism

    CN114782995A

  • Intelligent laboratory management system

    CN117332961A

Cited By

  • Laboratory intelligent comprehensive management method and system

    CN119359265A

  • Intelligent laboratory and automatic laboratory sample distribution system

    CN119398461A

  • Building energy consumption management system and method based on big data

    CN120387905A

  • Experiment operation automatic control method and device based on resource scheduling

    CN120672047A

  • Laboratory equipment management method and system based on artificial intelligence

    CN120996750A