Laboratory cloud monitoring platform and laboratory monitoring method

Abstract

The invention provides a laboratory cloud monitoring platform and a laboratory monitoring method, and belongs to the technical field, the laboratory cloud monitoring platform comprises an acquisition layer, the acquisition layer monitors the use condition of a laboratory by using a sensor, and a data transmission layer receives monitoring data from the acquisition layer by using a wireless communication technology and outputs the data. The data processing layer is used for optimizing the received data, analyzing the data and automatically calibrating the monitoring state of the sensor according to the received data, the data processing layer comprises a data preprocessing module, and the data preprocessing module is used for capturing the dynamic change of the received data to obtain high-quality data; the data analysis module is used for carrying out trend analysis according to the data processed by the data processing layer; and the automatic calibration module is used for carrying out automatic calibration processing on the sensor. The problem that in the actual operation process, the sensor may be affected by laboratory environment factors, and consequently the measurement precision is lowered or faults occur is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of laboratory equipment, and in particular relates to a laboratory cloud monitoring platform and a laboratory monitoring method. Background Art

[0002] Many research institutes and companies now have their own comprehensive laboratories, and a single organization may even have multiple laboratories. Real-time monitoring of these multiple laboratories has become a common problem in laboratory management. Generally, companies require laboratory personnel to constantly patrol the labs to understand the equipment status and experimental data. However, this is not possible for laboratory personnel to patrol the labs at all times, making it difficult to monitor the equipment status and data in real time. This creates inconvenience for laboratory personnel to understand the real-time status of the labs.

[0003] Therefore, in the existing technology for remote monitoring of laboratories, traditional monitoring systems are usually only able to perform simple data collection and storage, lacking the ability to deeply mine data and extract valuable information from massive amounts of monitoring data. They also lack the ability to predict and analyze historical data, making it impossible to provide early warning of potential problems. For example, patent announcement number CN118094461A discloses a multi-information fusion laboratory monitoring system and method. By fusing multiple information, applying advanced data processing algorithms, and integrating adaptive learning mechanisms, it improves the ability to identify, predict, and respond to potential laboratory security threats. Although multiple high-precision sensors are used, during actual operation, the sensors may be affected by laboratory environmental factors, resulting in reduced measurement accuracy or failure. To this end, a laboratory cloud monitoring platform and laboratory monitoring method are designed. Summary of the Invention

[0004] The embodiments of the present invention provide a laboratory cloud monitoring platform and a laboratory monitoring method, which solve the problem that sensors may be affected by laboratory environmental factors during actual operation, resulting in reduced measurement accuracy or malfunction.

[0005] In view of the above problems, the technical solution proposed by the present invention is:

[0006] The present invention provides a laboratory cloud monitoring platform, including a collection layer that uses sensors to monitor the use of the laboratory;

[0007] A data transmission layer, which uses wireless communication technology to receive the monitoring data from the collection layer and output the data;

[0008] The data processing layer is used to optimize and analyze the received data, and automatically calibrate the monitoring status of the sensor based on the received data. The data processing layer includes a data preprocessing module, which is used to capture the dynamic changes of the received data and perform learning and optimization on it to obtain high-quality data;

[0009] A data analysis module, which is used to perform trend analysis based on the data processed by the data processing layer to understand the development trend of laboratory usage;

[0010] An automatic calibration module is used to compare and analyze sensor data with reference standard data to obtain the status of the sensor and perform automatic calibration on the sensor;

[0011] The application layer provides a user interface for displaying monitoring data.

[0012] As a preferred technical solution of the present invention, the data preprocessing module includes a preprocessing unit, which is used to convert the original sensor data into a format suitable for spatiotemporal graph convolution processing to eliminate scale differences between the data;

[0013] A high-quality data processing unit, which uses a spatiotemporal graph convolutional adversarial network to capture spatiotemporal correlations in the data, generates high-quality data representations, and provides high-quality data for subsequent processing;

[0014] A model optimization unit, which uses a quantum annealing algorithm to optimize the parameter update process in federated learning to improve learning efficiency and model performance;

[0015] A data calibration unit uses an asymmetric federated learning calibration algorithm to perform local and global calibration on its own data to maintain consistency.

[0016] As a preferred technical solution of the present invention, the details of data processing performed by the high-quality data processing unit are as follows:

[0017] Step a: Convert the received sensor data into a spatiotemporal graph representation. Use a graph convolutional network to learn the relationship matrix between nodes, capture the spatial correlation of the sensor data, and combine it with a temporal convolutional network to process time series data. Output a feature representation that integrates spatial and temporal information, which is a high-quality data representation.

[0018] Step b: Generate the structure of the adversarial network. The generator learns the distribution of real sensor data, and the discriminator distinguishes between real data and generated data. The reconstruction loss and adversarial loss are combined to optimize the spatiotemporal graph convolutional adversarial network model. Through iterative training, the model achieves a balance between reconstructing real data and adversarial generated data.

[0019] The details of the data calibration performed by the data calibration unit are as follows:

[0020] In step c, considering the asymmetry between different sensors, a corresponding calibration algorithm is selected. Each data terminal performs local calibration based on its own data characteristics to eliminate individual differences. During the federated learning process, calibration parameters are exchanged and the asymmetric federated learning calibration algorithm is combined with the spatiotemporal graph convolutional adversarial network to further calibrate the data processed by the spatiotemporal graph convolutional adversarial network.

[0021] As a preferred technical solution of the present invention, the details of the optimization of the model of the data calibration unit by the model optimization unit are as follows:

[0022] In step d, the model parameters in federated learning are encoded into the state of quantum bits, the optimal solution is found in the parameter space through the quantum annealing algorithm, the optimized parameters are decoded, and used to update the model of the cloud monitoring platform.

[0023] As a preferred technical solution of the present invention, the automatic calibration module includes a parameter standard setting unit, which is set according to the reference standards fixed in the laboratory and the sensor;

[0024] An error comparison unit is used to compare the sensor data with the reference standard data, calculate the difference between the two, perform statistical analysis on the comparison results, and analyze the source of the error;

[0025] A calibration model unit is provided, wherein the calibration model unit establishes a calibration model of the sensor based on the error result of the error comparison unit, selects a suitable model according to the error characteristics, integrates the calibration model into the acquisition layer, and automatically applies the calibration model to correct the sensor data each time data is collected.

[0026] As a preferred technical solution of the present invention, the automatic calibration module uses pruning technology to simplify calculations, preliminarily prunes the existing calibration process, identifies and retains key steps, quantitatively analyzes the pruned process, evaluates the impact of each step, and further optimizes the process based on the quantitative results to obtain simplified automatic calibration steps.

[0027] As a preferred technical solution of the present invention, the data analysis module uses a combination of a multi-scale temporal graph network and a gated recurrent unit to perform trend analysis on the data processed by the data processing module. The specific trend analysis steps are as follows:

[0028] Step 1: Treat each time point in the time series as a node in the graph, define edges based on the proximity or correlation of time points, and convert the time series data into a graph structure;

[0029] Step 2: Divide the graph into multiple scales. Each scale focuses on relationships within a different time range. Design a gated recurrent layer for each scale to capture the temporal dependencies at that scale. Use the node embeddings learned by the multi-scale temporal graph network as input to the gated recurrent layer, which is then trained using historical data.

[0030] Step 3: The prediction data of the gated recurrent layers of different scales are fused and passed through one or more fully connected layers to form a prediction layer. The fused data is mapped to the target prediction space and linearly transformed. Nonlinear characteristics may be introduced through nonlinear activation functions. The output dimension and activation function of the last layer of the prediction layer are designed according to the task type. The output of the prediction layer is the prediction result of the model.

[0031] Step 4: Perform regular evaluation and optimization of the model and update the model regularly.

[0032] In another aspect, a laboratory monitoring method of a laboratory cloud monitoring platform comprises the following steps:

[0033] S1, install multiple temperature sensors, humidity sensors, and gas sensors inside the laboratory to collect temperature, humidity, and gas concentration data in different areas of the laboratory. At the same time, install multiple cameras to monitor the situation inside the laboratory in real time;

[0034] S2, transmitting the data collected by the sensor and the video image collected by the camera to the data processing layer through the data transmission layer, optimizing the real-time data through the data processing layer to provide a more accurate data basis for the data analysis module, and automatically calibrating the sensor through the automatic calibration module;

[0035] S3 uses a combination of a multi-scale temporal graph network and a gated recurrent unit to predict data trends and determine whether there are any abnormalities in the laboratory environment or equipment based on the prediction results;

[0036] S4, when an abnormal situation is detected, the cloud monitoring platform generates an alarm message and notifies relevant personnel through various means;

[0037] S5. At the same time, the real-time monitoring data is displayed through the user interface provided by the application layer, and the user can query the monitoring data within a specific time period according to needs.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] (1) The present invention uses graph convolutional networks and temporal convolutional networks to capture spatial correlation and temporal dynamic changes, providing more comprehensive data features, effectively filtering environmental noise, improving data stability and reliability, ensuring the consistency of sensor data through asymmetric federated learning calibration, reducing measurement deviations, and improving model training efficiency through quantum annealing algorithms, making the data more accurate and stable, and significantly reducing the measurement errors of sensors caused by environmental factors;

[0040] (2) The present invention reduces the impact of environmental factors on measurement results through automatic calibration and spatiotemporal graph coordination, ensuring data accuracy and reliability. It accurately predicts environmental change trends through multi-scale temporal graph networks and gated recurrent units, improves prediction accuracy, and can detect potential environmental anomalies earlier, providing a basis for taking timely measures.

[0041] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a block diagram of a laboratory cloud monitoring platform disclosed in the present invention;

[0043] Figure 2 It is a flow chart of a laboratory monitoring method of a laboratory cloud monitoring platform disclosed in the present invention;

[0044] Explanation of the accompanying drawings: 100, acquisition layer; 200, data transmission layer; 300, data processing layer; 301, data preprocessing module; 3011, preprocessing unit; 3012, high-quality data processing unit; 3013, model optimization unit; 3014, data calibration unit; 302, data analysis module; 303, automatic calibration module; 3031, reference standard setting unit; 3032, error comparison unit; 3033, calibration model unit; 400, application layer. DETAILED DESCRIPTION

[0045] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are also within the scope of protection of the present invention.

[0047] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0048] In the description of the present invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," "clockwise," "counterclockwise," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore should not be understood as limiting the present invention.

[0049] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Thus, features specified with "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0050] Example 1

[0051] Refer to the attached Figure 1 As shown, the present invention provides a technical solution: a laboratory cloud monitoring platform, including a collection layer 100, which uses sensors to monitor the use of the laboratory, collects laboratory environmental data such as temperature, humidity, and gas concentration in real time through the sensor module, and collects video images inside the laboratory in real time through the video monitoring module;

[0052] The data transmission layer 200 receives the monitoring data from the collection layer 100 using wireless communication technology and outputs the data;

[0053] The data processing layer 300 is used to optimize and analyze the received data, and automatically calibrate the monitoring status of the sensor based on the received data. The data processing layer 300 includes a data preprocessing module 301, which is used to capture the dynamic changes of the received data and perform learning and optimization on it to obtain high-quality data.

[0054] The data analysis module 302 is used to perform trend analysis based on the data processed by the data processing layer 300 to understand the development trend of laboratory usage;

[0055] Automatic calibration module 303, automatic calibration module 303 is used to compare and analyze sensor data with reference standard data, obtain the status of the sensor, and perform automatic calibration processing on the sensor;

[0056] Application layer 400: The application layer 400 provides a user interface for displaying monitoring data, alarm information, etc.

[0057] The embodiment of the present invention is also implemented through the following technical solutions.

[0058] In an embodiment of the present invention, the data preprocessing module 301 includes a preprocessing unit 3011, which is used to convert the raw sensor data into a format suitable for spatiotemporal graph convolution processing to eliminate scale differences between the data;

[0059] High-quality data processing unit 3012, the high-quality data processing unit 3012 uses the spatiotemporal graph convolutional adversarial network to capture the spatiotemporal correlation in the data, generate high-quality data representation, and provide high-quality data for subsequent processing;

[0060] Model optimization unit 3013, model optimization unit 3013 uses quantum annealing algorithm to optimize the parameter update process in federated learning, thereby improving learning efficiency and model performance;

[0061] The data calibration unit 3014 uses an asymmetric federated learning calibration algorithm to perform local and global calibration on its own data to maintain consistency.

[0062] In the embodiment of the present invention, the details of the data processing performed by the high-quality data processing unit 3012 are as follows:

[0063] Step a: Convert the received sensor data into a spatiotemporal graph representation, where nodes represent sensors and edges represent the spatiotemporal relationships between sensors. A graph convolutional network is used to learn the relationship matrix between nodes, capturing the spatial correlation of sensor data. A temporal convolutional network is then used to process time series data. The temporal convolutional network uses convolution kernels of varying lengths to capture short-term and long-term dynamic changes in the time series. After spatiotemporal convolution processing, the output is a feature representation that integrates spatial and temporal information, resulting in high-quality data representations because they contain both spatial and temporal correlation information.

[0064] Step b: Generate the structure of the adversarial network. The generator learns the distribution of real sensor data, and the discriminator distinguishes between real data and generated data to improve the generalization ability of the model. Reconstruction loss and adversarial loss are combined. The reconstruction loss ensures that the generated data is close to the real data at the feature level. The adversarial loss improves the generalization ability of the model through the game between the generator and the discriminator. Through this adversarial process, the authenticity and quality of the generated data are further ensured. The spatiotemporal graph convolutional adversarial network model is optimized, and optimization algorithms such as Adam are used to update the model parameters. Through iterative training, the model reaches a balance between reconstructing real data and adversarial generated data. The generated data at this balance point is a high-quality data representation.

[0065] The details of data calibration performed by the data calibration unit 3014 are as follows:

[0066] In step c, considering the possible asymmetry in data distribution and noise levels between different sensors, a corresponding calibration algorithm is selected. Each data end performs local calibration based on its own data characteristics to eliminate individual differences. In the federated learning process, global consistency calibration is achieved by exchanging calibration parameters instead of directly sharing data. This protects data privacy while achieving global calibration. The asymmetric federated learning calibration algorithm is combined with the spatiotemporal graph convolutional adversarial network to further calibrate the data processed by the spatiotemporal graph convolutional adversarial network, improving data quality and model performance.

[0067] The detailed steps of local calibration and global calibration are as follows:

[0068] Step c1: Local calibration. Each data end initializes local calibration parameters based on its own data characteristics. These parameters include but are not limited to scale factors, offsets, weights, etc. That is, each data end performs preliminary calibration based on its own data characteristics. The data end uses its own data to optimize the initialized calibration parameters, usually by minimizing a loss function such as mean square error or cross entropy loss, so that the calibration parameters are closer to the true distribution of its own data. The data end uploads the optimized local calibration parameters to the cloud monitoring platform and shares them with other data ends, providing a basis for global calibration. The cloud monitoring platform or participating decentralized nodes aggregate the received local calibration parameters. The aggregation method can be simple averaging, weighted averaging, or other more complex strategies such as federated averaging.

[0069] Step c2: Global calibration parameter generation. Based on the aggregated parameters, global calibration parameters are calculated. These global parameters represent the consensus of all data terminals and are used to achieve globally consistent calibration. The calculated global calibration parameters are distributed to each data terminal. After receiving the global calibration parameters, the data terminal updates its local calibration parameters. This can be achieved through direct replacement, weighted fusion, or other update strategies. The data terminal can use the updated parameters for further local optimization as needed to better adapt to the characteristics of its own data.

[0070] Step c3: The above process can be performed iteratively. That is, after multiple local optimizations and parameter exchanges, the data end gradually converges to the globally optimal calibration parameters. The iterative process is terminated by setting convergence conditions, such as when the parameter change is less than a certain threshold or when a preset number of iterations is reached.

[0071] The details of how the model optimization unit 3013 optimizes the model of the data calibration unit 3014 are as follows:

[0072] In step d, multiple data terminals store local data and do not directly share data, but only exchange model parameters. The cloud monitoring platform coordinates parameter exchange and model aggregation between clients. The model parameters in federated learning are encoded into quantum bit states. In other words, the model parameters are converted into a format that can be processed by quantum computing, namely quantum bit states. The quantum annealing algorithm is used to find the optimal solution in the parameter space, accelerate model convergence, and decode the optimized parameters to update the model of the cloud monitoring platform.

[0073] Among them, the details of finding the optimal solution are: initialize a quantum system so that it is in a specific initial state. This initial state is usually a superposition state, representing multiple possible positions in the parameter space. During the quantum annealing process, the system is subjected to a specific Hamiltonian. This Hamiltonian contains the specific energy landscape of the problem. The design of the Hamiltonian makes the low-energy state of the system correspond to the optimal solution of the optimization problem. Gradually adjust the parameters in the Hamiltonian so that the system gradually cools from the initial high-energy state to the low-energy state. This process simulates the evolution of the quantum system in the energy landscape, making it tend to the lowest energy state, that is, the optimal solution. After a period of quantum annealing, the quantum system is measured, and the measurement results will give one or more possible parameter combinations, which correspond to the potential optimal solutions to the optimization problem.

[0074] In an embodiment of the present invention, the automatic calibration module 303 includes a parameter standard setting unit, which is set according to a fixed reference standard in the laboratory and the sensor to ensure that its position and measurement environment are consistent with the sensor. The reference standard should be regularly calibrated and certified to maintain its accuracy.

[0075] The error comparison unit 3032 is used to compare the sensor data with the reference standard data, calculate the difference between the two, perform statistical analysis on the comparison results, identify systematic errors and random errors, and analyze the sources of errors, such as sensor drift and environmental factors;

[0076] The calibration model unit 3033 establishes a calibration model for the sensor based on the error result of the error comparison unit 3032, selects a suitable model according to the error characteristics, and integrates the calibration model into the acquisition layer 100 to achieve real-time or periodic automatic calibration. Each time data is collected, the calibration model is automatically applied to correct the sensor data. The automatically calibrated data is regularly compared with the reference standard data to verify the calibration effect. If the calibration effect is found to be poor, the calibration model is adjusted or the error analysis is performed again.

[0077] In an embodiment of the present invention, the automatic calibration module 303 uses pruning technology to simplify calculations, preliminarily prunes the existing calibration process, identifies and retains key steps, performs quantitative analysis on the pruned process, evaluates the impact of each step, and further optimizes the process based on the quantitative results to obtain simplified automatic calibration steps, such as adjusting parameters and reconstructing steps. Through iteration, the pruning and quantization strategies are continuously improved until a satisfactory simplification effect is achieved, thereby achieving automation of the simplified calibration process and reducing manual intervention. Through such a process, the pruning technology removes unimportant steps, which can speed up the calibration process and reduce resource consumption. The quantization technology can more accurately adjust and optimize parameters, effectively simplifying the calibration process, improving the accuracy and efficiency of calibration, and reducing the complexity and time cost of the operation.

[0078] The high-quality data processing unit 3012 uses multi-threading technology to process multiple sensor data in parallel, the model optimization unit 3013 uses multi-threading technology to accelerate the optimization process, and the data calibration unit 3014 uses distributed computing resources to achieve parallel computing of global parameter aggregation;

[0079] When the data is complex, the automatic calibration module 303 and the data preprocessing module 301 use the cloud monitoring platform to continuously monitor the trigger conditions. When the trigger conditions are met, the simplified processing steps are automatically started. By setting the performance threshold of the calibration process, such as calibration time, resource consumption, etc., when the actual performance exceeds the threshold, the simplified processing is triggered and the accuracy of the calibration results is monitored. When the accuracy drops to a certain level, the simplified processing is triggered to optimize the parameters. The application layer 400 provides data visualization function to generate detailed analysis reports to help staff understand the trigger conditions and the effects of the simplified processing.

[0080] In an embodiment of the present invention, the data analysis module 302 uses a combination of a multi-scale temporal graph network and a gated recurrent unit to perform trend analysis on the data processed by the data processing module. The specific trend analysis steps are as follows:

[0081] Step 1: Treat each time point in the time series as a node in the graph. Define edges based on the proximity or correlation of time points. Edges of different lengths can be set to represent relationships at different time scales. This converts the time series data into a graph structure, capturing the temporal dependencies and structural information in the data.

[0082] Step 2: Divide the graph into multiple scales. Each scale focuses on relationships within a different time range. Design a gated recurrent layer for each scale to capture the temporal dependencies at that scale. Use the node embeddings learned by the multi-scale temporal graph network as input to the gated recurrent layer, which is then trained using historical data.

[0083] Step three: Fuse the prediction data from gated recurrent layers of different scales. Methods such as weighted averaging can be used to form a prediction layer through one or more fully connected layers. The fully connected layer is a linear transformation that maps the fused data to the target prediction space, performs a linear transformation, and may introduce nonlinear characteristics through a nonlinear activation function. Each fully connected layer has a set of learnable weights and bias parameters. After multiplying the input data by the weight matrix and adding the bias vector, the result of the linear transformation is obtained. The output dimension and activation function of the last layer of the prediction layer are designed according to the task type. The output of the prediction layer is the prediction result of the model.

[0084] Step 4: Perform regular evaluation and optimization of the model and update the model regularly.

[0085] For example, suppose we have a simple prediction layer consisting of a fully connected layer and a ReLU activation function for regression tasks. The input features are current temperature, humidity, and pressure values ​​X(25.0, 60.0, 101.3), the weight matrix W is (0.1 0.05 0.02), and the bias vector b is 0.5. Then, after linear transformation, the prediction result of the prediction layer is 25.0*0.1+60.0*0.05+101.3*0.02+0.5=8.026. The predicted temperature value at the next time point is 8.026, which is obviously unreasonable (because the temperature cannot suddenly drop so low). This indicates that the model may need further adjustment and training, and the threshold should be compared with the predicted value for subsequent early warning processing.

[0086] Example 2

[0087] Refer to the attached Figure 2 As shown, an embodiment of the present invention further provides a laboratory monitoring method of a laboratory cloud monitoring platform, comprising the following steps:

[0088] S1, install multiple temperature sensors, humidity sensors, and gas sensors inside the laboratory to collect temperature, humidity, and gas concentration data in different areas of the laboratory. At the same time, install multiple cameras to monitor the situation inside the laboratory in real time;

[0089] S2, the data collected by the sensor and the video image collected by the camera are transmitted to the data processing layer 300 through the data transmission layer 200, and the real-time data is optimized and processed to provide a more accurate data basis for the data analysis module 302. At the same time, the sensor is automatically calibrated by the automatic calibration module 303;

[0090] S3 uses a combination of a multi-scale temporal graph network and a gated recurrent unit to predict data trends and determine whether there are any abnormalities in the laboratory environment or equipment based on the prediction results;

[0091] S4: When an abnormal situation is detected, the cloud monitoring platform generates an alarm and notifies relevant personnel through various means (such as SMS, email, APP push, etc.). The alarm information includes detailed information such as the abnormality type, occurrence time, and specific location;

[0092] S5. At the same time, the user interface provided by the application layer 400 displays real-time monitoring data, historical data, alarm information, etc. Users can query monitoring data within a specific time period according to their needs, and multiple query methods are supported (such as by time, by area, etc.).

[0093] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0094] It should be understood that the specific order or hierarchy of steps in the disclosed processes is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of the present disclosure. The accompanying method claims present elements of the various steps in an exemplary order and are not intended to be limited to a specific order or hierarchy.

[0095] In the foregoing detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that embodiments of the claimed subject matter require more features than are expressly recited in each claim. On the contrary, as reflected in the appended claims, the invention comprises less than all the features of any individual disclosed embodiment. The appended claims are therefore hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0096] Those skilled in the art will also appreciate that the various illustrative logic blocks, modules, circuits, and algorithmic steps described in conjunction with the embodiments herein may be implemented as electronic hardware, computer software, or a combination thereof. In order to clearly illustrate the interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps described above are generally described around their functions. Whether such functions are implemented as hardware or software depends on the specific application and the design constraints imposed on the entire system. A skilled person may implement the described functions in an adaptable manner for each specific application, but such implementation decisions should not be interpreted as departing from the scope of protection of this disclosure.

[0097] The steps of the methods or algorithms described in conjunction with the embodiments herein may be directly embodied as hardware, software modules executed by a processor, or a combination thereof. The software module may be located in a RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, register, hard disk, removable disk, CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is connected to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium may also be an integral part of the processor. The processor and storage medium may be located in an ASIC. The ASIC may be located in a user terminal. Of course, the processor and storage medium may also be present in a user terminal as discrete components.

[0098] For software implementation, the techniques described in this application can be implemented using modules (e.g., procedures, functions, etc.) that perform the functions of the application. These software codes can be stored in a memory unit and executed by a processor. The memory unit can be implemented within the processor or external to the processor. In the latter case, it is communicatively coupled to the processor via various means, which are well known in the art.

[0099] The foregoing description includes examples of one or more embodiments. Of course, it is not possible to describe all possible combinations of components or methods for the purposes of describing the above embodiments, but one of ordinary skill in the art will recognize that the various embodiments may be further combined and arranged. Therefore, the embodiments described herein are intended to encompass all such changes, modifications and variations that fall within the scope of the appended claims. Furthermore, to the extent the term "comprising" is used in the specification or claims, the term is intended to be encompassed in a manner similar to the term "including," as if "including" were used as a transitional word in the claims. Furthermore, any use of the term "or" in the specification of the claims is intended to mean a "non-exclusive or."

Claims

1. A laboratory cloud monitoring platform, characterized in that: It includes a collection layer (100), wherein the collection layer (100) uses sensors to monitor the use of the laboratory; A data transmission layer (200), wherein the data transmission layer (200) receives the monitoring data from the collection layer (100) using wireless communication technology and outputs the data; A data processing layer (300), the data processing layer (300) is used to optimize the received data, analyze the data, and automatically calibrate the monitoring state of the sensor according to the received data. The data processing layer (300) includes a data preprocessing module (301), the data preprocessing module (301) is used to capture the dynamic changes of the received data, and perform learning optimization on the data to obtain high-quality data; A data analysis module (302), the data analysis module (302) is used to perform trend analysis based on the data processed by the data processing layer (300) to grasp the development trend of laboratory usage; An automatic calibration module (303), the automatic calibration module (303) is used to compare and analyze sensor data with reference standard data, obtain the state of the sensor, and perform automatic calibration processing on the sensor; The application layer (400) provides a user interface for displaying monitoring data.

2. A laboratory cloud monitoring platform according to claim 1, characterized in that: The data preprocessing module (301) includes a preprocessing unit (3011), and the preprocessing unit (3011) is used to convert the original sensor data into a format suitable for spatiotemporal graph convolution processing to eliminate scale differences between the data; A high-quality data processing unit (3012), wherein the high-quality data processing unit (3012) uses a spatiotemporal graph convolutional adversarial network to capture spatiotemporal correlations in the data, generate high-quality data representations, and provide high-quality data for subsequent processing; A model optimization unit (3013), wherein the model optimization unit (3013) utilizes a quantum annealing algorithm to optimize the parameter update process in federated learning to improve learning efficiency and model performance; A data calibration unit (3014) is provided, wherein the data calibration unit (3014) uses an asymmetric federated learning calibration algorithm to perform local and global calibration on its own data to maintain consistency.

3. A laboratory cloud monitoring platform according to claim 2, characterized in that: The details of data processing performed by the high-quality data processing unit (3012) are as follows: Step a: Convert the received sensor data into a spatiotemporal graph representation. Use a graph convolutional network to learn the relationship matrix between nodes, capture the spatial correlation of the sensor data, and combine it with a temporal convolutional network to process time series data. Output a feature representation that integrates spatial and temporal information, which is a high-quality data representation. Step b: Generate the structure of the adversarial network. The generator learns the distribution of real sensor data, and the discriminator distinguishes between real data and generated data. The reconstruction loss and adversarial loss are combined to optimize the spatiotemporal graph convolutional adversarial network model. Through iterative training, the model achieves a balance between reconstructing real data and adversarial generated data. The details of the data calibration performed by the data calibration unit (3014) are as follows: In step c, considering the asymmetry between different sensors, the corresponding calibration algorithm is selected. Each data end performs local calibration based on its own data characteristics to eliminate individual differences. During the federated learning process, by exchanging calibration parameters and combining the asymmetric federated learning calibration algorithm with the spatiotemporal graph convolutional adversarial network, the data processed by the spatiotemporal graph convolutional adversarial network is further calibrated.

4. A laboratory cloud monitoring platform according to claim 3, characterized in that: The details of the optimization of the model of the data calibration unit (3014) by the model optimization unit (3013) are as follows: In step d, the model parameters in federated learning are encoded into the state of quantum bits, the optimal solution is found in the parameter space through the quantum annealing algorithm, the optimized parameters are decoded, and used to update the model of the cloud monitoring platform.

5. A laboratory cloud monitoring platform according to claim 4, characterized in that: The automatic calibration module (303) includes a parameter standard setting unit, which is set according to a fixed reference standard in the laboratory and the sensor; An error comparison unit (3032), the error comparison unit (3032) is used to compare the sensor data with the reference standard data, calculate the difference between the two, perform statistical analysis on the comparison results, and analyze the source of the error; A calibration model unit (3033) is provided, wherein the calibration model unit (3033) establishes a calibration model of the sensor based on the error result of the error comparison unit (3032), selects a suitable model according to the error characteristics, integrates the calibration model into the acquisition layer (100), and automatically applies the calibration model to correct the sensor data each time data is collected.

6. A laboratory cloud monitoring platform according to claim 5, characterized in that: The automatic calibration module (303) uses pruning technology to simplify calculations, performs preliminary pruning on the existing calibration process, identifies and retains key steps, performs quantitative analysis on the pruned process, evaluates the impact of each step, and further optimizes the process based on the quantitative results to obtain simplified automatic calibration steps.

7. A laboratory cloud monitoring platform according to claim 6, characterized in that: The data analysis module (302) uses a combination of a multi-scale temporal graph network and a gated recurrent unit to perform trend analysis on the data processed by the data processing module. The specific trend analysis steps are as follows: Step 1: Treat each time point in the time series as a node in the graph, define edges based on the proximity or correlation of time points, and convert the time series data into a graph structure; Step 2: Divide the graph into multiple scales. Each scale focuses on relationships within a different time range. Design a gated recurrent layer for each scale to capture the temporal dependencies at that scale. Use the node embeddings learned by the multi-scale temporal graph network as input to the gated recurrent layer, which is then trained using historical data. Step 3: The prediction data of the gated recurrent layers of different scales are fused and passed through one or more fully connected layers to form a prediction layer. The fused data is mapped to the target prediction space and linearly transformed. Nonlinear characteristics may be introduced through nonlinear activation functions. The output dimension and activation function of the last layer of the prediction layer are designed according to the task type. The output of the prediction layer is the prediction result of the model. Step 4: Perform regular evaluation and optimization of the model and update the model regularly.

8. A laboratory monitoring method for a laboratory cloud monitoring platform, applied to a laboratory cloud monitoring platform according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1, install multiple temperature sensors, humidity sensors, and gas sensors inside the laboratory to collect temperature, humidity, and gas concentration data in different areas of the laboratory. At the same time, install multiple cameras to monitor the situation inside the laboratory in real time; S2, transmitting the data collected by the sensor and the video image collected by the camera to the data transmission layer (200), optimizing the real-time data through the data processing layer (300), providing a more accurate data basis for the data analysis module (302), and automatically calibrating the sensor through the automatic calibration module (303); S3 uses a combination of a multi-scale temporal graph network and a gated recurrent unit to predict data trends and determine whether there are any abnormalities in the laboratory environment or equipment based on the prediction results; S4, when an abnormal situation is detected, the cloud monitoring platform generates an alarm message and notifies relevant personnel through various means; S5. At the same time, the real-time monitoring data is displayed through the user interface provided by the application layer (400), and the user can query the monitoring data within a specific time period according to the needs.

Images