Computer laboratory intelligent detection device based on Internet of Things

Abstract

The invention discloses a computer laboratory intelligent detection device based on Internet of Things. The computer laboratory intelligent detection device comprises a data collection module; a data preprocessing module; a prediction and decision module; an adaptive resource scheduling module; an execution and feedback module; the method is used for continuously optimizing the prediction precision and learning circulation of a resource allocation strategy, the CNN is used for identifying an abnormal mode in an equipment image, and potential faults of equipment can be predicted in advance, so that preventive maintenance measures are taken, and it is ensured that the equipment is in a good state. The RNN is used for capturing the time sequence features in the equipment operation data, the trend change of the equipment operation state can be identified, and the accuracy of equipment state prediction is further improved. Through adaptive resource scheduling, the resources can be flexibly allocated according to actual demands of a laboratory, the overall resource utilization rate is improved, the service life of equipment is prolonged, and delay caused by equipment faults is reduced. And the equipment maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of laboratory automation management, and in particular to an intelligent detection device for a computer laboratory based on the Internet of Things. Background Art

[0002] Laboratories are crucial venues for scientific research and technological development, often equipped with expensive and complex instruments and equipment. As scientific research continues to advance, laboratories face increasing resource demands and equipment maintenance pressures. To address these challenges, many laboratories have begun adopting digital and intelligent approaches to manage equipment and resources.

[0003] Traditional laboratory management often relies on static resource allocation strategies, failing to dynamically adjust resource allocation based on actual needs. This can lead to some equipment being overloaded while others are idle. Maintenance often relies on regular inspections or emergency repairs after a failure. This approach is not only costly but can also result in prolonged equipment downtime during a failure. The lack of an effective resource scheduling mechanism leads to underutilization of laboratory resources, impacting scientific research efficiency and output.

[0004] Therefore, an intelligent detection device for computer laboratories based on the Internet of Things is proposed. Summary of the Invention

[0005] The purpose of the present invention is to address the shortcomings of the prior art, where laboratory management is often based on static resource allocation strategies, unable to dynamically adjust resource allocation according to actual needs, resulting in some equipment being overloaded while other equipment is idle. Maintenance methods mostly rely on regular inspections or emergency repairs after a fault occurs. This approach is not only costly but may also cause equipment to be down for a long time during a fault. The lack of an effective resource scheduling mechanism leads to the underutilization of laboratory resources, affecting scientific research efficiency and output. Therefore, an intelligent detection device for computer laboratories based on the Internet of Things is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] An intelligent detection device for a computer laboratory based on the Internet of Things, comprising:

[0008] Data collection module: used to collect data about equipment operating status from laboratory equipment and user activities;

[0009] Data preprocessing module: used for cleaning, standardizing and feature extraction of the collected data;

[0010] Prediction and decision-making module: used to predict possible future equipment problems and resource requirements, and generate maintenance plans and resource allocation strategies based on the prediction results;

[0011] Adaptive resource scheduling module: used to dynamically adjust the allocation of laboratory resources based on the information provided by the prediction and decision module;

[0012] Execution and feedback module: used to execute the maintenance plan and resource allocation strategy, and collect execution results as feedback to further optimize the prediction and decision module and resource scheduling module:

[0013] A learning loop for continuous optimization of forecast accuracy and resource allocation strategies.

[0014] The above technical solution further includes:

[0015] Preferably, the data collection module includes:

[0016] Sensors and monitoring equipment for real-time monitoring of operating parameters of laboratory equipment;

[0017] User behavior recorder, used to record user operation history and resource requests.

[0018] Preferably, the prediction and decision module includes:

[0019] A device status prediction submodule, which analyzes historical device operating data based on deep learning algorithms, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to predict potential future problems.

[0020] The resource demand prediction submodule is used to predict the resource demand of the laboratory in the future;

[0021] Among them, the equipment status prediction submodule can predict potential equipment failures in advance through machine learning models for preventive maintenance.

[0022] Preferably, the device status prediction submodule includes:

[0023] A data labeling submodule is used to add labels to the historical operation data to form a training data set required for supervised learning;

[0024] A model training submodule, configured to train the deep learning model using the labeled historical operating data;

[0025] The model evaluation submodule is used to verify the prediction accuracy of the deep learning model and make model selection or parameter adjustments accordingly.

[0026] Preferably, the adaptive resource scheduling module includes:

[0027] Real-time monitoring submodule, used to continuously monitor the load conditions in the laboratory;

[0028] Priority management submodule, used to ensure that critical tasks receive sufficient resources;

[0029] Load balancing submodule, used to avoid resource overload or idleness;

[0030] The adaptive resource scheduling module automatically adjusts the resource allocation strategy according to the real-time load situation in the laboratory to ensure that each user can obtain the best service quality.

[0031] Preferably, the adaptive resource scheduling module includes:

[0032] A resource reservation submodule, configured to reserve corresponding computing resources based on the resource demand forecast provided by the forecasting and decision-making module;

[0033] Dynamic adjustment submodule, used to dynamically allocate resources based on actual usage within the laboratory to maximize resource utilization;

[0034] The resource recovery submodule is used to release idle resources in time after the task is completed so that they can be used by other tasks.

[0035] Preferably, the execution and feedback module includes:

[0036] User interface submodule, used to provide users with a visual interface of system status;

[0037] The feedback collection submodule is used to collect the execution results and input them as feedback to the prediction and decision module and the resource scheduling module.

[0038] Preferably, the learning cycle for continuously optimizing prediction accuracy and resource allocation strategy includes:

[0039] Use forecast results to guide resource scheduling decisions;

[0040] Collect execution results as feedback;

[0041] The deep learning model and resource scheduling strategy are optimized based on the feedback results.

[0042] Preferably, the device state prediction submodule uses a convolutional neural network (CNN) and a recurrent neural network (RNN) to predict the device state, wherein:

[0043] CNN is used to extract spatial features in the equipment operating status, including signs of wear in equipment images; the CNN model is formalized as a function:

[0044] f CNN (X;θ CNN )

[0045] Where X represents the input image data, θCNN Represents the parameters of the CNN model;

[0046] The CNN model is trained by minimizing the cross entropy loss function:

[0047] L CNN (f CNN (X;θ CNN ),y)

[0048] To achieve this, where y represents the target output, L CNN is the cross entropy loss function:

[0049] L CNN =-∑ i y i log(f CNN (X i θ CNN ))

[0050] The training process of the CNN model updates the parameters θ by gradient descent method CNN ;

[0051] RNN is used to capture the time series characteristics of the equipment's operating status, including the changing trend of the equipment's vibration signal. The RNN model is formalized as a function:

[0052] f RNN (Y;θ RNN )

[0053] Where Y represents time series data, θ RNN Represents the parameters of the RNN model;

[0054] The RNN model is trained by minimizing the mean squared error loss function:

[0055] L RNN (f RNN (Y;θ RNN ),y)

[0056] To achieve this, where Y represents the target output, L RNN is the mean square error loss function:

[0057]

[0058] The training process of the RNN model also updates the parameters θ by gradient descent method RNN .

[0059] Preferably, the device status prediction submodule includes:

[0060] CNN model training submodule, used to train the CNN model to identify abnormal patterns in device images; the training process updates the parameters θ through the gradient descent methodCNN , to minimize the cross entropy loss function:

[0061] L CNN (f CNN (X;θ CNN ),y)

[0062] The convolutional layers in the CNN model use convolution operations:

[0063]

[0064] To calculate the output, represents the output of the i,jth position of the lth convolutional layer, b (l) represents the bias term, represents the weight of the convolution kernel of the lth layer, represents the activation output of the l-1)th layer, * represents the convolution operation;

[0065] The RNN model training submodule is used to train the RNN model to identify the time correlation in the device operation data; the training process can also update the parameter θ by gradient descent. RNN , to minimize the mean square error loss function:

[0066] L RNN (f RNN (Y;θ RNN ),y)

[0067] The RNN unit in the RNN model uses the state update equation:

[0068]

[0069] To update the hidden state, where h t represents the hidden state at time step t, x t represents the input at time step t, W hh and represents the weight matrix, b h represents the bias term, σ(·) represents the activation function;

[0070] The model fusion submodule is used to integrate the prediction results of CNN and RNN to improve the accuracy of device status prediction; this model fusion adopts the form of weighted average:

[0071] P=w1·f CNN (X;θ CNN )+w2·f RNN (Y;θ RNN )

[0072] Where w1 and w2 are the weight coefficients of CNN and RNN prediction results respectively, and w1+w2=1.

[0073] Preferably, the adaptive resource scheduling module further includes:

[0074] The priority adjustment submodule is used to dynamically adjust the task priority according to the device status and the urgency of the task; priority adjustment is achieved by defining a priority function:

[0075] P(t)=α·S(t)+β·U(t)

[0076] Where S(t) represents the device status score, U(t) represents the urgency of the task, α and β are adjustment coefficients, and α+β=1; the device status score S(t) can be calculated based on the prediction results f of the CNN and RNN models. CNN (X;θ CNN ) and f RNN (Y;θ RNN ) to calculate;

[0077] The load forecasting submodule is used to predict the resource demand and load status of the laboratory in the future. The load forecasting is done by using the ARIMA model f ARIMA (t), using historical data Z = {z1,z2,...,z T} to predict future load z T+1 ,z T+2 ,..., where z t represents the load at the t-th time point, and T represents the last known time point;

[0078] The resource reservation strategy optimization submodule is used to optimize the resource reservation strategy to balance resource utilization and user satisfaction. The resource reservation strategy optimization solves the constrained optimization problem:

[0079] min θ C(θ)stR(θ)≥r min

[0080] Where C(θ) represents the cost function, R(θ) represents the resource utilization, and r min is the minimum resource utilization threshold; during the optimization process, C(θ) and R(θ) are obtained through simulation or actual operation respectively;

[0081] Resource utilization R(θ) can be expressed as:

[0082]

[0083] Where N and M represent the number of tasks and resource units respectively, a ij Indicates whether task i is assigned to resource unit j, d i represents the demand of task i, r jIndicates the availability of resource unit j.

[0084] The present invention has the following beneficial effects:

[0085] 1. In this invention, CNN is used to identify abnormal patterns in equipment images, enabling early prediction of potential equipment failures, enabling preventive maintenance measures to ensure optimal equipment condition. RNN is used to capture time series features in equipment operation data, identifying trends in equipment operating status and further improving the accuracy of equipment status prediction. Furthermore, adaptive resource scheduling ensures that resources can be flexibly allocated based on the actual needs of the laboratory, improving overall resource utilization.

[0086] 2. Predictive maintenance in this invention is more efficient and cost-effective than traditional periodic inspections or emergency repairs after a failure. Furthermore, adaptive resource scheduling ensures that resources are properly planned for the equipment's lifecycle, extending its service life and reducing delays caused by equipment failures. This reduces equipment maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 This is a system structure diagram of an Internet of Things-based intelligent detection device for computer laboratories proposed by the present invention;

[0088] Figure 2 This is a flow chart of an intelligent detection device for computer laboratories based on the Internet of Things in the present invention. DETAILED DESCRIPTION

[0089] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the 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.

[0090] Example 1

[0091] like Figure 1-Figure 2 As shown, the present invention proposes an intelligent detection device for computer laboratories based on the Internet of Things, comprising:

[0092] Data collection module: used to collect data about equipment operating status from laboratory equipment and user activities;

[0093] Data preprocessing module: used to clean, standardize and extract features of collected data;

[0094] Prediction and decision-making module: used to predict possible future equipment problems and resource requirements, and generate maintenance plans and resource allocation strategies based on the prediction results;

[0095] Adaptive resource scheduling module: used to dynamically adjust the allocation of laboratory resources based on the information provided by the prediction and decision-making module;

[0096] Execution and Feedback Module: used to execute maintenance plans and resource allocation strategies, and collect execution results as feedback to further optimize the prediction and decision-making module and resource scheduling module:

[0097] A learning loop for continuous optimization of forecast accuracy and resource allocation strategies.

[0098] In one embodiment, the data collection module includes:

[0099] Sensors and monitoring equipment for real-time monitoring of operating parameters of laboratory equipment;

[0100] User behavior recorder, used to record user operation history and resource requests.

[0101] In one embodiment, the prediction and decision module includes:

[0102] The device status prediction submodule analyzes historical device operating data based on deep learning algorithms, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to predict potential future problems.

[0103] The resource demand prediction submodule is used to predict the resource demand of the laboratory in the future;

[0104] Among them, the equipment status prediction submodule can predict potential equipment failures in advance through machine learning models for preventive maintenance.

[0105] In one embodiment, the device status prediction submodule includes:

[0106] The data annotation submodule is used to add labels to historical operation data to form the training data set required for supervised learning;

[0107] The model training submodule is used to train the deep learning model using labeled historical running data;

[0108] The model evaluation submodule is used to verify the prediction accuracy of the deep learning model and make model selection or parameter adjustments accordingly.

[0109] In one embodiment, the adaptive resource scheduling module includes:

[0110] Real-time monitoring submodule, used to continuously monitor the load conditions in the laboratory;

[0111] Priority management submodule, used to ensure that critical tasks receive sufficient resources;

[0112] Load balancing submodule, used to avoid resource overload or idleness;

[0113] Among them, the adaptive resource scheduling module automatically adjusts the resource allocation strategy according to the real-time load conditions in the laboratory to ensure that each user can obtain the best service quality.

[0114] In one embodiment, the adaptive resource scheduling module includes:

[0115] The resource reservation submodule is used to reserve corresponding computing resources based on the resource demand forecast provided by the prediction and decision module;

[0116] Dynamic adjustment submodule, used to dynamically allocate resources based on actual usage within the laboratory to maximize resource utilization;

[0117] The resource recovery submodule is used to release idle resources in time after the task is completed so that they can be used by other tasks.

[0118] In one embodiment, the execution and feedback module includes:

[0119] User interface submodule, used to provide users with a visual interface of system status;

[0120] The feedback collection submodule is used to collect the execution results and input them as feedback to the prediction and decision module and the resource scheduling module.

[0121] In one embodiment, a learning loop for continuously optimizing prediction accuracy and resource allocation strategy includes:

[0122] Use forecast results to guide resource scheduling decisions;

[0123] Collect execution results as feedback;

[0124] Optimize deep learning models and resource scheduling strategies based on feedback results.

[0125] In one embodiment, the device state prediction submodule uses a convolutional neural network (CNN) and a recurrent neural network (RNN) to predict the device state, where:

[0126] CNN is used to extract spatial features in the equipment operating status, including signs of wear in equipment images; the CNN model is formalized as a function:

[0127] f CNN (X;θ CNN )

[0128] Where X represents the input image data, θ CNN Represents the parameters of the CNN model;

[0129] The CNN model is trained by minimizing the cross entropy loss function:

[0130] L CNN (f CNN (X;θ CNN ),y)

[0131] To achieve this, where y represents the target output, L CNN is the cross entropy loss function:

[0132] L CNN =-∑ i y i log(f CNN (X i θ CNN ))

[0133] The training process of the CNN model updates the parameters θ by gradient descent method CNN ;

[0134] RNN is used to capture the time series characteristics of the equipment's operating status, including the changing trend of the equipment's vibration signal. The RNN model can be formalized as a function:

[0135] f RNN (Y;θ RNN )

[0136] Where Y represents time series data, θ RNN Represents the parameters of the RNN model;

[0137] The RNN model is trained by minimizing the mean squared error loss function:

[0138] L RNN (f RNN (Y;θ RNN ),y)

[0139] To achieve this, where Y represents the target output, L RNN is the mean square error loss function:

[0140]

[0141] The training process of the RNN model can also update the parameters θ by gradient descent method RNN .

[0142] In one embodiment, the device status prediction submodule includes:

[0143] CNN model training submodule, used to train the CNN model to identify abnormal patterns in device images; the training process updates the parameters θ through the gradient descent method CNN , to minimize the cross entropy loss function:

[0144] L CNN (f CNN (X;θ CNN ),y)

[0145] The convolutional layers in the CNN model use convolution operations:

[0146]

[0147] To calculate the output, represents the output of the i,jth position of the lth convolutional layer, b (l) represents the bias term, represents the weight of the convolution kernel of the lth layer, represents the activation output of the l-1)th layer, * represents the convolution operation;

[0148] The RNN model training submodule is used to train the RNN model to identify the time correlation in the device operation data; the training process can also update the parameter θ by gradient descent. RNN , to minimize the mean square error loss function:

[0149] L RNN (f RNN (Y;θ RNN ),y)

[0150] The RNN unit in the RNN model uses the state update equation:

[0151] h t =σ(W hh h t-1 +W xh x t +b h )

[0152] To update the hidden state, where h t represents the hidden state at time step t, x t represents the input at time step t, W hh and W xh represents the weight matrix, b h represents the bias term, σ(·) represents the activation function;

[0153] The model fusion submodule is used to integrate the prediction results of CNN and RNN to improve the accuracy of device status prediction; this model fusion adopts the form of weighted average:

[0154] P=w1·fCNN (X;θ CNN )+w2·f RNN (Y;θ RNN )

[0155] Where w1 and w2 are the weight coefficients of CNN and RNN prediction results respectively, and w1+w2=1.

[0156] In one embodiment, the adaptive resource scheduling module further includes:

[0157] The priority adjustment submodule is used to dynamically adjust the task priority according to the device status and the urgency of the task; priority adjustment is achieved by defining a priority function:

[0158] P(t)=α·S(t)+β·U(t)

[0159] Where S(t) represents the device status score, U(t) represents the urgency of the task, α and β are adjustment coefficients, and α+β=1; the device status score S(t) can be calculated based on the prediction results f of the CNN and RNN models. CNN (X;θ CNN ) and f RNN (Y;θ RNN ) to calculate;

[0160] The load forecasting submodule is used to predict the resource demand and load status of the laboratory in the future. The load forecasting is done by using the ARIMA model f ARIMA (t), using historical data Z = {z1,z2,...,z T} to predict future load z T+1 ,z T+2 ,..., where z t represents the load at the t-th time point, and T represents the last known time point;

[0161] The resource reservation strategy optimization submodule is used to optimize the resource reservation strategy to balance resource utilization and user satisfaction. The resource reservation strategy optimization solves the constrained optimization problem:

[0162] min θ C(θ)stR(θ)≥r min

[0163] Where C(θ) represents the cost function, R(θ) represents the resource utilization, and r min is the minimum resource utilization threshold; during the optimization process, C(θ) and R(θ) are obtained through simulation or actual operation respectively;

[0164] Resource utilization R(θ) can be expressed as:

[0165]

[0166] Where N and M represent the number of tasks and resource units respectively, a ij Indicates whether task i is assigned to resource unit j, d i represents the demand of task i, r j Indicates the availability of resource unit j.

[0167] In one embodiment, the initial setup:

[0168] Hardware installation: Install various sensors (such as temperature sensors and vibration sensors) on key equipment and set up surveillance cameras to cover the entire laboratory area.

[0169] Software deployment: Install data collection software, data preprocessing software, prediction and decision-making software, and adaptive resource scheduling software on the lab's servers. Also, install the user interface application on lab members' computers and phones.

[0170] Data initialization: Collect basic information of the equipment (such as model, manufacturer, maintenance records, etc.) and enter it into the system database.

[0171] In one embodiment, daily operations: Equipment status monitoring:

[0172] Data collection: The system begins to regularly collect data from sensors, such as the SEM's operating temperature and vibration frequency, and obtains images of the device's appearance from the camera.

[0173] Data preprocessing: The data is automatically cleaned to remove invalid or abnormal data points, and the image is preprocessed, such as grayscale and scaling.

[0174] In one embodiment, equipment failure prediction: equipment status prediction:

[0175] Image analysis: The CNN model analyzes the SEM surface images captured by the camera to determine whether there are cracks or other anomalies.

[0176] Time series analysis: The RNN model analyzes the historical operation data of the SEM and predicts its possible future failures.

[0177] Model fusion: The results of the two models are combined to obtain the final fault prediction result.

[0178] In one embodiment, resource demand forecasting: The resource demand forecasting submodule forecasts the laboratory's demand for specific equipment in the next week based on historical resource usage.

[0179] In one embodiment, automatic resource scheduling:

[0180] Real-time monitoring: The system continuously monitors the status and load of the equipment and adjusts resource allocation based on the real-time status of the equipment and predicted results.

[0181] Priority management: When a new experimental project is submitted, the system automatically calculates the priority based on the urgency and importance of the project and adjusts resource allocation based on the priority.

[0182] Load balancing: By dynamically adjusting the workload of devices, it can prevent some devices from being overloaded while others are idle.

[0183] In one embodiment, the user interacts with:

[0184] User interface: Lab members can view equipment status, prediction results, and resource allocation through a web application or mobile application.

[0185] Task submission: Users can submit experimental tasks through the app and view the progress and estimated completion time of the tasks.

[0186] Notification system: When equipment experiences an anomaly or requires maintenance, the system will send notifications to administrators; when a task is about to be completed or is ready for the next step, reminders will also be sent to relevant users.

[0187] In one embodiment, feedback and improvements:

[0188] User feedback: Users can make suggestions on aspects such as the system’s prediction accuracy and resource allocation rationality.

[0189] Model optimization: Based on user feedback and the system's own monitoring results, we continuously optimize the CNN and RNN models, as well as the resource scheduling algorithm.

[0190] In one embodiment, the implementation has the following effects:

[0191] Reduced failure rate: By predicting potential equipment problems in advance and performing maintenance in a timely manner, the equipment failure rate is significantly reduced.

[0192] Improved resource utilization: The intelligent scheduling system enables more reasonable allocation of laboratory resources and improves overall resource utilization.

[0193] Improved experimental efficiency: Due to more reasonable resource allocation, experimental projects can be completed more quickly, which promotes the output of scientific research results.

[0194] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent detection device for computer laboratories based on the Internet of Things, characterized in that: include: Data collection module: used to collect data about equipment operating status from laboratory equipment and user activities; Data preprocessing module: used for cleaning, standardizing and feature extraction of the collected data; Prediction and decision-making module: used to predict possible future equipment problems and resource requirements, and generate maintenance plans and resource allocation strategies based on the prediction results; Adaptive resource scheduling module: used to dynamically adjust the allocation of laboratory resources based on the information provided by the prediction and decision module; Execution and feedback module: used to execute the maintenance plan and resource allocation strategy, and collect execution results as feedback to further optimize the prediction and decision module and resource scheduling module: and a learning loop for continuously optimizing prediction accuracy and resource allocation strategy.

2. The intelligent detection device for computer laboratories based on the Internet of Things according to claim 1, characterized in that: The prediction and decision module includes: A device status prediction submodule, which analyzes historical device operating data based on deep learning algorithms, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to predict potential future problems. The resource demand prediction submodule is used to predict the resource demand of the laboratory in the future; The equipment status prediction submodule can predict potential equipment failures in advance through machine learning models for preventive maintenance.

3. The intelligent detection device for computer laboratories based on the Internet of Things according to claim 2, characterized in that: The device status prediction submodule includes: The data annotation submodule is used to add labels to historical operation data and form the training data set required for supervised learning; A model training submodule, configured to train the deep learning model using labeled historical operating data; The model evaluation submodule is used to verify the prediction accuracy of the deep learning model and make model selection or parameter adjustments accordingly.

4. The intelligent detection device for computer labs based on Internet of Things according to claim 1, characterized in that: The adaptive resource scheduling module includes: Real-time monitoring submodule, used to continuously monitor the load conditions in the laboratory; Priority management submodule, used to ensure that critical tasks receive resources; Load balancing submodule, used to avoid resource overload or idleness; The adaptive resource scheduling module automatically adjusts the resource allocation strategy according to the real-time load conditions in the laboratory.

5. The intelligent detection device for computer laboratories based on the Internet of Things according to claim 4, characterized in that: The adaptive resource scheduling module includes: A resource reservation submodule, configured to reserve corresponding computing resources based on the resource demand forecast provided by the forecasting and decision-making module; Dynamic adjustment submodule, used to dynamically allocate resources based on actual usage within the laboratory; The resource recovery submodule is used to release idle resources in time after the task is completed.

6. The intelligent detection device for computer laboratories based on the Internet of Things according to claim 1, characterized in that: The execution and feedback module includes: User interface submodule, used to provide users with a visual interface of system status; The feedback collection submodule is used to collect the execution results and input them as feedback to the prediction and decision module and the resource scheduling module.

7. The intelligent detection device for computer labs based on Internet of Things according to claim 1, characterized in that: The learning loop for continuously optimizing forecast accuracy and resource allocation strategy includes: Use forecast results to guide resource scheduling decisions; Collect execution results as feedback; Optimize deep learning models and resource scheduling strategies based on feedback results.

8. The intelligent detection device for computer labs based on Internet of Things according to claim 2, characterized in that: The device state prediction submodule uses convolutional neural network (CNN) and recurrent neural network (RNN) to predict the device state, where: CNN is used to extract spatial features in the equipment operating status, including signs of wear in equipment images; the CNN model is formalized as a function: f CNN (X;θ CNN ) Where X represents the input image data, θ CNN Represents the parameters of the CNN model; The CNN model is trained by minimizing the cross entropy loss function: L CNN (f CNN (X:θ CNN ),y) To achieve this, where y represents the target output, L CNN is the cross entropy loss function: L CNN =-∑ i y i log(f CNN (X i ;θ CNN )) The training process of the CNN model updates the parameters θ by gradient descent method CNN ; RNN is used to capture the time series characteristics of the equipment's operating status, including the changing trend of the equipment's vibration signal. The RNN model is formalized as a function: f RNN (Y;θ RNN ) Where Y represents time series data, θ RNN Represents the parameters of the RNN model; The RNN model is trained by minimizing the mean squared error loss function: L RNN (f RNN (Y;θ RNN ),y) To achieve this, where Y represents the target output, L RNN is the mean square error loss function: The training process of the RNN model also updates the parameters θ by gradient descent method RNN .

9. The intelligent detection device for computer laboratories based on the Internet of Things according to claim 8, characterized in that: The device status prediction submodule includes: CNN model training submodule, used to train the CNN model to identify abnormal patterns in device images; the training process updates the parameters θ through the gradient descent method CNN , to minimize the cross entropy loss function: L CNN (f CNN (X:θ CNN ),y) The convolutional layers in the CNN model use convolution operations: To calculate the output, represents the output of the i,jth position of the lth convolutional layer, b (l) represents the bias term, represents the weight of the convolution kernel of the lth layer, represents the activation output of the l-1)th layer, * represents the convolution operation; The RNN model training submodule is used to train the RNN model to identify the time correlation in the device operation data; the training process also updates the parameter θ by gradient descent. RNN , to minimize the mean square error loss function: L RNN (f RNN (Y;θ RNN ),y) The RNN unit in the RNN model uses the state update equation: h t =σ(W hh h t-1 +W xh x t +b h ) To update the hidden state, where h t represents the hidden state at time step t, x t represents the input at time step t, W hh and W xh represents the weight matrix, b h represents the bias term, σ(·) represents the activation function; The model fusion submodule is used to integrate the prediction results of CNN and RNN to improve the accuracy of device status prediction; this model fusion adopts the form of weighted average: P=w1·f CNN (X;θ CNN )+w2·f RNN (Y;θ RNN ) Where w1 and w2 are the weight coefficients of CNN and RNN prediction results respectively, and w1+w2=1.

10. The intelligent detection device for computer labs based on Internet of Things according to claim 4, characterized in that: The adaptive resource scheduling module also includes: The priority adjustment submodule is used to dynamically adjust the task priority according to the device status and the urgency of the task; priority adjustment is achieved by defining a priority function: P(t)=α·S(t)+β·U(t) Where S(t) represents the device status score, U(t) represents the urgency of the task, α and β are adjustment coefficients, and α+β=1; the device status score S(t) is calculated based on the prediction results f of the CNN and RNN models. CNN (X;θ CNN ) and f RNN (Y;θ RNN ) to calculate; The load forecasting submodule is used to predict the resource demand and load status of the laboratory in the future. The load forecasting is done by using the ARIMA model f ARIMA (t), using historical data Z = {z1,z2,...,z T } to predict future load z T+1 ,z T+2 ,..., where z t represents the load at the t-th time point, and T represents the last known time point; The resource reservation strategy optimization submodule is used to optimize the resource reservation strategy to balance resource utilization and user satisfaction. The resource reservation strategy optimization solves the constrained optimization problem: minutes θ C(θ)stR(θ)≥r min Where C(θ) represents the cost function, R(θ) represents the resource utilization, and r min is the minimum resource utilization threshold; during the optimization process, C(θ) and R(θ) are obtained through simulation or actual operation respectively; Resource utilization R(θ) is expressed as: Where N and M represent the number of tasks and resource units respectively, a ij Indicates whether task i is assigned to resource unit j, d i represents the demand of task i, r j Indicates the availability of resource unit j.

Images